diff --git a/.gitattributes b/.gitattributes index bed0738c7eeb449bca98b5d2f33c89a1ee56349a..b6d5e013e46c62737adea1c3d96dae771169bd77 100644 --- a/.gitattributes +++ b/.gitattributes @@ -58,3 +58,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text # Video files - compressed *.mp4 filter=lfs diff=lfs merge=lfs -text *.webm filter=lfs diff=lfs merge=lfs -text +repro_bundle/probeshift_honest.pdf filter=lfs diff=lfs merge=lfs -text +repro_bundle/probeshift_main.pdf filter=lfs diff=lfs merge=lfs -text diff --git a/repro_bundle/CODE_SPEC.md b/repro_bundle/CODE_SPEC.md new file mode 100644 index 0000000000000000000000000000000000000000..f0ca1224ff7d18ac248a29e155f2313739ff0cca --- /dev/null +++ b/repro_bundle/CODE_SPEC.md @@ -0,0 +1,428 @@ +这是一个工程架构设计任务,不是 ADR 决策。我直接输出完整的实验代码架构 spec。下面是为《Probe Stability Score》设计的、可在单卡 4090 上运行的完整工程化 spec。 + +--- + +# Probe Stability Score · 实验代码架构 Spec + +设计目标:一次性激活抽取 + 缓存复用,所有探针/指标计算只读缓存;严控显存与 GPU·h(≤30 GPU·h);单条命令复现最小可行实验(MVE)。 + +--- + +## 1. 目录 / 文件树 + +``` +probe_stability/ +├── README.md +├── pyproject.toml # 依赖 + 入口 console_scripts +├── Makefile # make mve / make full / make figures +├── configs/ +│ ├── base.yaml # 全局:路径、dtype、seed 列表、CI 参数 +│ ├── models.yaml # 7 个模型的 HF id、层数、d_model、最佳 batch +│ ├── datasets.yaml # 8 个任务:HF id、字段映射、label 空间、采样数 +│ ├── shifts.yaml # back-translation / domain / length 三类 shift 配置 +│ ├── probes.yaml # logreg / mass_mean / mlp / control 超参 +│ ├── metrics.yaml # Score 双分量权重、bootstrap、多重检验方法 +│ └── experiments/ +│ ├── mve.yaml # 最小可行:pythia-70m + SST-2 + 1 shift + 2 seed +│ └── full.yaml # 全量矩阵 +├── env/ +│ └── download_models.py # 预下载所有 HF 权重到本地缓存(离线复现) +├── src/probe_stability/ +│ ├── __init__.py +│ ├── cli.py # Typer/argparse 主入口,子命令路由 +│ ├── config.py # OmegaConf 加载 + dataclass schema 校验 +│ ├── registry.py # MODEL/DATASET/PROBE/SHIFT 注册表(字符串→构造器) +│ ├── utils/ +│ │ ├── seeding.py # set_all_seeds(seed) +│ │ ├── logging.py # rich/structlog,JSONL run log +│ │ ├── io.py # shard 路径生成、原子写、manifest 读写 +│ │ ├── gpu.py # 显存监控、autocast、空缓存、OOM 自动降 batch +│ │ └── hashing.py # config/input 指纹 → 缓存 key、幂等跳过 +│ ├── data/ +│ │ ├── loaders.py # 8 个数据集统一加载为 Example schema +│ │ ├── schema.py # Example / DistributionSet dataclass +│ │ ├── sampling.py # 分层下采样、固定 split、length buckets 切分 +│ │ └── shifts/ +│ │ ├── base.py # Shift 抽象基类 +│ │ ├── backtranslation.py # NLLB / opus-mt 本地往返翻译 +│ │ ├── domain.py # Amazon multi-domain / IMDB↔SST-2 域迁移 +│ │ └── length.py # 短/中/长 token 分桶 +│ ├── fidelity/ +│ │ └── mnli_filter.py # 本地 DeBERTa-v3-MNLI 过滤 label flip +│ ├── extract/ +│ │ ├── activations.py # 核心:逐层 residual-stream 抽取 + 流式落盘 +│ │ ├── hooks.py # 各架构 residual hook 适配(Pythia/GPT2/Qwen2) +│ │ └── pooling.py # last-token / mean pooling +│ ├── cache/ +│ │ ├── store.py # ShardStore:read/write float16 分片 + manifest +│ │ └── manifest.py # 全局缓存索引(parquet),完整性校验 +│ ├── probes/ +│ │ ├── base.py # Probe 抽象:fit/predict/direction +│ │ ├── logreg.py # sklearn / torch LogReg +│ │ ├── mass_mean.py # difference-of-means 方向探针 +│ │ ├── mlp.py # 2-layer MLP (torch) +│ │ └── control.py # random-label control-task 包装器 +│ ├── eval/ +│ │ ├── train_eval.py # 读缓存→训练→IID/OOD 评估 单元 +│ │ ├── directions.py # 概念方向、跨分布 cosine 旋转 +│ │ └── selectivity.py # selectivity = acc(real) - acc(control) +│ ├── score/ +│ │ └── stability.py # 双分量 Probe Stability Score 计算 +│ ├── stats/ +│ │ ├── bootstrap.py # bootstrap CI +│ │ ├── correlation.py # Spearman(Score ↔ OOD drop) +│ │ ├── effect_size.py # Cliff's delta +│ │ └── multiple_testing.py # Holm / Benjamini-Hochberg +│ └── pipeline/ +│ ├── stage_extract.py # 阶段1:激活抽取(写缓存) +│ ├── stage_probe.py # 阶段2:探针训练+评估(读缓存) +│ ├── stage_score.py # 阶段3:Score 聚合 +│ └── stage_stats.py # 阶段4:统计检验 + 表格 +├── scripts/ +│ ├── run_mve.sh # 一条命令复现 MVE +│ ├── run_full.sh +│ └── make_figures.py # 论文图(cosine 旋转、Score↔drop 散点) +├── tests/ +│ ├── test_cache_roundtrip.py +│ ├── test_shapes.py +│ ├── test_probes_smoke.py +│ └── test_score_math.py +└── artifacts/ # 运行产物(git-ignored) + ├── cache/ # 激活分片 + ├── results/ # parquet 结果表 + ├── figures/ + └── logs/ +``` + +--- + +## 2. 各模块职责 + 关键函数签名 + +### 2.1 config / registry + +```python +# config.py +@dataclass +class RunConfig: + models: list[str]; datasets: list[str]; shifts: list[str] + probes: list[str]; seeds: list[int] + cache_dir: Path; results_dir: Path + pooling: str = "last_token" # last_token | mean + dtype: str = "float16" + max_examples_per_dist: int = 4000 # 控制规模/显存 + extract_batch: dict[str, int] = ... # 每模型 batch 覆盖 + +def load_config(exp_path: str, overrides: list[str]) -> RunConfig: ... + +# registry.py +def register(kind: str, name: str): ... # 装饰器 +def build_model(name: str) -> "LMWrapper": ... +def build_dataset(name: str) -> "DistributionSet": ... +def build_probe(name: str, **kw) -> "Probe": ... +def build_shift(name: str) -> "Shift": ... +``` + +### 2.2 data + +```python +# schema.py +@dataclass(frozen=True) +class Example: + uid: str; text: str; label: int; meta: dict # meta 含 domain/length_bucket + +@dataclass +class DistributionSet: + name: str # e.g. "sst2" + distribution: str # "iid" | "bt_de" | "domain_books" | "len_long" + examples: list[Example] + num_labels: int + +# loaders.py +def load_dataset(name: str, split: str, max_n: int, seed: int) -> DistributionSet: ... +# 支持: sst2 imdb amazon_multi ag_news dbpedia ud_pos truthfulqa_bin counterfact_bin + +# sampling.py +def stratified_subsample(ds: DistributionSet, n: int, seed: int) -> DistributionSet: ... +def length_buckets(ds: DistributionSet, tokenizer, edges=(32,128)) -> dict[str, DistributionSet]: ... + +# shifts/base.py +class Shift(ABC): + name: str + @abstractmethod + def apply(self, ds: DistributionSet) -> DistributionSet: ... + # 产出 distribution 字段已改写、label 暂保留(待 fidelity 过滤) + +# shifts/backtranslation.py +class BackTranslationShift(Shift): + def __init__(self, pivot="deu_Latn", model="facebook/nllb-200-distilled-600M", + batch_size=32, device="cuda"): ... + def apply(self, ds) -> DistributionSet: ... # text→pivot→en,流式 batch +``` + +### 2.3 fidelity + +```python +# mnli_filter.py +class MNLIFidelityFilter: + def __init__(self, model="MoritzLaurer/DeBERTa-v3-base-mnli", thresh=0.5): ... + def keep_mask(self, src_texts: list[str], shifted_texts: list[str]) -> np.ndarray: + """双向蕴含判定:保留语义未翻转(label-preserving)样本,返回 bool mask""" + def filter(self, src: DistributionSet, shifted: DistributionSet) -> DistributionSet: ... +``` + +### 2.4 extract(核心) + +```python +# hooks.py +def residual_hook_points(model, arch: str) -> list[tuple[int, nn.Module]]: + """返回 [(layer_idx, module)],module 输出即该层 residual-stream。 + Pythia(GPTNeoX): gpt_neox.layers[i]; GPT2: transformer.h[i]; Qwen2: model.layers[i]""" + +# pooling.py +def pool(hidden: Tensor, attn_mask: Tensor, mode: str) -> Tensor: + """(B,T,D)->(B,D);last_token 取最后非 pad 位,mean 做 mask 平均""" + +# activations.py +class ActivationExtractor: + def __init__(self, model_name: str, dtype="float16", device="cuda", + pooling="last_token", max_len=256): ... + + @torch.inference_mode() + def extract(self, ds: DistributionSet, store: "ShardStore", + batch_size: int, layers: list[int] | None = None) -> ShardMeta: + """单次前向抓全层 residual,逐 batch pool→cast fp16→流式 append 落盘。 + 返回写入分片的元信息(n, layers, d_model)。""" +``` + +抽取核心逻辑:注册 forward hook 抓全部层 → 一次前向 → 每层 pool → fp16 → 立即追加写盘,**不在显存/内存累积全量激活**。 + +### 2.5 cache + +```python +# store.py 分片单位 = (model, dataset, distribution);层维进 .npz 同文件多 key 或单层一文件 +class ShardStore: + def __init__(self, root: Path, dtype="float16"): ... + def shard_path(self, model, dataset, distribution, layer, seed_split="all") -> Path: ... + def open_writer(self, model, dataset, distribution, n_layers, d_model, n: int): ... + def append(self, layer_arrays: dict[int, np.ndarray], labels, uids): ... # 流式 + def finalize(self): ... # 写 X.npy / y.npy / uids + def load(self, model, dataset, distribution, layer) -> tuple[np.ndarray, np.ndarray, list[str]]: + """mmap 读取 X(fp16), y(int8), uids""" + def exists(self, model, dataset, distribution, layer) -> bool: ... # 幂等跳过 + +# manifest.py +def update_manifest(root: Path, meta: dict): ... # 追加到 manifest.parquet +def verify_cache(root: Path) -> list[str]: ... # 返回缺失/损坏 shard +``` + +### 2.6 probes + +```python +# base.py +class Probe(ABC): + @abstractmethod + def fit(self, X: np.ndarray, y: np.ndarray) -> "Probe": ... + @abstractmethod + def predict(self, X: np.ndarray) -> np.ndarray: ... + @abstractmethod + def direction(self) -> np.ndarray | None: # 概念方向(单位向量),MLP 返回 None 或一阶近似 + ... + +# logreg.py LogRegProbe(C=1.0, max_iter=1000) —— 内部 X 升精到 float32 训练 +# mass_mean.py MassMeanProbe —— direction = (μ_pos - μ_neg)/||·||;阈值=中点投影 +# mlp.py MLPProbe(hidden=256, epochs=30, lr=1e-3, device="cuda") —— direction=输入层雅可比近似 +# control.py +class ControlTaskProbe(Probe): + def __init__(self, inner: Probe, seed: int): + """random-label control:固定随机标签置换后训练同结构探针""" +``` + +### 2.7 eval / score / stats + +```python +# train_eval.py +def fit_and_eval(probe_name, store, model, dataset, + train_dist="iid", eval_dists=("iid","bt_de",...), + seed=0) -> ProbeResult: + """读 train_dist 缓存→fit→在各 eval_dist 上算 acc;返回 acc 字典 + 方向向量""" + +# directions.py +def cosine_rotation(dir_iid: np.ndarray, dir_ood: np.ndarray) -> float: ... # 1 - cos +def aggregate_directions(results: list[ProbeResult], by="layer") -> pd.DataFrame: ... + +# selectivity.py +def selectivity(real_acc: float, control_acc: float) -> float: ... + +# score/stability.py +def probe_stability_score(acc_iid, acc_ood, dir_iid, dir_ood, + w_perf=0.5, w_geom=0.5) -> dict: + """双分量: + perf_component = acc_ood / acc_iid (性能保持度,∈[0,1]) + geom_component = cos(dir_iid, dir_ood) (方向稳定度,∈[-1,1]→clip[0,1]) + score = w_perf*perf + w_geom*geom + 返回 {score, perf_component, geom_component}""" + +# stats/ +def bootstrap_ci(values, fn=np.mean, n=1000, alpha=0.05, seed=0) -> tuple[float,float,float]: ... +def spearman_with_ci(score, ood_drop, n_boot=1000) -> dict: ... # rho + CI + p +def cliffs_delta(a, b) -> float: ... +def holm_bonferroni(pvals) -> np.ndarray: ... +def benjamini_hochberg(pvals, q=0.05) -> np.ndarray: ... +``` + +### 2.8 pipeline + +```python +# stage_extract.py +def run_extract(cfg: RunConfig): + for model in cfg.models: + ext = ActivationExtractor(model, ...) + for ds_name in cfg.datasets: + base = load_dataset(ds_name, "test", cfg.max_examples_per_dist, seed=0) + dists = [base] + [filter_fidelity(s.apply(base), base) for s in shifts] + for d in dists: + if store.exists(model, ds_name, d.distribution, layer=0): continue # 幂等 + ext.extract(d, store, batch_size=cfg.extract_batch[model]) + del ext; free_gpu() # 关键:释放当前模型再载下一个 + +# stage_probe.py / stage_score.py / stage_stats.py 全程只读缓存,CPU/小 GPU 即可 +``` + +--- + +## 3. 数据流与缓存格式 + +### 3.1 数据流 + +``` +原始数据集 ──load_dataset──► DistributionSet(iid) + │ + shift.apply ──┼──► bt_de / domain_x / len_* (候选) + │ + MNLI fidelity 过滤 ─┘──► label-preserving DistributionSet + │ + ActivationExtractor(单模型常驻)│ 一次前向抓全层 → pool → fp16 → 流式 + ▼ + ShardStore 落盘 (阶段1结束) + │ (此后 GPU 几乎闲置) + fit_and_eval(读缓存) ──────┼──► ProbeResult{acc[dist], direction} + ▼ + stability.py ──► Score(双分量) + ▼ + stats ──► Spearman/CI/Holm-BH/Cliff's δ ──► results/*.parquet ──► figures +``` + +### 3.2 缓存目录与 shard 命名 + +分片粒度:`(model, dataset, distribution)` 一组目录,层维拆文件,便于按层流式与按需读取。 + +``` +artifacts/cache/ +└── {model}/{dataset}/{distribution}/ + ├── X_layer{LL}.npy # 每层一个,float16,shape (N, d_model) + ├── y.npy # int8, shape (N,) + ├── uids.npy # 把论文重心从"发现探针不稳 + 比较探针类型"(已被占)收窄到: +> **"一个针对 linear/concept probe 方向几何、完全免新标注、仅 forward-pass 的双分量(dispersion ⊕ augmentation-consistency)合成分数,首次在训练前 a priori 预测探针在 label-preserving 语义偏移(paraphrase/domain/length)下的 OOD 迁移性,并通过 shift-type × model-size 矩阵 + 与 SIP/Fragility/Dispersion-Score/mass-mean 的 head-to-head baseline 对比,证明该分数提供超越单分量与现有几何/谱预测器的增量。"** + +护城河三要素,缺一即被判 incremental: +1. **双分量合成 + 互补性消融**(无人合做两分量); +2. **a priori 预测 ↔ 验证闭环**(用 PSS 预测 C1 矩阵真实 drop;Dies/RAPTOR/SIP 都停在诊断或 ID 判据); +3. **完全免触碰 OOD 数据/新标注**(对比 Truthfulness Spectrum 的 OOD reference probe 需求)。 +将 C1 降为"被预测的真值",C3/C4 降为外部效度,C5 降为 sanity check。 + +--- + +## 4. 必引清单(must-cite,漏引会被直接攻击) + +**绝对必引(high-threat,逐篇做"我们 vs 他们"切割):** +1. Dies, Maynard, Savcisens & Eliassi-Rad 2025, *Representational Stability of Truth in LLMs*(2511.19166)——最近邻,rotation+dispersion 已合做。 +2. Gao et al. 2026, *RAPTOR: Ridge-Adaptive Logistic Probes*(2602.00158)——dispersion 分量 + size ladder。 +3. Lysnæs-Larsen, Eggen & Strümke 2025, *Probing the Probes*(2511.04312)——augmentation-robustness 分量。 +4. Huang 2025, *Spectral Identifiability for Interpretable Probe Geometry*(2511.16288)——a priori 探针稳定性判据。 +5. Reblitz-Richardson 2026, *When Probing Accuracy Saturates, Fragility Resolves*(2606.11375)——forward-pass 崩溃点。 +6. Marks & Tegmark 2023/24, *The Geometry of Truth*(2310.06824)——mass-mean vs LogReg(C3 基石)。 +7. Xie et al. 2023, *Dispersion Score*(NeurIPS, 2303.15488)——同名同质 OOD 预测器。 +8. Liang et al. 2025, *LLM Knowledge is Brittle*(2510.11905)——C1 三类 shift 逐字一致。 +9. Truthfulness Spectrum 2026(2602.20273)——白化方向相似度预测 OOD(C2 预测目标最近邻)。 +10. Chou et al. 2026, *Diagnosing Generalization Failures from Geometry Markers*(ICLR 2026 / 2603.01879)——ID-only 预测 OOD 范式。 + +**必引背景(定位用):** Hewitt & Liang 2019;Belinkov 2022;Kumar/Ravfogel et al. 2022(NeurIPS);Voita & Titov 2020;Azizian et al. 2025(2506.08572);Bürger 2024(2404.18865)/ Bao 2025(2506.00823);LEEP 2020 / LogME 2021;Deng et al. 2023(Confidence & Dispersity);Baek et al. 2024(Agreement-on-the-Line);Zhang 2026 SAE Crystal Ball;Biderman et al. 2023(Pythia)+ 2505.21399;NLI-filter 系列(C5)。 + +> 注:所有 2025–2026 编号(2511.x / 2602.x / 2603.x / 2606.x / 2602.20273 / 2605.27958 等)venue 与作者多为"待核实",投稿前必须在 arXiv/OpenReview 逐条复核标题、作者、是否已被 ACML/NeurIPS/ICLR/ACL 接收——prior-art 效力取决于此。 + +--- + +## 5. 必比 baseline 清单(must-have baselines,不比会被质疑) + +**A. a priori 探针稳定性 / 可靠性预测器(同定位,直接 head-to-head,核心)** +1. **SIP / Spectral Identifiability**(eigengap 几何判据)— 必比,否则"首个 a priori 预测器"主张崩塌。 +2. **RAPTOR directional stability**(K 次 ablation mean |cos|)— 即单独的 dispersion 分量,用作消融对照。 +3. **Fragility**(临界噪声崩溃点)— forward-pass 稳定性对照。 +4. **Augmentation-robustness**(Probing the Probes)— 即单独的 augmentation-consistency 分量,用作消融对照。 + +**B. 通用免标注 OOD 性能预测器(证明探针方向几何 ≠ 输出/流形几何)** +5. **Dispersion Score**(Xie 2023)— 同名,必跑作 baseline,否则被判换皮。 +6. **Confidence & Dispersity**(Deng 2023)/ **Agreement-on-the-Line**(Baek 2024)— 至少跑其一。 +7. **Geometry Markers**(effective manifold dim / utility)— ID-only 几何预测对照。 + +**C. 探针估计子(C3 内部对照)** +8. **mass-mean / difference-of-means**(Marks & Tegmark)— 必含。 +9. **LogReg** 与 **MLP** 探针 — C3 三方比较的另两方。 + +**D. 方法学对照(C1/C5)** +10. **control-task / selectivity probe**(Hewitt & Liang)— 作为"已知探针诊断"的对照基线,证明 PSS 提供 selectivity 之外的信息。 +11. **普通 cosine vs Mahalanobis-白化 cosine**(Truthfulness Spectrum)— 方向相似度度量的对照,回应"普通 cosine 不够"的已知结论。 + +**消融(内部 must-have):** dispersion-only / augmentation-consistency-only / 双分量合成 三档,证明互补增量;以及"PSS 预测的 OOD drop vs 实际 OOD drop"的相关性/排序能力(Spearman ρ、Kendall τ)作为闭环验证主结果。 + +--- + +**一句话结论:** 现状是 crowded;唯一安全的护城河是把 C2 收窄成"双分量合成 + a priori 预测↔验证闭环 + 完全免 OOD 标注 + 针对 probe 方向几何",并以 SIP / Fragility / RAPTOR-stability / Dispersion-Score / Augmentation-robustness 五个 baseline 的 head-to-head 增量来支撑;C1 降为被预测真值,C3/C4 降为外部效度,C5 降为 sanity check。**关键风险点:Dies 2025(2511.19166)必须拿到全文确认它是否已做"预测 OOD"闭环——若已做,C2 护城河将进一步收窄,需立即二次检索核实。** + + +## 二、Novelty 红队(最强攻击 + 必需防守) + +# ACML 苛刻 Reviewer 意见:Reject(Probe Stability Score,PSS) + +我已通读定位文档。下面分两档给出我作为对抗性 reviewer 准备的攻击,每条都点名撞车对象/设计漏洞,并配上作者"活下来"所需的最小防守动作。 + +--- + +## 一、致命级攻击(任一条不解决即建议 Reject) + +### F1. 与 Dies et al. 2025(2511.19166)的"组合新颖性"被整体抹平 +**攻击点:** 作者的全部卖点是"rotation + dispersion 双视角"。Dies 2025 已经把 representational stability of truth 用 rotation + dispersion 合做了。作者文档自己承认"这是最危险的单篇"。如果 Dies 已经做了"用 ID 几何预测 OOD 行为"的闭环,那 C2 护城河的(1)(2)两要素同时塌掉,PSS 只剩"换了个加权求和的公式 + 多了几个概念",这是教科书级的 incremental。 +**这撞了什么:** 直接撞 Dies 2025 的方法核心,且作者尚未确认对方是否已做预测闭环。 +**必需防守:** +- 拿到 Dies 2025 全文,做一张**逐能力对照表**(rotation? dispersion? 二者合成单一分数? a priori 预测 OOD? 免 OOD 数据? 跨多概念? size ladder?),逐格标 Y/N。 +- 必须证明至少一格是"他们 N、我们 Y"且该格是**预测闭环或合成分数**,而非边角(概念数量、模型数量这类不算)。 +- 在同一 benchmark 上**复现 Dies 作为 baseline 并跑输/跑平**——若跑不赢,论文死。 + +### F2. "首个 a priori 预测器"主张直接被 SIP / Fragility / Truthfulness Spectrum 证伪 +**攻击点:** 文档同时列了 SIP(2511.16288,部署前 a priori 几何判据)、Fragility(2606.11375,forward-pass 崩溃点)、Truthfulness Spectrum(2602.20273,白化方向相似度预测 OOD AUROC,R²≈0.98)。"训练前/免 OOD 预测探针稳定性"这一 framing 已被至少三方占据。任何"首次"措辞都可被一句话击穿。 +**这撞了什么:** SIP + Fragility + Truthfulness Spectrum,三方夹击 framing 新颖性。 +**必需防守:** +- **删除一切"first / novel framing"措辞**,改为"a label-free composite that empirically outperforms SIP / Fragility / Dispersion-Score on probe-direction OOD prediction"。 +- 三者**全部作为 head-to-head baseline 跑数**,主结果用 Spearman ρ / Kendall τ 报告 PSS 的增量,且增量需带置信区间/显著性检验(bootstrap)。 +- 特别针对 Truthfulness Spectrum 的 R²≈0.98:若对方已能 R²=0.98 预测 OOD,作者必须说明 PSS 在"完全不碰 OOD reference probe"约束下还能逼近此精度,否则"免 OOD 数据"这条护城河不值钱。 + +### F3. 与 Xie 2023 Dispersion Score 同名同质,涉嫌"换皮" +**攻击点:** Xie 2023(NeurIPS)已有 label-free + forward-pass + feature dispersion 预测 OOD acc,名字都一样。reviewer 第一反应是"这就是把 Dispersion Score 套到 probe 方向上"。 +**这撞了什么:** Xie 2023,命名与机制双重撞车。 +**必需防守:** +- **改名**,避免 "Dispersion" 作主词根,杜绝第一眼误判。 +- 用实验证明"probe 方向几何 ≠ feature 流形 dispersion":在同一数据上跑 Xie 的 Dispersion Score,展示二者预测目标/相关性显著分离(例如 Dispersion Score 在 paraphrase shift 上失效而 PSS 不失效)。 + +### F4. 预测目标循环论证 / 闭环自证(实验设计漏洞) +**攻击点:** "用 PSS 预测 C1 矩阵里的真实 drop"——但 C1 矩阵的 shift、概念、size 配置若与构造 PSS / 调权重所用的是同一批数据,则是在**训练集上验证**,Spearman ρ 没有意义。双分量"合成"必然涉及一个加权/聚合超参,若该超参在评测集上选,闭环就是过拟合。 +**这撞了什么:** 实验设计的数据泄漏 / 无 held-out。 +**必需防守:** +- 严格 **split**:权重/聚合超参只在 dev 概念-shift 组合上选,报告结果只在**完全 held-out 的概念 × shift-type × size 组合**上算。 +- 提供 **leave-one-concept-out / leave-one-shift-type-out** 交叉验证,证明 PSS 跨概念、跨 shift-type 泛化,而非记忆 C1 矩阵。 + +### F5. C3 几乎是 Marks & Tegmark 2023 的复现 +**攻击点:** "mass-mean vs LogReg 在 OOD 下谁更稳"结论已发表。加一个 MLP、换 label-preserving shift,不足以构成独立贡献。若作者把 C3 列为并列主贡献,reviewer 直接判 partial reproduction。 +**这撞了什么:** Marks & Tegmark 2023。 +**必需防守:** 按文档自己的结论,**C3 降格为支撑性发现/外部效度**,且只在"PSS 能否预测三类 estimator 谁更稳"这一从属问题下出现,绝不并列。 + +--- + +## 二、重要级攻击(不致命但累加可压到 Reject) + +### M1. 互补性消融若不显著,双分量合成失去存在理由 +**攻击点:** PSS 的核心增量主张是"aug-consistency 在 dispersion 之外有增量"。若 dispersion-only 与双分量合成的 ρ 无显著差异,则合成是噱头。 +**必需防守:** 三档消融(dispersion-only / aug-only / 合成)必须报告**配对显著性检验 + 效应量**,且合成显著优于两个单分量。否则坦诚降级为单分量方法。 + +### M2. C4(size ladder)无独立贡献,且有反例风险 +**攻击点:** "越大越稳/越一致"已被多篇占;且存在"越大越脆"反例(Pressure-Testing Deception 2605.27958)。若作者下"越大越稳"结论而不处理反例,会被打成 cherry-picking。 +**必需防守:** C4 仅作交叉维度;明确测试并讨论"PSS 预测有效性是否随 size 单调",主动引用并复现/反驳反例,不下普适性结论。 + +### M3. 度量选择(普通 cosine)已被已知结论否定 +**攻击点:** Park 2024 LRH 用 cosine 有前提;Truthfulness Spectrum 已表明需 Mahalanobis 白化。用裸 cosine 度量 rotation 会被质疑度量不当。 +**必需防守:** 加入**白化 cosine vs 裸 cosine** 的消融,说明 PSS 的稳健性不依赖度量取巧。 + +### M4. baseline 覆盖不足即被判"未与最近邻比较" +**攻击点:** 文档列出的 must-have baseline(SIP / RAPTOR-stability / Fragility / Aug-robustness / Dispersion-Score / mass-mean / Geometry Markers / control-task)若缺任一核心项,reviewer 可直接以"未与显然相关工作比较"拒稿。 +**必需防守:** 至少 A 组四个 a priori 预测器(SIP / RAPTOR-stability / Fragility / Aug-robustness)+ Xie Dispersion Score **全跑**;B/C 组至少各跑一个。缺项需在 limitation 明确说明理由。 + +### M5. shift 的 label-preserving 保证仅靠单个 DeBERTa-MNLI,可信度不足 +**攻击点:** C5 用本地 DeBERTa-MNLI 审计 label fidelity。单模型审计、未引 overgenerate-and-filter / roundtrip(Falsesum 2022、DISCO 2023)文献,会被质疑审计本身不可靠且不知情。 +**必需防守:** C5 明确降为 sanity check;主动引用 NLI-filter 系列;报告 MNLI 审计的**通过率与人工抽检一致性**,证明 shift 确为 label-preserving。 + +### M6. 跨概念泛化样本量与统计功效存疑 +**攻击点:** 若"多概念"只有 3–5 个概念,Spearman ρ 的置信区间会很宽,"跨多概念优于 truth 单概念"的卖点站不住。 +**必需防守:** 概念数足够大(建议 ≥10),报告 ρ 的 bootstrap CI;若概念少,撤回"跨多概念"作为新颖性主张。 + +### M7. prior-art 编号未核实,存在 prior-art 效力风险(双刃) +**攻击点:** 大量 2025–2026 arXiv 编号"待核实"。reviewer 若发现作者**未引用**已正式接收的某篇最近邻(如 Dies/RAPTOR 若已中 NeurIPS/ICLR),漏引即攻击点;反之若作者拿"未发表 arXiv"当挡箭牌排除 prior art,也会被驳。 +**必需防守:** 投稿前逐条核实标题/作者/venue/接收状态;对所有 high-threat 篇目无论是否正式发表均做切割讨论,不以"preprint"为由回避。 + +--- + +## 三、给作者的一句话总结(survival path) + +**当前状态 crowded,唯一活路:** 删掉所有"first/novel framing"措辞,把贡献彻底收窄为"label-free 双分量合成分数 + 严格 held-out 的 a priori 预测↔验证闭环 + 完全免 OOD 数据约束下对 SIP/Fragility/Aug-robustness/Dispersion-Score 五个 baseline 的可显著性检验的增量",并以 **F1(Dies 全文对照表跑赢)+ F4(无泄漏 split)+ M1(互补性显著)** 三者同时成立为生死线。C1 降为被预测真值,C3/C4 降为外部效度,C5 降为 sanity check。三条生死线缺一,我维持 Reject。 + + +## 三、原始检索明细(10 角度) + + +### 角度: 探针方法学与 control task(linear probing / diagnostic classifiers / selectivity / cont... + +- **Hewitt & Liang, 2019** — *Designing and Interpreting Probes with Control Tasks* (EMNLP 2019) [威胁:medium] + - 概念上撞 C1 的核心论点(探针准确率会高估表征中真正编码的信息)与 C2/C3 的探针选择哲学。但它用『随机标签控制 + selectivity』这一轴,而非『label-preserving 输入分布偏移 + 方向 cosine 旋转 + dispersion/augmentation-consistency』。是我们必须正面区分的奠基工作,但范式不同。 +- **Belinkov, 2022** — *Probing Classifiers: Promises, Shortcomings, and Advances* (Computational Linguistics (MIT Press), 48(1)) [威胁:low] + - 背景层面覆盖 C1/C2/C5 所处问题空间(探针可靠性诊断),但不提供我们提议的具体度量(分布偏移下的方向旋转矩阵、免标注稳定性分数)。是定位 novelty 的必引背景,本身不抢占。 +- **Pimentel et al., 2020** — *Information-Theoretic Probing for Linguistic Structure* (ACL 2020) [威胁:low] + - 与 C1『准确率作为表征质量代理不可靠』同源动机,但走信息论方向,不涉及分布偏移、方向稳定性或免标注预测。背景相关。 +- **Marks & Tegmark, 2023/2024** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (COLM 2024(arXiv 2310.06824)) [威胁:high] + - 直接、强烈撞 C3:mass-mean vs LogReg 在 OOD 下相对可迁移性的系统比较正是该文核心结果之一。C3 若不显著扩展(如加入 MLP、跨 shift type × model size 的系统矩阵、把比较绑定到我们的稳定性分数上)将被认为已被做过。 +- **Anonymous / 待核实, 2026** — *Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations* (arXiv 2605.27958(2026,预印本,待核实正式 venue)) [威胁:high] + - 高度撞 C1(探针准确率在 label-preserving 风格偏移下崩溃,高估稳定性)与 C4(model size ladder 对探针 OOD 稳定性的影响,且给出反直觉结论)。也部分触及 C3(线性 vs MLP)。区别:它未度量概念方向的 cosine 旋转,未提免标注 a-priori 稳定性分数,且把 augmentation 当作『修复手段』而非我们 C2 中的『预测信号』。这是本角度对 C1+C4 威胁最大的近期工作。 +- **待核实, 2025** — *Calibrating LLM Confidence by Probing Perturbed Representation Stability (CCPS)* (arXiv 2505.21772(2025,预印本)) [威胁:medium] + - 撞 C2 的『表征/方向稳定性作为可靠性信号』直觉,且同样是 forward-pass 轻量方案。但关键差异:扰动作用于隐状态(非 label-preserving 输入增广/paraphrase),目标是 per-answer 置信度校准而非『训练前预测探针 OOD 迁移性』,也无 dispersion(重采样方向离散度)分量。相关但目标错位。 +- **待核实, 2025** — *Probing the Geometry of Truth: Consistency and Generalization of Truth Directions in LLMs Across Logical Transformations and QA Tasks* (ACL Findings 2025(arXiv 2506.00823)) [威胁:medium] + - 撞 C1 的方向一致性部分:它确实在『变换后概念方向是否保持』这一轴上做了实证。差异:变换是逻辑/任务层面而非 paraphrase/domain/length 的 label-preserving 表层偏移;未给出 accuracy drop × cosine 旋转 的跨 shift×scale 矩阵,也未提免标注稳定性分数。是 C1 方向旋转主张需正面区分的工作。 +- **Ravichander et al. / 待核实, 2022** — *Probing Classifiers are Unreliable for Concept Removal and Detection* (NeurIPS 2022) [威胁:low] + - 概念上支撑 C1 的批判性前提(探针准确率高估表征中概念的真实存在/稳定性),但聚焦因果性与 concept removal,不涉及分布偏移下的方向稳定性度量或免标注预测分数。背景相关。 +- **待核实, 2026** — *Diagnosing Generalization Failures from Representational Geometry Markers* (arXiv 2603.01879(2026,预印本)) [威胁:medium] + - 强烈撞 C2 的方法论框架(免 OOD 标注、用 ID 侧量化指标 a-priori 预测 OOD 迁移性)。差异:用 manifold 几何指标而非探针方向 dispersion/augmentation-consistency,且为视觉模型、非语言探针、不比较探针类型。是 C2『免标注预测 OOD 迁移性』这一卖点的概念竞品,需说明我们的双分量方向指标与之正交。 +- **待核实, 2022** — *Predicting Fine-Tuning Performance with Probing* (EMNLP 2022(arXiv 2210.07352,待核实)) [威胁:low] + - 与 C2『用探针侧信号预测下游/迁移表现』同向,但预测目标是 fine-tuning 性能而非探针自身 OOD 稳定性,且不涉及方向 dispersion/augmentation-consistency。弱相关。 + +> 评估: 已联网检索。本角度的结论:你们的五条贡献没有被任何单篇完整抢占,但每条都已被『部分前置』,且有两个高危撞车点。 + +最大撞车点(高危): +1) C3(mass-mean vs LogReg 的 OOD 相对可迁移性比较)几乎被 Marks & Tegmark, The Geometry of Truth(COLM 2024)直接做过——他们正是提出 mass-mean 并系统对比其与 LogReg/CCS 在跨数据集 transfer 下更鲁棒。C3 单独看 novelty 很弱,必须靠『加入 MLP 第三方、跨 shift type × model size 的系统矩阵、并把比较结果绑定到你们的稳定性分数』来续命,否则会被审稿人直接判定重复。 +2) C1+C4 的核心实证(label-preserving 风格偏移下探针准确率崩塃、且随 model size 出现非单调/逆向 scaling、augmentation 可缓解)与 2026 的 Pressure-Testing Deception Probes 高度重叠。你们相对它的唯一硬区分是『概念方向 cosine 旋转矩阵』这一几何量(它只看 AUROC 崩塃,不看方向旋转)以及把 augmentation 当预测信号而非修复手段——这两点必须在论文里前置强调,否则 C1/C4 会被认为是该文的子集。 + +仍然空白(你们真正的护城河): +- C2 的具体形式——『dispersion(同分布 K 次重采样方向离散度)+ augmentation-consistency(label-preserving 增广后方向一致性)』的双分量、免新标注、纯 forward-pass、用于训练前 a-priori 预测探针 OOD 迁移性——在本角度未找到完全对应工作。CCPS(隐状态扰动做置信度校准)与 Geometry Markers(ID 几何预测 OOD,视觉域)分别占了『稳定性信号』和『免标注 a-priori 预测』两个相邻 niche,但都不是『探针方向的重采样离散度 + 增广一致性』。这是你们最干净的 novelty,应作为主卖点突出,并显式与这两篇划界。 +- C5(用本地 DeBERTa-MNLI 做 label-fidelity 审计来保证 shift 是 label-preserving)在本角度未见任何探针文献采用;这是方法学增量(虽不足以单独成贡献),建议作为支撑 C1 严谨性的卖点而非独立 claim。 + +行动建议:把论文 framing 从『我们发现探针不稳定 + 我们比较探针类型』(已被占)转向『我们提出免标注、训练前可计算的方向级稳定性分数,并证明它能预测 OOD 迁移』(C2 为核心),把 C1/C3/C4 降格为支撑性诊断证据,并在 related work 中点名 Marks & Tegmark、Pressure-Testing Deception Probes、CCPS、Geometry Markers 做正面切割。注意:多篇 2026 预印本(2605.27958、2603.01879、ICLR 2026 kHVzEjThKE)venue 与作者仍待核实,投稿前请二次确认。 + +### 角度: 探针到底测了什么、探针可靠性争议(amnesic probing / information-theoretic probing / probe vs mode... + +- **Elazar, Ravfogel, Jacovi & Goldberg, 2021 (arXiv 2020)** — *Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals* (TACL 2021) [威胁:medium] + - 与 C1 的动机层强重叠:都在质疑探针准确率高估了特征的真实地位。但 Amnesic Probing 质疑的是'被模型使用与否(causal usage)',而非'在 label-preserving 输入分布偏移下是否稳定(robustness/OOD)';它需要反事实擦除而非纯 forward-pass,也不给 a priori 预测分数。 +- **Voita & Titov, 2020** — *Information-Theoretic Probing with Minimum Description Length* (EMNLP 2020) [威胁:medium] + - 与 C1+C2 概念层重叠:同样主张'accuracy 不足以衡量特征被编码的可靠程度',并提出替代度量。但 MDL 衡量的是同分布下的编码努力/压缩,不涉及 paraphrase/domain/length 偏移,也不预测 OOD 迁移、不含方向稳定性或增广一致性。是你度量动机的最重要先验之一,需在 related work 明确区分。 +- **Hewitt & Liang, 2019** — *Designing and Interpreting Probes with Control Tasks* (EMNLP 2019) [威胁:medium] + - 与 C1 动机重叠(accuracy 会高估)且与 C3 间接相关(线性 vs 非线性探针的可信度差异,呼应你 mass-mean/LogReg/MLP 比较)。但 selectivity 衡量的是'探针自身容量混淆',不是输入分布偏移下的稳定性,也无方向旋转、无 OOD 预测分数。 +- **Belinkov, 2022** — *Probing Classifiers: Promises, Shortcomings, and Advances* (Computational Linguistics (MIT Press) 2022) [威胁:medium] + - 与 C1 高度重叠在'问题陈述'层面——你要量化的'accuracy 高估特征稳定性'正是该综述列出的核心缺陷之一。威胁不在于它做了你的实验,而在于审稿人会用它说'这个问题早已是共识'。需要你用具体的偏移矩阵+方向旋转把'已知缺陷'落成'可量化诊断'来体现增量。 +- **Marks & Tegmark, 2023/2024** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (COLM 2024 (arXiv 2310.06824)) [威胁:high] + - 对 C3 是最直接威胁:'mass-mean vs LogReg 的相对(跨分布)可迁移性比较'核心结论它已给出,且方法上引入了 difference-of-means 探针的现代用法。你的 C3 若仅复述'mass-mean 更可迁移'会被判为已知。差异点:它聚焦 truth 这一概念、用跨数据集泛化而非系统化的 paraphrase/domain/length label-preserving 偏移矩阵,也无 MLP、无 model-size ladder、无 a priori 稳定性分数。 +- **Gao, Zhu, Zeng, Zhao, Wang, Ruan & Ding, 2026** — *RAPTOR: Ridge-Adaptive Logistic Probes* (arXiv 2602.00158 (2026, venue 待核实)) [威胁:high] + - 对 C2 的 dispersion 分量是最近、最危险的撞车:它已经把'对训练数据扰动下的方向 cosine 稳定性'作为正式评测指标。区别在于 RAPTOR 把方向稳定性当作改进探针训练的目标/评估,而非'训练前、免标注、用来预测 OOD 迁移的双分量预测分数',也不含 augmentation-consistency 分量、不做 paraphrase/domain 偏移迁移、不做 size ladder。务必引用并清楚切割:你是 a priori 预测器,它是 robustness 评估+方法改进。 +- **Boxo, Neelappa & Raval, 2026** — *Linear Probes Rely on Textual Evidence (leakage mitigation in LLM probes)* (ICML 2026 (arXiv 2509.21344, venue 待核实)) [威胁:medium] + - 与 C1 动机强重叠的'近作':直接给出'探针准确率虚高、表征不稳健'的量化证据,会被用来质疑你 C1 的新颖性。但它走的是 token-level 证据泄漏路径,不是 label-preserving 输入分布偏移,也无方向旋转、无 a priori 分数、无 mass-mean/size 比较。是必须 cite 并区分的同期工作。 +- **Anonymous/待核实, 2025-2026** — *No Answer Needed: Predicting LLM Answer Accuracy from Question-Only Linear Probes* (OpenReview (待核实)) [威胁:low] + - 与 C2 的'仅 forward-pass、事前预测'范式精神相近,可能被审稿人援引为'事前探针预测'已有先例。但它预测的是答案正确性,不是探针自身的 OOD 可迁移性,目标对象完全不同,无方向稳定性/增广一致性。 +- **Ravichander, Belinkov & Hovy, 2021** — *Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance?* (EACL 2021) [威胁:medium] + - 与 C1 动机重叠:再次证明 accuracy 会高估特征的真实地位。但关注'任务相关性'而非分布偏移稳健性,无方向旋转、无预测分数、无 C3/C4。属背景到中等的反例支撑文献。 + +> 评估: 已联网检索(WebSearch/WebFetch 可用)。本角度结论如下。 + +最大撞车点(按风险排序): +1) C3 撞 Marks & Tegmark《The Geometry of Truth》(COLM 2024, high)。'mass-mean(difference-of-means)vs LogReg 在跨分布下更可迁移'这一核心结论已被发表,且 mass-mean 探针的现代用法即源于此。C3 不能停在'mass-mean 更稳';必须靠'系统化 shift type × model size 矩阵 + 加入 MLP + label-fidelity 审计'来制造增量,否则会被直接判为已知。 +2) C2 的 dispersion 分量撞 RAPTOR(arXiv 2026, high)。'对训练数据扰动的方向 cosine 稳定性'已被它作为正式 directional-robustness 指标。你的护城河必须落在 RAPTOR 没有的三点:(a) 这是训练前、免新标注的 a priori 预测器,目标是预测 OOD 迁移性而非改进探针;(b) 双分量,额外含 augmentation-consistency;(c) 跨 paraphrase/domain/length 偏移与 size ladder 的预测有效性验证。强烈建议把'我们的分数能 a priori 预测 C1 矩阵里的 accuracy drop / 方向旋转'做成核心实验,这是与所有先验最干净的切割。 +3) C1 的'问题本身'被大量经典+近作占据(Belinkov 综述、Hewitt&Liang、Ravichander、Voita&Titov,以及 2026 的 leakage 论文)。'探针准确率高估特征可靠性'已是领域共识级命题,审稿人极可能说'not new'。C1 的新颖性只能来自'量化方式'——即把已知缺陷落成 accuracy drop + 概念方向 cosine 旋转 的双轴矩阵,且明确限定在 label-preserving 输入分布偏移(而非 token 泄漏、控制任务或因果擦除)。务必在 related work 用一句话各自切割上述四篇。 + +仍有空白(你的真实可辩护新颖性): +- 没有任何一篇把'同分布重采样方向离散度 + label-preserving 增广方向一致性'组合成单一、免标注、训练前的探针稳定性预测分数,并验证其对 OOD 迁移的预测力。这是 C2 最干净的空白,应作为论文第一卖点。 +- 没有人用'概念方向 cosine 旋转'作为跨 paraphrase/domain/length 偏移的稳定性轴(现有方向-cosine 工作都在'数据扰动/seed'维度,如 RAPTOR;偏移维度是空白)。C1 的方向旋转矩阵因此有真增量。 +- C4(model-size ladder 对探针 OOD 稳定性的影响)在本角度文献里基本无人系统做(本角度多为单模型/单规模),是相对安全的贡献,但单独不足以撑论文。 +- C5(本地 NLI/DeBERTa-MNLI 做 label-fidelity 审计以保证 shift 是 label-preserving)在本角度未见对应工作,属方法学卫生措施而非独立新颖点,审稿人不会视为贡献但会因其缺失而质疑 C1 的有效性——保留但别当卖点。 + +行动建议:把叙事重心从 C1(易被判已知)转到 C2(a priori 免标注预测器)为主轴,用 C1 矩阵作为'被预测的真值'、C3/C4 作为外部效度验 + +### 角度: 线性表示假设 / 真值几何 / 概念方向(linear representation hypothesis, geometry of truth, concep... + +- **Liang et al. (2025), 作者名待核实** — *LLM Knowledge is Brittle: Truthfulness Representations Rely on Superficial Resemblance* (arXiv:2510.11905(2025-10);会议归属待核实) [威胁:high] + - C1 几乎完全重叠(accuracy drop + 概念方向 cosine 旋转 across shift type,且 shift 类型 paraphrase/domain/length 与我们逐字一致);很可能也涉及 C4(多模型/规模)。未见 C2(无 forward-pass 先验预测分数)、未明确 C5(NLI 审计)、C3 仅部分(用了多种探针但未见系统的 mass-mean/LogReg/MLP 可迁移性对比)。 +- **Bao, Zhang et al. (2025)** — *Probing the Geometry of Truth: Consistency and Generalization of Truth Directions in LLMs Across Logical Transformations and Question Answering Tasks* (Findings of ACL 2025(arXiv:2506.00823);有公开代码) [威胁:high] + - C1 中等重叠(跨变换的方向一致性/泛化,但其变换是逻辑变换/QA,而非我们的 paraphrase/domain/length 分布偏移);C4 重叠(模型能力/规模越强方向越一致);C3 部分(讨论简单 vs 复杂探针是否必要)。未涉及 C2 的先验预测分数,未涉及 C5。 +- **作者待核实 (2026)** — *How Context Shapes Truth: Geometric Transformations of Statement-level Truth Representations in LLMs* (arXiv:2601.06599(2026-01)) [威胁:high] + - C1 中-高重叠(方向旋转角 θ 作为核心度量,与我们的概念方向 cosine 旋转同构,只是触发因素是上下文而非 paraphrase/domain/length);C4 高度重叠(明确给出 model size × 方向变化的结论)。未涉及 C2 的免标注先验稳定性分数,未涉及 C3、C5。 +- **Marks & Tegmark (2023/2024)** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (arXiv:2310.06824;COLM 2024(待核实)) [威胁:high] + - C3 高度重叠(正是 mass-mean vs LogReg vs CCS 的可迁移性/鲁棒性比较的来源,且已声称 mass-mean 对分布偏移更鲁棒——我们的 C3 很大程度是在复刻+扩展);C4 部分(规模才出现线性结构);C1 背景(transfer 实验)。未涉及 C2、C5。 +- **Aithal, Kashyap et al. (2021)** — *Robustness to Augmentations as a Generalization Metric* (arXiv:2101.06459(NeurIPS 2020 Predicting Generalization in Deep Learning 竞赛亚军方案)) [威胁:medium] + - C2 中等重叠:augmentation-consistency 作为 label-free 泛化预测量的核心思想已被提出(但对象是分类网络整体泛化,而非探针方向在 OOD 的可迁移性,也无 dispersion 分量、无 LLM 概念方向语境)。是我们 C2「双分量分数」中 augmentation-consistency 一支的最直接先验思想。 +- **作者待核实 (2024)** — *Truth-value judgment in language models: 'truth directions' are context sensitive* (arXiv:2404.18865(2024)) [威胁:medium] + - C1 中等(为「探针/方向不稳定」提供早期证据,但未做 accuracy drop × 方向旋转的系统矩阵,也无 paraphrase/domain/length 三类偏移设计);C4 弱相关。未涉及 C2/C3/C5。 +- **作者待核实 (2025)** — *Are Sparse Autoencoders Useful? / SAE probes vs logistic regression baselines* (arXiv:2502.16681(2025)) [威胁:medium] + - C3 中等(探针方法系统比较的范式与基线选择,虽未含 mass-mean/MLP 的 OOD 专门对比);C1 弱相关。未涉及 C2、C4(规模)、C5。 +- **作者待核实 (2025)** — *The Impact of Off-Policy Training Data on Probe Generalisation* (arXiv:2511.17408(2025)) [威胁:low] + - C1 弱-中(探针泛化受分布影响,但聚焦训练数据 on/off-policy 而非输入端 label-preserving 偏移,也无方向旋转度量);未涉及 C2/C3/C4/C5。 +- **Park, Choe & Veitch (2023/2024)** — *The Linear Representation Hypothesis and the Geometry of Large Language Models* (ICML 2024(arXiv:2311.03658)) [威胁:low] + - 全为背景(C1 方向 cosine 度量的理论依据);不直接做探针稳定性/OOD 实验。无 C2-C5 重叠。 + +> 评估: 已联网检索(WebSearch + WebFetch arXiv 全文)。诚实结论:本角度下我们的 novelty 处于高度拥挤区,C1、C3、C4 基本已被抢占,真正未被占领的护城河只剩 C2 的「双分量、免标注、训练前先验预测」整体打包,以及 C5 的方法学细节。 + +最大撞车点(逐条): +- C1 = 最危险。arXiv:2510.11905「LLM Knowledge is Brittle」几乎逐字做了我们 C1 要做的事:in-distribution 探针准确率高估真值特征鲁棒性,且在 paraphrase/domain/length(与我们三类偏移完全一致)下度量 accuracy drop + 概念方向 cosine 漂移。再叠加 2601.06599 把「方向旋转角 θ × model size」这条结论也做了。我们不能再把 C1 当作主贡献,最多作为复现+扩展,且必须显式区分于这两篇(例如:我们是 shift-type × model-size 的完整矩阵化、并把方向旋转与 accuracy drop 解耦量化)。务必注意 2510.11905 的「superficial resemblance」叙事会被审稿人直接拿来质疑 C1 的新颖性。 +- C3 = 已被 Marks & Tegmark(2310.06824)定调:mass-mean vs LogReg vs CCS、且已声称 mass-mean「对分布偏移更鲁棒、更因果」。我们的 C3 若只比 mass-mean/LogReg/MLP 的 OOD 可迁移性,基本是复刻;增量必须落在 MLP(非线性)这一未被充分比较的轴上,或与 C2 分数的相关性验证上,否则审稿人会判为已知结论。 +- C4 = 2506.00823 与 2601.06599 都已给出「模型越大/越强→方向越一致/靠方向变化区分」的规模结论;C4 单独不构成贡献,只能作为辅助维度。 +- C5(NLI label-fidelity 审计)= 检索未见同角度有人明确用本地 NLI(DeBERTa-MNLI)审计 shift 的 label-preserving 性;但这是方法学卫生措施,难以单独支撑 novelty。 + +仍留白(可主打的差异化): +1) C2 的整体打包是目前唯一未见先例的核心。检索到的最接近先验是 2101.06459(augmentation-robustness 作为 label-free 泛化度量),但那是「分类网络整体泛化」非「探针方向 OOD 可迁移性」,且没有 dispersion(同分布 K 次重采样方向离散度)这一分量、没有 LLM 概念方向语境。建议把论文重心明确转向 C2:强调它是 a priori、forward-pass-only、双分量、专门预测探针方向 OOD 迁移的分数,并实证它与 C1 观测到的 accuracy drop / 方向旋转高度相关(把 C1/C3/C4 降级为支撑 C2 的证据,而非并列贡献)。 +2) 必须补充与 2101.06459 的区分与(若可能)对照实验,否则 C2 的 augmentation-consistency 分量会被指为已知思想的迁移。 +3) dispersion 分量(重采样方向离散度作为稳定性预测量)在本角度检索中未见直接对应,是 C2 内最干净的新点,建议作为卖点之一并单独消融。 + +未能确认的点(建议进一步精读全文):2510.11905 是否做了 mass-mean/LogReg/MLP 系统对比(WebFetch 显示 + +### 角度: 概念擦除与方向估计:INLP / RLACE / LEACE / mass-mean(difference-of-means)/ steering vector... + +- **Marks & Tegmark, 2023** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (COLM 2024 (arXiv:2310.06824)) [威胁:high] + - C3 直接重叠(mass-mean vs LogReg 的相对可迁移性正是它的核心卖点之一);C1 部分重叠(展示了探针跨数据集迁移 accuracy 的变化)。但它把 mass-mean 当作'更好'来推销,没有把跨分布 accuracy 高估当成系统性诊断问题,也无方向旋转矩阵、无 a priori 预测分数、无 NLI 审计。 +- **Anonymous/Hazratian & Zia (待核实作者), 2026** — *The Truthfulness Spectrum Hypothesis* (arXiv:2602.20273(2026-02,预印本,待核实正式 venue)) [威胁:high] + - C2 最强威胁:它正面解决了'预测探针 OOD 迁移性',且明确指出普通 cosine 不够、需白化——这会侵蚀你 C2 中'方向一致性/旋转'分量的新意。C4 也重叠(model size ladder)。关键差异(你的护城河):它的预测器需要在 OOD 数据上训练一个 reference probe 才能算 MCS,不是 forward-pass-only / a priori / 免新标注;它评估的是真值这一单一概念域、用 LogReg、无 paraphrase/length/domain 的 label-preserving shift 框架、无 NLI 审计、无 dispersion(K 次同分布重采样)分量。 +- **Ravfogel, Goldberg et al. (待核实完整作者), 2022** — *Probing Classifiers are Unreliable for Concept Removal and Detection* (NeurIPS 2022 (arXiv:2207.04153)) [威胁:medium] + - C1 的基石性背景:它确立了'探针准确率可被虚假相关误导、高估概念存在'这一核心命题。你 C1 的增量在于把它具体化为 label-preserving 分布偏移(paraphrase/domain/length)下的 accuracy drop × cosine 旋转矩阵 × model size,而该文不做分布偏移矩阵、不做方向旋转量化、不做预测分数。属于必须正面区分并引用的 prior art。 +- **Tan, Chanin et al. (待核实完整作者), 2024** — *Analysing the Generalisation and Reliability of Steering Vectors* (NeurIPS 2024 (arXiv:2407.12404)) [威胁:medium] + - C1/C3 中等重叠:steering vector = mass-mean 方向,该文已展示这类方向对输入改动的脆性(与你'稳定性被高估'同源),并触及可控性可被某些信号预测。差异:它面向 steering 干预效果(behavioural)而非探针分类 accuracy/方向旋转矩阵;无 a priori forward-pass 双分量分数;无 NLI label-fidelity 框架;模型规模非主轴。 +- **(待核实作者), 2025** — *Understanding (Un)Reliability of Steering Vectors in Language Models* (arXiv:2505.22637(2025,预印本,待核实 venue)) [威胁:medium] + - C2 中高威胁:它的'cosine 一致性预测有效性'+ d' 可分性指标,在思想上非常接近你 C2 的两个分量(方向一致性 + 离散度/可分性),且都是从激活几何 a priori 推断。差异:它针对 steering 有效性而非探针 OOD accuracy;无 dispersion(K 次同分布重采样)正式定义;无 paraphrase/domain/length 的 label-preserving shift 协议;无 NLI 审计;无 mass-mean/LogReg/MLP 三者系统比较;无 model size ladder。 +- **Méloux, Portet & Peyrard, 2025** — *Mechanistic Interpretability as Statistical Estimation: A Variance Analysis* (arXiv:2510.00845(2025,预印本,待核实 venue)) [威胁:medium] + - C2 dispersion 分量的概念前身:'用 K 次重采样的方向离散度衡量探针稳定性'这一核心机制与你高度同构。差异(待全文确认):它把 dispersion 当作'可信度/可复现性'诊断,未必把它接到'a priori 预测 OOD 迁移 accuracy'这一具体目标上,且未必含 augmentation-consistency 第二分量、label-preserving shift 协议与 NLI 审计。建议优先精读全文判定是否需降为护城河外或升为 high。 +- **Azizian, Kirchhof, Ndiaye, Bethune, Klein, Ablin, Cuturi, 2025** — *The Geometries of Truth Are Orthogonal Across Tasks* (arXiv:2506.08572(2025,预印本,待核实 venue)) [威胁:medium] + - C1/C3 重叠:它正是'方向旋转/正交 ↔ accuracy 掉落'的量化,与你 C1 矩阵中'cosine 旋转'一列同源,并涉及多探针类型。差异:它跨任务(task)而非你强调的 label-preserving 同任务输入分布偏移(paraphrase/domain/length);无 forward-pass a priori 预测分数;无 NLI 审计;无 model size ladder 为主轴。 +- **Ravfogel, Twiton, Goldberg, Cotterell (待核实), 2022** — *Linear Adversarial Concept Erasure (RLACE)* (ICML 2022 (arXiv:2201.12091)) [威胁:low] + - 背景相关:为你'方向估计本身不稳定'提供方法学锚点(RLACE 优化不稳),但它不研究 label-preserving 分布偏移下探针 accuracy 高估,不提预测分数,不做 paraphrase/NLI。属应引用的方向估计可靠性脉络,但不构成 novelty 撞车。 +- **Belrose, Schneider-Joseph, Ravfogel, Cotterell, Raff, Biderman, 2023** — *LEACE: Perfect Linear Concept Erasure in Closed Form* (NeurIPS 2023 (arXiv:2306.03819)) [威胁:low] + - 背景相关:concept erasure 角度的 SOTA 锚点,但聚焦'擦除得多干净'而非'探针方向在 label-preserving 偏移下是否稳定/可迁移',与 C1–C5 无直接方法撞车。 +- **Ravfogel, Elazar, Goldberg, Cotterell (待核实), 2020** — *Null It Out: Guarding Protected Attributes by Iterative Nullspace Projection (INLP)* (ACL 2020 (arXiv:2004.07667)) [威胁:low] + - 背景相关:与 RLACE 一道支撑'线性方向估计本身脆弱'的论点,但不涉及分布偏移下的探针 accuracy 高估、预测分数、paraphrase/NLI。 + +> 评估: 已联网检索(WebSearch/WebFetch)。重要可靠性提示:WebFetch 的小模型摘要出现过明显的"附和我的检查清单全中"幻觉——我已对每篇下载 PDF 抽样核验,确认 arXiv:2602.03951(TorRicc)其实是 CIFAR 视觉模型 + 谱复杂度/Ricci 曲率做无监督 checkpoint 选择,与本角度几乎无关(故未列入,顶多算 label-free OOD 预测哲学的远亲);2510.00845 与 2602.20273 的"全中"部分也按怀疑态度下调或标注待核实。 + +最大撞车点(按威胁排序): +1) C2(a priori 预测探针 OOD 迁移性)是你被抢占风险最高的贡献。arXiv:2602.20273(Truthfulness Spectrum, 2026-02)已用 Mahalanobis-白化后的方向相似度近完美预测 OOD AUROC(R²≈0.98),并明确说普通 cosine 不够——这几乎正面命中你"用方向一致性预测迁移"的卖点,且跨 model size。你唯一且关键的护城河是:它的预测器需要在 OOD 数据上额外训练 reference probe(需 OOD 标注),而你主张 forward-pass-only / 免新标注 / 训练前 a priori,且含 dispersion(同分布 K 次重采样)这一不需要任何 shift 数据的分量。务必在论文里把"是否需要触碰目标分布/新标注"作为核心区分轴写死,并直接对比 MCS。 +2) C3(mass-mean vs LogReg vs MLP 的 OOD 相对可迁移性)已被 Marks & Tegmark(2023)实质性占据 mass-mean vs LogReg 这一半;你的增量只剩"加入 MLP + 在系统化 label-preserving shift 矩阵下比较"。建议把 C3 从"比较"重新定位为"在你的 shift × size 矩阵框架内的受控复现+扩展",否则单独作为贡献偏弱。 +3) C1(accuracy 高估 + 方向旋转矩阵)在'探针不可靠/会高估'(NeurIPS'22)与'方向跨任务正交 ↔ 迁移失败'(2506.08572)之间被两头夹。但没有一篇把它组织成 shift type(paraphrase/domain/length)× model size 的 accuracy-drop × cosine-rotation 双矩阵,这个具体形态仍可主张为新颖呈现。 +4) C2 的 dispersion 分量思想已有前身(2510.00845 把方向重采样离散度当可靠性指标;2505.22637 用激活差 cosine 一致性 + d' 预测 steering 有效性)。"双分量(dispersion + augmentation-consistency)合成一个分数、专门用于训练前预测探针 OOD"这个组合目前未见完全重合者,是你最干净的立足点。 + +仍存的空白(= 你可主张的真正 novelty): +- forward-pass-only / 免任何 OOD 数据与新标注 的 a priori 预测(vs 2602.20273 需 OOD reference probe); +- dispersion + augmentation-consistency 双分量的显式合成与互补性消融; +- label-preserving 的 paraphrase/domain/length shift 受控协议,尤其 C5 + +### 角度: 探针在分布偏移/OOD 下的泛化与失效(probing under distribution shift; robustness of probing clas... + +- **Xie, Wei, Feng, Cao, An (2023)** — *On the Importance of Feature Separability in Predicting Out-Of-Distribution Error (Dispersion Score)* (NeurIPS 2023(arXiv:2303.15488,待核实最终录用会议)) [威胁:high] + - 对 C2 威胁最大:与我们 PSS 的第一个分量(dispersion/方向离散度)在名称、机制(feature dispersion)、目标(免标注+仅前向预测 OOD 性能)上几乎正面撞车。差异:他们用于一般图像分类器的整体精度估计,不是 concept-direction/线性探针,不含 augmentation-consistency 第二分量,也不做 a priori(他们是在给定 OOD 测试集上估精度,我们是训练前预测探针迁移性)。 +- **Levinstein & Herrmann / Bürger 等 (2024)（'truth directions are context sensitive'）** — *Truth-value judgment in language models: 'truth directions' are context sensitive* (arXiv:2404.18865(待核实正式 venue)) [威胁:high] + - 对 C1 威胁高:正是'探针准确率高估了概念方向稳定性、在 label-preserving 上下文/改述偏移下方向漂移'这一核心观察。差异:聚焦真值方向单一概念、未给出 accuracy-drop × direction-rotation 跨 shift×模型规模 的系统矩阵,也未提出预测性 score(不碰 C2)。 +- **Azizian, Kirchhof, Ndiaye, Bethune, Klein, Ablin, Cuturi (2025)** — *The Geometries of Truth Are Orthogonal Across Tasks* (arXiv:2506.08572(待核实)) [威胁:high] + - 对 C1/C3 中等偏高:用方向几何(正交=cosine 旋转)量化探针跨分布不可迁移,与 C1 的'方向旋转'指标同源。差异:跨任务而非 label-preserving 输入偏移(paraphrase/domain/length),且无免标注预测 score、无 mass-mean/LogReg/MLP 比较、无规模阶梯。 +- **Marks & Tegmark (2023)** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (arXiv:2310.06824(后续 COLM/会议版本待核实)) [威胁:high] + - 对 C3 威胁高:已直接做了 mass-mean vs LogReg 的跨数据集(可视为 OOD)相对迁移性比较,且结论是 difference-of-means 更稳。我们的 C3 若只停在'mass-mean vs LogReg vs MLP 谁更稳'会被认为增量;需靠 MLP 纳入 + 系统化 shift 类型 + 与 PSS 预测对齐来区分。差异:无 MLP,无 paraphrase/length 偏移设计,无预测 score。 +- **多篇 (2025-2026)：'generalization of LLM truth directions on conversational formats'(2505.09807)、'Probing the Geometry of Truth: Consistency and Generalization across Logical Transformations'(2506.00823)、'Testing the Limits of Truth Directions'(2604.03754)** — *Truth-direction generalization across formats / logical transformations 系列* (arXiv(2025-2026,均待核实)) [威胁:medium] + - 对 C1/C4 中等:覆盖了 length/format 这类 label-preserving 偏移并量化泛化,与 C1 的 length shift、与 C4 的'结构变化下稳定性'有重叠。差异:仍局限于真值任务,无统一的 accuracy-drop×rotation 矩阵,无免标注 a priori score,规模阶梯不是主轴。 +- **Müller 等 / 待核实 (2025)** — *Intermediate Layer Classifiers for OOD Generalization (ICLR 2025)* (ICLR 2025(proceedings.iclr.cc,待核实作者)) [威胁:medium] + - 对 C1/C4 中等:直接研究线性探针 OOD 泛化、层与规模因素,且量化 accuracy 在 shift 下的保持率。差异:视觉/表征学习设定为主,概念方向旋转、免标注预测 score、NLI label-fidelity 审计均不涉及。 +- **Hewitt & Liang (2019)** — *Designing and Interpreting Probes with Control Tasks (selectivity)* (EMNLP 2019) [威胁:low] + - 对 C1/C2 低-中:是'探针准确率会误导、需要额外诊断量'这一论点的奠基性背景,与我们 PSS 作为'诊断量'同属一脉,但 selectivity 针对记忆/复杂度而非分布偏移稳定性,且非免标注预测 OOD 迁移。 +- **Belinkov (2022)** — *Probing Classifiers: Promises, Shortcomings, and Advances* (Computational Linguistics 48(1)) [威胁:low] + - 对 C1 低:为我们的动机(探针指标不可尽信)提供权威背景,但不提供任何偏移矩阵、score 或方法比较,属定位/引用价值。 +- **Ravichander, Belinkov, Hovy (2021)** — *Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance?* (EACL 2021) [威胁:low] + - 对 C1 低:支撑'准确率高估特征有用性/稳定性'的总论点,但聚焦任务相关性而非分布偏移下的方向稳定性,无 score 提案。 + +> 评估: 已联网检索。本角度下你们最大的撞车点有两处,均需在论文中正面处理: + +1) C2(免标注、仅前向的双分量 PSS)与 Xie et al. 2023 的 "Dispersion score" 高度同名同质——同样是 label-free、forward-pass、用 feature dispersion 预测分布偏移下性能。区分点必须明确且可被审稿人一眼看到:(a) 我们是对'线性/概念探针的方向稳定性'而非'整体分类精度'打分;(b) 我们是训练前 a priori 预测探针迁移性,而非在给定 OOD 测试集上估精度;(c) 我们加了第二个正交分量 augmentation-consistency,且需用消融证明它在 dispersion 之外提供增量。否则 C2 会被定性为 Dispersion score 的换皮迁移。建议直接把 Dispersion score 作为 baseline 跑对比。 + +2) C1(准确率高估方向稳定性 + accuracy-drop×cosine-rotation 矩阵)的核心现象已被 "truth directions" 系列(Levinstein/Bürger 2024、Azizian et al. 2025 正交性、conversational-format/logical-transformation 系列 2025-2026)在真值方向这一特例上反复证明,包括用方向旋转/正交性量化。你们的真正新意只能落在'系统性矩阵'(shift type × model size × probe type)+ 跨多概念(不止真值)+ 与可预测 score 挂钩,而非'发现方向会漂移'这一已知事实。C3(mass-mean vs LogReg)也已被 Marks & Tegmark 2023 部分做掉,纯比较是增量;靠纳入 MLP、统一 shift 设计、并把比较结果用 PSS 预测出来才有区分度。 + +仍存在的空白(你们可主打的护城河):没有任何一篇把(i)免标注 a priori 探针稳定性 score、(ii)dispersion+augmentation-consistency 双分量、(iii)label-preserving 偏移的 accuracy-drop×direction-rotation 系统矩阵、(iv)mass-mean/LogReg/MLP 三探针×规模阶梯、(v)本地 NLI(DeBERTa-MNLI)做 shift 的 label-fidelity 审计——五者整合在统一框架里。尤其 C5(NLI 审计保证 shift label-preserving)在本角度文献中几乎无人配套使用,是相对干净的方法学贡献点,但单独 novelty 不足以支撑论文,应作为严谨性卖点而非核心 claim。 + +诚实结论:C1、C2、C3 单独看都已被相邻工作不同程度抢占(C2 最危险,直接撞 Dispersion score 之名与机制),论文的可辩护新颖性集中在'a priori 预测 + 双分量 + 系统矩阵 + NLI 审计'的整体打包与'方向稳定性视角的 score 化'上,务必在 related work 显式区分上述四篇 high-threat 工作并用实验(Dispersion score、mass-mean 作 baseline)证明增量。 + +### 角度: 表示/方向的稳定性与旋转:线性探针方向在 seed/重采样/扰动/分布偏移下的不稳定性与旋转(cosine 漂移)、direction consistency、... + +- **Gao, Zhu, Zeng, Zhao, Wang, Ruan, Ding — 2026** — *RAPTOR: Ridge-Adaptive Logistic Probes* (arXiv:2602.00158 (preprint;投稿目标未知,待核实)) [威胁:high] + - 与 C2 的 dispersion 分量高度重叠(其 eq.9 几乎就是我们的'同分布重采样方向离散度',只是用 1−|cos| 还是 |cos| 的符号约定差异);与 C4(model size ladder 影响方向稳定性)直接重叠;与 C3 部分重叠(比较不同探针估计子的方向稳定性,但不是 mass-mean/LogReg/MLP 这条具体轴线);与 C1 在动机表述上重叠(accuracy 高估稳定性)。缺口:没有 augmentation-consistency 分量、不做 label-preserving 输入偏移(paraphrase/length/domain)、不构建 accuracy-drop×cosine-rotation 矩阵、不把分数当作训练前 a priori 的 OOD 迁移预测器(它的稳定性是诊断而非预测 OOD)。 +- **Lysnæs-Larsen, Eggen, Strümke (NTNU) — 2025** — *Probing the Probes: Methods and Metrics for Concept Alignment* (arXiv:2511.04312 (待核实是否已被会议接收)) [威胁:high] + - 与 C2 的 augmentation-consistency 分量高度重叠:同样是'label-preserving 增广后方向一致性、forward-pass-only、归一化到 0–1'的探针质量分数。区别:这是视觉/ResNet50/CAV 场景,增广是图像变换而非文本 paraphrase/length/domain;不含同分布重采样 dispersion 分量;不把两分量合成单一 Probe Stability Score;不预测 OOD 迁移;不比较 mass-mean/LogReg/MLP;不变 model size。 +- **Huang (William Hao-Cheng) — 2025** — *Spectral Identifiability for Interpretable Probe Geometry* (arXiv:2511.16288 (preprint,待核实)) [威胁:high] + - 与 C2 的核心动机/定位撞车最危险的一篇:同样是'在训练/部署前 a priori 预测探针方向可靠性'的免标注判据。但机制不同——它用 spectral/eigengap 几何而非 dispersion+augmentation 双分量;聚焦 in-distribution 可识别性而非显式预测 OOD/label-preserving shift 迁移;不构建 accuracy-drop×rotation 矩阵;不比较探针族;model size 未系统化。属'同一空白被另一种方法占住',对 C2 的'首个 a priori 探针稳定性预测器'新颖性主张构成实质威胁。 +- **Marks & Tegmark — 2023/2024** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (COLM 2024(待核实);arXiv:2310.06824) [威胁:high] + - 与 C3 直接、强重叠:正是'mass-mean vs LogReg(+CCS)在跨数据集/OOD 下相对可迁移性'的系统比较,且给出 mass-mean 更可迁移的结论——这会削弱 C3 的新颖性,需明确我们相对它的增量(加入 MLP、label-preserving 输入偏移、规模 ladder、以及把可迁移性与稳定性分数挂钩)。与 C4 部分相关(它有 scale 讨论)。不覆盖 C2 的稳定性分数,也不做 paraphrase/length 偏移。 +- **Bürger / Levinstein 等(Truth-value judgment)— 2024,及 Probing the Geometry of Truth — 2025** — *Truth-value judgment in language models: 'truth directions' are context sensitive (2404.18865);Probing the Geometry of Truth: Consistency and Generalization of Truth Directions across Logical Transformations and QA (2506.00823)* (arXiv:2404.18865 / arXiv:2506.00823(会议状态待核实)) [威胁:high] + - 与 C1 核心现象强重叠:同时记录了'准确率下降'与'方向旋转/近正交'这两件事,正是我们 C1 想系统化的 accuracy-drop + direction-rotation。区别:它们的偏移是逻辑/上下文/否定与跨域,而非我们刻意构造的 label-preserving paraphrase/length/domain 矩阵,也未把它组织成 shift-type × model-size 的二维矩阵,且未提出稳定性分数。仍显著压缩 C1 '系统量化高估'的新颖空间。 +- **Kramár, Engels, Wang, Chughtai, Shah, Nanda, Conmy (Google DeepMind)— 2026** — *Building Production-Ready Probes for Gemini* (arXiv:2601.11516 (cs.LG, preprint)) [威胁:medium] + - 与 C1 在'准确率高估稳定性 / 输入长度偏移下崩溃'上重叠,且是 length-shift 的高质量证据点。但偏工程/安全部署视角,不量化方向 cosine 旋转、不做 paraphrase/domain 矩阵、不提稳定性分数、不比较探针族、不做规模 ladder。主要威胁 C1 的'长度偏移导致高估'这一子论点的首发性。 +- **Ding & Koehn 等(On Model Stability as a Function of Random Seed)— 2019** — *On Model Stability as a Function of Random Seed* (CoNLL 2019 (aclanthology K19-1087)) [威胁:low] + - 背景层面支撑 C2 的 dispersion 动机(同条件多次训练方向/性能会漂移),但针对的是任务模型而非线性探针方向、不涉及 cosine 旋转、不预测 OOD、不提稳定性分数。仅作 related work 引用。 +- **Ding, Denain & Steinhardt(Grounding Representation Similarity with Statistical Testing);Davari 等(Reliability of CKA, ICLR 2023);Deconfounded CKA — 2021–2023** — *Grounding Representation Similarity with Statistical Testing / Reliability of CKA as a Similarity Measure / Deconfounded Representation Similarity* (NeurIPS 2021 / ICLR 2023 / 2022(arXiv:2108.01661, 2202.00095 等)) [威胁:low] + - 方法论背景:为'用相似度/方向一致性衡量表示稳定性'提供工具与告诫,与 C2 用 cosine 度量方向一致性的思路同源。但研究对象是整层表示相似度而非单个探针概念方向、不针对 OOD 探针迁移预测、不涉及 label-preserving 文本偏移。低威胁,但建议引用以表明我们知道相似度度量的可靠性陷阱。 +- **Ravichander, Belinkov, Hovy 等(probing 方法论批评线)— 2021–2022** — *Probing-Classifier 方法论批评(Probing the Probing / Belinkov 2022 survey 等),及 Probing Classifiers are Unreliable for Concept Removal and Detection (NeurIPS 2022)* (NeurIPS 2022 / CL 2022 等(待核实具体条目)) [威胁:medium] + - 与 C1 的'准确率高估特征稳定性/存在性'动机重叠(高估的是 presence/causal,我们强调的是 across-shift stability),并被 RAPTOR 直接引用为'accuracy≠可靠方向'的依据。属强背景:压缩 C1 '高估'论点的概念新颖性,但未做 shift×size 矩阵、无旋转量化、无稳定性分数。 + +> 评估: 已联网检索(WebSearch/WebFetch + 本地读取 RAPTOR PDF),结论可信。 + +最大撞车点(按威胁排序): +1) C2 的两个分量已被分别占住,且各自都强。RAPTOR(2602.00158)的 directional-stability 度量(eq.9,K 次 ablation 后方向 mean |cos|)几乎就是我们的 dispersion 分量;"Probing the Probes"(2511.04312)的 Augmentation Robustness(label-preserving 变换、forward-pass-only、0–1)几乎就是我们的 augmentation-consistency 分量。也就是说 C2 的"双分量"里每一个分量单独看都已有人做过,真正可主张的新颖性只剩:(a) 把两者合成单一 Probe Stability Score、(b) 把它定位为训练前 a priori 的 OOD 迁移预测器(而非事后诊断)、(c) 用在文本 LLM 的 paraphrase/length/domain label-preserving 偏移上。 + +2) C2 的"a priori 预测探针可靠性"定位本身也已被 Spectral Identifiability(2511.16288)抢占——它同样是免标注、部署前可验证的判据,只是用 eigengap 几何而非我们的双分量。这对"首个 a priori 探针稳定性预测器"的措辞最危险,必须改写主张、并做正面 baseline 对比(我们的分数 vs SIP vs RAPTOR-stability 谁更能预测真实 OOD drop)。 + +3) C1 的核心现象(准确率下降 + 方向旋转)在 truth-direction 文献(2404.18865、2506.00823)里已被同时记录(约 25% OOD drop、跨域方向近正交),length 偏移崩溃在 Gemini probes(2601.11516)里也有强证据。C1 仍可主张的是"系统化的 shift-type × model-size 二维矩阵 + 同一概念跨多种 label-preserving 偏移的受控量化",而非"首次发现高估"。 + +4) C3 与 Marks & Tegmark(Geometry of Truth)直接撞:mass-mean vs LR(+CCS)的跨数据集可迁移性比较已有,且结论是 mass-mean 更可迁移。我们的增量必须明确为:加入 MLP 探针、用受控 label-preserving 输入偏移而非换数据集、跨规模 ladder、并把可迁移性与我们的稳定性分数关联。 + +仍存的空白(可成为真正卖点): +- 没有任何一篇把 dispersion + augmentation-consistency 合成单一分数,并实证证明该分数能在训练前预测 label-preserving OOD 下的 accuracy-drop 与 cosine-rotation(即把稳定性分数当 OOD 迁移的预测器并报告相关性/排序能力)。这是 C2 最干净的立足点。 +- 没有一篇把 (accuracy drop × concept-direction cosine rotation) 组织成 shift-type(paraphrase/domain/length)× model-size 的完整矩阵(C1 的呈现形式仍是空白)。 +- mass-mean vs LogReg vs + +### 角度: 可迁移性的免训练事前预测(transferability estimation: LEEP/LogME/H-score/GBC 及其在 probe 方向上的类比... + +- **Nguyen, Hassner, Seeger, Archambeau (2020)** — *LEEP: A New Measure to Evaluate Transferability of Learned Representations* (ICML 2020) [威胁:medium] + - 与 C2 的方法论范式(免训练 + 仅 forward-pass 预测可迁移性)直接对齐,是 C2 明确自认的类比来源。但目标完全不同:LEEP 预测的是【整模型迁移/微调后准确率】,需要源标签头/伪标签,且不涉及探针方向(direction)的稳定性,更不涉及 label-preserving 分布偏移下的方向旋转。C2 的新意正是把这一范式迁移到 probe 方向几何上并提出 dispersion+augmentation-consistency 两个新分量。 +- **You, Liu, Wang, Long (2021)** — *LogME: Practical Assessment of Pre-trained Models for Transfer Learning* (ICML 2021) [威胁:medium] + - 与 C2 范式同源(免训练打分预测下游可用性),且 LogME 本质就是在特征上做线性证据评估,与 probe 场景天然接近,因此是范式撞车点。差异:LogME 用【目标任务标签】拟合,属于事后/需要目标标注的估计,预测对象是 in-distribution 下游准确率,而非 OOD 下方向稳定性;不做重采样离散度,也不做增广一致性。C2 的『完全免新标注 + 专测 OOD 方向稳定』是区分点。 +- **Bao, Li, Huang, et al. (2019)** — *An Information-Theoretic Approach to Transferability in Task Transfer Learning (H-score)* (ICIP 2019) [威胁:low] + - 与 C2 共享『用特征几何/统计量免训练预测迁移』的思路。但 H-score 仍是任务级、需目标标签的可迁移性度量,与 probe 方向稳定性、label-preserving 输入偏移无关。属于范式背景而非具体抢占。 +- **Zhang, Wang, Wang, Chai, Yin, Lin, Wang (2026)** — *SAE as a Crystal Ball: Interpretable Features Predict Cross-domain Transferability of LLMs without Training* (arXiv 2603.02908(预印本,待核实最终去向)) [威胁:high] + - 这是本角度下时间最近、立意最接近 C2 的工作:同样主张【免训练 + 仅 forward-pass + 训练/微调前 a priori 预测 LLM 跨域(OOD)迁移性】。撞车点在于『事前可迁移性预测』这一核心卖点的 LLM 表征版本已被占。但机制不同——它用 SAE 可解释特征维度的偏移,而非 C2 的『同分布重采样方向离散度 + label-preserving 增广方向一致性』;预测对象是 SFT 后整体迁移,而非线性探针方向几何的 OOD 稳定性。C2 仍有方法差异,但 novelty 表述需明显避让此文。 +- **(作者待核实) (2025)** — *Feature Space Perturbation: A Panacea to Enhanced Transferability Estimation* (arXiv 2502.16471(待核实)) [威胁:medium] + - 与 C2 的第二分量(augmentation-consistency:扰动/增广后方向一致性)在直觉上最接近——都把『对扰动的鲁棒性』当作可迁移性信号。但它扰动的是特征空间用于模型选择评分,不是 label-preserving 输入增广,也不针对 probe 方向的旋转或 OOD 稳定性。是 augmentation-consistency 思想的近邻先例,需在 related work 中正面区分。 +- **Kirch, Dower, Skapars, Yannakoudakis, Lubana, Krasheninnikov (2025)** — *The Impact of Off-Policy Training Data on Probe Generalisation* (arXiv 2511.17408(待核实)) [威胁:medium] + - 与 C1(探针在分布偏移下泛化失败)主题高度相关,也触及『探针 OOD 迁移性』这一对象。但它是【诊断性/因果研究】(探究什么影响泛化),而非提出免训练的 a priori 预测分数,因此不抢 C2 的预测器贡献;对 C1 构成部分主题重叠(探针 OOD 不稳定的现象记录),但没有 C1 的 shift type × model size 矩阵 + 方向 cosine 旋转量化。 +- **Lee, Lee, Shin, et al. (2021)** — *Improving Transferability of Representations via Augmentation-Aware Self-Supervision (AugSelf)* (NeurIPS 2021) [威胁:low] + - 与 C2 augmentation-consistency 分量共享『增广与可迁移性挂钩』的底层假设,提供了『增广不变性≠总是好』的重要注意点(对 C2 增广一致性作为稳定性代理的有效性边界有警示意义)。但它是表征学习方法,不预测 probe 方向稳定性。属背景与潜在 limitation 讨论来源。 + +> 评估: 已联网检索(WebSearch/WebFetch 可用)。 + +【本角度最大撞车点】两层: +1) 范式层(medium):LEEP/LogME/H-score/GBC 早已确立『免训练、仅前向、用特征统计量事前预测可迁移性』这一完整范式。C2 不能宣称『免标注/免训练事前预测可迁移性』本身为新——必须把 C2 明确定位为『该范式在 linear probe 方向几何 + label-preserving 输入偏移下的 OOD 专用类比』,并强调新的预测量(方向离散度、增广方向一致性)和新的预测目标(方向稳定性而非微调准确率)。 + +2) 最近的具体抢占(high):Zhang et al. 2026《SAE as a Crystal Ball》是威胁最大的一篇——它已经在 LLM 表征上主张『免训练、仅 forward-pass、训练前 a priori 预测跨域(OOD)迁移性』,并给出强相关性证据。C2 与它的核心卖点高度撞句。差异点(C2 用 probe 方向的重采样离散度+增广一致性,目标是 probe 方向几何稳定性;该文用 SAE 特征维度偏移,目标是 SFT 整体迁移)是可辩护的,但论文 novelty 句必须显式引用并区分此文,否则审稿人极可能据此判定 incremental。建议把 C2 的卖点收窄为『首个针对 linear probe 方向(direction)在 label-preserving 偏移下 OOD 稳定性的免标注前向预测分数,且双分量分别捕捉 epistemic(重采样离散度)与 invariance(增广一致性)两类不稳定来源』。 + +【留白/可主张的空白】(a) 现有 transferability estimators 全部预测『迁移后准确率/性能』,无一预测『概念方向的几何稳定性(cosine 旋转)』——C2 的预测目标是真空地带;(b) 没有工作把『同分布 K 次重采样方向离散度』作为 probe 稳定性的事前信号——dispersion 分量看起来确属新;(c) 没有工作把 label-preserving 增广方向一致性 + NLI 标签保真审计(C5)接到 transferability 预测上;(d) 把 mass-mean/LogReg/MLP 探针类型(C3)与 model size(C4)纳入同一可迁移性预测框架,在本角度文献中未见。 + +【行动建议】Related Work 必须正面引用并区分:LEEP、LogME、H-score(范式),Zhang 2026 SAE Crystal Ball(最近 LLM-OOD 事前预测,重点避让),Feature Space Perturbation 2502.16471(扰动一致性的近邻),Kirch 2025 off-policy probe generalisation(探针 OOD 现象,支撑 C1 但需声明其非预测器)。最大风险是 C2 与 SAE Crystal Ball 的卖点表述撞车,务必收窄措辞。注:部分 arXiv 编号(2603.x/2602.x/2601.x)和作者列表为检索片段所得,投稿前需逐条人工核实标题/作者/venue。 + +### 角度: label-preserving 文本增广(back-translation / EDA / contrast sets / counterfactual au... + +- **Lysnæs-Larsen, Eggen & Strümke, 2025** — *Probing the Probes: Methods and Metrics for Concept Alignment* (arXiv preprint (cs.AI), 待核实是否已投会议) [威胁:high] + - 对 C1 与 C2 构成最直接的概念性撞车:C1 的核心论点'准确率高估特征稳定性'与该文'准确率不可靠'几乎同义;C2 的 augmentation-consistency 分量与其'augmentation robustness'度量同源。差异:该文偏视觉/spatial 概念定位,未做文本 paraphrase/domain/length 的 shift 矩阵,未提出免标注的 a priori OOD 迁移性'预测分数',也未含 dispersion 分量。 +- **Kaushik, Hovy & Lipton, 2020 (arXiv 2019)** — *Learning the Difference that Makes a Difference with Counterfactually-Augmented Data* (ICLR 2020) [威胁:medium] + - 为 C1 提供核心思想基础:分类器对 label-preserving 编辑敏感、依赖虚假特征。但其对象是端到端分类器而非探针的概念方向,且不量化 cosine 旋转、不提稳定性分数,也无 NLI 审计。属强背景威胁而非方法抢占。 +- **Gardner et al., 2020** — *Evaluating Models Local Decision Boundaries via Contrast Sets* (Findings of EMNLP 2020) [威胁:medium] + - 与 C1 的'准确率高估'叙事高度同构(标准评测 vs 扰动评测的 gap)。差异:面向端到端模型而非探针/概念方向,扰动多为改标签的对比而非纯 label-preserving 分布偏移,无 forward-pass 稳定性分数、无概念旋转度量。 +- **Wei & Zou, 2019** — *EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks* (EMNLP-IJCNLP 2019) [威胁:low] + - 提供 C2 增广分量与 C5 审计动机的工具与'增广未必真 label-preserving'的证据。本身不涉及探针稳定性或 NLI 审计,属背景。 +- **多篇 (e.g. Falsesum 2022; On-the-fly Denoising 2022; DISCO 2023; A Synthetic Data Approach for NLI 2024)** — *NLI/entailment 作为增广质量过滤器(overgenerate-and-filter / roundtrip consistency)* (ACL/EMNLP/NAACL 系列, 各篇待核实) [威胁:high] + - 对 C5 构成方法层面的强威胁——'用 NLI 核验 label 是否被保留'已是公认做法,DeBERTa-MNLI 做 fidelity 审计难算新方法,只能算工程实例化。C5 应降级为方法学保障而非核心新颖性主张。 +- **待核实 (检索命中, OOD probing transferability)** — *关于 probing 在 distribution shift 下迁移性的研究(binary vs multiclass probe OOD F1 显著下降)* (待核实) [威胁:medium] + - 直接触及 C1 的实证核心(探针 OOD 掉点)并部分触及 C3(探针类型对比)。需进一步核实该文是否已给出按 shift type × model size 的矩阵;若有则 C1/C3 威胁升至 high。当前证据为 medium。 +- **Sen et al. / Huang et al. 等, 2023 (待核实)** — *Counterfactually-Augmented SNLI Training Data Does Not Yield Better Generalization* (openreview / EMNLP 系列, 待核实) [威胁:low] + - 与 C1/C5 相关:佐证'增广的 label 质量与稳定性需被审计'的动机,但不提出探针稳定性分数,属支持性背景。 + +> 评估: 未联网核验全部 venue 的部分以"待核实"标注;以下基于本轮真实检索(WebSearch)结果与既有知识。 + +最大撞车点(按威胁排序): +1) C5 基本已被抢占。"用 NLI/entailment 模型核验增广/paraphrase 是否 label-preserving、不一致就过滤"是 2022–2024 的成熟范式(overgenerate-and-filter、roundtrip consistency、DeBERTa-MNLI 判同义)。把它作为核心新颖性会被审稿人直接驳回。建议把 C5 明确定位为"方法学保障/sanity check",而非贡献点,并主动引用这些过滤工作以示知情。 + +2) C1+C2 的概念框架面临 "Probing the Probes" (2025) 的正面威胁。该文已独立提出:探针准确率不可靠、需用 "augmentation robustness" 度量概念对齐。这与你们"准确率高估稳定性"(C1)和"augmentation-consistency 分量"(C2)在叙事和度量上同源。务必把它作为最近邻工作正面对比,凸显差异:(a) 你们在文本 NLP 上系统化 paraphrase/domain/length × model size 的偏移矩阵 + 概念方向 cosine 旋转;(b) 你们的 dispersion 分量(同分布 K 次重采样离散度)它没有;(c) 关键差异化——你们做的是 a priori、免新标注、训练前预测探针 OOD 迁移性的分数,而它停留在事后诊断度量。这条"预测性 + 双分量(dispersion ⊕ augmentation-consistency)"是目前看最可能站得住的真新颖点。 + +3) C1 的"准确率高估真实能力"叙事不是新发现:contrast sets (Gardner 2020)、counterfactually-augmented data (Kaushik 2020) 早已确立同构论点,且检索到的 OOD probing transferability 工作已实证探针 OOD 掉点。你们的增量必须落在"专门针对线性探针的概念方向、用 cosine 旋转量化、并跨 shift type × scale 成矩阵",而非"发现了掉点"本身。 + +仍存的空白(可主打的真新颖性): +- 双分量、forward-pass-only、免新标注、训练前 a priori 预测探针 OOD 迁移性的"分数"——未见同形态工作;dispersion(同分布重采样方向离散度)作为稳定性信号在文本探针语境下尚属空白。 +- "accuracy drop 与概念方向 cosine 旋转解耦"的矩阵化呈现(同一 shift 下准确率稳但方向已旋转)——未见直接对应工作,是较强卖点。 +- C3(mass-mean vs LogReg vs MLP 在 OOD 下相对可迁移性的系统比较)与 C4(model size ladder 对探针 OOD 稳定性影响)在本角度检索中未见被系统覆盖,威胁较低,可作为扎实的实证贡献,但需另行核查"OOD probing transferability"系列是否已部分触及 C3/C4。 + +诚实结论:C5 应放弃作为新颖性、改为方法保障;C1 需从"现象发现"转为"探针概念方向的量化方式";C2 是核心命脉,但必须在 "Probing the Probes" 之上明确切出"预测性 + dispersion 分量"的差异,否则有被判为增量的风险。 + +### 角度: 小语言模型/Pythia/GPT-2 上的可解释性与探针 size-ladder 研究(scaling of representations、Pythia in... + +- **Reblitz-Richardson, 2026** — *When Probing Accuracy Saturates, Fragility Resolves: A Complementary Metric for LLM Pre-Training Analysis* (arXiv 2606.11375 (cs.CL),待核实是否已投会议) [威胁:high] + - 对 C2 与 C1 威胁最大。C2:它同样是 forward-pass、免新标注、对‘准确率会高估稳定性’做补救的探针稳定性度量,与你的 dispersion 分量(对扰动的离散度/崩溃点)精神高度重合,只是它用激活噪声而非 label-preserving 增广/重采样。C1:与你‘准确率高估特征稳定性’的核心动机几乎同款表述。差异:它是噪声注入而非分布偏移(paraphrase/domain/length),不做方向 cosine 旋转矩阵,也不做训练前 a priori 预测 OOD 迁移。 +- **Huang, 2025** — *Spectral Identifiability for Interpretable Probe Geometry* (arXiv 2511.16288 (stat.ML),待核实) [威胁:high] + - 对 C2 威胁高。这正是‘训练前/无需新标注、用内在几何量 a priori 预测探针会不会失稳’的同类工作,与你 Probe Stability Score 的预测性目标直接撞车。差异:用 eigengap/Fisher 谱判据而非 dispersion+augmentation-consistency 双分量;主要在合成数据上验证,未明确做 Pythia/GPT-2 size-ladder、未做 label-preserving NLP 偏移、未给 shift×size 准确率+方向旋转矩阵。 +- **Anonymous/待核实, 2025** — *Factual Self-Awareness in Language Models: Representation, Robustness, and Scaling* (arXiv 2505.21399,待核实) [威胁:high] + - 对 C1+C4 威胁高。它已经做了‘Pythia size-ladder × 输入扰动 → 探针准确率下降’的矩阵式分析,且强调‘对表层噪声稳健、对语义偏移敏感’,与你 C1 的 accuracy-drop across shift×size 直接重叠,并覆盖 C4 的规模阶梯。差异:只测扰动下的准确率,不测概念方向 cosine 旋转;不提出预测性 stability score(C2);扰动非严格 NLI 审计的 label-preserving 集合(C5);单一任务(事实自知)。 +- **待核实, 2026** — *Linear Probe Accuracy Scales with Model Size and Benefits from Multi-Layer Ensembling* (arXiv 2604.13386,待核实) [威胁:medium] + - 对 C4 中度、C1/C3 低度。它直接做了‘探针准确率随 model size 提升’的 scaling 结论(C4 核心),并触及方向跨层旋转(与你的 cosine 旋转测量相邻但你的是跨 shift 旋转、它是跨层旋转)。差异:模型族非 Pythia/GPT-2;不做 label-preserving 分布偏移、不做免标注预测性 stability score、不比较 mass-mean/LogReg/MLP。 +- **Chou, Kirsanov, Yang & Chung, 2026** — *Diagnosing Generalization Failures from Representational Geometry Markers* (ICLR 2026) [威胁:medium] + - 对 C2 中度。这是‘用内在几何标记、免 OOD 标注、事先预测泛化失败’的同范式工作,与你 stability score 的预测性主张方法论撞车。差异:领域是视觉而非 LLM/Pythia,不用 augmentation-consistency/dispersion 双分量,不涉及 NLP label-preserving 偏移、方向旋转或 size-ladder。可作为‘几何预测泛化’这一思路的强先例,削弱 C2 的范式新颖性。 +- **Marks & Tegmark, 2023** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (arXiv 2310.06824(后续亦见会议版,待核实)) [威胁:high] + - 对 C3 威胁高、C4 低。它已经做了 mass-mean vs LogReg(vs CCS)在‘跨数据集/分布’下相对可迁移性的系统比较,结论是 mass-mean 更稳——这正是你 C3 的核心命题之一(只是它没纳入 MLP 探针,且‘OOD’指跨数据集而非 paraphrase/domain/length label-preserving 偏移)。差异:不做 size-ladder 矩阵、不做方向旋转量化、不提出预测性 stability score。你的 C3 若不强调与 MLP 的三方比较 + label-preserving shift 设定,很容易被认为是其复现。 +- **待核实(EmergentMind/相关综述指向), 2023** — *Gaussian Process Probes (GPP) for Uncertainty-Aware Probing* (arXiv 2305.18213,待核实) [威胁:low] + - 对 C2 低度背景相关。同样关注‘探针给出的判断有多可信/稳定’,但走贝叶斯不确定性路线,不预测 OOD 迁移、不用增广一致性/重采样离散度、不涉及 size-ladder 或方向旋转。可作为‘探针可靠性度量’的相关背景引用。 +- **Biderman et al., 2023** — *Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling* (ICML 2023) [威胁:low] + - 背景相关(C4 的实验底座)。本身不主张探针 OOD 稳定性结论,但你的 C4 实验若用 Pythia 则必须引用;它界定了‘size-ladder’这一研究范式,使单纯‘随规模变化’的发现难称新颖。 + +> 评估: 联网完成。本角度下你们面临的撞车风险偏高,且分散在 C1/C2/C3/C4 四条主张上,需要重新收紧定位。 + +最大的三个撞车点: +1) C1(准确率高估稳定性 + Pythia size-ladder × 输入扰动)已被 "Factual Self-Awareness"(2505.21399, 2025)在 Pythia 全 size-ladder 上基本做掉:它已给出‘扰动 × 规模 → 探针准确率下降’的矩阵,并明确‘对表层稳健、对语义偏移敏感’。你们若仅停留在 accuracy-drop,几乎是复现。你们尚存的差异——概念方向 cosine 旋转(direction rotation)、严格用本地 NLI 审计的 label-preserving shift 集合(C5)、覆盖 paraphrase/domain/length 三类偏移而非零散扰动——必须被抬成 C1 的主卖点,否则 C1 会被审稿人判为已有。 + +2) C2(免标注、forward-pass-only、训练前预测 OOD 迁移的探针稳定性 score)是你们最值钱也最危险的主张:同一‘事前预测探针可靠性/泛化失败’的范式已有至少三篇——Spectral Identifiability(2511.16288, eigengap/Fisher 判据)、Diagnosing Generalization Failures from Geometry Markers(ICLR 2026,几何量预测 OOD,免标签)、以及 Fragility(2606.11375,forward-pass 的崩溃点度量)。范式新颖性已被占。你们真正未被占的是具体机制:dispersion(同分布 K 次重采样方向离散度)+ augmentation-consistency(label-preserving 增广后方向一致性)这个‘双分量、基于方向而非准确率/谱/流形维度’的组合,且明确用于 NLP 探针的 a priori OOD 预测。建议把 C2 的 novelty 严格锚定在‘双分量方向一致性度量’及其相对上述谱/几何/噪声度量的预测增益对比(强烈建议做 head-to-head baseline:SIP、Fragility、manifold dimensionality 都要作为对照)。 + +3) C3(mass-mean vs LogReg vs MLP 的 OOD 相对可迁移性)中,mass-mean vs LogReg(vs CCS)的跨分布比较已由 Geometry of Truth(Marks & Tegmark 2023)做过,结论 mass-mean 更稳。你们的增量只剩:加入 MLP 探针构成三方比较 + 在 label-preserving paraphrase/domain/length 偏移(而非跨数据集)下重做 + 与 size-ladder 交叉。务必如此措辞,否则 C3 易被视为复现。 + +留有的空白(可作为你们真正的差异化护城河): +- 没有任何一篇把‘概念方向 cosine 旋转矩阵’作为 shift×size 的系统量化主轴——这是 C1 的真空白,应重点强化。 +- 没有人用‘重采样离散度 + 增广方向一致性’这个双分量、纯方向几何的免标注 score 来预测 NLP 探针 OOD 迁移,且没人把它和谱/流形/噪声类预测器做过对比——这是 C2 的真空白,但必须靠 baseline 对比来证明增益。 +- C4 单独看价值最低:‘探针随规模更稳/更准’已被多篇(2 + +### 角度: Probe reliability / calibration / trustworthiness — "when not to trust probes" a... + +- **Dies, Maynard, Savcisens & Eliassi-Rad, 2025** — *Representational Stability of Truth in Large Language Models* (arXiv:2511.19166 (preprint; venue待核实)) [威胁:high] + - Directly overlaps C1 (concept-direction rotation under paraphrase/perturbation) and C2 (dispersion-across-resamples as a stability signal). This is the single closest paper to our core mechanism — it already operationalizes 'direction rotation under rephrasing' + 'dispersion across resamples'. Differs in that it targets one concept (truth), does not frame the metric as an a-priori label-free OOD-transfer predictor, and lacks C3/C4/C5. +- **Kumar, Tan et al., 2022** — *Probing Classifiers are Unreliable for Concept Removal and Detection* (NeurIPS 2022 (arXiv:2207.04153)) [威胁:high] + - Overlaps C1 (probe accuracy overstates feature quality/stability) and the spirit of C2 (a label-free quality metric for probes). Establishes the 'do not trust probe accuracy' premise our paper depends on, and pre-empts the framing novelty of 'a metric to gauge probe quality.' Differs: focuses on concept removal/spurious correlation, not label-preserving distribution shift, not direction rotation, no OOD-transfer prediction, no estimator/scale ladder. +- **Lysnæs-Larsen, Eggen & Strümke, 2025** — *Probing the Probes: Methods and Metrics for Concept Alignment* (arXiv:2511.04312 (preprint; cs.AI)) [威胁:high] + - Overlaps C2's augmentation-consistency component directly (they call it 'augmentation robustness' as a probe-quality metric) and C1 (accuracy overstates faithfulness). Strong threat to our 'augmentation-consistency as a probe-reliability signal' framing. Differs: vision/CAV-centric, no dispersion-across-resamples component, no a-priori OOD-transfer prediction, no NLP paraphrase/domain shift, no scale ladder. +- **Liu, Chen, Cheng & He, 2024** — *On the Universal Truthfulness Hyperplane Inside LLMs* (EMNLP 2024 (arXiv:2407.08582)) [威胁:medium] + - Overlaps C1 (documents probe accuracy dropping OOD / cross-domain) and partially C3 (compares probing setups across datasets). Establishes the empirical 'probes don't transfer OOD' result we want to extend. Differs: requires labeled OOD datasets to measure transfer (no label-free a-priori predictor), no direction-rotation metric, no dispersion/augmentation score, no scale ladder. +- **Marks & Tegmark, 2023** — *The Geometry of Truth: Emergent Linear Structure in LLM Representations of True/False Datasets* (arXiv:2310.06824 (later COLM/workshop线; venue待核实)) [威胁:high] + - Directly overlaps C3 — it is the canonical mass-mean vs LogReg generalization comparison. Our C3 must position as an extension (adding MLP, label-preserving shift types, a stability-score lens) rather than a first comparison. Differs: no MLP, no paraphrase/length/domain shift taxonomy, no a-priori stability score, no scale ladder. +- **Baek, Raghunathan et al., 2024** — *Predicting the Performance of Foundation Models via Agreement-on-the-Line* (arXiv:2404.01542 (NeurIPS 2024线; venue待核实)) [威胁:medium] + - Overlaps C2's goal (label-free, forward-pass-only prediction of OOD performance) — establishes a strong prior-art baseline for 'predict OOD reliability without OOD labels.' Reviewers will demand comparison. Differs: uses inter-model agreement rather than direction dispersion/augmentation consistency; not probe/concept-direction specific; no rotation metric, no estimator/scale study. +- **Deng, Zheng et al., 2023** — *Confidence and Dispersity Speak: Characterising Prediction Matrix for Unsupervised Accuracy Estimation* (arXiv:2302.01094 (ICML 2023线; venue待核实)) [威胁:medium] + - Overlaps C2 conceptually — a label-free, dispersion-flavored predictor of OOD accuracy. Notably it uses the word 'dispersity,' close to our 'dispersion' component, which weakens the terminology/concept novelty of that sub-claim. Differs: operates on output prediction matrices, not on probe-direction geometry; no concept-direction rotation, no augmentation-consistency, not probe-faithfulness framed. +- **Belrose et al. / SAE-probe baseline studies, 2025** — *SAE probes underperform the baseline of logistic regression (SAE Features for Classifications and Transferability line)* (arXiv:2502.16681 / 2502.11367 (preprints)) [威胁:low] + - Background-to-medium overlap with C3 (probe-estimator transferability comparison under shift). Adds to the crowded space of 'which probe estimator transfers best' that our C3 sits in. Differs: SAE-vs-LogReg focus, no mass-mean/MLP triad, no a-priori stability score, no rotation metric. +- **Chou, Kirsanov, Yang & Chung, 2026** — *Diagnosing Generalization Failures from Representational Geometry Markers* (ICLR 2026 (arXiv:2603.01879)) [威胁:medium] + - Overlaps C2's high-level claim (predict OOD reliability from ID-only signal, no OOD labels). A conceptual competitor for the 'a-priori predictor' framing. Differs: vision/ImageNet, manifold-geometry not probe-direction dispersion/augmentation, no NLP paraphrase shift, no estimator/scale ladder. Adjacent companion: 'Representation Geometry as a Diagnostic for OOD Robustness' (arXiv:2602.03951) makes the same ID-only-predicts-OOD claim. +- **Belinkov, 2022** — *Probing Classifiers: Promises, Shortcomings, and Advances* (Computational Linguistics (MIT Press) 48(1)) [威胁:low] + - Background for C1/C5 — frames the 'probes can mislead' problem and the need for controls (relevant to our NLI label-fidelity audit in C5). Differs: survey, no quantitative shift matrix, no stability score, no OOD predictor. + +> 评估: 已联网检索(WebSearch/WebFetch可用)。结论:本角度对我们novelty构成实质威胁,但没有任何一篇把我们的完整组合(C2双分量 a-priori 免标注 forward-pass 探针稳定性分数 = dispersion + augmentation-consistency,用于训练前预测探针OOD迁移性)做完整。 + +最大撞车点(按优先级): +1) Representational Stability of Truth in LLMs (Dies et al., 2025-11) 是头号威胁——它已经把"概念方向在改写/扰动下的旋转"与"跨重采样的dispersion"两件事合在一起做了。这几乎覆盖了我们C1的旋转矩阵和C2的dispersion分量。我们必须把它当作最近的prior art正面引用并明确切割:它(a)只针对truth单一概念、(b)没有把这个量当成"训练前、免新标注的OOD迁移性预测器"来validate、(c)没有augmentation-consistency第二分量、(d)没有C3/C4/C5。建议把C2的卖点从"测量稳定性"上移到"用稳定性分数a-priori预测下游OOD迁移,并做预测-验证闭环",这是它没做的。 + +2) 关于C2的"免标注预测OOD"框架,存在一整条成熟的unsupervised-OOD-accuracy-prediction文献(Agreement-on-the-Line 2024;Confidence-and-Dispersity 2023——后者甚至用了"dispersity"一词,直接削弱我们"dispersion"术语新颖性;Representation-Geometry-as-Diagnostic 2026;Geometry Markers ICLR 2026)。这些不是探针专用,但reviewer几乎一定会要求对比/区分。我们的护城河必须是:作用在探针方向几何(不是输出预测矩阵或流形几何)、专为label-preserving语义shift设计、且做probe-faithfulness而非泛OOD acc。 + +3) C2的augmentation-consistency分量已被 "Probing the Probes" (2025-11) 以"augmentation robustness"之名作为探针质量度量提出(视觉/CAV域)。需引用并以"NLP label-preserving paraphrase + 与dispersion组合 + 作为预测器"区分。 + +4) C3(mass-mean vs LogReg)已被 Geometry of Truth (Marks & Tegmark 2023) 做过,是canonical对比,不能宣称首次;加MLP和label-preserving shift分类法是我们的增量,但增量较薄,建议把C3降格为支撑性发现而非主贡献。 + +5) C1的"探针accuracy高估特征稳定性/faithfulness"premise已被 Kumar et al. NeurIPS 2022 与 Belinkov 2022 survey 牢固确立,不可作为novelty主张,只能作为motivation。 + +留有的空白(我们仍可主张的真正新意): +- 把 dispersion + augmentation-consistency 显式组合成单一双分量分数,并在 shift-type × model-size 二维上验证它能a-priori排序探 + +--- + +## 四、独立验证记录(2026-06-22,直接抓取 arXiv abstract 复核) + +> 由 Claude 主回路用 WebFetch 逐条核对工作流给出的高危引用,确认其**真实存在**(并非 agent 幻觉)。 + +| arXiv ID | 实际标题 / 作者 | 核实结论 | 对我们的影响 | +|---|---|---|---| +| 2511.19166 | *Representational and Behavioral Stability of Truth in LLMs*(Dies, Maynard, Savcisens, Eliassi-Rad)— P-StaT 框架 | ✅ 真实 | 测 truth 信念在语义重述下的稳定性(fictional vs synthetic),**未见**双分量免标注 a-priori 预测闭环。最近邻,但 C2 仍有缝。 | +| 2602.00158 | *RAPTOR: Ridge-Adaptive Logistic Probes*(Gao 等) | ✅ 真实 | ridge 自适应 logistic 探针 + 概念向量 + stability;偏方法,非直接 OOD 预测。须作 baseline。 | +| 2510.11905 | *LLM Knowledge is Brittle: Truthfulness Representations Rely on Superficial Resemblance*(Haller, Ibrahim, Kirichenko, Sagun, Bell @ Meta) | ✅ 真实 | 仅覆盖 **paraphrase/reformulation + 仅 truthfulness**;**工作流高估**——未明确做 domain/length/多概念。C1 的"paraphrase×domain×length×多概念×drop⊥rotation 解耦矩阵"仍可区分。 | +| 2303.15488 | *On the Importance of Feature Separability in Predicting OOD Error*(Xie, Wei, Feng, Cao, An, NeurIPS 2023) | ✅ 真实 | label-free + feature dispersion 预测 OOD acc。标题非"Dispersion Score"但机制重叠 → **避免用 "Dispersion" 作主名 + 必跑对照**。 | +| 2602.20273 | *The Truthfulness Spectrum Hypothesis*(Ying, Ravfogel, Kriegeskorte, Hase) | ✅ 真实 | **Mahalanobis cosine 预测跨域泛化 R²=0.98** —— C2 最强竞品。但它需"探针间"相似度(隐含 OOD 侧 reference probe);我们 PSS 是 **IID-only**,这是唯一硬区分。 | +| 2511.16288 | *Spectral Identifiability for Interpretable Probe Geometry (SIP)*(W. Huang) | ✅ 真实 | eigengap a-priori 探针稳定性判据;偏理论+合成实验。必作 baseline。 | + +**结论**:该子领域(probe stability / truth directions)是 2025Q4–2026Q2 的高热赛道,多为 Meta/Northeastern 等强组工作。纯 novelty 竞争**高风险**。但所有竞品都可转化为 baseline —— 这为"benchmark-first"重定位提供了基础。 diff --git a/repro_bundle/PROPOSAL.md b/repro_bundle/PROPOSAL.md new file mode 100644 index 0000000000000000000000000000000000000000..d5797d4e137cabb621d8dacaddcfe1879c1c7a45 --- /dev/null +++ b/repro_bundle/PROPOSAL.md @@ -0,0 +1,159 @@ +# ProbeShift —— ACML 2026 主会投稿提案(benchmark-first 定位) + +> **定位决策(2026-06-22):** 经深度 prior-art 核查 + 直连 arXiv 验证(见 [PRIOR_ART.md](PRIOR_ART.md)), +> `probe stability / truth direction` 是 2025Q4–2026Q2 高热赛道,纯 novelty 竞争高风险。 +> 因此**把基准(benchmark)作为第一贡献**,所有竞品转化为在统一基准上评测的 baseline; +> PAC(我们的 label-free 复合分数)**降为"有竞争力的 IID-only 入选项,不宣称 SOTA"**。 +> 这样即使 PAC 不夺冠,"系统基准 + 统一评测"仍独立成立 —— 这是当前最稳的 accept 路线。 +> predictor-first 的备选 framing 仍保留在 git 历史 / `PROPOSAL_predictor_framing.md`。 + +--- + +## 1. 标题与一句话卖点 + +**中文标题:**《ProbeShift:线性探针方向在语义保真偏移下 OOD 稳定性的系统免标注基准,及对训练前预测信号的统一评测》 + +**英文标题:** *ProbeShift: A Systematic, Label-free Benchmark for the OOD Directional Stability of Linear Probes under Label-preserving Semantic Shift, with a Unified Evaluation of A-priori Predictors* + +**一句话卖点:** 我们**不**主张"首次发现探针不稳",也**不**主张某个新预测器最强;我们提供该方向**第一个系统、免新标注、跨多概念 × 多偏移类型 × 多模型规模 × 多探针类型**的探针 OOD 方向稳定性基准,并在统一基准上对 **5 个现有训练前 label-free 信号**(SIP / RAPTOR-stability / Fragility / augmentation-robustness / Xie-dispersion)做首次 apples-to-apples 评测,诚实回答"**到底什么信号能在训练前预测探针会不会迁移**"。 + +--- + +## 2. Abstract(中文,约 190 词) + +线性探针被广泛用作 LLM 内部表示的"免标注读出器",但其 in-distribution(ID)准确率系统性高估了它在语义保真偏移(paraphrase / domain / length)下的方向稳定性。近期涌现的诊断与预测工作各自为政、benchmark 与协议互不可比:有的只度量不稳(Dies et al. 2025 的 P-StaT、Haller et al. 2025),有的需触碰 OOD 数据做白化(Truthfulness Spectrum 2026),有的只给单一几何判据(SIP eigengap、Fragility、RAPTOR directional stability、Probing-the-Probes augmentation-robustness)。本文提供 **ProbeShift**:一个完全免新标注、随激活缓存一并发布、可在数分钟内复现的探针 OOD 方向稳定性基准,系统覆盖 concept × shift-type × model-size × estimator,并将 accuracy-drop 与 direction-rotation **解耦**为两维 ground truth。在此基准上,我们首次对 5 个现有训练前 label-free 预测信号做统一 head-to-head 评测(Spearman ρ / Kendall τ,leave-one-concept/shift-out 交叉验证),给出"何种信号预测探针 OOD 稳定性"的可比结论。我们另提供一个简单的 IID-only 复合分数 **PAC**(dispersion ⊕ augmentation-consistency)作为入选项,诚实报告其在何处带来增量、何处现有单分量信号已足够。全部实验受限于单卡 4090 ≤200 GPU·h、API <$100、零新人工标注。 + +--- + +## 3. Introduction:研究缺口与贡献 + +### 3.1 缺口 +探针被当作真实性/欺骗/概念检测的免标注评测器,但在分布偏移下方向几何崩塌。实践者需要**部署前**判断"这个探针方向能否迁移",理想情况下还**不应需要 OOD 数据**。问题在于:相关信号(SIP/Fragility/RAPTOR/aug-robustness/Xie/Truth-Spectrum/P-StaT)散落在**不同概念、不同模型、不同偏移定义、不同评测口径**下,**无人在一个统一基准上比较它们**,导致"到底什么能预测探针 OOD 稳定性"这一实践问题无答案。 + +### 3.2 贡献(formal,编号) +- **(G1,主)ProbeShift 基准。** concept(≥12)× shift-type(paraphrase/domain/length)× model-size(Pythia ladder + GPT-2 + Qwen)× estimator(LogReg/mass-mean/MLP)的网格,记录**解耦的** accuracy-drop 与 direction cosine-rotation 作为 ground truth。完全免新标注;随**激活缓存**发布,reviewer 数分钟可复现。 +- **(G2,主)统一 a-priori 预测信号评测。** 首次在同一基准上 head-to-head 评测 SIP(eigengap)、RAPTOR directional stability、Fragility、augmentation-robustness、Xie feature-dispersion 五个现有 label-free 训练前信号对探针 OOD 方向 drop 的预测力(Spearman ρ / Kendall τ + held-out CV),并与"需 OOD 数据"的 Truth-Spectrum 白化 cosine 作上界对照。**诚实结论而非夺冠**。 +- **(G3,次)PAC 入选项。** 一个简单、IID-only、免 OOD 数据的双分量复合分数(dispersion ⊕ augmentation-consistency),作为基准上的一个 entry;做互补性消融,报告其增量边界。**不宣称 SOTA。** +- **(支撑)** estimator × size 的外部效度;基于本地 NLI 的 label-fidelity sanity check(主动引用 overgenerate-and-filter 文献)。 + +> **关键 framing:** G1+G2 即使 PAC(G3)毫无增量也独立成立 —— 这是把"赛道很挤"从威胁转成贡献("我们统一比较了所有相关工作")的核心。 + +--- + +## 4. Related Work(逐能力对照表 + 逐篇切割) + +下表中所有 high-threat 文献均经直连 arXiv 核实(见 PRIOR_ART.md §四);ProbeShift 的定位是**容纳**它们为 baseline,而非取代。 + +| 能力维度 | Dies/P-StaT | SIP | Fragility | RAPTOR | Probing-Probes | Xie | Truth-Spectrum | **ProbeShift(本文)** | +|---|---|---|---|---|---|---|---|---| +| 统一可比基准 | N | N | N | N | N | N | N | **Y(G1)** | +| 多预测器 head-to-head | N | N | N | N | N | N | N | **Y(G2)** | +| accuracy-drop ⊥ rotation 解耦 | 部分 | N | N | N | N | N | 部分 | **Y** | +| 跨多概念(≥12) | N(truth) | 合成 | N | concept | 视觉 | 视觉/分类 | 5 truth | **Y** | +| 跨 size ladder | N | N | Y | Y | N | N | N | **Y** | +| 完全免 OOD 数据 | N | Y | Y | Y | Y | Y | N | **Y** | +| 随激活缓存发布、分钟级复现 | N | N | N | N | N | N | N | **Y** | + +**逐篇切割(正文必写):** P-StaT 触碰 OOD 扰动、仅 truth、不预测特定 drop → 我们将其 perturbation 纳入 ground truth 并复现为 baseline;SIP/Fragility/RAPTOR/aug-robustness 各为单一信号 → 我们统一评测之;Xie 预测整体 acc 而非探针方向几何 → §7.4 分离实验证明二者机制不同;Truth-Spectrum 需 OOD covariance → 列为"放宽约束上界"。**全文删除一切 "first predictor / novel framing" 措辞,只保留 "first unified benchmark + evaluation"。** + +--- + +## 5. Formal Claims(可证伪 + 预注册阈值,看到 held-out 前冻结) + +- **H1(基准有效性):** ID accuracy 高估方向稳定性。判定:跨配置 Spearman ρ(ID-acc, OOD-retention) 的 bootstrap 95% CI 上界 < 0.5。 +- **H2(G2 主结论):** 存在至少一个现有 label-free 信号能 held-out 预测 OOD 方向 drop。判定:至少一个 baseline 的 held-out ρ 95% CI 下界 > 0.4;**报告全部信号的排名表**(这是 G2 的产出,无论谁赢)。 +- **H3(G3 互补性):** 若 PAC 入选,合成需显著优于其两单分量。判定:配对 Wilcoxon(Holm 后 p<0.05)+ Cliff's δ ≥ 0.33;**否则诚实降级为单分量 entry**。 +- **H4(外部效度):** 排名结论跨 estimator 与跨 size 稳健;不下"越大越稳"普适结论,显式检验 Pressure-Testing-Deception(2605.27958)的 inverse-scaling 是否为 training-distribution artifact。 +- **H5(sanity):** shift 确为 label-preserving。判定:NLI 审计通过率 ≥95%,作者内部抽检(每 shift-type 50 条,**非新众包**)一致率 ≥90%。 + +--- + +## 6. Method:基准构造 + PAC 定义 + +### 6.1 基准 ground truth(G1) +对每个 (concept c, shift s, size m, estimator e):在 ID 训练集得探针方向 **w̄**;在 OOD(shift)集得 **w_s**。记录两维: +- **accuracy-drop** = ID_acc − OOD_acc(用 ID 探针在 OOD 集上预测)。 +- **direction-rotation** = 1 − |cos(**w̄**, **w_s**)|(在 OOD 集上重拟合方向,度量几何旋转)。 +解耦报告二者(回应"准确率与方向不是一回事")。 + +### 6.2 a-priori 预测信号(G2,全部 label-free、IID-only) +统一接口下实现并评测: +- **SIP**:LDA 相关子空间的 eigengap / Fisher 误差(越大越稳)。 +- **RAPTOR directional stability**:K 次 bootstrap 方向 mean|cos|(= dispersion 分量)。 +- **Fragility**:对 ID 激活注入递增各向同性噪声,临界崩溃 σ(越大越稳)。 +- **augmentation-robustness**:label-preserving 文本增广后方向一致性(= aug 分量)。 +- **Xie feature-dispersion**:inter-class 特征离散度(预测整体 acc 的通用信号,作机制对照)。 +- **(上界对照)Truth-Spectrum 白化 cosine**:Mahalanobis 白化(注:需 OOD covariance,故列为放宽约束上界)。 + +### 6.3 PAC 入选项(G3) +PAC(c,ℓ,e) = σ( α·z(1−D_disp) + (1−α)·D_aug ),α 为唯一聚合超参,**仅在 dev 概念-shift 组合上选**,选定冻结。D_disp 为方向重采样离散度,D_aug 为增广一致性,白化仅用 **ID** covariance(守住免 OOD 约束)。 + +### 6.4 防泄漏(生死线) +所有 G2/G3 结果只在**完全 held-out 的 concept × shift-type × size 组合**上算;主结果用 **leave-one-concept-out / leave-one-shift-type-out** 双交叉验证;split 清单随码发布。 + +### 6.5 Shift 构造 + label-fidelity(sanity) +paraphrase(LLM API 改写)/ domain(领域措辞改写)/ length(扩缩写),逐字对齐 Haller et al. 2025 三类定义。overgenerate-and-filter:API 过量生成 → 本地 DeBERTa-v3-MNLI 双向 entailment(roundtrip)过滤。主动引用 Falsesum 2022 / DISCO 2023。 + +--- + +## 7. Experiments + +### 7.1 数据/模型 +- **概念 ≥12**:truth、sentiment、formality、toxicity、language-ID、tense、negation、topic(≥5 类)等,覆盖 factual + stylistic,保证统计功效。 +- **size ladder**:Pythia 70M–2.8B + GPT-2 small/medium + Qwen2.5-0.5B(单卡可载)。 +- 激活一次前向缓存,所有信号共享缓存。 + +### 7.2 主实验 +1. 构建 ProbeShift ground-truth 矩阵(G1)。 +2. 计算全部 a-priori 信号(含 PAC)。 +3. 报告每个信号 held-out **Spearman ρ / Kendall τ** + bootstrap CI + 排名表(G2)。 + +### 7.3 关键消融/分离 +- **互补性(H3):** dispersion-only / aug-only / PAC 合成,配对 Wilcoxon + Cliff's δ + Holm。 +- **方向几何 ≠ 流形 dispersion:** paraphrase shift 上对比 PAC 与 Xie,预期 Xie 失效、PAC 不失效。 +- **度量消融:** 裸 cosine vs ID-白化 cosine。 +- **size 效应:** 预测有效性随 size 是否单调;检验 2605.27958 artifact。 + +### 7.4 统计 +每配置 **5 seeds**;相关系数报 **bootstrap(10k)95% CI**;多重比较 **Holm-Bonferroni**;效应量 **Cliff's δ / Cohen's d**;预注册阈值 §5。 + +--- + +## 8. Expected Results 与负结果叙事 + +**预期产出(无论谁赢都成立):** ProbeShift 排名表给出"现有信号谁能预测探针 OOD 方向 drop";很可能发现单一信号(如 SIP 或 aug-robustness)在某些 shift 类型上强、在另一些上弱,而**没有单信号通吃** —— 这本身就是有价值的 G2 结论。 + +**诚实负结果(预先写入):** +- 若 PAC 无显著增量 → 降为"现有单分量已足够 + PAC 仅作统一框架下的参照",G1+G2 不受影响。 +- 若所有信号 held-out ρ 都低 → 这是更强的 G2 结论:"探针 OOD 稳定性目前不可被现有 label-free 信号可靠预测",指明 open problem(对 benchmark 论文是加分而非减分)。 +- 若 size 出现 inverse scaling → 不下普适结论,引 2605.27958 解释为 artifact。 + +--- + +## 9. Reproducibility +发布:激活缓存、探针权重、全部 baseline + PAC 实现、预注册阈值、shift 数据集 + NLI 通过率、seed 列表、leave-one-out split 清单、bootstrap 脚本、一键复现入口。 + +## 10. Budget(单卡 4090 ≤200 GPU·h,API <$100) + +| 项目 | 估算 | 说明 | +|---|---|---| +| 激活缓存(12 概念 × 7 模型)| ~40 GPU·h | 一次前向,全 baseline 复用 | +| 探针训练(3 estimator × bootstrap × 5 seed)| ~50 GPU·h | LogReg/mass-mean 极廉,MLP 主成本 | +| Augmentation 重训(3 shift)| ~45 GPU·h | 增广激活 + 重训 | +| Baselines(SIP/Fragility/RAPTOR/aug/Xie/Geometry)| ~35 GPU·h | Fragility 噪声扫描最贵 | +| P-StaT baseline 复现 | ~15 GPU·h | 用其公开 code/data | +| 缓冲 | ~15 GPU·h | 合计 **~200 GPU·h** | +| **API(3 shift × 12 概念 × overgenerate 3×)** | **<$80** | GPT-class API;NLI 过滤本地零成本 | + +## 11. Risk & 退路 + +| 风险 | 缓解 / 退路 | +|---|---| +| benchmark 被指"只是把已有数据拼一起" | 强调三个新东西:**解耦 drop⊥rotation**、**跨 ≥12 概念统一口径**、**首次多预测器 head-to-head**;且随缓存发布、分钟级复现是实打实的社区价值。 | +| 数据泄漏/循环验证(F4) | held-out + 双交叉验证 + split 随码发布。 | +| PAC 无增量(M1) | 已 framing 为次要 entry,G1+G2 独立成立;诚实降级。 | +| size inverse-scaling 反例(M2) | 仅交叉维度,复现/检验 2605.27958 artifact,不下普适结论。 | +| baseline 覆盖不足(M4) | A 组 4 信号 + Xie 全跑,B/C 各 ≥1;缺项写 limitation。 | +| P-StaT 修订版加入预测闭环 | 不影响本 framing(我们是基准+评测,不与单篇争"首个预测器");投稿前二次复核其 venue/版本。 | + +**已核实文献(投稿前复核 venue):** Dies/P-StaT [2511.19166](https://arxiv.org/abs/2511.19166) · SIP [2511.16288](https://arxiv.org/abs/2511.16288) · Fragility [2606.11375](https://arxiv.org/abs/2606.11375) · RAPTOR [2602.00158](https://arxiv.org/abs/2602.00158) · Probing-the-Probes [2511.04312](https://arxiv.org/abs/2511.04312) · Xie [2303.15488](https://arxiv.org/abs/2303.15488) · Truth-Spectrum [2602.20273](https://arxiv.org/abs/2602.20273) · Haller/Brittle [2510.11905](https://arxiv.org/abs/2510.11905) · Pressure-Testing-Deception [2605.27958](https://arxiv.org/abs/2605.27958) diff --git a/repro_bundle/PROPOSAL_predictor_framing.md b/repro_bundle/PROPOSAL_predictor_framing.md new file mode 100644 index 0000000000000000000000000000000000000000..250b3584f4a0c2ca8a357d0334df780aa4c9bc5b --- /dev/null +++ b/repro_bundle/PROPOSAL_predictor_framing.md @@ -0,0 +1,219 @@ +# [备选 framing] Predictor-first 提案(PAC 作主贡献) + +> 此为 predictor-first 备选;当前主选为 benchmark-first(见 PROPOSAL.md)。保留以备 PAC 增量显著时升级 framing。 + +--- + +All verified. The counterexample paper (2605.27958) actually concludes the inverse-scaling pattern is a *training-distribution artifact*, not genuine — which I can use to frame M2 carefully. I now have fully verified ground truth. Writing the complete proposal. + +# 《Probe Stability Score》ACML 2026 主会投稿提案(完整版) + +> **撰写前置说明(已完成 prior-art 核验)**:本提案撰写前已对定位文档第 4 节"必引清单"中全部 10 篇 high-threat 文献逐条在 arXiv 核实标题、作者与机制。核实结果直接改写了若干防守策略,关键修正如下(投稿前仍须复核 venue/接收状态): +> - **Dies et al. 2025(2511.19166)** 真实标题为 *Representational **and Behavioral** Stability of Truth in LLMs*,框架名 **P-StaT**。**核实确认:它是"主动扰动数据 + 度量 retraction(最高 32.7%)"的评测框架,既未把 rotation+dispersion 合成单一分数,也未做 a priori OOD 预测,且全程触碰 OOD 数据、仅限 truth 单概念。** 这是 F1 生死线的关键有利事实。 +> - **Xie 2023(2303.15488)** 真实标题为 *On the Importance of Feature Separability in Predicting OOD Error*;其分数为 "feature dispersion",**并未正式命名为 "Dispersion Score"**,且预测目标是整体 test accuracy 而非探针方向几何。F3 改名 + 机制切割成立。 +> - **Truthfulness Spectrum(2602.20273)** 确认其 R²=0.98 依赖 **test-set covariance(即 OOD reference)** 做 Mahalanobis 白化;PSS "免 OOD 数据"护城河有效。 +> - **Probing the Probes(2511.04312)** 真实副标题 *Methods and Metrics for Concept Alignment*,其 augmentation-robustness 面向 **视觉 CAV / 空间概念(translation invariance)**,非文本 paraphrase。切割成立。 +> - **Liang/Brittle(2510.11905)** 第一作者实为 **Patrick Haller et al.**,范围 4 LLM 家族 × 5 数据集 × 3 探针方法,只做 truth。 +> - **Pressure-Testing Deception(2605.27958)** 结论是"inverse scaling 是 training-distribution artifact 而非真实 scale 现象"——M2 据此改为"不下普适 scaling 结论,并复现该 artifact 解释"。 +> - RAPTOR(2602.00158)、SIP(2511.16288)、Fragility(2606.11375)、Marks & Tegmark(2310.06824)均核实属实。 + +--- + +## 1. 标题与一句话卖点 + +**中文标题:** 《探针锚定一致性分数:一种完全免 OOD 标注、仅前向传播的双分量复合指标,用于训练前 a priori 预测线性探针方向在语义保真偏移下的几何稳定性》 + +**英文标题:** *Probe Anchoring Consistency (PAC): A Label-free, Forward-pass Composite for A-priori Prediction of Linear-probe Directional Stability under Label-preserving Semantic Shift* + +> **重要改名(回应 F3):** 弃用 "Probe Stability Score / Dispersion" 作主词根,改用 **Probe Anchoring Consistency(PAC)**,避免与 Xie 2023 的 feature dispersion 第一眼撞名。"dispersion" 仅作内部分量技术名出现,绝不作总分数名。下文 PAC = 原 PSS。 + +**一句话卖点:** 我们不主张"首次发现探针不稳",而主张提供一个**在严格 held-out、完全不触碰 OOD 数据的约束下,对 SIP / Fragility / Augmentation-robustness / Xie-dispersion / mass-mean 五个 baseline 取得可经 bootstrap 显著性检验之增量**的 label-free 复合预测器,并闭环验证它能排序真实 OOD drop。 + +--- + +## 2. Abstract(中文,约 190 词) + +线性探针(linear/concept probe)被广泛用于读取 LLM 内部表示,但其 in-distribution(ID)准确率系统性高估了它在语义保真偏移(paraphrase / domain / length)下的方向稳定性。已有工作或停留在"诊断不稳"(Dies et al. 2025 的 P-StaT、Haller et al. 2025),或需触碰 OOD 数据做白化(Truthfulness Spectrum 2026,R²=0.98),或仅给单分量几何判据(SIP、Fragility、RAPTOR directional stability、Probing-the-Probes augmentation-robustness)。我们提出 **Probe Anchoring Consistency(PAC)**:一个仅用前向传播、完全免新标注与免 OOD 数据的双分量复合分数,分量一为方向重采样离散度(dispersion),分量二为文本增广一致性(augmentation-consistency),二者经 held-out 校准聚合为单一分数。我们的核心主张不是 framing 新颖性,而是:(i) 两分量**合成**相对各自单分量有统计显著增量(配对检验 + effect size);(ii) PAC 能在训练前 a priori **排序** 一个 shift-type × model-size × concept 矩阵中的真实 OOD drop(Spearman ρ / Kendall τ,leave-one-concept/​shift-out 交叉验证);(iii) 在零 OOD 数据约束下,对上述五个 baseline 取得 bootstrap 显著的预测增量。全部实验受限于单卡 4090 ≤200 GPU·h。 + +--- + +## 3. Introduction:研究缺口与贡献列表 + +### 3.1 研究缺口 + +线性探针正被当作"免标注评测器"用于真实性、欺骗、概念检测(Marks & Tegmark 2023;Pressure-Testing Deception 2026)。但探针在 clean benchmark 上 AUROC > 0.96,而在分布偏移下方向几何崩塌(Haller et al. 2025;2605.27958)。实践者需要一个**部署前**就能判断"这个探针方向能否迁移"的指标,且理想情况下**不应需要 OOD 数据**(否则与直接在 OOD 上评测无异)。 + +现有候选各缺一角: +- **诊断型**(Dies/P-StaT、Fragility):度量不稳,但不预测特定 OOD drop,且 P-StaT 触碰 OOD 扰动数据。 +- **需 OOD reference 型**(Truthfulness Spectrum):R²=0.98 但需 test-set covariance 白化。 +- **单分量 a priori 型**(SIP eigengap、RAPTOR mean|cos|、Probing-the-Probes augmentation):各占一个分量或一个几何判据,无人**合成两分量**并验证互补增量。 +- **通用免标注 OOD 预测**(Xie 2023 feature dispersion、Deng 2023、Baek 2024):预测整体 accuracy / 流形几何,非**探针方向几何**。 + +**缺口一句话:** 无人在"双分量合成 + 完全免 OOD 数据 + a priori 预测 ↔ 验证闭环 + 针对探针方向几何"四约束同时成立下,给出经显著性检验、跑赢上述 baseline 的预测器。 + +### 3.2 贡献列表(formal,编号;已按红队收窄) + +- **(主贡献 G1)** 提出 **PAC**——dispersion ⊕ augmentation-consistency 的 label-free、forward-pass、免 OOD 数据复合分数,并在**严格 held-out(leave-one-concept-out / leave-one-shift-type-out)**下证明其 a priori 排序真实 OOD directional drop 的能力(Spearman ρ / Kendall τ + bootstrap CI)。 +- **(主贡献 G2)** **互补性消融**:证明双分量合成相对 dispersion-only、augmentation-only 有**配对显著**增量(Wilcoxon + Cliff's δ),即 augmentation-consistency 在 dispersion 之外携带增量信息。 +- **(主贡献 G3)** **Head-to-head baseline 增量**:在零 OOD 数据约束下,PAC 对 SIP / Fragility / RAPTOR-stability / Probing-the-Probes-augmentation / Xie-feature-dispersion 取得 bootstrap 显著的预测增量;并与需 OOD 的 Truthfulness-Spectrum 白化 cosine 做"放宽约束上界"对照。 +- **(支撑发现 S1=原 C1,降为被预测真值)** 构建 shift-type × model-size × concept 的 **accuracy-drop / cosine-rotation 解耦二维矩阵**,作为 PAC 预测的 ground-truth,而非独立贡献。 +- **(外部效度 S2=原 C3+C4,降级)** mass-mean / LogReg / MLP 三类 estimator 与 size ladder 仅作为"PAC 预测有效性是否跨 estimator、跨 size 成立"的外部效度检验,不并列。 +- **(Sanity check S3=原 C5,降级)** 基于本地 NLI 的 label-fidelity 审计协议,定位为方法学 sanity check,主动引用 NLI-filter 文献。 + +--- + +## 4. Related Work(含逐 claim 定位表 + 逐篇切割) + +### 4.1 主题归类(略写,完整见附录引用) +A. 探针准确率不可靠 / faithfulness;B. 真值几何 / 方向旋转 under shift;C. 探针方向稳定性度量;D. 训练前 a priori 判据;E. 免标注 OOD 性能预测;F. transferability estimation;G–I. estimator 比较 / size ladder / NLI 审计。 + +### 4.2 与每篇最接近工作的逐能力对照表(直接回应 F1/F2/F3) + +| 能力维度 | Dies/P-StaT (2511.19166) | SIP (2511.16288) | Fragility (2606.11375) | RAPTOR (2602.00158) | Probing-Probes (2511.04312) | Xie (2303.15488) | Truth-Spectrum (2602.20273) | **PAC(本文)** | +|---|---|---|---|---|---|---|---|---| +| rotation 分量 | Y | N | N | N | N | N | Y(cos) | **Y** | +| dispersion 分量 | Y | 谱gap | 噪声崩点 | Y(mean\|cos\|) | N | feature流形 | N | **Y** | +| augmentation-consistency 分量 | N | N | N | N | Y(视觉CAV) | N | N | **Y** | +| **两分量合成单一分数** | **N** | N | N | N | N | N | N | **Y** | +| a priori **预测特定 OOD drop** | **N(只度量retraction)** | N(只给ID判据) | N(只给崩点) | N(只诊断) | N | Y(整体acc) | Y(需OOD cov) | **Y(探针方向drop)** | +| **完全免 OOD 数据** | **N** | Y | Y | Y | Y | Y | **N(需test cov)** | **Y** | +| 预测目标=探针方向几何 | 部分 | ID可识别 | 崩点 | 诊断 | CAV对齐 | **N(流形)** | Y | **Y** | +| 跨多概念(≥10) | N(truth) | 合成 | N | concept | 视觉 | 视觉/分类 | 5 truth类型 | **Y(目标≥12)** | +| size ladder | N | N | Y | Y | N | N | N | **Y** | +| **预测↔验证闭环** | **N** | N | N | N | N | 部分 | 部分 | **Y** | + +**逐篇 one-paragraph 切割(论文正文必写):** +- **vs Dies/P-StaT:** 经核实其为评测框架,度量 retraction 而**不预测**,主动**触碰 OOD 扰动数据**,且**不合成单一分数**、仅 truth。PAC 的"合成 + a priori 预测 + 免 OOD"三格对其为 N→Y。我们将 P-StaT 的 perturbation 集合作为 ground-truth drop 来源之一(被预测真值),并复现其作为 baseline。 +- **vs SIP / Fragility:** 二者是单分量 a priori 判据(eigengap / 崩点),不预测特定 shift 下的 drop 排序。二者作为 baseline 跑数;PAC 主张是**增量**而非取代,删除一切 "first" 措辞。 +- **vs Xie 2023:** 已核实非 "Dispersion Score" 命名、预测整体 acc。我们改名规避撞名,并实验证明"探针方向几何 ≠ feature 流形 dispersion"(见 §7.4 分离实验)。 +- **vs Truthfulness Spectrum:** 其 R²=0.98 **依赖 OOD test covariance**。PAC 在更苛约束(零 OOD)下逼近;我们诚实将白化 cosine 列为"放宽约束的上界对照",而非声称跑赢。 + +--- + +## 5. Formal Claims & Hypotheses(可证伪 + 预注册阈值) + +所有阈值在**看到 held-out 测试结果前**冻结,写入预注册文件随代码发布。 + +- **H1(原 C1,被预测真值):** 对每个 (concept, shift-type, size),探针 ID accuracy 与 OOD directional retention(1−rotation)弱相关。**判定:** 若跨配置 Spearman ρ(ID-acc, OOD-retention) 的 bootstrap 95% CI 上界 < 0.5,则 H1 成立(准确率确实高估稳定性)。 +- **H2(主,G1 闭环):** PAC 在 held-out 配置上排序真实 OOD directional drop。**判定:** held-out Spearman ρ(PAC, −drop) ≥ 0.6 且 bootstrap 95% CI 下界 > 0.4。 +- **H3(主,G2 互补性):** 合成 PAC 显著优于两个单分量。**判定:** 配对 Wilcoxon(合成 vs dispersion-only,以及 vs augmentation-only)p < 0.05(Holm 校正后),且 Cliff's δ ≥ 0.33(中等以上效应)。**若不成立,坦诚降级为单分量方法(回应 M1)。** +- **H4(主,G3 baseline 增量):** PAC 对每个 a priori baseline 的预测 ρ 有正增量。**判定:** Δρ(PAC − baseline) 的 bootstrap 95% CI 下界 > 0,对 SIP/Fragility/RAPTOR-stab/Aug-robust/Xie 五者分别检验,Holm 校正。 +- **H5(外部效度,S2):** PAC 的预测有效性跨 estimator(mass-mean/LogReg/MLP)与跨 size 成立。**判定:** 各 estimator 子集与各 size 子集的 ρ 均 > 0.5;**不下"越大越稳"普适结论**,并显式测试 2605.27958 的 inverse-scaling 是否为 training-distribution artifact。 +- **H6(sanity,S3):** 构造的 shift 确为 label-preserving。**判定:** NLI 审计通过率 ≥ 95%,且与人工抽检(每 shift-type 抽 50 条,作者内部标注,**非新众包标注**)一致率 ≥ 90%。 + +--- + +## 6. Method:PAC 的精确定义 + +### 6.1 记号 +设隐状态层 ℓ、概念 c、估计子 e。在 ID 训练集 D 上得探针方向 **w**(单位向量)。语义保真增广算子集 𝒜 = {paraphrase, domain-reframe, length}(label-preserving)。 + +### 6.2 分量一:Dispersion(方向重采样离散度,免 OOD) +对 D 做 B 次 bootstrap 重采样,得方向 {**w**_b}。以 mean direction **w̄** 为锚: + + D_disp(c,ℓ,e) = 1 − (1/B) Σ_b |cos(**w**_b, **w̄**)| + +值越小越稳。**度量消融(回应 M3):** 同时计算裸 cosine 与 Mahalanobis 白化 cosine(白化协方差仅用 **ID** covariance,绝不用 OOD,以守住免 OOD 约束),报告 PAC 对度量选择的稳健性。 + +### 6.3 分量二:Augmentation-consistency(文本增广一致性,免 OOD、免标注) +对 ID 样本施加 𝒜 中算子得增广版本(**仍是 ID 语义、标签由构造保持**,经 §6.4 审计),在增广前向激活上**重训**探针得 **w**_a: + + D_aug(c,ℓ,e) = 1 − (1/|𝒜|) Σ_{a∈𝒜} |cos(**w**_a, **w̄**)| + +关键:这里增广是对 **ID 数据**做 label-preserving 变换以探测方向脆性,**不是 OOD 测试数据**——这正是与 Truthfulness Spectrum 触碰 OOD 的区别。 + +### 6.4 合成(单一分数,held-out 校准) + PAC(c,ℓ,e) = σ( α·z(D_disp) + (1−α)·z(D_aug) ) +其中 z(·) 为基于 **dev 配置** 统计量的标准化,α∈[0,1] 为唯一聚合超参。 + +**严格防泄漏(回应 F4,生死线):** +- α **只在 dev 概念-shift 组合上**用 grid {0,0.1,…,1} 选定,选定后冻结。 +- 所有报告结果只在**完全 held-out 的 concept × shift-type × size 组合**上计算。 +- 主结果用 **leave-one-concept-out** 与 **leave-one-shift-type-out** 双交叉验证,证明 PAC 跨概念、跨 shift-type 泛化而非记忆 S1 矩阵。 + +### 6.5 Shift 构造与 label-fidelity 协议(S3,sanity) +- 三类 label-preserving shift:**paraphrase**(LLM API 改写)、**domain**(改写为新领域措辞)、**length**(扩写/缩写)。逐字对齐 Haller et al. 2025 的三类定义以便 head-to-head。 +- **生成-过滤管线(overgenerate-and-filter,主动引用 Falsesum 2022 / DISCO 2023):** API 过量生成 → 本地 DeBERTa-v3-MNLI 双向 entailment 过滤(roundtrip consistency)→ 仅保留双向 entailment 的样本。 +- 审计定位为 **sanity check**,报告通过率与作者内部抽检一致率(**零新众包标注**,遵守约束)。 + +--- + +## 7. Experiments + +### 7.1 数据集与模型 +- **概念(≥12,回应 M6 统计功效):** truth(Cities/Counterfact 风格)、sentiment、formality、toxicity、language-ID、tense、negation、topic(≥5 类)等,涵盖 factual 与 stylistic,确保"跨多概念"卖点有功效;报告每个 ρ 的 bootstrap CI。 +- **模型 / size ladder:** Pythia 全 ladder(70M–2.8B,单卡可载)+ Gemma-2-2B(若显存允许),对齐 2505.21399 / 2605.27958 以复现/检验 scaling。 +- **激活提取:** 一次前向缓存隐状态,全部 baseline 共享缓存(省 GPU)。 + +### 7.2 Baselines(纳入 must-have,回应 M4) +- **A 组 a priori(全跑):** SIP(eigengap)、RAPTOR directional stability(K 次 ablation mean|cos|,= dispersion-only 对照)、Fragility(临界噪声崩点)、Probing-the-Probes augmentation-robustness(= augmentation-only 对照,适配为文本增广)。 +- **B 组通用免标注 OOD:** Xie feature-dispersion(必跑,证明非换皮)、Deng 2023 Confidence&Dispersity 或 Baek 2024 Agreement-on-the-Line(至少一)、Geometry Markers(effective manifold dim)。 +- **C 组 estimator:** mass-mean、LogReg、MLP。 +- **D 组方法学:** Hewitt & Liang control-task/selectivity probe;裸 cosine vs Mahalanobis 白化 cosine(Truthfulness-Spectrum 风格,作为放宽 OOD 约束的上界对照)。 + +### 7.3 主实验 +1. 构建 S1 矩阵(真实 OOD accuracy-drop / cosine-rotation,解耦)。 +2. 计算所有 a priori 预测器(含 PAC)。 +3. 报告 held-out **Spearman ρ / Kendall τ**(PAC vs 真实 drop),双交叉验证。 + +### 7.4 关键消融与分离实验 +- **互补性(H3):** dispersion-only / augmentation-only / 合成 三档,配对 Wilcoxon + Cliff's δ + Holm 校正。 +- **方向几何 ≠ 流形 dispersion(回应 F3):** 在 paraphrase shift 上对比 PAC 与 Xie feature-dispersion 的预测相关性,预期 Xie 在 paraphrase 上失效(因流形整体不变但方向旋转),PAC 不失效——以图证明机制分离。 +- **度量消融(M3):** 裸 vs 白化 cosine。 +- **size 效应(M2):** PAC 预测有效性随 size 是否单调;复现/检验 2605.27958 的 inverse-scaling-as-artifact。 + +### 7.5 统计方案 +- 每配置 **5 seeds**;所有相关系数报 **bootstrap(10k)95% CI**;多重比较 **Holm-Bonferroni**;效应量报 **Cliff's δ / Cohen's d**;预注册阈值见 §5。 + +--- + +## 8. Expected Results 与可能的负结果叙事 + +**预期正结果:** H2(ρ≈0.6–0.75 held-out)、H3(合成显著优,δ≈0.4)、H4(对 SIP/Fragility/Xie 有正增量,对 RAPTOR-stab 因其=dispersion-only 故增量主要来自 augmentation 分量)。 + +**可能负结果的诚实叙事(预先写入,回应 M1/M2):** +- **若 H3 不成立**(合成不优于 dispersion-only):坦诚降级为"单分量 dispersion 预测器 + 一项 augmentation 失败的负结果",仍贡献"augmentation 分量在文本探针上无增量"这一可发表 negative finding。 +- **若对某 baseline 无显著增量**:在 limitation 明确该 baseline 已足够,PAC 的价值退守到"统一框架 + 闭环验证"。 +- **若 size 出现 inverse scaling**:不下普适结论,引 2605.27958 解释为 training-distribution artifact 并复现其 style-augmented 恢复实验。 +- **若白化 cosine(用 OOD cov)大幅超过 PAC**:诚实陈述"免 OOD 的代价",将 PAC 定位为"约束受限场景下的最优",而非全局最优。 + +## 9. Reproducibility +- 发布:全部**激活缓存**、探针权重、PAC/baseline 实现、**预注册阈值文件**、shift 数据集 + NLI 审计通过率、seed 列表、leave-one-out split 清单、bootstrap 脚本。 +- 一键复现脚本 + 固定 commit;所有 baseline 在共享缓存上运行以保证可比。 + +## 10. Budget 明细(单卡 4090 ≤200 GPU·h,API <$100) + +| 项目 | 估算 | 说明 | +|---|---|---| +| 激活缓存(12 概念 × ladder 6 size × 缓存)| ~40 GPU·h | 一次性前向,全 baseline 复用 | +| 探针训练(3 estimator × bootstrap B=50 × 5 seed)| ~50 GPU·h | LogReg/mass-mean 极廉,MLP 主成本 | +| Augmentation 重训探针(3 shift × 上述)| ~45 GPU·h | label-preserving 增广激活 + 重训 | +| Baselines(SIP/Fragility/RAPTOR/Aug/Xie/Geometry)| ~35 GPU·h | Fragility 噪声扫描最贵 | +| Dies/P-StaT baseline 复现 | ~15 GPU·h | 用其公开 code/data | +| 缓冲 | ~15 GPU·h | 共计 **~200 GPU·h** | +| **API(paraphrase/domain/length 过量生成)** | **<$80** | 12 概念 × 3 shift × overgenerate 3x,GPT-class API;DeBERTa-MNLI 过滤本地零成本 | + +## 11. Risk & 退路(逐条对应红队致命攻击) + +| 攻击 | 缓解 / 退路 | +|---|---| +| **F1(Dies 抹平)** | 已核实 P-StaT 不预测、触碰 OOD、不合成、仅 truth → 逐能力对照表 3 格 N→Y(合成/预测/免OOD,均非边角);复现 P-StaT 作 baseline。**退路:** 若 P-StaT 修订版加入预测,立即二次检索并将 PAC 收窄至"免 OOD + 合成"双格。 | +| **F2(首个被证伪)** | **全文删除 first/novel framing**,改 "empirically outperforms ... under zero-OOD constraint";SIP/Fragility/Truth-Spectrum 全跑;增量带 bootstrap CI。 | +| **F3(Xie 换皮)** | 改名 PAC;§7.4 分离实验证明方向几何 ≠ 流形 dispersion(paraphrase 上 Xie 失效)。 | +| **F4(循环论证/泄漏)** | α 仅 dev 选;结果仅 held-out;leave-one-concept/shift-out 双交叉验证;split 清单随码发布。 | +| **F5(C3 复现)** | C3 降为 S2 外部效度,仅在"PAC 能否预测三 estimator 谁更稳"从属问题下出现,绝不并列。 | +| **M1** | H3 配对显著 + 效应量;不显著则降级 + negative finding。 | +| **M2** | C4 仅交叉维度;主动复现/检验 2605.27958 artifact,不下普适 scaling 结论。 | +| **M3** | 裸 vs 白化 cosine 消融(白化仅用 ID cov)。 | +| **M4** | A 组四 baseline + Xie 全跑,B/C 各 ≥1;缺项写 limitation。 | +| **M5** | C5 降 sanity;引 Falsesum/DISCO;报通过率 + 作者内部抽检一致率(零新众包)。 | +| **M6** | ≥12 概念;ρ 报 bootstrap CI;概念不足则撤回"跨多概念"主张。 | +| **M7** | 已逐条核实 10 篇 high-threat(本提案顶部);投稿前再复核 venue/接收状态,对所有篇目无论是否发表均切割。 | + +**三条生死线状态:** F1(对照表 3 格 N→Y + 复现跑赢)✓ 设计就绪;F4(无泄漏 split + 双交叉验证)✓ 设计就绪;M1(互补性显著或诚实降级)✓ 含退路。三者同时成立,具备从 Reject 翻盘的最小充分条件。 + +--- + +**关键风险提示(交付给通讯作者):** 唯一未闭合的风险仍是 **Dies/P-StaT 是否会在修订版加入预测闭环**。本提案已核实其当前版本不预测,但投稿前须二次检索其最新版与是否被 NeurIPS/ICLR 接收;若其新增预测,PAC 须立即收窄至"合成 + 免 OOD 数据"双格护城河,并强化 §7.4 的方向几何 vs 流形分离实验作为差异化主轴。 + +**已核实文献链接(供投稿前复核 venue):** Dies/P-StaT [2511.19166](https://arxiv.org/abs/2511.19166) · SIP [2511.16288](https://arxiv.org/abs/2511.16288) · Fragility [2606.11375](https://arxiv.org/abs/2606.11375) · RAPTOR [2602.00158](https://arxiv.org/abs/2602.00158) · Probing-the-Probes [2511.04312](https://arxiv.org/abs/2511.04312) · Xie [2303.15488](https://arxiv.org/abs/2303.15488) · Truthfulness-Spectrum [2602.20273](https://arxiv.org/abs/2602.20273) · Haller/Brittle [2510.11905](https://arxiv.org/abs/2510.11905) · Factual-Self-Awareness [2505.21399](https://arxiv.org/abs/2505.21399) · Pressure-Testing-Deception [2605.27958](https://arxiv.org/abs/2605.27958) \ No newline at end of file diff --git a/repro_bundle/README.md b/repro_bundle/README.md new file mode 100644 index 0000000000000000000000000000000000000000..5c8fc8c68a5564061fee4d0f67aeecc6752cce85 --- /dev/null +++ b/repro_bundle/README.md @@ -0,0 +1,90 @@ +# ProbeShift + +> *ProbeShift: A Systematic, Label-free Benchmark for the OOD Directional Stability of Linear +> Probes under Label-preserving Semantic Shift, with a Unified Evaluation of A-priori Predictors.* +> +> 投稿目标:ACML 2026 主会。定位:**benchmark-first**(见 [PROPOSAL.md](PROPOSAL.md))。 +> 预算锚:单卡 4090 ≤ 200 GPU·h、LLM API <$100、零新人工标注。 + +## 核心贡献(benchmark-first,见 PROPOSAL.md) + +- **G1(主)ProbeShift 基准**:concept(≥12)× shift(paraphrase/domain/length)× size(Pythia ladder + + GPT-2 + Qwen)× estimator(LogReg/mass-mean/MLP)的网格,记录 **解耦的** accuracy-drop 与 direction + cosine-rotation 作为 ground truth。完全免新标注;随激活缓存发布,分钟级复现。 +- **G2(主)统一 a-priori 预测信号评测**:在同一基准上 head-to-head 评测现有 label-free 训练前信号 + —— **SIP**(eigengap)、**RAPTOR-stability**、**Fragility**、**augmentation-robustness**、**Xie + feature-dispersion** —— 对探针 OOD 方向 drop 的预测力(Spearman/Kendall + leave-one-concept-out CV)。 +- **G3(次)PAC 入选项**:简单、IID-only、免 OOD 数据的双分量复合分数(dispersion ⊕ augmentation- + consistency),作为基准上的一个 entry,做互补性消融。**不宣称 SOTA。** +- **支撑**:estimator × size 外部效度;本地 NLI 的 label-fidelity sanity check。 + +> 关键 framing:G1+G2 即使 PAC 毫无增量也独立成立 —— 把"赛道很挤"从威胁转成贡献。 +> 所有竞品均经直连 arXiv 核实为真(见 [PRIOR_ART.md](PRIOR_ART.md) §四),并已转化为 baseline。 + +## Pipeline + +``` +extract 一次性抽取 residual-stream 激活,float16 分片缓存 (唯一大头 GPU) +eval 训练 LogReg/mass-mean/MLP 探针,IID/OOD accuracy + 方向旋转 (读缓存) -> G1 ground truth +predict 对每个 config 计算全部 a-priori 预测器(SIP/Fragility/.../PAC)(读缓存) -> G2 信号 +score PAC 双分量明细 + 互补性消融 (读缓存) -> G3 +stats 每个预测器 vs 真实 OOD drop 的 Spearman/Kendall + LOCO 排名表 (读缓存) -> G2 主结果 +``` + +**关键设计:激活只抽一次、落盘、之后全部读缓存** —— 这是整个项目守住算力预算 + 可复现的命门。 +缓存可直接随论文发布,reviewer 几分钟即可复现探针实验。 + +## 安装 + +```bash +cd probe_stability +python -m venv .venv && source .venv/bin/activate # 或用 conda +pip install -r requirements.txt +``` + +## 最小可行实验(M2 go/no-go,Day 3) + +一条命令在单模型单数据集上验证"探针脆性"是否存在(OOD drop + 方向旋转): + +```bash +python run_pipeline.py smoke --model EleutherAI/pythia-410m --dataset sst2 +``` + +若代表性概念上 **OOD drop < 5pt 或方向 cosine > 0.9**(探针根本不脆)→ 触发退路:切提案② +(Demonstration Order Illusion),M0/M1 基础设施可复用。 + +## 全量运行 + +```bash +python run_pipeline.py extract --all # 抽取所有 (model,dataset,distribution) 激活 +python run_pipeline.py eval --all # 训练探针 + IID/OOD + 旋转矩阵 (G1 ground truth) +python run_pipeline.py predict --all # 计算全部 a-priori 预测器 (G2 信号) +python run_pipeline.py score --all # PAC 双分量明细 + 互补性消融 (G3) +python run_pipeline.py stats # 多预测器 vs OOD drop 排名表 (G2 主结果;读全部结果) +``` + +## 目录结构 + +``` +probe_stability/ +├── README.md +├── requirements.txt +├── config.py # 模型/数据/shift 注册表 + 路径 + 超参 +├── cache.py # 激活分片的命名/保存/mmap 读取 + manifest +├── metrics.py # accuracy / 方向旋转 / Spearman / bootstrap CI / Cliff's delta / Holm-BH +├── probes.py # LogReg / mass-mean / MLP / control-task 探针 +├── stability_score.py # dispersion + augmentation-consistency +├── extract_activations.py # HF 模型 hooks,residual-stream 抽取 + float16 分片落盘 +├── run_pipeline.py # CLI 编排 M1–M5 +└── data/ + ├── __init__.py + ├── datasets.py # 统一加载 HF 数据集 -> (text, label, concept) + ├── shifts.py # label-preserving shift 构造(back-translation / domain / length) + └── label_fidelity.py # DeBERTa-MNLI 过滤 label-flip +``` + +## 状态 + +骨架代码:核心数据流、缓存、探针、Stability Score、指标已实现;以下处标注 `TODO` 待在 4090 上迭代: +back-translation 模型选型与显存调参、各数据集 domain-shift 配对、MLP 探针方向旋转的处理。 +`PROPOSAL.md` / `PRIOR_ART.md` 由研究工作流产出后补入。 diff --git a/repro_bundle/RESULTS.md b/repro_bundle/RESULTS.md new file mode 100644 index 0000000000000000000000000000000000000000..b85b71d32c05b95be71b74fd0691ef6abb815c2c --- /dev/null +++ b/repro_bundle/RESULTS.md @@ -0,0 +1,203 @@ +# ProbeShift —— 全量跑结果 + +## ⚠️ 关键诚实校正(对抗式评审驱动,2026-06-23) + +评审指出 EXCESS=naive−placebo 的差值构造会**机械制造负相关**(dispersion 高度预测 placebo,从被减数减去它→残差对 dispersion 必然趋负)。用**偏相关**(无差值)复核: + +| 预测器 | 偏相关 ρ(signal, para-rot \| placebo) | EXCESS ρ(差值) | +|---|---|---| +| pac | **+0.374** | −0.029 | +| augmentation_robustness | **+0.345** | +0.046 | +| raptor_stability | **+0.323** | −0.195(artifact) | +| whitened_cosine | −0.015 | −0.245 | + +**结论修正:** (1) "dispersion 主动误导/anti-predict" 是**差值 artifact**——偏相关下 dispersion 正 +0.323; +(2) 戏剧"反转"大幅缩水——偏相关下 aug(0.345)≈raptor(0.323),PAC 最高(0.374); +(3) cluster bootstrap(按 12 概念,修正伪重复)Δρ[aug−raptor] on EXCESS 仍 +0.241 [+0.118,+0.363] SIGNIFICANT, +但 EXCESS 目标本身被质疑。**诚实 headline:循环性真实重要;但 shift 脆性不是"完全不可预测",dispersion/aug 都有适度残余预测力(~0.32–0.37,PAC 最好),原"open problem + 反转"被差值构造夸大。** + +--- + +## ★★★ 原始结果:Option A,5 独立 seed,n=472(差值/EXCESS 口径,需配合上方校正解读) + +12 概念 × 7 模型 × **5 个完全独立的 seed**(各自独立抽样 + 独立回译,gold-standard 复现)。 +gpt2-medium/qwen 的少数 massive-activation 格子被容错跳过(每 seed ~4 个)。 + +### 四目标 + 配对显著性(n=472) +| 预测器 | acc-drop | naive-rot | **PLACEBO(IID-resample)** | **EXCESS(非循环)** | +|---|---|---|---|---| +| raptor_stability(纯dispersion) | -0.06 | 0.945 [.93,.95] | **0.966 [.96,.97]** | **-0.195 [-.29,-.12]** | +| pac | 0.12 | 0.839 | 0.799 | -0.029 [-.11,.06] | +| augmentation_robustness | 0.18 | 0.711 | 0.653 | **0.046 [-.04,.13]** | +| whitened_cosine_id | -0.08 | 0.769 | 0.814 | -0.245 [-.34,-.16] | +| fragility | 0.11 | 0.640 | 0.643 | -0.136 | +| xie_feature_dispersion | 0.06 | 0.458 | 0.464 | -0.117 | +| sip_eigengap | 0.00 | -0.072 | -0.057 | 0.020 | + +**配对检验(bootstrap 95% CI,全 SIGNIFICANT):** +- Δρ[**aug − raptor**] on EXCESS = **+0.241 [+0.170,+0.309]** ✓ +- Δρ[**pac − raptor**] on EXCESS = **+0.166 [+0.116,+0.215]** ✓ +- Δρ[raptor − aug] on naive-rot = +0.234 [+0.189,+0.284] ✓(naive 上 dispersion 赢——但循环) + +### 最终结论(5-seed,最严格、最诚实) +1. **naive rotation 的强预测是循环的**:raptor 对 naive 0.945,对**纯采样安慰剂 0.966(更高)** → 它预测的是采样噪声地板。红队 F4 循环性被自家对照坐实(多 seed,CI 极紧)。 +2. **真·shift 脆性(EXCESS)无人能正向预测**:最佳 aug 仅 0.046(CI 跨 0);dispersion/whitened **显著负相关**(-0.195/-0.245)→ 它们在诚实目标上**主动误导**。→ 这是有价值的 **open problem**。 +3. **反转显著(headline)**:在诚实的 EXCESS 上,**基于增广的信号显著优于基于离散度的**(Δρ aug−raptor=+0.241,5-seed 稳健),naive 上则相反。即"哪个信号有用"完全取决于是否扣掉采样噪声。 +4. acc-drop:多 seed 下 aug 0.18 弱显著但 LOCO 0.06(跨概念不泛化)。 + +### Tier 1 重分析(纯缓存,2026-06-23) +- **Shift-type 拆分**:paraphrase(n=472,raptor 0.945 但循环)、domain(n=65,弱:raptor 0.59)、 + length(n=30,全高:aug 0.95/pac 0.95/raptor 0.92)。预测器行为强烈依赖 shift 类型。 +- **Estimator 外部效度**:用 logreg 预测器去预测 **mass-mean 方向旋转** → 全部弱(xie 0.47 最高, + 其余 0.25-0.29)且 **LOCO≈0(跨概念不泛化)**。→ 可预测性是 estimator 特异的,非一般属性(重要 caveat)。 +- **Label-fidelity(NLI 通过率)**:整体 80.6%;counterfact 96%/sst2 89% 高,**dbpedia 50% / imdb 64% 偏低** + (实体/长文回译易翻车)→ 按 H7 标注,这两个数据集 rotation 解读需谨慎。 +- **Layer 稳健性(refit)**:paraphrase rotation 按相对深度 0.68→0.71→0.79→0.78,**所有层都普遍(非挑层产物)**。 +- **白化度量消融(M3,refit)**:用 ID-白化 cosine 作目标时,raptor 从裸 cosine 的 0.945 **掉到 0.698** + (whitened_cosine_id 0.851 居首)→ dispersion 的循环优势**部分由度量造成**(裸 cosine 忽略协方差); + 换白化度量后优势缩小。诚实的"结果依赖度量"caveat,非度量取巧。 + +### Tier 2 多-pivot 增广(de/fr/ru,5 seed × 7 模型,n=472/490) +- aug 在 EXCESS 上从 1-aug 的 0.046 仅升到 **0.070(CI 仍跨 0)** → 多样化增广**没把它推到显著正**。 + → **强化 open-problem**:即使用 de/fr/ru 多样增广,shift 特异脆性仍不可正向预测(排除"增广太弱"解释)。 +- 反转**对增广多样性稳健**:Δρ[aug−raptor] on EXCESS = **+0.247**(1-aug 是 +0.241),SIGNIFICANT。 +- 结论:Tier 2 是干净的鲁棒性确认——核心发现不依赖增广质量。 + +### Tier 3 #11 7B 抽检(pythia-6.9b,seed 0,7 数据集,n=7) +- naive-rot:raptor **+0.93**(仍循环);excess:raptor **−0.39**(仍反预测)、aug **+0.68**(更强正)。 +- → **循环性 + aug>dispersion 反转在 6.9B 仍成立且 aug 优势放大**(n=7 偏粗,方向一致)→ 堵死"仅小模型"质疑。 +- size-ladder(70M→6.9B):raptor naive 全程 0.83–0.95、excess 全程 ≤0;**可预测性结构对规模不变,无 inverse-scaling**。 + +### 图(figures/) +- fig1_circularity:预测器 × {naive / 安慰剂 / excess} 分组柱(killer figure:naive≈安慰剂=循环,excess≤0)。 +- fig2_sizeladder:raptor/aug/pac 的 ρ vs 规模(naive 与 excess 两面板)。 +- fig3_mechanism:raptor vs excess(负)对比 aug vs excess(正)散点——为什么 aug 行 dispersion 不行。 + +--- + +## ★★ 中间版:12 概念 / 84 configs(n=78,单 seed)+ 循环性对照 + +**配置:** 6 模型 × 14 数据集 / **12 概念**(sentiment, topic, truth, emotion, hate, irony, offensive, +subjectivity, spam, grammaticality, stance, counterfactual)。stance 因推文经 NLI label-fidelity 过滤后 +paraphrase 集为空,部分 config 缺 rotation(n=78)。 + +### 四目标对照(Spearman ρ,n=78) +| 预测器 | acc-drop | rotation(naive) | **IID-resample(安慰剂)** | **EXCESS(=para−IID, 非循环)** | EXCESS-LOCO | +|---|---|---|---|---|---| +| raptor_stability(纯 dispersion) | -0.07 | 0.951 | **0.975** | **-0.075** | -0.23 | +| pac(ours = disp ⊕ aug) | 0.09 | 0.895 | 0.859 | **0.124** | -0.03 | +| **augmentation_robustness** | 0.15 | 0.811 | 0.752 | **0.225** [0.00,0.41] | 0.09 | +| whitened_cosine_id | -0.13 | 0.741 | 0.790 | -0.186 | -0.14 | +| fragility | 0.09 | 0.685 | 0.685 | -0.015 | -0.04 | +| xie_feature_dispersion | -0.01 | 0.502 | 0.491 | -0.053 | -0.01 | +| sip_eigengap | 0.12 | -0.151 | -0.092 | -0.101 | 0.05 | + +### 决定性结论(改写论文故事) +1. **naive rotation 预测大部分是循环的**:raptor_stability 对 naive rotation 0.951,但对**安慰剂(纯 IID + 采样旋转)同样 0.975** → 它主要在预测采样噪声地板,而非 shift 脆性。**红队 F4 循环性质疑被我们自己的 + 对照证实。** +2. **真·shift 特异脆性(EXCESS rotation)几乎不可预测**:去掉采样噪声后,dispersion 彻底失效(-0.075), + 全场最高仅 augmentation_robustness 0.225(CI 下界贴 0)。→ 这是真正的 **open problem**(benchmark 论文加分)。 +3. **关键反转,支持 augmentation 组件**:在诚实的 EXCESS 目标上,**基于增广的信号 > 基于离散度的** + (aug 0.225 / pac 0.124 > raptor -0.075)。augmentation 直接探测 label-preserving 扰动下的方向变化, + 天然捕捉 shift 特异结构;纯 bootstrap dispersion 不能。**PAC 的 aug 分量恰在最该起作用处起作用。** +4. accuracy-drop 仍全场不可预测(≈0)。 + +> **论文主张应改为**:不是"我们提出最强预测器",而是"**我们用循环性对照揭示:现有探针稳定性信号对 OOD +> 方向旋转的强预测力大部分是采样噪声的循环假象;扣除后,shift 特异脆性基本不可预测,且只有基于 +> label-preserving 增广的信号有微弱(非循环)预测力**"。这是更难被攻击、更有 insight 的 contribution。 + +数据:`results_full_run/`(eval 234 行、predictors 78、stats_ranking 4 张表、audit 84)。 + +--- + +## 附:9 概念 / 66 configs(中间版,无循环性对照) + +**配置:** 6 模型(pythia-70m/160m/410m/1.4b + gpt2 + qwen2.5-0.5b)× 11 数据集 / **9 概念**: +sentiment(sst2,imdb)、topic(ag_news,dbpedia)、truth(counterfact)、emotion、hate、irony、 +offensive、subjectivity(subj)、spam。N_train=1500 / N_eval=800 / n_aug=1,单卡 4090。 +各数据集 IID 探针准确率均显著高于随机(dbpedia .96 / spam .99 / subj .93 / … / counterfact .55, +真值探针接近随机符合"小模型读真值难"的已知规律)。 + +### 目标①:预测 OOD accuracy-drop —— 全员 ρ≈0,CI 全跨 0、LOCO≈0 → **不可预测** +### 目标②:预测 OOD direction-rotation +| 预测器 | Spearman | 95% CI | LOCO(9概念留一) | +|---|---|---|---| +| **raptor_stability**(IID方向离散度) | **0.961** | [0.91, 0.98] | **0.879** | +| **pac(ours)** | **0.881** | [0.78, 0.93] | **0.810** | +| augmentation_robustness | 0.774 | [0.64, 0.86] | 0.691 | +| whitened_cosine_id | 0.764 | [0.60, 0.87] | **0.205** ⚠️不泛化 | +| fragility | 0.704 | [0.52, 0.84] | 0.388 | +| xie_feature_dispersion | 0.428 | [0.15, 0.64] | **0.030** ⚠️不泛化 | +| sip_eigengap | -0.135 | [-0.39, 0.14] | 0.192 | + +**9 概念版关键结论(比 3 概念版更硬):** +1. **解耦 + 可预测性不对称**(headline,稳健):rotation 高度可预测(raptor ρ=0.96,LOCO 0.88 跨 9 概念), + accuracy-drop 完全不可预测(全 ~0)。 +2. **LOCO 列成为强区分器**:raptor(0.88)/pac(0.81)/aug(0.69)**跨概念泛化**;而 whitened_cosine(0.21)、 + xie(0.03)样本内不错但**跨概念崩盘** → "哪些信号 robust、哪些是概念过拟合"本身是 G2 的有价值产出。 +3. **xie 跨概念失败**(LOCO 0.03)稳健复现 → feature 流形 dispersion ≠ probe 方向几何。 +4. **PAC 诚实**:competitive(第2,LOCO 0.81 稳健)但未超过单分量 raptor → H3 互补性仍不成立,如实报告。 + +--- + +## 附:初版 3 概念 / 30 configs(已被上面的 9 概念版取代,留作记录) + +**配置:** 6 模型 × 5 数据集(sentiment×2, topic×2, truth×1)= **30 configs**,3 概念。 +N_train=1500 / N_eval=800 / n_aug=1。单卡 4090。 + +## G2 主结果:哪个 a-priori 信号能预测探针 OOD 不稳定性? + +报告 `Spearman ρ(信号, −目标)`(越高=越能预测),bootstrap 95% CI,leave-one-concept-out(LOCO)。 + +### 目标一:OOD **accuracy-drop**(主要由 domain shift 驱动) +| 预测器 | Spearman | 95% CI | LOCO | +|---|---|---|---| +| augmentation_robustness | 0.109 | [-0.27, 0.43] | -0.47 | +| pac (ours) | -0.015 | [-0.35, 0.31] | -0.62 | +| xie_feature_dispersion | -0.029 | [-0.38, 0.33] | -0.05 | +| fragility | -0.082 | [-0.39, 0.24] | -0.26 | +| raptor_stability | -0.095 | [-0.37, 0.19] | -0.35 | +| whitened_cosine_id | -0.193 | [-0.48, 0.15] | -0.34 | +| sip_eigengap | -0.384 | [-0.60, -0.12] | +0.40 | + +→ **结论:accuracy-drop 基本不可被任何现有 label-free 几何信号 a-priori 预测**(所有 CI 跨 0; +sip 弱负相关但 LOCO 翻号、不稳)。这是一个有价值的"open problem"型结论(对 benchmark 论文是加分)。 + +### 目标二:OOD **direction-rotation**(主要由 paraphrase 驱动) +| 预测器 | Spearman | 95% CI | LOCO | +|---|---|---|---| +| **raptor_stability** | **0.969** | [0.89, 0.99] | **0.937** | +| **pac (ours)** | **0.818** | [0.60, 0.92] | 0.662 | +| whitened_cosine_id | 0.750 | [0.46, 0.90] | 0.592 | +| fragility | 0.682 | [0.39, 0.86] | 0.271 | +| xie_feature_dispersion | 0.677 | [0.39, 0.85] | **-0.052** | +| augmentation_robustness | 0.677 | [0.43, 0.83] | 0.493 | +| sip_eigengap | -0.525 | [-0.78, -0.10] | -0.305 | + +→ **结论:direction-rotation 高度可 a-priori 预测**;`raptor_stability`(= IID 方向 bootstrap 离散度) +近乎完美(ρ=0.969,LOCO 0.937)。 + +## 关键科学结论(支撑论文) + +1. **解耦 + 可预测性不对称(headline)**:两个 OOD 目标彻底分离——**方向旋转高度可预测(ρ≈0.97), + 准确率下降几乎不可预测(ρ≈0)**。这把 "accuracy ⊥ direction" 从"现象"升级为"可预测性结构差异", + 是 G1+G2 的核心卖点。 +2. **PAC 诚实结论**:PAC(双分量合成)排第二(0.818),但**未超过单分量 raptor_stability(0.969)** + → 互补性假设 H3 在此**不成立**:对旋转预测,augmentation 分量未在 dispersion 之外带来增量。 + benchmark-first 定位正确——PAC 只是入选项之一,我们如实报告"现有单分量已足够"。 +3. **xie 的跨概念失败**:xie 在样本内 0.677,但 **LOCO -0.05**(跨概念不泛化)→ 实证"feature 流形 + dispersion ≠ probe 方向几何",正好坐实 §7.4 的机制分离论点。 +4. **sip_eigengap 反相关**:eigengap 在此不是好的旋转预测器(甚至反号),与其理论定位形成对比,值得讨论。 + +## 待处理 / 注意(投稿前) + +- **循环性(对应红队 F4)**:dispersion(IID bootstrap 方向方差)预测 rotation(shift 引起的方向变化), + 二者都刻画"方向移动"。非平凡点在于 dispersion **完全不碰 OOD 数据**即可预测 shift 后旋转——论文须把 + 这点讲清,并补一个"dispersion 与 rotation 测量相互独立"的对照(如不同随机源、白化对照)以防"换皮"质疑。 +- n=30、仅 3 概念,LOCO 只有 3 组;论文版需扩到 ≥8-12 概念以增强 LOCO 统计功效。 +- n_aug=1(单 pivot 回译);多 pivot 多样化增广留待论文版,可能改变 augmentation_robustness/PAC 的表现。 +- accuracy-drop 的 domain shift 仅 sentiment 数据集有(topic/truth 无 domain 配对),其不可预测性结论需在 + 更广 domain 覆盖下复核。 + +数据存档:`results_full_run/`(eval.jsonl 90 行、predictors.jsonl 30、stats_ranking.jsonl、audit.jsonl)。 diff --git a/repro_bundle/analyze.py b/repro_bundle/analyze.py new file mode 100644 index 0000000000000000000000000000000000000000..ac7280f18df7d2d9b53ae326d181ee874a0d61d8 --- /dev/null +++ b/repro_bundle/analyze.py @@ -0,0 +1,214 @@ +"""Tier-1 re-analysis of cached results (no GPU, no refitting): reads results/*.jsonl and +produces the breakdowns the paper needs. Safe to run alongside / after the grid. + + python analyze.py shift # per shift-type predictor ranking (paraphrase/domain/length) + python analyze.py estimator # cross-estimator: do logreg-based predictors predict mass-mean rotation? + python analyze.py fidelity # label-fidelity (NLI flip-rate) summary per dataset + python analyze.py all +""" +from __future__ import annotations + +import sys + +import numpy as np + +import metrics +from config import DATASETS +from baselines import PREDICTORS +from run_pipeline import _rot_mag, _mean_ood_drop, _read_results, _rank_predictors + + +def _pred_index(preds): + return {(p["model"], p["dataset"], p["seed"]): p["predictors"] for p in preds} + + +def shift_breakdown(): + evals = _read_results("eval") + pidx = _pred_index(_read_results("predictors")) + print("\n########## SHIFT-TYPE BREAKDOWN (rotation target per shift) ##########") + for shift in ["paraphrase", "domain", "length"]: + rows = [] + for e in evals: + if e.get("probe") != "logreg": + continue + key = (e["model"], e["dataset"], e["seed"]) + if key not in pidx: + continue + mag = _rot_mag(e.get("dists", {}).get(shift, {}).get("rotation")) + if mag != mag: + continue + rows.append({"concept": DATASETS[e["dataset"]].concept, "rot": mag, **pidx[key]}) + if len(rows) < 8: + print(f"\n[{shift}] only {len(rows)} configs — skipped") + continue + _rank_predictors(rows, "rot", f"rotation under {shift.upper()} (n={len(rows)})") + + +def estimator_cross(): + """Do logreg-derived predictors also forecast the MASS-MEAN probe's rotation? (S2 external validity).""" + evals = _read_results("eval") + pidx = _pred_index(_read_results("predictors")) + print("\n########## ESTIMATOR EXTERNAL VALIDITY (predict mass-mean rotation) ##########") + rows = [] + for e in evals: + if e.get("probe") != "mass_mean": + continue + key = (e["model"], e["dataset"], e["seed"]) + if key not in pidx: + continue + rots = [_rot_mag(v.get("rotation")) for d, v in e.get("dists", {}).items() if d != "iid"] + rots = [r for r in rots if r == r] + if not rots: + continue + rows.append({"concept": DATASETS[e["dataset"]].concept, "rot": float(np.mean(rots)), **pidx[key]}) + if len(rows) < 8: + print(f"only {len(rows)} configs — skipped") + return + _rank_predictors(rows, "rot", f"mass-mean rotation (n={len(rows)})") + + +def size_ladder(): + """Tier 3: does predictability of rotation / excess change with model size? + (C4 / inverse-scaling check). Per model, correlate key predictors vs paraphrase rotation + and vs excess rotation across its (dataset x seed) configs. + """ + from config import MODELS + evals = _read_results("eval") + pidx = _pred_index(_read_results("predictors")) + keypreds = ["raptor_stability", "augmentation_robustness", "pac"] + # build per-model rows of (predictors, para_rot, excess) + by_model = {} + for e in evals: + if e.get("probe") != "logreg": + continue + key = (e["model"], e["dataset"], e["seed"]) + if key not in pidx: + continue + para = _rot_mag(e.get("dists", {}).get("paraphrase", {}).get("rotation")) + iids = _rot_mag(e.get("iid_split_rotation")) + if para != para: + continue + excess = para - iids if iids == iids else float("nan") + by_model.setdefault(e["model"], []).append({"para": para, "excess": excess, **pidx[key]}) + print("\n########## SIZE-LADDER: predictability vs model size ##########") + print(f"{'model':16s}{'params_M':>9s} | rho(predictor, target) for para-rot / excess") + for m in sorted(by_model, key=lambda x: MODELS[x].params_m if x in MODELS else 0): + rows = by_model[m] + pm = MODELS[m].params_m if m in MODELS else 0 + cells = [] + for tgt in ("para", "excess"): + y = np.array([r[tgt] for r in rows], dtype=float) + for kp in keypreds: + s = np.array([r.get(kp, np.nan) for r in rows], dtype=float) + ok = ~(np.isnan(s) | np.isnan(y)) + rho = metrics.spearman(s[ok], -y[ok])[0] if ok.sum() >= 6 else float("nan") + cells.append(f"{kp.split('_')[0][:4]}:{rho:+.2f}") + cells.append("|") + print(f"{m:16s}{pm:>9.0f} | {' '.join(cells)}") + print("(target order: para-rot [raptor aug pac] | excess [raptor aug pac])") + + +def _partial_spearman(x, y, z): + """Partial Spearman of (x,y) controlling for z: rank-transform then residualise on z.""" + from scipy.stats import rankdata + ok = ~(np.isnan(x) | np.isnan(y) | np.isnan(z)) + x, y, z = x[ok], y[ok], z[ok] + if len(x) < 8: + return float("nan"), 0 + rx, ry, rz = rankdata(x), rankdata(y), rankdata(z) + Z = np.c_[np.ones_like(rz), rz] + res = lambda a: a - Z @ np.linalg.lstsq(Z, a, rcond=None)[0] + ex, ey = res(rx), res(ry) + return float(np.corrcoef(ex, ey)[0, 1]), len(x) + + +def robust_circularity(): + """Reviewer rebuttal (R2 difference-score artefact, R1 pseudo-replication): + (a) PARTIAL Spearman of (signal, paraphrase-rotation | placebo) — does a signal predict + shift-rotation BEYOND the sampling-noise floor, without the naive-minus-placebo subtraction? + (b) CLUSTER bootstrap (resampling CONCEPTS, not cells) for Δρ[aug−raptor] on excess. + """ + evals = _read_results("eval") + pidx = _pred_index(_read_results("predictors")) + rows = [] + for e in evals: + if e.get("probe") != "logreg": + continue + k = (e["model"], e["dataset"], e["seed"]) + if k not in pidx: + continue + para = _rot_mag(e.get("dists", {}).get("paraphrase", {}).get("rotation")) + plac = _rot_mag(e.get("iid_split_rotation")) + if para != para or plac != plac: + continue + rows.append({"concept": DATASETS[e["dataset"]].concept, "para": para, "plac": plac, + "excess": para - plac, **pidx[k]}) + print("\n########## ROBUST CIRCULARITY CHECK (reviewer R1/R2 rebuttal) ##########") + print("PARTIAL Spearman rho(signal, paraphrase-rot | placebo) [the artefact-free version]") + print(f"{'predictor':26s}{'partial-rho':>12s}{'naive-rho':>11s}{'excess-rho':>11s}") + para = np.array([r["para"] for r in rows]); plac = np.array([r["plac"] for r in rows]) + exc = np.array([r["excess"] for r in rows]) + for name in PREDICTORS: + s = np.array([r.get(name, np.nan) for r in rows], float) + pr, n = _partial_spearman(s, para, plac) + nr = metrics.spearman(s[~np.isnan(s)], -para[~np.isnan(s)])[0] if (~np.isnan(s)).sum() >= 8 else float("nan") + er = metrics.spearman(s[~np.isnan(s)], -exc[~np.isnan(s)])[0] if (~np.isnan(s)).sum() >= 8 else float("nan") + print(f"{name:26s}{-pr:>12.3f}{nr:>11.3f}{er:>11.3f}") + print("(partial-rho>0 => signal predicts shift-rotation beyond the sampling floor, non-circular," + " no difference-score subtraction)") + + # cluster bootstrap by concept for Delta rho[aug - raptor] on excess + concepts = np.array([r["concept"] for r in rows]) + uc = list(np.unique(concepts)) + sa = np.array([r.get("augmentation_robustness", np.nan) for r in rows], float) + sb = np.array([r.get("raptor_stability", np.nan) for r in rows], float) + rng = np.random.default_rng(0) + def drho(mask): + m = mask & ~(np.isnan(sa) | np.isnan(sb) | np.isnan(exc)) + if m.sum() < 8: + return np.nan + return metrics.spearman(sa[m], -exc[m])[0] - metrics.spearman(sb[m], -exc[m])[0] + base = drho(np.ones(len(rows), bool)) + boots = [] + for _ in range(2000): + pick = rng.choice(uc, len(uc), replace=True) + mask = np.isin(concepts, pick) + # build resampled arrays by concept blocks + idx = np.concatenate([np.where(concepts == c)[0] for c in pick]) + m = ~(np.isnan(sa[idx]) | np.isnan(sb[idx]) | np.isnan(exc[idx])) + if m.sum() >= 8: + boots.append(metrics.spearman(sa[idx][m], -exc[idx][m])[0] - metrics.spearman(sb[idx][m], -exc[idx][m])[0]) + lo, hi = np.percentile(boots, [2.5, 97.5]) + sig = "SIGNIFICANT" if lo > 0 else "n.s." + print(f"\nCLUSTER bootstrap (resample {len(uc)} CONCEPTS): Δρ[aug−raptor] on EXCESS = {base:+.3f}" + f" 95%CI [{lo:+.3f},{hi:+.3f}] -> {sig}") + print("(this is the pseudo-replication-corrected version of the headline paired test)") + + +def fidelity_summary(): + audit = _read_results("audit") + print("\n########## LABEL-FIDELITY (NLI flip-rate of paraphrase shift) ##########") + by_ds = {} + for a in audit: + by_ds.setdefault(a["dataset"], []).append(a.get("flip_rate", float("nan"))) + print(f"{'dataset':18s}{'mean flip-rate':>16s}{'pass-rate':>12s}") + for ds in sorted(by_ds): + fr = float(np.nanmean(by_ds[ds])) + print(f"{ds:18s}{fr:>16.3f}{1 - fr:>12.3f}") + allfr = [v for vs in by_ds.values() for v in vs if v == v] + if allfr: + print(f"{'OVERALL':18s}{np.mean(allfr):>16.3f}{1 - np.mean(allfr):>12.3f}") + + +if __name__ == "__main__": + mode = sys.argv[1] if len(sys.argv) > 1 else "all" + if mode in ("shift", "all"): + shift_breakdown() + if mode in ("estimator", "all"): + estimator_cross() + if mode in ("size", "all"): + size_ladder() + if mode in ("robust", "all"): + robust_circularity() + if mode in ("fidelity", "all"): + fidelity_summary() diff --git a/repro_bundle/analyze_refit.py b/repro_bundle/analyze_refit.py new file mode 100644 index 0000000000000000000000000000000000000000..58fd195f3122f28231291b4a0a800d997d0a03c0 --- /dev/null +++ b/repro_bundle/analyze_refit.py @@ -0,0 +1,111 @@ +"""Tier-1 refit analyses (read cache, refit probes). Run on the server pointing PROBE_CACHE +at a seed cache. Reads predictors from results/ (default predictors_1aug.jsonl, the +A run, since Tier 2 rewrites predictors.jsonl). + + PROBE_CACHE=/root/rivermind-fs/cache_seed0 python analyze_refit.py layer + PROBE_CACHE=/root/rivermind-fs/cache_seed0 python analyze_refit.py whitened [predfile] + +layer : mean paraphrase rotation per (relative) layer over a sample — is the rotation a + property of the chosen layer or pervasive? (defends "not layer-cherry-picked") +whitened : recompute paraphrase rotation with ID-whitened (Mahalanobis) cosine and re-rank the + predictors — does the circularity / ranking conclusion survive a better metric? (M3) +""" +from __future__ import annotations + +import json +import sys +from pathlib import Path + +import numpy as np +from scipy import linalg + +import cache +import metrics +from config import DATASETS, MODELS, EXTRACT, RESULTS_DIR +from probes import make_probe +from run_pipeline import _select_layer, _rank_predictors +from baselines import PREDICTORS + +SAMPLE_DS = ["sst2", "ag_news", "counterfact", "emotion", "subj"] + + +def _read(name): + p = RESULTS_DIR / name + return [json.loads(l) for l in p.read_text().splitlines() if l.strip()] if p.exists() else [] + + +def _dir(X, y, nl, W=None): + if W is not None: + X = X @ W + return make_probe("logreg", num_labels=nl, seed=0, max_iter=500).fit(X, y).direction + + +def layer_curve(models=None, dsets=None): + models = models or list(MODELS) + dsets = dsets or SAMPLE_DS + print("\n########## LAYER ROBUSTNESS: paraphrase rotation by relative depth ##########") + buckets = {q: [] for q in [0.1, 0.25, 0.5, 0.75, 0.9, 1.0]} + for m in models: + for ds in dsets: + if not (cache.exists(m, ds, "train") and cache.exists(m, ds, "paraphrase")): + continue + nl = DATASETS[ds].num_labels + meta = cache.load_meta(m, ds, "train") + Xtr, ytr, _ = cache.load_shard(m, ds, "train") + Xpa, ypa, _ = cache.load_shard(m, ds, "paraphrase") + nl_layers = meta["n_layers"] + for q in buckets: + L = max(1, min(nl_layers - 1, int(q * (nl_layers - 1)))) + da = _dir(np.asarray(Xtr[:, L, :], np.float32), ytr, nl) + db = _dir(np.asarray(Xpa[:, L, :], np.float32), ypa, nl) + buckets[q].append(_rotmag(da, db)) + print(f"{'rel-depth':>10s}{'mean rotation (1-cos)':>24s}{'n':>5s}") + for q in sorted(buckets): + v = [x for x in buckets[q] if x == x] + if v: + print(f"{q:>10.2f}{np.mean(v):>24.3f}{len(v):>5d}") + print("(rotation present across depths -> not an artefact of the IID-selected layer)") + + +def _rotmag(da, db): + try: + return 1.0 - metrics.mean_class_cosine(da, db) + except Exception: + return 1.0 - float(np.cos(metrics.subspace_principal_angle(da, db))) + + +def whitened_target(predfile="predictors_1aug.jsonl"): + evals = _read("eval.jsonl") + preds = {(p["model"], p["dataset"], p["seed"]): p["predictors"] for p in _read(predfile)} + print(f"\n########## METRIC ABLATION: ID-whitened paraphrase rotation as target ##########") + print(f"(predictors from {predfile}; recomputing whitened rotation by refit on cache)") + rows = [] + for e in evals: + if e.get("probe") != "logreg": + continue + m, ds, seed = e["model"], e["dataset"], e["seed"] + if (m, ds, seed) not in preds: + continue + if not (cache.exists(m, ds, "train") and cache.exists(m, ds, "paraphrase")): + continue + nl = DATASETS[ds].num_labels + layer = e.get("layer") or _select_layer(m, ds, nl, seed) + Xtr, ytr, _ = cache.load_shard(m, ds, "train", layer=layer) + Xpa, ypa, _ = cache.load_shard(m, ds, "paraphrase", layer=layer) + C = np.cov(np.asarray(Xtr, np.float64), rowvar=False) + 1e-3 * np.eye(Xtr.shape[1]) + W = linalg.fractional_matrix_power(C, -0.5).real.astype(np.float32) + da = _dir(Xtr, ytr, nl, W=W) + db = _dir(Xpa, ypa, nl, W=W) + rows.append({"concept": DATASETS[ds].concept, "wrot": _rotmag(da, db), **preds[(m, ds, seed)]}) + if len(rows) < 8: + print(f"only {len(rows)} configs — skipped") + return + _rank_predictors(rows, "wrot", f"WHITENED paraphrase rotation (n={len(rows)})") + + +if __name__ == "__main__": + mode = sys.argv[1] if len(sys.argv) > 1 else "layer" + if mode == "layer": + layer_curve() + elif mode == "whitened": + whitened_target(sys.argv[2] if len(sys.argv) > 2 else "predictors_1aug.jsonl") diff --git a/repro_bundle/auto_pipeline.sh b/repro_bundle/auto_pipeline.sh new file mode 100644 index 0000000000000000000000000000000000000000..be94f94be152047fa6448bc9fcdcb4f942ae1325 --- /dev/null +++ b/repro_bundle/auto_pipeline.sh @@ -0,0 +1,24 @@ +#!/bin/bash +# Server-side autonomous orchestrator: advances the pipeline WITHOUT depending on the +# laptop-side monitoring. Waits for the A run to finish, then auto-produces Tier-0 stats +# and Tier-1 re-analyses. Each milestone writes a marker to auto.log. +cd /root/rivermind-data/probe_stability +AUTO=/root/rivermind-data/auto.log +export OMP_NUM_THREADS=4 OPENBLAS_NUM_THREADS=4 MKL_NUM_THREADS=4 + +echo "AUTO_START $(date +%F_%H:%M:%S)" >> $AUTO +# 1) wait for the multi-seed A run to finish +while ! grep -q A_DONE /root/rivermind-data/A.log 2>/dev/null; do sleep 60; done +echo "A_DONE_SEEN $(date +%F_%H:%M:%S)" >> $AUTO + +# 2) Tier 0: final 4-target tables + paired significance (re-run stats cleanly to a file) +python3 -u run_pipeline.py stats > /root/rivermind-data/tier0_stats.txt 2>&1 +echo "TIER0_STATS_DONE $(date +%F_%H:%M:%S)" >> $AUTO + +# 3) Tier 1: shift-type breakdown + estimator cross-validity + label-fidelity +python3 -u analyze.py all > /root/rivermind-data/tier1.txt 2>&1 +echo "TIER1_DONE $(date +%F_%H:%M:%S)" >> $AUTO + +# 4) snapshot results for safety +cp -r results /root/rivermind-data/results_snapshot_A 2>/dev/null +echo "AUTO_COMPLETE $(date +%F_%H:%M:%S)" >> $AUTO diff --git a/repro_bundle/baselines.py b/repro_bundle/baselines.py new file mode 100644 index 0000000000000000000000000000000000000000..a85a1d17c5288d7a230c277c2dc36a5847e3547f --- /dev/null +++ b/repro_bundle/baselines.py @@ -0,0 +1,153 @@ +"""A-priori, label-free probe-stability predictors evaluated by ProbeShift (G2). + +Every predictor maps IID training activations (+ labels, + optional augmentation +activations) to a scalar where **higher = predicted more OOD-stable**, so all can be +correlated against (negated) OOD drop under one convention. All are computed WITHOUT any +OOD data and WITHOUT new annotation, except `whitened_cosine_upper` which is flagged as an +OOD-using upper-bound reference (Truthfulness-Spectrum style). + +Implemented (each a verified competitor — see PRIOR_ART.md §四): + sip_eigengap SIP (2511.16288) — LDA eigengap / Fisher error + raptor_stability RAPTOR (2602.00158) — bootstrap direction mean|cos| (= dispersion) + fragility Fragility (2606.11375) — critical isotropic-noise collapse sigma + augmentation_robustness Probing-the-Probes — direction consistency under augmentation + xie_feature_dispersion Xie 2023 (2303.15488) — inter-class feature dispersion + pac ours (G3) — composite of dispersion + augmentation + whitened_cosine_upper Truth-Spectrum style — ID/OOD-whitened cosine (upper-bound ref) +""" +from __future__ import annotations + +import numpy as np +from scipy import linalg + +from config import STABILITY +from probes import make_probe +from metrics import accuracy +from stability_score import dispersion, augmentation_consistency, combine + + +# -------------------------------------------------------------------------------------- +# SIP — spectral identifiability (eigengap of the task-relevant LDA subspace) +# -------------------------------------------------------------------------------------- +def sip_eigengap(X: np.ndarray, y: np.ndarray, num_labels: int, ridge: float = 1e-3, **_): + X = np.asarray(X, dtype=np.float64) + classes = np.unique(y) + mu = X.mean(0) + Sb = np.zeros((X.shape[1], X.shape[1])) + Sw = np.zeros_like(Sb) + for c in classes: + Xc = X[y == c] + d = (Xc.mean(0) - mu)[:, None] + Sb += len(Xc) * (d @ d.T) + Sw += np.cov(Xc, rowvar=False) * (len(Xc) - 1) + Sw += ridge * np.eye(X.shape[1]) + # generalised eigenvalues of (Sb, Sw); top (num_labels-1) span the relevant subspace + eig = np.sort(linalg.eigvalsh(Sb, Sw))[::-1] + k = max(1, num_labels - 1) + if len(eig) <= k: + return float("nan") + gap = eig[k - 1] - eig[k] # gap after the relevant subspace + fisher_err = np.sqrt(X.shape[1] / len(X)) # ~ estimation error scale + return float(gap / (eig[0] + 1e-12) / (fisher_err + 1e-12)) + + +# -------------------------------------------------------------------------------------- +# RAPTOR-style directional stability == 1 - dispersion +# -------------------------------------------------------------------------------------- +def raptor_stability(X, y, num_labels, probe_kind="logreg", seed=0, **kw): + disp = dispersion(np.asarray(X, np.float32), np.asarray(y), num_labels, + probe_kind=probe_kind, k=STABILITY.k_bootstrap, seed=seed, **kw) + return float(1.0 - disp) + + +# -------------------------------------------------------------------------------------- +# Fragility — critical isotropic-noise collapse level +# -------------------------------------------------------------------------------------- +def fragility(X, y, num_labels, probe_kind="logreg", seed=0, + sigmas=(0.0, 0.25, 0.5, 1.0, 1.5, 2.0, 3.0), **kw): + X = np.asarray(X, dtype=np.float32) + rng = np.random.default_rng(seed) + p = make_probe(probe_kind, num_labels=num_labels, seed=seed, **kw).fit(X, y) + iid_acc = accuracy(y, p.predict(X)) + chance = 1.0 / num_labels + thresh = chance + 0.5 * (iid_acc - chance) # halfway from chance to IID + scale = X.std() + critical = sigmas[-1] + for s in sigmas: + Xn = X + s * scale * rng.standard_normal(X.shape).astype(np.float32) + if accuracy(y, p.predict(Xn)) < thresh: + critical = s + break + return float(critical) # higher = more robust = more stable + + +# -------------------------------------------------------------------------------------- +# augmentation-robustness == augmentation_consistency component +# -------------------------------------------------------------------------------------- +def augmentation_robustness(X, y, aug_X_list, num_labels, probe_kind="logreg", seed=0, **kw): + return float(augmentation_consistency( + np.asarray(X, np.float32), np.asarray(y), [np.asarray(a, np.float32) for a in aug_X_list], + num_labels, probe_kind=probe_kind, seed=seed, **kw)) + + +# -------------------------------------------------------------------------------------- +# Xie 2023 — inter-class feature dispersion (predicts overall acc, mechanism contrast) +# -------------------------------------------------------------------------------------- +def xie_feature_dispersion(X, y, num_labels, **_): + X = np.asarray(X, dtype=np.float64) + classes = np.unique(y) + mus = np.stack([X[y == c].mean(0) for c in classes]) # [C,H] + inter = np.mean([np.linalg.norm(mus[i] - mus[j]) + for i in range(len(mus)) for j in range(i + 1, len(mus))]) + return float(inter / (X.std() + 1e-12)) + + +# -------------------------------------------------------------------------------------- +# PAC (ours, G3) — composite of dispersion + augmentation-consistency (IID-only) +# -------------------------------------------------------------------------------------- +def pac(X, y, aug_X_list, num_labels, probe_kind="logreg", seed=0, **kw): + disp = dispersion(np.asarray(X, np.float32), np.asarray(y), num_labels, + probe_kind=probe_kind, k=STABILITY.k_bootstrap, seed=seed, **kw) + aug = augmentation_consistency( + np.asarray(X, np.float32), np.asarray(y), + [np.asarray(a, np.float32) for a in aug_X_list], num_labels, + probe_kind=probe_kind, seed=seed, **kw) + return float(combine(disp, aug, STABILITY)) + + +# -------------------------------------------------------------------------------------- +# Whitened-cosine upper bound (Truth-Spectrum style) — uses covariance; flagged. +# -------------------------------------------------------------------------------------- +def whitened_cosine_upper(X, y, num_labels, probe_kind="logreg", seed=0, cov=None, **kw): + """If `cov` (e.g. OOD covariance) is supplied, this becomes the OOD-using upper bound; + with cov=None it whitens by ID covariance (the metric-ablation variant).""" + X = np.asarray(X, dtype=np.float64) + C = np.cov(X, rowvar=False) if cov is None else np.asarray(cov, np.float64) + C += 1e-3 * np.eye(C.shape[0]) + W = linalg.fractional_matrix_power(C, -0.5).real + return raptor_stability((X @ W).astype(np.float32), y, num_labels, + probe_kind=probe_kind, seed=seed, **kw) + + +# Registry: name -> (callable, needs_aug). All are higher = more stable. +PREDICTORS = { + "sip_eigengap": (sip_eigengap, False), + "raptor_stability": (raptor_stability, False), + "fragility": (fragility, False), + "augmentation_robustness": (augmentation_robustness, True), + "xie_feature_dispersion": (xie_feature_dispersion, False), + "pac": (pac, True), + "whitened_cosine_id": (whitened_cosine_upper, False), +} + + +def compute_all(X_iid, y_iid, aug_X_list, num_labels, probe_kind="logreg", seed=0) -> dict: + out = {} + for name, (fn, needs_aug) in PREDICTORS.items(): + try: + out[name] = fn(X_iid, y_iid, aug_X_list, num_labels, probe_kind=probe_kind, seed=seed) \ + if needs_aug else fn(X_iid, y_iid, num_labels, probe_kind=probe_kind, seed=seed) + except Exception as e: # keep the grid going; record the failure + out[name] = float("nan") + out[name + "__error"] = str(e) + return out diff --git a/repro_bundle/cache.py b/repro_bundle/cache.py new file mode 100644 index 0000000000000000000000000000000000000000..e3739bd1feb6a56a2648db89e985766787af7ea7 --- /dev/null +++ b/repro_bundle/cache.py @@ -0,0 +1,89 @@ +"""Activation-shard cache. + +One shard = activations for a single (model, dataset, distribution) triple. +Stored as a directory of `.npy` files so that the (potentially multi-GB) activation +tensor can be memory-mapped and sliced by layer without loading everything into RAM. + +Layout: + cache//// + acts.npy float16 [N, L+1, H] (L+1 = embeddings + L transformer layers) + labels.npy int64 [N] + ids.npy int64 [N] (stable example ids, for cross-shard alignment) + meta.json {model, dataset, distribution, pooling, n, n_layers, hidden, dtype} +""" +from __future__ import annotations + +import json +from pathlib import Path + +import numpy as np + +from config import CACHE_DIR + + +def shard_dir(model_key: str, dataset_key: str, distribution: str) -> Path: + return CACHE_DIR / model_key / dataset_key / distribution + + +def exists(model_key: str, dataset_key: str, distribution: str) -> bool: + d = shard_dir(model_key, dataset_key, distribution) + return (d / "acts.npy").exists() and (d / "meta.json").exists() + + +def save_shard( + model_key: str, + dataset_key: str, + distribution: str, + acts: np.ndarray, # [N, L+1, H] float16 + labels: np.ndarray, # [N] + ids: np.ndarray, # [N] + pooling: str, +) -> Path: + d = shard_dir(model_key, dataset_key, distribution) + d.mkdir(parents=True, exist_ok=True) + acts = np.ascontiguousarray(acts.astype(np.float16)) + np.save(d / "acts.npy", acts) + np.save(d / "labels.npy", labels.astype(np.int64)) + np.save(d / "ids.npy", ids.astype(np.int64)) + meta = { + "model": model_key, + "dataset": dataset_key, + "distribution": distribution, + "pooling": pooling, + "n": int(acts.shape[0]), + "n_layers": int(acts.shape[1]), # includes embedding layer + "hidden": int(acts.shape[2]), + "dtype": "float16", + } + (d / "meta.json").write_text(json.dumps(meta, indent=2)) + return d + + +def load_meta(model_key: str, dataset_key: str, distribution: str) -> dict: + return json.loads((shard_dir(model_key, dataset_key, distribution) / "meta.json").read_text()) + + +def load_shard( + model_key: str, + dataset_key: str, + distribution: str, + layer: int | None = None, + mmap: bool = True, +) -> tuple[np.ndarray, np.ndarray, np.ndarray]: + """Return (acts, labels, ids). + + If `layer` is given, acts is [N, H] for that layer (0 = embeddings, 1..L = layers). + Otherwise acts is the full [N, L+1, H] tensor (mmap'd by default — do not mutate). + """ + d = shard_dir(model_key, dataset_key, distribution) + mode = "r" if mmap else None + acts = np.load(d / "acts.npy", mmap_mode=mode) + labels = np.load(d / "labels.npy") + ids = np.load(d / "ids.npy") + if layer is not None: + acts = np.asarray(acts[:, layer, :], dtype=np.float32) # materialise one layer + # sanitise fp16-overflow inf/nan (e.g. Qwen "massive activations") -> finite. + # No-op for models without overflow (pythia/gpt2), so existing results are unchanged. + if not np.isfinite(acts).all(): + acts = np.nan_to_num(acts, nan=0.0, posinf=65504.0, neginf=-65504.0) + return acts, labels, ids diff --git a/repro_bundle/config.py b/repro_bundle/config.py new file mode 100644 index 0000000000000000000000000000000000000000..4707ed5fae48bf9766bd4314c2d9b94c06147a40 --- /dev/null +++ b/repro_bundle/config.py @@ -0,0 +1,198 @@ +"""Central configuration: model registry, dataset registry, shift specs, paths, hyper-params. + +Importing this module is dependency-light (no torch / transformers) so it can be used by +analysis scripts on a laptop. Heavy imports live in extract_activations.py / data/*. +""" +from __future__ import annotations + +import os +from dataclasses import dataclass, field +from pathlib import Path + +# -------------------------------------------------------------------------------------- +# Paths +# -------------------------------------------------------------------------------------- +ROOT = Path(__file__).resolve().parent +CACHE_DIR = Path(os.environ.get("PROBE_CACHE", ROOT / "cache")) # activation shards +RESULTS_DIR = Path(os.environ.get("PROBE_RESULTS", ROOT / "results")) # metrics, tables +DATA_DIR = Path(os.environ.get("PROBE_DATA", ROOT / "data_cache")) # shifted text cache + +for _d in (CACHE_DIR, RESULTS_DIR, DATA_DIR): + _d.mkdir(parents=True, exist_ok=True) + + +# -------------------------------------------------------------------------------------- +# Models — the size ladder (claim C4). Keys are short ids used in shard filenames. +# -------------------------------------------------------------------------------------- +@dataclass(frozen=True) +class ModelSpec: + key: str # short id, used on disk (no slashes) + hf_name: str # HF hub id + family: str # for grouping in mixed-effects / size ladder + params_m: float # parameter count in millions (for the scale axis) + + +MODELS: dict[str, ModelSpec] = { + m.key: m for m in [ + ModelSpec("pythia-70m", "EleutherAI/pythia-70m", "pythia", 70), + ModelSpec("pythia-160m", "EleutherAI/pythia-160m", "pythia", 160), + ModelSpec("pythia-410m", "EleutherAI/pythia-410m", "pythia", 410), + ModelSpec("pythia-1.4b", "EleutherAI/pythia-1.4b", "pythia", 1400), + ModelSpec("pythia-6.9b", "EleutherAI/pythia-6.9b", "pythia", 6900), # Tier 3: 7B spot-check + ModelSpec("gpt2", "gpt2", "gpt2", 124), + ModelSpec("gpt2-medium", "gpt2-medium", "gpt2", 355), + ModelSpec("qwen2.5-0.5b", "Qwen/Qwen2.5-0.5B", "qwen", 500), + ] +} + +# Default ladder for the headline scale experiment (cheapest -> most expensive). +SIZE_LADDER = ["pythia-70m", "pythia-160m", "pythia-410m", "pythia-1.4b"] + + +# -------------------------------------------------------------------------------------- +# Datasets — concept-bearing classification tasks (claim families C1/C3). +# `concept` groups datasets that probe the SAME underlying concept (for the Spearman +# generalisation analysis across concept families). +# -------------------------------------------------------------------------------------- +@dataclass(frozen=True) +class DatasetSpec: + key: str + hf_path: str + hf_config: str | None + split: str + text_field: str + label_field: str + concept: str # "sentiment" | "topic" | "pos" | "truth" + num_labels: int + # Domain-shift partner: another DatasetSpec key whose inputs share the label space + # but come from a different domain (used as the `domain` OOD distribution). + domain_partner: str | None = None + + +# +# NOTE: datasets>=5.0 dropped legacy short names AND script-based datasets (no more +# trust_remote_code). Use fully-qualified `namespace/name` Hub ids that ship parquet. +# Verified loadable 2026-06-22: nyu-mll/glue, stanfordnlp/imdb, fancyzhx/ag_news, +# fancyzhx/dbpedia_14, fancyzhx/amazon_polarity. +DATASETS: dict[str, DatasetSpec] = { + d.key: d for d in [ + # --- sentiment concept --- + DatasetSpec("sst2", "nyu-mll/glue", "sst2", "train", "sentence", "label", "sentiment", 2, + domain_partner="imdb"), + DatasetSpec("imdb", "stanfordnlp/imdb", None, "train", "text", "label", "sentiment", 2, + domain_partner="sst2"), + DatasetSpec("amazon_polarity", "fancyzhx/amazon_polarity", None, "train", "content", + "label", "sentiment", 2, domain_partner="sst2"), + # --- topic concept --- + DatasetSpec("ag_news", "fancyzhx/ag_news", None, "train", "text", "label", "topic", 4, + domain_partner="dbpedia"), + DatasetSpec("dbpedia", "fancyzhx/dbpedia_14", None, "train", "content", "label", "topic", 14), + # --- syntactic concept (POS) --- + # WARNING: universal_dependencies is script-based -> NOT loadable under datasets>=5.0. + # TODO(4090): swap to a parquet POS source (e.g. a CoNLL-2003 POS parquet) before using. + DatasetSpec("ud_pos", "universal-dependencies/universal_dependencies", "en_ewt", "train", + "tokens", "upos", "pos", 17), + # --- truthfulness / factual concept (derived, label from existing fields) --- + # TODO(4090): truthful_qa generation may be script-based; counterfact-tracing ships parquet. + DatasetSpec("truthfulqa", "truthfulqa/truthful_qa", "generation", "validation", "question", + "label", "truth", 2), + DatasetSpec("counterfact", "NeelNanda/counterfact-tracing", None, "train", "prompt", + "label", "truth", 2), + # --- extra concepts for cross-concept (LOCO) power; all parquet, verified 2026-06-22 --- + DatasetSpec("emotion", "dair-ai/emotion", None, "train", "text", "label", "emotion", 6), + DatasetSpec("tweet_hate", "cardiffnlp/tweet_eval", "hate", "train", "text", "label", + "hate", 2), + DatasetSpec("tweet_irony", "cardiffnlp/tweet_eval", "irony", "train", "text", "label", + "irony", 2), + DatasetSpec("tweet_offensive", "cardiffnlp/tweet_eval", "offensive", "train", "text", + "label", "offensive", 2), + DatasetSpec("subj", "SetFit/subj", None, "train", "text", "label", "subjectivity", 2), + DatasetSpec("spam", "ucirvine/sms_spam", "plain_text", "train", "sms", "label", "spam", 2), + # --- 3 more concepts -> 12 total (verified loadable 2026-06-22) --- + DatasetSpec("cola", "nyu-mll/glue", "cola", "train", "sentence", "label", "grammaticality", 2), + DatasetSpec("stance", "cardiffnlp/tweet_eval", "stance_feminist", "train", "text", "label", + "stance", 3), + DatasetSpec("amazon_cf", "mteb/amazon_counterfactual", "en", "train", "text", "label", + "counterfactual", 2), + ] +} + + +# -------------------------------------------------------------------------------------- +# Distribution shifts (label-preserving). Each (dataset, distribution) pair is one +# activation shard. "iid" is the in-distribution train/test of the dataset itself. +# -------------------------------------------------------------------------------------- +SHIFTS = ["iid", "paraphrase", "domain", "length"] + +# Back-translation pivot for the `paraphrase` SHIFT (rotation ground truth) — keep single & +# fixed (de) so rotation measurements are consistent across runs. +BT_PIVOTS = ["de"] + +# Tier 2: the AUGMENTATION predictor uses MULTIPLE pivots for genuine paraphrastic diversity +# (de/fr/ru). With PROBE_N_AUG=3 each aug uses a different pivot, so augmentation-consistency +# reflects robustness to varied surface forms (not the single-pivot artefact of the 1-aug run). +AUG_PIVOTS = ["de", "fr", "ru"] + +# Length buckets (token counts) for the `length` shift: train on SHORT, eval on LONG. +LENGTH_BUCKETS = {"short": (0, 32), "long": (96, 512)} + + +# -------------------------------------------------------------------------------------- +# Extraction hyper-params +# -------------------------------------------------------------------------------------- +@dataclass +class ExtractConfig: + pooling: str = "mean" # "mean" (masked) | "last" (last non-pad token) + max_length: int = 256 + batch_size: int = 32 # lower for pythia-1.4b on 24GB + dtype: str = "float16" + n_train: int = 2000 # examples used to FIT probes per seed (subsampled from pool) + n_eval: int = 1000 # eval examples per (dataset, distribution) per seed (from pool) + n_train_pool: int = 4000 # examples EXTRACTED for train (multi-seed subsamples from this) + n_eval_pool: int = 2000 # examples EXTRACTED for eval/shifts (subsampled per seed) + device: str = "cuda" + + +# -------------------------------------------------------------------------------------- +# Probe / Stability-Score / stats hyper-params +# -------------------------------------------------------------------------------------- +@dataclass +class ProbeConfig: + probe_types: tuple[str, ...] = ("logreg", "mass_mean", "mlp") + l2: float = 1.0 # inverse-C handled inside probes.py + mlp_hidden: int = 128 + mlp_epochs: int = 50 + layer_selection: str = "iid_val" # pick layer by IID val acc, never post-hoc on OOD + + +@dataclass +class StabilityConfig: + k_bootstrap: int = 8 # K resamples for dispersion (multi-seed compensates) + n_aug: int = 5 # label-preserving augmentations for consistency + # combine(dispersion, aug_consistency) -> scalar score; weights are a hyper-param we + # report sensitivity for. Pre-registered hypothesis: combination > dispersion alone. + w_dispersion: float = 0.5 + w_consistency: float = 0.5 + + +@dataclass +class StatsConfig: + seeds: tuple[int, ...] = (0, 1, 2, 3, 4) # >=5; covers init / resample / augmentation + n_bootstrap: int = 1000 + fdr_method: str = "holm" # "holm" | "bh" + + +# Singletons used across the pipeline (override per-run in run_pipeline.py if needed). +EXTRACT = ExtractConfig() +PROBE = ProbeConfig() +STABILITY = StabilityConfig() +STATS = StatsConfig() + +# Env overrides — let a small pilot run cheaply without editing code, e.g. +# PROBE_N_TRAIN=600 PROBE_N_EVAL=400 PROBE_N_AUG=2 PROBE_BATCH=64 python run_pipeline.py ... +EXTRACT.n_train = int(os.environ.get("PROBE_N_TRAIN", EXTRACT.n_train)) +EXTRACT.n_eval = int(os.environ.get("PROBE_N_EVAL", EXTRACT.n_eval)) +EXTRACT.n_train_pool = int(os.environ.get("PROBE_N_TRAIN_POOL", EXTRACT.n_train_pool)) +EXTRACT.n_eval_pool = int(os.environ.get("PROBE_N_EVAL_POOL", EXTRACT.n_eval_pool)) +EXTRACT.batch_size = int(os.environ.get("PROBE_BATCH", EXTRACT.batch_size)) +STABILITY.n_aug = int(os.environ.get("PROBE_N_AUG", STABILITY.n_aug)) diff --git a/repro_bundle/data/__init__.py b/repro_bundle/data/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..13e9594663725ccfbe23499499e45bfa5944c673 --- /dev/null +++ b/repro_bundle/data/__init__.py @@ -0,0 +1 @@ +"""Data subpackage: dataset loading, label-preserving shifts, label-fidelity audit.""" diff --git a/repro_bundle/data/datasets.py b/repro_bundle/data/datasets.py new file mode 100644 index 0000000000000000000000000000000000000000..1b5218ad0c80c83b6e5404beac2e17c9d13edeca --- /dev/null +++ b/repro_bundle/data/datasets.py @@ -0,0 +1,121 @@ +"""Load HF datasets into a unified record format. + +A loaded split is a dict: + {"texts": list[str], "labels": np.ndarray[int], "ids": np.ndarray[int]} + +`ids` are stable per (dataset, split) so the SAME example can be aligned across the +iid / paraphrase / length shards (paraphrase keeps the id of its source). + +Standard text-classification datasets are loaded directly. POS / truthfulness datasets +are *derived* into a probing format (see builder functions). Derived builders are marked +TODO where a modelling choice should be revisited on the 4090. +""" +from __future__ import annotations + +import numpy as np + +from config import DATASETS, DatasetSpec + + +def _subsample(texts, labels, n, seed): + rng = np.random.default_rng(seed) + n = min(n, len(texts)) + # stratified-ish: shuffle then take n (good enough; balance checked downstream) + idx = rng.permutation(len(texts))[:n] + return [texts[i] for i in idx], np.asarray(labels)[idx], idx.astype(np.int64) + + +# -------------------------------------------------------------------------------------- +# Builders for derived / non-trivial datasets +# -------------------------------------------------------------------------------------- +def _build_ud_pos(spec, split, n, seed): + """Word-in-context POS probe: pick one target token per sentence, mark it with [], + label = its UPOS tag. Simple, deterministic target selection (first content word). + TODO(4090): consider sampling multiple target positions per sentence for more data. + """ + from datasets import load_dataset + ds = load_dataset(spec.hf_path, spec.hf_config, split=split) + texts, labels = [], [] + for ex in ds: + toks, tags = ex["tokens"], ex["upos"] + if not toks: + continue + j = min(range(len(toks)), key=lambda i: (tags[i] in (0,), i)) # first non-punct-ish + marked = " ".join(toks[:j] + ["[" + toks[j] + "]"] + toks[j + 1:]) + texts.append(marked) + labels.append(int(tags[j])) + return _subsample(texts, labels, n, seed) + + +def _build_truthfulqa(spec, split, n, seed): + """Derive a binary truth probe from TruthfulQA: (question + correct answer) -> 1, + (question + incorrect answer) -> 0. Label comes from existing fields (no annotation). + """ + from datasets import load_dataset + ds = load_dataset(spec.hf_path, spec.hf_config, split=split) + texts, labels = [], [] + for ex in ds: + q = ex["question"] + for a in ex.get("correct_answers", [])[:1]: + texts.append(f"Q: {q} A: {a}"); labels.append(1) + for a in ex.get("incorrect_answers", [])[:1]: + texts.append(f"Q: {q} A: {a}"); labels.append(0) + return _subsample(texts, labels, n, seed) + + +def _build_counterfact(spec, split, n, seed): + """Derive true/false statements from counterfact-tracing prompt + target_true/false.""" + from datasets import load_dataset + ds = load_dataset(spec.hf_path, split=split) + texts, labels = [], [] + for ex in ds: + prompt = ex.get("prompt") or ex.get("relation_prefix", "") + t, f = ex.get("target_true"), ex.get("target_false") + if t: + texts.append(f"{prompt} {t}".strip()); labels.append(1) + if f: + texts.append(f"{prompt} {f}".strip()); labels.append(0) + return _subsample(texts, labels, n, seed) + + +_DERIVED = { + "ud_pos": _build_ud_pos, + "truthfulqa": _build_truthfulqa, + "counterfact": _build_counterfact, +} + + +# -------------------------------------------------------------------------------------- +# Public API +# -------------------------------------------------------------------------------------- +def load(dataset_key: str, n: int, split: str | None = None, seed: int = 0) -> dict: + spec: DatasetSpec = DATASETS[dataset_key] + split = split or spec.split + + if dataset_key in _DERIVED: + texts, labels, ids = _DERIVED[dataset_key](spec, split, n, seed) + else: + from datasets import load_dataset + ds = load_dataset(spec.hf_path, spec.hf_config, split=split) + texts = list(ds[spec.text_field]) + labels = list(ds[spec.label_field]) + texts, labels, ids = _subsample(texts, labels, n, seed) + + return {"texts": texts, "labels": np.asarray(labels, dtype=np.int64), "ids": ids} + + +def load_train_eval(dataset_key: str, n_train: int, n_eval: int, seed: int = 0 + ) -> tuple[dict, dict]: + """Disjoint IID train / eval draws from the dataset's train split. + + If the dataset is smaller than n_train + n_eval, split PROPORTIONALLY so eval is never + empty (small datasets like tweet_eval/stance_feminist have <1500 rows). + """ + full = load(dataset_key, n=n_train + n_eval, split=None, seed=seed) + n = len(full["texts"]) + cut = n_train if n >= n_train + n_eval else max(1, int(n * n_train / (n_train + n_eval))) + tr = {"texts": full["texts"][:cut], "labels": full["labels"][:cut], + "ids": full["ids"][:cut]} + ev = {"texts": full["texts"][cut:], "labels": full["labels"][cut:], + "ids": full["ids"][cut:]} + return tr, ev diff --git a/repro_bundle/data/label_fidelity.py b/repro_bundle/data/label_fidelity.py new file mode 100644 index 0000000000000000000000000000000000000000..81ca8d84e6a652d7fc794bba42bc1f91629104e8 --- /dev/null +++ b/repro_bundle/data/label_fidelity.py @@ -0,0 +1,66 @@ +"""Label-fidelity audit (claim C5). + +A shift is only valid if it is *label-preserving*. We verify this WITHOUT new human +annotation using a local NLI model (DeBERTa-v3-MNLI): a paraphrase should be mutually +entailing with its source; we drop pairs that flip. We report the flip rate so reviewers +can see the shifts are clean. + +This is the standard "are your counterfactuals actually label-preserving?" defence. +Zero API cost — the NLI model runs locally. +""" +from __future__ import annotations + +import numpy as np + +_NLI = None +_NLI_LABELS = None # index order of {entailment, neutral, contradiction} + + +def _get_nli(model_name: str = "MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli"): + global _NLI, _NLI_LABELS + if _NLI is None: + import torch + from transformers import AutoModelForSequenceClassification, AutoTokenizer + tok = AutoTokenizer.from_pretrained(model_name) + model = AutoModelForSequenceClassification.from_pretrained(model_name) + if torch.cuda.is_available(): + model = model.to("cuda").half() + model.eval() + _NLI = (tok, model) + _NLI_LABELS = model.config.id2label # {0:'entailment',...} varies by checkpoint + return _NLI + + +def _entail_prob(premises: list[str], hypotheses: list[str], batch_size: int = 32) -> np.ndarray: + """P(entailment) for each (premise, hypothesis) pair.""" + import torch + tok, model = _get_nli() + device = next(model.parameters()).device + ent_idx = [i for i, v in _NLI_LABELS.items() if "entail" in v.lower()][0] + probs = [] + for i in range(0, len(premises), batch_size): + enc = tok(premises[i:i + batch_size], hypotheses[i:i + batch_size], + return_tensors="pt", padding=True, truncation=True, max_length=256).to(device) + with torch.no_grad(): + logits = model(**enc).logits.float() + p = torch.softmax(logits, dim=-1)[:, ent_idx] + probs.extend(p.cpu().numpy().tolist()) + return np.asarray(probs) + + +def paraphrase_keep_mask(originals: list[str], shifted: list[str], thresh: float = 0.5 + ) -> np.ndarray: + """Boolean mask: keep pairs that are mutually entailing (label preserved). + + TODO(4090): tune `thresh` on a small dev set of known paraphrases; report flip rate + and a sensitivity curve over thresh in the paper appendix. + """ + fwd = _entail_prob(originals, shifted) + bwd = _entail_prob(shifted, originals) + keep = (fwd >= thresh) & (bwd >= thresh) + return keep + + +def audit(originals: list[str], shifted: list[str], thresh: float = 0.5) -> dict: + keep = paraphrase_keep_mask(originals, shifted, thresh) + return {"keep_mask": keep, "flip_rate": float(1.0 - keep.mean()), "n": len(originals)} diff --git a/repro_bundle/data/shifts.py b/repro_bundle/data/shifts.py new file mode 100644 index 0000000000000000000000000000000000000000..787e34a563b8590d163a2852669627248a6438b9 --- /dev/null +++ b/repro_bundle/data/shifts.py @@ -0,0 +1,146 @@ +"""Label-preserving distribution shifts. + +Each shift maps an IID record dict -> a new record dict with the SAME labels/ids but +shifted *inputs*. The probe is trained on IID train; these produce the OOD eval sets +(claim C1). They are also reused as the augmentations for augmentation_consistency +(claim C2) — augmentation == a mild label-preserving shift of the TRAIN set. + +Shifts implemented: + * paraphrase : round-trip back-translation (Marian opus-mt en->pivot->en). + * domain : swap in the `domain_partner` dataset (same label space). + * length : keep only the LONG token-length bucket (train uses SHORT bucket). + +All shifted text is cached to disk (data_cache/) so the expensive back-translation runs +once and is $0 thereafter (local models, zero API). +""" +from __future__ import annotations + +import hashlib +import json +from pathlib import Path + +import numpy as np + +from config import AUG_PIVOTS, BT_PIVOTS, DATASETS, DATA_DIR, LENGTH_BUCKETS +from data import datasets as ds_mod + + +# -------------------------------------------------------------------------------------- +# Disk cache for shifted text (keyed by content hash so re-runs are free) +# -------------------------------------------------------------------------------------- +def _cache_path(tag: str, key: str) -> Path: + h = hashlib.md5(key.encode()).hexdigest()[:12] + p = DATA_DIR / tag + p.mkdir(parents=True, exist_ok=True) + return p / f"{h}.json" + + +def _load_cached(tag, key): + p = _cache_path(tag, key) + return json.loads(p.read_text()) if p.exists() else None + + +def _save_cached(tag, key, obj): + _cache_path(tag, key).write_text(json.dumps(obj)) + + +# -------------------------------------------------------------------------------------- +# paraphrase via back-translation +# -------------------------------------------------------------------------------------- +_BT_CACHE: dict = {} + + +def _get_marian(src: str, tgt: str): + """Lazy-load + memoise a Marian opus-mt translation model.""" + name = f"Helsinki-NLP/opus-mt-{src}-{tgt}" + if name not in _BT_CACHE: + from transformers import MarianMTModel, MarianTokenizer + import torch + tok = MarianTokenizer.from_pretrained(name) + model = MarianMTModel.from_pretrained(name) + if torch.cuda.is_available(): + model = model.to("cuda").half() + _BT_CACHE[name] = (tok, model) + return _BT_CACHE[name] + + +def _translate(texts: list[str], src: str, tgt: str, batch_size: int = 32) -> list[str]: + import torch + tok, model = _get_marian(src, tgt) + device = next(model.parameters()).device + out: list[str] = [] + for i in range(0, len(texts), batch_size): + batch = texts[i:i + batch_size] + enc = tok(batch, return_tensors="pt", padding=True, truncation=True, + max_length=256).to(device) + with torch.no_grad(): + gen = model.generate(**enc, max_new_tokens=256) + out.extend(tok.batch_decode(gen, skip_special_tokens=True)) + return out + + +def paraphrase(record: dict, pivot: str | None = None) -> dict: + pivot = pivot or BT_PIVOTS[0] + key = f"{pivot}:" + "␟".join(record["texts"]) + cached = _load_cached("paraphrase", key) + if cached is None: + mid = _translate(record["texts"], "en", pivot) + back = _translate(mid, pivot, "en") + cached = back + _save_cached("paraphrase", key, cached) + return {"texts": cached, "labels": record["labels"], "ids": record["ids"]} + + +# -------------------------------------------------------------------------------------- +# domain shift +# -------------------------------------------------------------------------------------- +def domain(dataset_key: str, n: int, seed: int = 0) -> dict | None: + partner = DATASETS[dataset_key].domain_partner + if partner is None: + return None + if DATASETS[partner].num_labels != DATASETS[dataset_key].num_labels: + # label spaces differ (e.g. ag_news 4 vs dbpedia 14) -> domain shift undefined. + # TODO(4090): map to shared coarse classes, or restrict `domain` to sentiment. + return None + return ds_mod.load(partner, n=n, seed=seed) + + +# -------------------------------------------------------------------------------------- +# length shift +# -------------------------------------------------------------------------------------- +def _token_len(t: str) -> int: + return len(t.split()) + + +def length_bucket(record: dict, bucket: str) -> dict: + lo, hi = LENGTH_BUCKETS[bucket] + keep = [i for i, t in enumerate(record["texts"]) if lo <= _token_len(t) < hi] + return { + "texts": [record["texts"][i] for i in keep], + "labels": record["labels"][np.asarray(keep, dtype=int)], + "ids": record["ids"][np.asarray(keep, dtype=int)], + } + + +# -------------------------------------------------------------------------------------- +# dispatcher +# -------------------------------------------------------------------------------------- +def make_shift(distribution: str, dataset_key: str, iid_record: dict, n: int, seed: int = 0): + """Return the eval record for a given distribution, or None if not applicable.""" + if distribution == "iid": + return iid_record + if distribution == "paraphrase": + return paraphrase(iid_record) + if distribution == "domain": + return domain(dataset_key, n=n, seed=seed) + if distribution == "length": + return length_bucket(iid_record, "long") + raise ValueError(f"unknown distribution {distribution}") + + +def make_augmentations(dataset_key: str, train_record: dict, n_aug: int) -> list[dict]: + """n_aug label-preserving augmentations of the TRAIN set for augmentation_consistency. + Uses DISTINCT back-translation pivots (AUG_PIVOTS = de/fr/ru) for genuine diversity; + cycles if n_aug > len(AUG_PIVOTS).""" + pivots = (AUG_PIVOTS * ((n_aug // len(AUG_PIVOTS)) + 1))[:n_aug] + return [paraphrase(train_record, pivot=p) for p in pivots] diff --git a/repro_bundle/extract_activations.py b/repro_bundle/extract_activations.py new file mode 100644 index 0000000000000000000000000000000000000000..317d0edd12c67e0d91256facec7e804586cbd219 --- /dev/null +++ b/repro_bundle/extract_activations.py @@ -0,0 +1,79 @@ +"""Residual-stream activation extraction (pipeline stage M1 — the only big GPU cost). + +For each (model, dataset, distribution) we run a single forward pass over the texts with +`output_hidden_states=True`, masked-mean-pool (or last-token-pool) each layer, cast to +float16, and write one cache shard. Everything downstream reads the cache. + +Memory discipline (24 GB 4090): + * model loaded in float16 + * inference under torch.no_grad() + * per-batch hidden states pooled and moved to CPU immediately (we never keep [B,L,T,H]) + * batch_size tuned per model in config (drop for pythia-1.4b) +""" +from __future__ import annotations + +import numpy as np + +import cache +from config import EXTRACT, MODELS, ExtractConfig + + +def _pool(hidden_stack, attention_mask, pooling: str): + """hidden_stack: [B, L+1, T, H] ; attention_mask: [B, T] -> [B, L+1, H].""" + import torch + if pooling == "mean": + mask = attention_mask[:, None, :, None] # [B,1,T,1] + summed = (hidden_stack * mask).sum(dim=2) # [B,L+1,H] + counts = mask.sum(dim=2).clamp(min=1) # [B,1,H]->broadcast + return summed / counts + if pooling == "last": + # index of last non-pad token per example + lengths = attention_mask.sum(dim=1).long() - 1 # [B] + idx = lengths[:, None, None, None].expand(-1, hidden_stack.size(1), 1, hidden_stack.size(3)) + return torch.gather(hidden_stack, 2, idx).squeeze(2) # [B,L+1,H] + raise ValueError(f"unknown pooling {pooling}") + + +def extract_texts(model_key: str, texts: list[str], cfg: ExtractConfig) -> np.ndarray: + """Return pooled activations [N, L+1, H] (float16) for the given texts.""" + import torch + from transformers import AutoModel, AutoTokenizer + + spec = MODELS[model_key] + tok = AutoTokenizer.from_pretrained(spec.hf_name) + if tok.pad_token is None: + tok.pad_token = tok.eos_token + dtype = getattr(torch, cfg.dtype) + device = cfg.device if torch.cuda.is_available() else "cpu" + model = AutoModel.from_pretrained( + spec.hf_name, torch_dtype=dtype, output_hidden_states=True + ).to(device).eval() + + chunks = [] + with torch.no_grad(): + for i in range(0, len(texts), cfg.batch_size): + batch = texts[i:i + cfg.batch_size] + enc = tok(batch, return_tensors="pt", padding=True, truncation=True, + max_length=cfg.max_length).to(device) + out = model(**enc) + hs = torch.stack(out.hidden_states, dim=1) # [B, L+1, T, H] + pooled = _pool(hs, enc.attention_mask, cfg.pooling) # [B, L+1, H] + chunks.append(pooled.to(torch.float16).cpu().numpy()) + del out, hs, pooled, enc + + del model + if torch.cuda.is_available(): + torch.cuda.empty_cache() + return np.concatenate(chunks, axis=0) + + +def extract_shard(model_key: str, dataset_key: str, distribution: str, record: dict, + cfg: ExtractConfig = EXTRACT, overwrite: bool = False): + """Extract + cache one shard. `record` = {texts, labels, ids}.""" + if len(record["texts"]) == 0: + return None # empty record (e.g. paraphrase fully filtered out) -> skip, don't crash + if cache.exists(model_key, dataset_key, distribution) and not overwrite: + return cache.shard_dir(model_key, dataset_key, distribution) + acts = extract_texts(model_key, record["texts"], cfg) + return cache.save_shard(model_key, dataset_key, distribution, + acts, record["labels"], record["ids"], cfg.pooling) diff --git a/repro_bundle/fig1_circularity.pdf b/repro_bundle/fig1_circularity.pdf new file mode 100644 index 0000000000000000000000000000000000000000..53b8157402234c67bcb14df01efb427840278970 Binary files /dev/null and b/repro_bundle/fig1_circularity.pdf differ diff --git a/repro_bundle/fig2_sizeladder.pdf b/repro_bundle/fig2_sizeladder.pdf new file mode 100644 index 0000000000000000000000000000000000000000..a8e4bd1dafd46e96321240ae1ab1a0b7baaf01a0 Binary files /dev/null and b/repro_bundle/fig2_sizeladder.pdf differ diff --git a/repro_bundle/fig3_mechanism.pdf b/repro_bundle/fig3_mechanism.pdf new file mode 100644 index 0000000000000000000000000000000000000000..6f35dc887dc60b774a0f0178202095d541245eb9 Binary files /dev/null and b/repro_bundle/fig3_mechanism.pdf differ diff --git a/repro_bundle/figures.py b/repro_bundle/figures.py new file mode 100644 index 0000000000000000000000000000000000000000..a06bc44ed481e20789c66b548260ba3fa789bc1f --- /dev/null +++ b/repro_bundle/figures.py @@ -0,0 +1,117 @@ +"""Generate the paper's headline figures from cached results (no GPU). + python figures.py # writes figures/*.pdf + *.png from results_A/ +""" +from __future__ import annotations + +import json +from pathlib import Path + +import numpy as np +import matplotlib +matplotlib.use("Agg") +import matplotlib.pyplot as plt + +import metrics +from config import MODELS +from run_pipeline import (_para_rotation, _iid_split_rotation, _mean_ood_rotation, _mean_ood_drop) + +OUT = Path("figures"); OUT.mkdir(exist_ok=True) +PREDS = ["raptor_stability", "pac", "augmentation_robustness", "whitened_cosine_id", + "fragility", "xie_feature_dispersion", "sip_eigengap"] +LBL = {"raptor_stability": "RAPTOR\n(dispersion)", "pac": "PAC (ours)", + "augmentation_robustness": "aug-robust", "whitened_cosine_id": "whitened-cos", + "fragility": "Fragility", "xie_feature_dispersion": "Xie-disp", "sip_eigengap": "SIP"} + + +def load_rows(results_dir, predfile="predictors.jsonl"): + R = Path(results_dir) + evals = [json.loads(l) for l in (R / "eval.jsonl").read_text().splitlines() if l.strip()] + preds = {(p["model"], p["dataset"], p["seed"]): p["predictors"] + for p in (json.loads(l) for l in (R / predfile).read_text().splitlines() if l.strip())} + rows = [] + for e in evals: + if e.get("probe") != "logreg": + continue + k = (e["model"], e["dataset"], e["seed"]) + if k not in preds: + continue + para, iids = _para_rotation(e), _iid_split_rotation(e) + excess = (para - iids) if (para == para and iids == iids) else np.nan + rows.append({"model": e["model"], "naive": _mean_ood_rotation(e), "placebo": iids, + "excess": excess, "drop": _mean_ood_drop(e), **preds[k]}) + return rows + + +def rho(rows, pred, target): + s = np.array([r.get(pred, np.nan) for r in rows], float) + y = np.array([r[target] for r in rows], float) + ok = ~(np.isnan(s) | np.isnan(y)) + return metrics.spearman(s[ok], -y[ok])[0] if ok.sum() >= 8 else np.nan + + +def fig_circularity(rows): + targets = [("naive", "naive rotation"), ("placebo", "IID-resample\n(placebo)"), + ("excess", "EXCESS\n(shift-specific)")] + x = np.arange(len(PREDS)); w = 0.26 + fig, ax = plt.subplots(figsize=(9, 4.2)) + colors = ["#4C72B0", "#999999", "#C44E52"] + for i, (t, name) in enumerate(targets): + vals = [rho(rows, p, t) for p in PREDS] + ax.bar(x + (i - 1) * w, vals, w, label=name, color=colors[i]) + ax.axhline(0, color="k", lw=0.8) + ax.set_xticks(x); ax.set_xticklabels([LBL[p] for p in PREDS], fontsize=8) + ax.set_ylabel("Spearman ρ (predictor, −target)") + ax.set_title("Predicting probe direction rotation: naive ρ is circular (≈ placebo);\n" + "on EXCESS dispersion turns anti-predictive while augmentation is least-bad") + ax.legend(fontsize=8, loc="lower left"); ax.set_ylim(-0.45, 1.0) + fig.tight_layout(); fig.savefig(OUT / "fig1_circularity.pdf"); fig.savefig(OUT / "fig1_circularity.png", dpi=150) + plt.close(fig) + + +def fig_sizeladder(rows): + models = sorted({r["model"] for r in rows}, key=lambda m: MODELS[m].params_m if m in MODELS else 0) + xs = [MODELS[m].params_m for m in models if m in MODELS] + fig, axes = plt.subplots(1, 2, figsize=(10, 4)) + for ax, tgt, title in [(axes[0], "naive", "naive rotation"), (axes[1], "excess", "EXCESS rotation")]: + for p, c in [("raptor_stability", "#4C72B0"), ("augmentation_robustness", "#C44E52"), ("pac", "#55A868")]: + ys = [rho([r for r in rows if r["model"] == m], p, tgt) for m in models if m in MODELS] + ax.plot(xs, ys, "o-", label=LBL[p].replace("\n", " "), color=c) + ax.axhline(0, color="k", lw=0.6); ax.set_xscale("log") + ax.set_xlabel("params (M, log)"); ax.set_ylabel("Spearman ρ"); ax.set_title(title) + ax.legend(fontsize=8) + fig.suptitle("Size-ladder: predictability structure is scale-invariant (no inverse-scaling)") + fig.tight_layout(); fig.savefig(OUT / "fig2_sizeladder.pdf"); fig.savefig(OUT / "fig2_sizeladder.png", dpi=150) + plt.close(fig) + + +def fig_mechanism(rows): + """Why augmentation > dispersion on EXCESS: dispersion (IID bootstrap) is unrelated/anti to + the shift-specific component; augmentation (a label-preserving perturbation) tracks it.""" + fig, axes = plt.subplots(1, 2, figsize=(9, 4), sharey=True) + for ax, p, c in [(axes[0], "raptor_stability", "#4C72B0"), + (axes[1], "augmentation_robustness", "#C44E52")]: + s = np.array([r.get(p, np.nan) for r in rows], float) + y = np.array([r["excess"] for r in rows], float) + ok = ~(np.isnan(s) | np.isnan(y)) + s, y = s[ok], y[ok] + ax.scatter(s, y, s=8, alpha=0.3, color=c) + if len(s) > 2: + b, a = np.polyfit(s, y, 1) + xs = np.linspace(s.min(), s.max(), 20) + ax.plot(xs, a + b * xs, color="k", lw=1.5) + r = metrics.spearman(s, -y)[0] + ax.set_title(f"{LBL[p].replace(chr(10),' ')}\nρ(score,−excess)={r:+.2f}") + ax.set_xlabel(f"{p} (a-priori stability)"); ax.axhline(0, color="grey", lw=0.6) + axes[0].set_ylabel("EXCESS rotation (shift-specific)") + fig.suptitle("Mechanism: dispersion is blind to shift-specific fragility; augmentation is not") + fig.tight_layout(); fig.savefig(OUT / "fig3_mechanism.pdf"); fig.savefig(OUT / "fig3_mechanism.png", dpi=150) + plt.close(fig) + + +if __name__ == "__main__": + rows = load_rows("results_A/results") + print(f"loaded {len(rows)} configs") + fig_circularity(rows) + fig_sizeladder(rows) + fig_mechanism(rows) + print("wrote figures/fig1_circularity, fig2_sizeladder, fig3_mechanism (.pdf/.png)") diff --git a/repro_bundle/figures/fig1_circularity.pdf b/repro_bundle/figures/fig1_circularity.pdf new file mode 100644 index 0000000000000000000000000000000000000000..38888d540e925797c988cba9a15065bb397acd96 Binary files /dev/null and b/repro_bundle/figures/fig1_circularity.pdf differ diff --git a/repro_bundle/figures/fig1_circularity.png b/repro_bundle/figures/fig1_circularity.png new file mode 100644 index 0000000000000000000000000000000000000000..c319b3688b30dfd48d412e55bfef4c2cbbca9565 --- /dev/null +++ b/repro_bundle/figures/fig1_circularity.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13c3d2f3653f38fb99e0744d1b341c97b2c28b03039926b99bb084d2f7d0d8a9 +size 67445 diff --git a/repro_bundle/figures/fig2_sizeladder.pdf b/repro_bundle/figures/fig2_sizeladder.pdf new file mode 100644 index 0000000000000000000000000000000000000000..00a7a95dbe87294db0d80bf6766c4d0c036e94c4 Binary files /dev/null and b/repro_bundle/figures/fig2_sizeladder.pdf differ diff --git a/repro_bundle/figures/fig2_sizeladder.png b/repro_bundle/figures/fig2_sizeladder.png new file mode 100644 index 0000000000000000000000000000000000000000..aaa01003b35f40f2acb2bd59747f167fad8e47e7 --- /dev/null +++ b/repro_bundle/figures/fig2_sizeladder.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca21cf1e5f9f1a5868022a1d9e76e7315f65fc3368a7483c951dd4169827c425 +size 128401 diff --git a/repro_bundle/figures/fig3_mechanism.pdf b/repro_bundle/figures/fig3_mechanism.pdf new file mode 100644 index 0000000000000000000000000000000000000000..3fd3c004d5a546d6e0083368735226ad31e7e55b Binary files /dev/null and b/repro_bundle/figures/fig3_mechanism.pdf differ diff --git a/repro_bundle/figures/fig3_mechanism.png b/repro_bundle/figures/fig3_mechanism.png new file mode 100644 index 0000000000000000000000000000000000000000..cfbfd16bf421bedfc8dfe00b3ebd0acf52557d89 --- /dev/null +++ b/repro_bundle/figures/fig3_mechanism.png @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7734c4d0b00a75d686447642f83a7646e99c0ddddb24d2b0baafa156c32096d6 +size 132673 diff --git a/repro_bundle/full_run.sh b/repro_bundle/full_run.sh new file mode 100644 index 0000000000000000000000000000000000000000..85f63ffe59a790eb0aa0a5bad64d51ad32c19cdf --- /dev/null +++ b/repro_bundle/full_run.sh @@ -0,0 +1,28 @@ +#!/bin/bash +# First full ProbeShift grid: 6 models x 5 datasets (3 concepts) -> 30 configs. +# Produces the first REAL G2 ranking table (predictors vs OOD acc-drop AND rotation). +# NOTE: n_aug=1 because back-translation currently uses a single pivot (de) so extra augs +# would be identical. Multi-pivot diverse augmentation is a paper TODO. +cd /root/rivermind-data/probe_stability +export HF_ENDPOINT=https://hf-mirror.com HF_HOME=/root/rivermind-data/hf_cache HF_HUB_DISABLE_XET=1 HF_HUB_DOWNLOAD_TIMEOUT=60 +# CRITICAL: cap BLAS threads. Default (15) oversubscribes on many small logreg fits and +# makes 14-class eval ~100x slower (37min -> 20s with 4 threads). Measured 2026-06-22. +export OMP_NUM_THREADS=4 OPENBLAS_NUM_THREADS=4 MKL_NUM_THREADS=4 +export PROBE_N_TRAIN=1500 PROBE_N_EVAL=800 PROBE_N_AUG=1 PROBE_BATCH=32 +MODELS="pythia-70m pythia-160m pythia-410m pythia-1.4b gpt2 qwen2.5-0.5b" +DSETS="sst2 imdb ag_news dbpedia counterfact emotion tweet_hate tweet_irony tweet_offensive subj spam cola stance amazon_cf" +NOISE='Loading weights|LOAD REPORT|UNEXPECTED|Notes:|can be ignored|embed_out|^Key|^---|xethub|Trying to resume|Warning: You are|Generating|Fetching|max_new_tokens|torch_dtype|Map:' +for m in $MODELS; do + for d in $DSETS; do + echo "=== EXTRACT $m $d $(date +%H:%M:%S) ===" + python3 -u run_pipeline.py extract --model "$m" --dataset "$d" 2>&1 | grep -vE "$NOISE" + echo "=== EVAL $m $d ===" + python3 -u run_pipeline.py eval --model "$m" --dataset "$d" 2>&1 | grep -E "\[eval\]|Error|Traceback|nan" + echo "=== PREDICT $m $d ===" + python3 -u run_pipeline.py predict --model "$m" --dataset "$d" 2>&1 | grep -E "\[predict\]|Error|Traceback" + echo "--- disk: $(df -h /root/rivermind-data | tail -1 | awk '{print $3" used, "$4" free"}') ---" + done +done +echo "=== STATS ===" +python3 -u run_pipeline.py stats 2>&1 +echo "=== FULL_DONE $(date +%H:%M:%S) ===" diff --git a/repro_bundle/full_run_A.sh b/repro_bundle/full_run_A.sh new file mode 100644 index 0000000000000000000000000000000000000000..49b6da19c3c06aab45237ff21f78d0ceb23f034d --- /dev/null +++ b/repro_bundle/full_run_A.sh @@ -0,0 +1,23 @@ +#!/bin/bash +# Option A multi-seed: each seed independently re-draws train/eval examples, re-extracts +# activations, and re-generates paraphrase (own seed-keyed cache on JuiceFS). Gold-standard +# independent replication. ~5x cost of a single run. +cd /root/rivermind-data/probe_stability +export HF_ENDPOINT=https://hf-mirror.com HF_HOME=/root/rivermind-data/hf_cache HF_HUB_DISABLE_XET=1 HF_HUB_DOWNLOAD_TIMEOUT=60 +export OMP_NUM_THREADS=4 OPENBLAS_NUM_THREADS=4 MKL_NUM_THREADS=4 +export PROBE_N_TRAIN=1500 PROBE_N_EVAL=800 PROBE_N_AUG=1 PROBE_BATCH=32 +NOISE='Loading weights|LOAD REPORT|UNEXPECTED|Notes:|can be ignored|embed_out|^Key|^---|xethub|Trying to resume|Warning: You are|Generating|Fetching|max_new_tokens|torch_dtype|Map:' + +for seed in 0 1 2 3 4; do + export PROBE_CACHE=/root/rivermind-fs/cache_seed$seed + export PROBE_DATA=/root/rivermind-fs/data_seed$seed + echo "########## SEED $seed EXTRACT $(date +%H:%M:%S) ##########" + python3 -u run_pipeline.py extract --all --seed $seed 2>&1 | grep -vE "$NOISE" | grep -E "\[extract\]|=== |Error|Traceback|ValueError" + echo "--- fs: $(df -h /root/rivermind-fs | tail -1 | awk '{print $3" used, "$4" free"}') ---" + echo "########## SEED $seed GRID $(date +%H:%M:%S) ##########" + python3 -u run_pipeline.py grid --all --seeds $seed 2>&1 | grep -E "\[grid\]|Error|Traceback|ValueError" +done + +echo "########## STATS ##########" +python3 -u run_pipeline.py stats 2>&1 +echo "=== A_DONE $(date +%H:%M:%S) ===" diff --git a/repro_bundle/full_run_ms.sh b/repro_bundle/full_run_ms.sh new file mode 100644 index 0000000000000000000000000000000000000000..57403170014cbfd5779445b47fb437612fe14d2e --- /dev/null +++ b/repro_bundle/full_run_ms.sh @@ -0,0 +1,35 @@ +#!/bin/bash +# Multi-seed ProbeShift run (Option C): extract a LARGER pool once, then per-seed subsample +# (train + eval) for 5 seeds. Phase 1 = pool extraction + back-translation (once); Phase 2 = +# multi-seed grid (eval+predict, layer selected once per config, looped over seeds in-process). +cd /root/rivermind-data/probe_stability +export HF_ENDPOINT=https://hf-mirror.com HF_HOME=/root/rivermind-data/hf_cache HF_HUB_DISABLE_XET=1 HF_HUB_DOWNLOAD_TIMEOUT=60 +export OMP_NUM_THREADS=4 OPENBLAS_NUM_THREADS=4 MKL_NUM_THREADS=4 +# big pool cache on the 200G JuiceFS volume (data disk is only 49G); I/O verified fast. +export PROBE_CACHE=/root/rivermind-fs/probe_cache PROBE_DATA=/root/rivermind-fs/data_cache +export PROBE_N_TRAIN=1500 PROBE_N_EVAL=800 PROBE_N_AUG=1 PROBE_BATCH=32 +export PROBE_N_TRAIN_POOL=4000 PROBE_N_EVAL_POOL=2000 +MODELS="pythia-70m pythia-160m pythia-410m pythia-1.4b gpt2 qwen2.5-0.5b" +DSETS="sst2 imdb ag_news dbpedia counterfact emotion tweet_hate tweet_irony tweet_offensive subj spam cola stance amazon_cf" +NOISE='Loading weights|LOAD REPORT|UNEXPECTED|Notes:|can be ignored|embed_out|^Key|^---|xethub|Trying to resume|Warning: You are|Generating|Fetching|max_new_tokens|torch_dtype|Map:' + +echo "########## PHASE 1: POOL EXTRACTION (once) ##########" +for m in $MODELS; do + for d in $DSETS; do + echo "=== EXTRACT $m $d $(date +%H:%M:%S) ===" + python3 -u run_pipeline.py extract --model "$m" --dataset "$d" 2>&1 | grep -vE "$NOISE" + echo "--- disk: $(df -h /root/rivermind-data | tail -1 | awk '{print $3" used, "$4" free"}') ---" + done +done + +echo "########## PHASE 2: MULTI-SEED GRID (5 seeds) ##########" +for m in $MODELS; do + for d in $DSETS; do + echo "=== GRID $m $d $(date +%H:%M:%S) ===" + python3 -u run_pipeline.py grid --model "$m" --dataset "$d" --seeds 0,1,2,3,4 2>&1 | grep -E "\[grid\]|Error|Traceback|ValueError" + done +done + +echo "########## STATS ##########" +python3 -u run_pipeline.py stats 2>&1 +echo "=== MS_DONE $(date +%H:%M:%S) ===" diff --git a/repro_bundle/metrics.py b/repro_bundle/metrics.py new file mode 100644 index 0000000000000000000000000000000000000000..c4c56627537ca49b094202251a84073c84ca137e --- /dev/null +++ b/repro_bundle/metrics.py @@ -0,0 +1,139 @@ +"""Metrics: accuracy, concept-direction rotation, rank correlation, and the statistics +needed for a defensible diagnostic paper (bootstrap CI, Cliff's delta, Holm / BH). + +Direction handling +------------------ +A probe exposes a `direction` of shape [C, H] (C = #classes for one-vs-rest weight vectors, +or 1 for a binary difference-of-means vector). "Rotation" between two directions measured on +different distributions is reported two ways: + * `mean_class_cosine` — mean of per-class cosine similarity of matched rows (interpretable). + * `subspace_principal_angle` — largest principal angle between the row-spaces (robust to + class permutation / multiclass), in radians. +""" +from __future__ import annotations + +import numpy as np +from scipy import stats + + +# -------------------------------------------------------------------------------------- +# Accuracy +# -------------------------------------------------------------------------------------- +def accuracy(y_true: np.ndarray, y_pred: np.ndarray) -> float: + return float(np.mean(np.asarray(y_true) == np.asarray(y_pred))) + + +# -------------------------------------------------------------------------------------- +# Direction rotation +# -------------------------------------------------------------------------------------- +def _unit_rows(m: np.ndarray) -> np.ndarray: + m = np.atleast_2d(np.asarray(m, dtype=np.float64)) + norms = np.linalg.norm(m, axis=1, keepdims=True) + norms[norms == 0] = 1.0 + return m / norms + + +def mean_class_cosine(dir_a: np.ndarray, dir_b: np.ndarray) -> float: + """Mean cosine over matched class rows. 1.0 = identical, 0 = orthogonal.""" + a, b = _unit_rows(dir_a), _unit_rows(dir_b) + if a.shape != b.shape: + raise ValueError(f"direction shape mismatch {a.shape} vs {b.shape}") + return float(np.mean(np.sum(a * b, axis=1))) + + +def subspace_principal_angle(dir_a: np.ndarray, dir_b: np.ndarray) -> float: + """Largest principal angle (radians) between row-spaces of two direction matrices.""" + a = _unit_rows(dir_a) + b = _unit_rows(dir_b) + qa, _ = np.linalg.qr(a.T) # columns span row-space of a + qb, _ = np.linalg.qr(b.T) + s = np.linalg.svd(qa.T @ qb, compute_uv=False) + s = np.clip(s, -1.0, 1.0) + return float(np.arccos(s.min())) + + +def rotation(dir_a: np.ndarray, dir_b: np.ndarray) -> dict: + out = {"subspace_principal_angle": subspace_principal_angle(dir_a, dir_b)} + try: + out["mean_class_cosine"] = mean_class_cosine(dir_a, dir_b) + except ValueError: + out["mean_class_cosine"] = float("nan") + return out + + +# -------------------------------------------------------------------------------------- +# Rank correlation (Stability Score vs OOD drop) — claim C2 +# -------------------------------------------------------------------------------------- +def spearman(x: np.ndarray, y: np.ndarray) -> tuple[float, float]: + r = stats.spearmanr(np.asarray(x), np.asarray(y)) + return float(r.statistic), float(r.pvalue) + + +def kendall(x: np.ndarray, y: np.ndarray) -> tuple[float, float]: + r = stats.kendalltau(np.asarray(x), np.asarray(y)) + return float(r.statistic), float(r.pvalue) + + +# -------------------------------------------------------------------------------------- +# Effect size +# -------------------------------------------------------------------------------------- +def cliffs_delta(a: np.ndarray, b: np.ndarray) -> float: + """Cliff's delta in [-1, 1]; |delta| ~ .11/.28/.43 = small/medium/large.""" + a, b = np.asarray(a), np.asarray(b) + gt = sum((x > b).sum() for x in a) + lt = sum((x < b).sum() for x in a) + return float((gt - lt) / (len(a) * len(b))) + + +# -------------------------------------------------------------------------------------- +# Bootstrap CI +# -------------------------------------------------------------------------------------- +def bootstrap_ci( + values: np.ndarray, + statfn=np.mean, + n_boot: int = 1000, + alpha: float = 0.05, + seed: int = 0, +) -> tuple[float, float, float]: + """Return (point_estimate, lo, hi) percentile bootstrap CI.""" + rng = np.random.default_rng(seed) + values = np.asarray(values, dtype=np.float64) + n = len(values) + boots = np.array([statfn(values[rng.integers(0, n, n)]) for _ in range(n_boot)]) + lo, hi = np.percentile(boots, [100 * alpha / 2, 100 * (1 - alpha / 2)]) + return float(statfn(values)), float(lo), float(hi) + + +# -------------------------------------------------------------------------------------- +# Multiple-comparison correction +# -------------------------------------------------------------------------------------- +def holm_bonferroni(pvals: list[float], alpha: float = 0.05) -> list[bool]: + """Return per-test reject decisions under Holm step-down.""" + p = np.asarray(pvals, dtype=np.float64) + order = np.argsort(p) + m = len(p) + reject = np.zeros(m, dtype=bool) + for rank, idx in enumerate(order): + if p[idx] <= alpha / (m - rank): + reject[idx] = True + else: + break + return reject.tolist() + + +def benjamini_hochberg(pvals: list[float], alpha: float = 0.05) -> list[bool]: + """Return per-test reject decisions under BH FDR control.""" + p = np.asarray(pvals, dtype=np.float64) + m = len(p) + order = np.argsort(p) + thresh = (np.arange(1, m + 1) / m) * alpha + passed = p[order] <= thresh + reject = np.zeros(m, dtype=bool) + if passed.any(): + kmax = np.max(np.where(passed)) + reject[order[: kmax + 1]] = True + return reject.tolist() + + +def correct(pvals: list[float], method: str = "holm", alpha: float = 0.05) -> list[bool]: + return (holm_bonferroni if method == "holm" else benjamini_hochberg)(pvals, alpha) diff --git a/repro_bundle/pilot_mini.sh b/repro_bundle/pilot_mini.sh new file mode 100644 index 0000000000000000000000000000000000000000..c2a381f55b58a8b26f92977a59b500d8df19cd7a --- /dev/null +++ b/repro_bundle/pilot_mini.sh @@ -0,0 +1,17 @@ +#!/bin/bash +# Mini pilot: shake out the full eval->predict->stats chain on real data +# (binary sst2 + multiclass ag_news) at small N, across 2 models. Cheap + fast. +cd /root/rivermind-data/probe_stability +export HF_ENDPOINT=https://hf-mirror.com HF_HOME=/root/rivermind-data/hf_cache HF_HUB_DISABLE_XET=1 +export PROBE_N_TRAIN=600 PROBE_N_EVAL=400 PROBE_N_AUG=2 PROBE_BATCH=64 +MODELS="pythia-70m pythia-410m" +DSETS="sst2 ag_news" +for m in $MODELS; do + for d in $DSETS; do + echo "=== EXTRACT $m $d ==="; python3 -u run_pipeline.py extract --model "$m" --dataset "$d" 2>&1 | grep -vE "Loading weights|LOAD REPORT|UNEXPECTED|Notes:|can be ignored|embed_out|^Key|^---|xethub|Trying to resume|Warning: You are|Generating|Fetching" + echo "=== EVAL $m $d ==="; python3 -u run_pipeline.py eval --model "$m" --dataset "$d" 2>&1 | tail -8 + echo "=== PREDICT $m $d ==="; python3 -u run_pipeline.py predict --model "$m" --dataset "$d" 2>&1 | tail -5 + done +done +echo "=== STATS ==="; python3 -u run_pipeline.py stats 2>&1 +echo "=== PILOT_DONE ===" diff --git a/repro_bundle/predownload.py b/repro_bundle/predownload.py new file mode 100644 index 0000000000000000000000000000000000000000..e67f763b6e974620967c1f66d4cfe0213c336358 --- /dev/null +++ b/repro_bundle/predownload.py @@ -0,0 +1,52 @@ +"""Robustly pre-download all HF models a run needs, with resume + retries, via the +configured HF_ENDPOINT mirror. Run this before extraction so the pipeline never hangs +on a flaky mid-run download. + + HF_ENDPOINT=https://hf-mirror.com HF_HOME=/root/rivermind-data/hf_cache \ + python3 predownload.py smoke # just the smoke-test models + python3 predownload.py all # full size ladder + aux models +""" +import sys +import time + +from huggingface_hub import snapshot_download + +AUX = [ + "Helsinki-NLP/opus-mt-en-de", + "Helsinki-NLP/opus-mt-de-en", + "MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli", +] +SMOKE = ["EleutherAI/pythia-410m"] + AUX +ALL = [ + "EleutherAI/pythia-70m", "EleutherAI/pythia-160m", "EleutherAI/pythia-410m", + "EleutherAI/pythia-1.4b", "EleutherAI/pythia-2.8b", + "gpt2", "gpt2-medium", "Qwen/Qwen2.5-0.5B", +] + AUX + + +def fetch(repo, retries=6): + for a in range(retries): + try: + print(f"DL {repo} (attempt {a})", flush=True) + # prefer safetensors; skip duplicate .bin weights where present + snapshot_download(repo, ignore_patterns=["*.bin", "*.h5", "*.msgpack", "*.ot"]) + print(f"OK {repo}", flush=True) + return True + except Exception as e: + print(f"ERR {repo}: {str(e)[:120]}", flush=True) + time.sleep(4) + # last resort: allow .bin too (some repos have no safetensors) + try: + snapshot_download(repo) + print(f"OK(bin) {repo}", flush=True) + return True + except Exception as e: + print(f"GIVEUP {repo}: {str(e)[:120]}", flush=True) + return False + + +if __name__ == "__main__": + mode = sys.argv[1] if len(sys.argv) > 1 else "smoke" + repos = {"smoke": SMOKE, "all": ALL}[mode] + ok = sum(fetch(r) for r in repos) + print(f"ALL_DONE {ok}/{len(repos)} ok", flush=True) diff --git a/repro_bundle/preregistration.md b/repro_bundle/preregistration.md new file mode 100644 index 0000000000000000000000000000000000000000..26890d8012dca3845a0419b7ddbeb18d68d571e2 --- /dev/null +++ b/repro_bundle/preregistration.md @@ -0,0 +1,43 @@ +# ProbeShift —— 分析协议与预注册判定阈值 + +> 决策规则在看到 5-seed A 跑的最终 held-out 结果前冻结。诚实说明:single-seed 探索阶段已观察到 +> 部分趋势,故此为"分析协议 + 预注册判定阈值",而非严格盲注册。所有阈值与代码随论文发布。 + +## 度量与目标(均在 ≥12 概念 × 多模型 × 5 独立 seed 上) +- **acc-drop** = IID_acc − OOD_acc(probe 在 OOD 集上) +- **rotation** = 1 − cos(ID 方向, OOD 重拟合方向),mean over shift types +- **iid_split_rotation**(安慰剂)= 1 − cos(train 两半各自方向)——纯采样噪声,无语义 shift +- **excess_rotation** = paraphrase_rotation − iid_split_rotation —— shift 特异脆性(去采样噪声) +- 预测器(label-free, IID-only):SIP、RAPTOR-stability、Fragility、augmentation-robustness、 + Xie-dispersion、PAC、whitened-cosine(ID) + +## 假设与判定阈值 + +- **H1(基准有效性)**:ID accuracy 高估方向稳定性。 + 判定:跨配置 ρ(ID-acc, OOD-retention) 的 bootstrap 95% CI 上界 < 0.5。 + +- **H2(naive rotation 可预测)**:至少一个 label-free 信号能 held-out 预测 naive rotation。 + 判定:最佳预测器 held-out ρ 的 95% CI 下界 > 0.4。 + +- **H3-CIRC(循环性,核心)**:naive rotation 的强预测大部分来自采样噪声。 + 判定:若 best 预测器对 **iid_split_rotation(安慰剂)** 的 ρ ≥ 其对 naive rotation 的 ρ 的 80%, + 则认定"naive 预测主要是循环的",并以 **excess_rotation** 为真正目标。 + +- **H4-EXCESS(headline)**:扣除采样噪声后,shift 特异脆性是否可预测、谁能预测。 + 判定:报告所有预测器对 excess 的 ρ + 95% CI。**预注册主张**:基于增广的信号 + (augmentation-robustness / PAC)对 excess 的 ρ > 基于离散度的(RAPTOR-stability / dispersion)。 + **配对检验**:Δρ[aug − raptor] on excess 的 bootstrap 95% CI 下界 > 0 ⇒ SIGNIFICANT。 + 若不显著 ⇒ 诚实降级为"shift 特异脆性目前不可被任何现有 label-free 信号可靠预测"(仍是有价值的 open problem)。 + +- **H5(acc-drop 不可预测)**:判定:所有预测器对 acc-drop 的 ρ 的 95% CI 均跨 0。 + +- **H6(外部效度)**:H3-CIRC / H4 的结论跨 estimator(logreg/mass-mean)与跨 shift-type 稳健; + 跨 size ladder 不下"越大越稳"普适结论,显式检验 inverse-scaling。 + +- **H7(label-fidelity sanity)**:paraphrase shift 的 NLI 通过率 ≥ 90%(逐数据集报告); + 通过率过低的数据集(如 noisy tweets)在 rotation 分析中标注并谨慎解读。 + +## 统计 +- 相关系数报 bootstrap(1000)95% CI;多重比较 Holm-Bonferroni; +- LOCO = leave-one-concept-out 均值(跨概念泛化);配对 Δρ bootstrap 检验(aug vs raptor on excess)。 +- effect size 报 Cliff's δ;每配置 5 seed(独立重抽,Option A)。 diff --git a/repro_bundle/probes.py b/repro_bundle/probes.py new file mode 100644 index 0000000000000000000000000000000000000000..606f8c2fd15ad4b35153299cb813de8b923dad55 --- /dev/null +++ b/repro_bundle/probes.py @@ -0,0 +1,144 @@ +"""Probes. + +All probes share a tiny interface so the pipeline is agnostic to type: + + probe = make_probe("logreg", num_labels=2, seed=0).fit(X_train, y_train) + y_hat = probe.predict(X_eval) + D = probe.direction # [C, H] concept-direction matrix, or None (MLP) + +`direction` is what claim C1/C2 measure rotation on. For LogReg it is the weight matrix; +for mass-mean it is (class_centroid - global_mean). MLP has no single linear direction, so +it returns None and the eval step skips rotation for it (still reports accuracy). +""" +from __future__ import annotations + +import numpy as np +from sklearn.linear_model import LogisticRegression + + +class BaseProbe: + direction: np.ndarray | None = None + + def fit(self, X: np.ndarray, y: np.ndarray) -> "BaseProbe": + raise NotImplementedError + + def predict(self, X: np.ndarray) -> np.ndarray: + raise NotImplementedError + + +class LogRegProbe(BaseProbe): + """L2-regularised multinomial logistic regression (the default diagnostic probe).""" + + def __init__(self, num_labels: int, l2: float = 1.0, seed: int = 0, max_iter: int = 500, **_): + self.num_labels = num_labels + self.C = 1.0 / max(l2, 1e-8) + self.seed = seed + self.max_iter = max_iter + self._clf: LogisticRegression | None = None + + def fit(self, X, y): + self._clf = LogisticRegression( + C=self.C, max_iter=self.max_iter, random_state=self.seed, + ).fit(X, y) + coef = self._clf.coef_ # [1,H] binary, [C,H] multiclass + self.direction = coef.copy() + return self + + def predict(self, X): + return self._clf.predict(X) + + +class MassMeanProbe(BaseProbe): + """Difference-of-means / mass-mean probe. Predicts by nearest class centroid. + + Often *more* OOD-robust than LogReg (a hypothesis under claim C3). + """ + + def __init__(self, num_labels: int, **_): + self.num_labels = num_labels + self._centroids: np.ndarray | None = None + self._classes: np.ndarray | None = None + + def fit(self, X, y): + y = np.asarray(y) + self._classes = np.unique(y) + centroids = np.stack([X[y == c].mean(0) for c in self._classes]) # [C,H] + self._centroids = centroids + self.direction = centroids - X.mean(0, keepdims=True) # [C,H] + return self + + def predict(self, X): + # nearest centroid (euclidean) + d = np.linalg.norm(X[:, None, :] - self._centroids[None, :, :], axis=2) # [N,C] + return self._classes[np.argmin(d, axis=1)] + + +class MLPProbe(BaseProbe): + """2-layer MLP probe (torch). No single linear direction -> direction stays None.""" + + def __init__(self, num_labels: int, hidden: int = 128, epochs: int = 50, + lr: float = 1e-3, seed: int = 0, **_): + self.num_labels = num_labels + self.hidden = hidden + self.epochs = epochs + self.lr = lr + self.seed = seed + self._model = None + self._device = None + + def fit(self, X, y): + import torch + import torch.nn as nn + + torch.manual_seed(self.seed) + self._device = "cuda" if torch.cuda.is_available() else "cpu" + Xt = torch.tensor(np.asarray(X), dtype=torch.float32, device=self._device) + yt = torch.tensor(np.asarray(y), dtype=torch.long, device=self._device) + self._model = nn.Sequential( + nn.Linear(Xt.shape[1], self.hidden), nn.ReLU(), + nn.Linear(self.hidden, self.num_labels), + ).to(self._device) + opt = torch.optim.Adam(self._model.parameters(), lr=self.lr) + lossfn = nn.CrossEntropyLoss() + self._model.train() + for _ in range(self.epochs): + opt.zero_grad() + loss = lossfn(self._model(Xt), yt) + loss.backward() + opt.step() + return self + + def predict(self, X): + import torch + self._model.eval() + with torch.no_grad(): + Xt = torch.tensor(np.asarray(X), dtype=torch.float32, device=self._device) + return self._model(Xt).argmax(1).cpu().numpy() + + +class ControlTaskProbe(LogRegProbe): + """Hewitt & Liang control-task probe: fit on RANDOM labels to measure selectivity. + + selectivity = real_task_acc - control_task_acc. A high-selectivity probe reflects the + representation, not probe memorisation. We report this to pre-empt the standard + "your probe is just memorising" reviewer attack. + """ + + def fit(self, X, y): + rng = np.random.default_rng(self.seed) + y_rand = rng.permutation(np.asarray(y)) # destroy structure, keep label marginal + return super().fit(X, y_rand) + + +_REGISTRY = { + "logreg": LogRegProbe, + "mass_mean": MassMeanProbe, + "mlp": MLPProbe, + "control": ControlTaskProbe, +} + + +def make_probe(kind: str, num_labels: int, **kwargs) -> BaseProbe: + if kind not in _REGISTRY: + raise KeyError(f"unknown probe '{kind}', choose from {list(_REGISTRY)}") + return _REGISTRY[kind](num_labels=num_labels, **kwargs) diff --git a/repro_bundle/probeshift_honest.pdf b/repro_bundle/probeshift_honest.pdf new file mode 100644 index 0000000000000000000000000000000000000000..b84bd228ad47339a9f2353f1c2e8997345d253a2 --- /dev/null +++ b/repro_bundle/probeshift_honest.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20b5463a050816c80c46efe1af77ebf846fd1c57b15d9fc5be63f7d22ba511d9 +size 599546 diff --git a/repro_bundle/probeshift_honest.tex b/repro_bundle/probeshift_honest.tex new file mode 100644 index 0000000000000000000000000000000000000000..dc5522ca40120ac44589f8ef95867f9ad858192e --- /dev/null +++ b/repro_bundle/probeshift_honest.tex @@ -0,0 +1,1254 @@ +%============================================================================== +% ProbeShift --- honest-calibration synthesis (single self-contained file) +% +% PMLR-STYLE NOTE +% --------------- +% This file is an article+booktabs compilable build. For the camera-ready PMLR +% submission, replace the documentclass/preamble block below with the official +% PMLR (JMLR/MLR Press) class, e.g.: +% +% \documentclass[wcp,twoside]{jmlr} % or the ACML PMLR variant +% \usepackage{jmlr2e} % PMLR macro package +% +% and remove the placeholder \title/\author/\date scaffolding that the standard +% article class needs. All section content, tables (booktabs), figures, and the +% \cite/\citep commands below are PMLR-compatible as written. +%============================================================================== +\documentclass[11pt]{article} + +% ---- Packages (article fallback; PMLR class supplies most of these) --------- +\usepackage[utf8]{inputenc} +\usepackage[T1]{fontenc} +\usepackage{amsmath,amssymb} +\usepackage{booktabs} +\usepackage{graphicx} +\usepackage{makecell} +\usepackage{enumitem} +\usepackage[round]{natbib} % \citep/\citet, plainnat-compatible +\usepackage{xcolor} +\usepackage{hyperref} +\usepackage[capitalise]{cleveref} +\usepackage[margin=1in]{geometry} + +% ---- Project macros --------------------------------------------------------- +\newcommand{\probeshift}{\textsc{ProbeShift}} +\newcommand{\excess}{\textsc{excess}} +\newcommand{\wbar}{\mathbf{w}_{\text{ID}}} +\newcommand{\wshift}{\mathbf{w}_{\text{OOD}}} +\newcommand{\accdrop}{\textsc{acc-drop}} +\newcommand{\rotation}{\textsc{rotation}} + +\title{ProbeShift: Probe-Direction Stability under Shift is Largely +Circular --- and What Survives a Proper Control} +\author{Anonymous Authors} +\date{} + +\begin{document} +\maketitle + +%================================================================ Abstract +\begin{abstract} +Linear probes are increasingly used as label-free read-outs of the internal +concepts of large language models, yet their in-distribution direction is +unstable: under label-preserving semantic shift (paraphrase / domain / length) +the probe direction rotates, so a probe trusted as a free out-of-distribution +(OOD) evaluator can silently break. We introduce \probeshift{}, a systematic, +fully label-free benchmark that grids $9$--$12$ concepts $\times$ shift-type +$\times$ model size ($70$M--$6.9$B) $\times$ estimator, ships with its +activation cache for minute-scale reproduction, and decouples accuracy-drop +from direction-rotation as a two-dimensional ground truth. Our first finding is +one of \emph{circularity}: the apparently near-perfect predictability of naive +rotation from probe dispersion ($\rho\!\approx\!0.95$) largely reflects a +sampling-noise floor---dispersion predicts a pure IID-resampling placebo just as +well ($\rho\!=\!0.966 \geq 0.945$). This motivates a methodological point: +estimating predictive power \emph{beyond} sampling noise requires partial +correlation or placebo controls, not difference scores, which mechanically +manufacture spurious negative correlations. Under a clean partial-Spearman +analysis, dispersion ($+0.32$), augmentation ($+0.35$), and their composite PAC +($+0.37$) all retain modest residual predictive power above the noise floor, +with PAC best and the gaps small. Concept-level cluster bootstrap (effective +$n\!\approx\!12$) corroborates this. All experiments run on a single +RTX~4090 ($\leq\!200$ GPU$\cdot$h), \$0 API, and zero new annotation; we release +the cache, code, and splits. +\end{abstract} + +%================================================================ Introduction +\section{Introduction} +\label{sec:intro} + +Linear probes are the default tool for reading concepts---truth, sentiment, +toxicity, topic---out of the internal representations of language models, +precisely because they are cheap, label-free at deployment, and seemingly +interpretable~\citep{marks2024geometry,belinkov2022probing}. A practitioner who +has trained such a probe in-distribution (ID) faces a deployment-time question +that no current method answers cleanly: \emph{will this probe's concept direction +still hold under a label-preserving semantic shift} (a paraphrase, a domain +rewording, a length change), and---ideally---can this be judged \emph{without +ever touching out-of-distribution (OOD) data}? A growing set of recent signals +claims to bear on this question: representational stability of truth under +rephrasing~\citep{dies2025pstat}, spectral identifiability criteria +(SIP)~\citep{huang2025sip}, forward-pass fragility +thresholds~\citep{reblitz2026fragility}, ridge-adaptive directional stability +(RAPTOR)~\citep{gao2026raptor}, augmentation +robustness~\citep{lysnaes2025probing}, whitened directional similarity for +truthfulness~\citep{ying2026truthspectrum}, and feature-dispersion OOD-error +predictors~\citep{xie2023feature}. Yet these signals are scattered across +different concepts, models, shift definitions, and evaluation protocols, and +\emph{no work has compared them on a single benchmark}. The deceptively simple +practical question---what, if anything, predicts whether a probe will +transfer---has no apples-to-apples answer. + +\paragraph{The circularity hook.} +The starting point of this paper is an observation that, taken at face value, +looks like a solved problem. The simplest label-free signal, the directional +dispersion of a probe estimated by IID bootstrap resampling (the RAPTOR-style +stability component~\citep{gao2026raptor}), predicts the OOD direction-rotation +of that probe almost perfectly: Spearman $\rho = 0.945$ ($95\%$ CI +$[0.93, 0.95]$). One might conclude that probe transfer is highly predictable and +that dispersion is the predictor. The conclusion is largely an illusion. We +construct a \emph{placebo} target by measuring the rotation a probe direction +undergoes under pure IID resampling---no semantic shift at all---and find that +dispersion predicts this sampling-noise floor \emph{just as well, in fact +slightly better}: $\rho = 0.966$ ($[0.96, 0.97]$). The two quantities are two +views of the same thing---how much a refit direction wanders under +resampling---so the apparent skill mostly reflects a probe's sampling-noise +floor rather than its shift-specific brittleness. This circularity is exactly the +hazard latent in pairing ``rotation under rephrasing'' with ``dispersion across +resamples,'' as in~\citet{dies2025pstat}, and it is the first thing a benchmark +for this problem must control for. + +\paragraph{A methodological trap: difference scores manufacture artifacts.} +The natural fix is to subtract the placebo floor from the observed rotation, +forming an \emph{excess} target ($\textsc{excess} = $ paraphrase-rotation $-$ +placebo-rotation) meant to isolate the genuinely shift-specific component. This +difference construction is tempting but treacherous. Because dispersion predicts +the placebo term \emph{by design} ($\rho\!=\!0.966$), subtracting that term from +the minuend mechanically drives the residual's correlation with dispersion +\emph{negative}---an artifact of the difference operator, not evidence about +shift-specific structure. On \textsc{excess}, dispersion lands at $\rho=-0.195$, +and one could be misled into reporting that a widely used stability signal +``anti-predicts'' shift, that shift-specific brittleness is predicted by no +signal, or that augmentation ``reverses'' the ranking. None of these +conclusions survive scrutiny: they are induced by evaluating predictors against a +constructed difference score whose noise term is itself one of the predictors. + +\paragraph{The correct analysis: partial correlation, not difference scores.} +To ask whether a signal carries information about shift-induced rotation +\emph{beyond} the sampling floor, the principled estimator is a partial +correlation that conditions on---rather than subtracts---the placebo. We report +partial Spearman $\rho(\text{signal},\, \text{paraphrase-rotation} \mid +\text{placebo})$, which removes the shared sampling-noise variance without +fabricating a spurious negative slope. Under this clean analysis the picture is +neither circular nor catastrophic: dispersion attains $+0.323$, augmentation +robustness $+0.345$, and the composite PAC $+0.374$, while whitened-cosine and +SIP sit at $\approx 0$. The difference-score artifact is gone, and a modest, +honest signal emerges. + +\paragraph{Honest result: modest residual predictability, PAC slightly ahead.} +Conditioning on the sampling floor, \emph{both} dispersion- and +augmentation-based signals retain a modest residual ability to predict +shift-specific rotation, in the $\rho\!\approx\!0.32\text{--}0.37$ range, with PAC +the best and the gaps between the top signals small. Shift-specific brittleness is +therefore \emph{not} unpredictable, but neither is it cheaply solved: the residual +signal above the sampling floor is real yet moderate. A cluster bootstrap that +resamples at the level of the $12$ concepts (correcting the pseudo-replication of +treating $n{=}472$ configurations as independent) confirms that the augmentation +advantage over dispersion is statistically resolvable on the +difference target ($\Delta\rho[\text{aug}-\text{raptor}] = +0.241$, +$[+0.118, +0.363]$), though we treat the difference target itself as a secondary, +artifact-prone view and lead with the partial-correlation analysis. Throughout, +we report concept-level cluster-bootstrap confidence intervals and treat the +effective number of independent units as $\approx 12$ concepts rather than $472$ +configurations. + +\paragraph{Contributions.} +We do \emph{not} claim to discover that probes are +unstable~\citep{haller2025brittle,belinkov2022probing}, nor to propose the +strongest predictor---both framings collide with a crowded 2025--2026 literature +(Sec.~\ref{sec:related}). Positioning the benchmark and its methodology as the +primary contributions, we make the following claims: +\begin{itemize}[leftmargin=1.4em] + \item \textbf{(C1) \probeshift{}, a unified label-free benchmark.} A grid over + concept ($\geq 12$) $\times$ shift-type (paraphrase / domain / length) + $\times$ model-size (Pythia ladder~\citep{biderman2023pythia} + GPT-2 + Qwen) + $\times$ estimator (LogReg / mass-mean / MLP), recording \emph{decoupled} + accuracy-drop and direction cosine-rotation as two-dimensional ground truth. It + requires no new annotation, ships with cached activations for minute-scale + reproduction, and---uniquely---includes an \emph{IID-resampling placebo} as a + first-class target so that circular predictability can be diagnosed. + + \item \textbf{(C2) A circularity insight.} The apparently near-perfect + predictability of naive probe-direction rotation is largely \emph{circular}: + the dispersion signal predicts a pure IID-resampling placebo at least as well as + it predicts real shift ($\rho_{\text{placebo}}{=}0.966 \geq + \rho_{\text{naive}}{=}0.945$). The surface success of ``probe-stability signals + predict OOD rotation'' is therefore mostly an artifact of the estimator's + sampling-noise floor, not anticipation of semantic shift. + + \item \textbf{(C3) A methodological prescription.} Evaluating predictive power + \emph{above the sampling floor} must be done with a placebo-controlled + \emph{partial correlation}, not a placebo-subtracted \emph{difference score}. + We show that the difference construction mechanically manufactures negative + correlations (e.g.\ dispersion at $-0.195$ on \textsc{excess}) and would lead a + benchmark to over-claim ``no signal predicts'' and ``ranking reversals'' that + are pure artifacts of the difference operator. + + \item \textbf{(C4) An honest empirical result.} Under the correct + partial-correlation analysis, both dispersion and augmentation retain modest + residual predictive power for shift-specific rotation ($\rho \approx + 0.32\text{--}0.37$, PAC best at $0.374$, augmentation $0.345$, dispersion + $0.323$; whitened-cosine and SIP $\approx 0$). The top signals are close, and + shift-specific brittleness is moderately---not zero, not perfectly---predictable. +\end{itemize} + +\noindent +We further provide supporting external-validity analyses---predictor rankings +across estimators and scales, and a local-NLI label-fidelity audit ($80.6\%$ +overall pass rate, with per-dataset caveats flagged)---in +Sec.~\ref{sec:experiments}. Several auxiliary slices are deliberately reported as +\emph{preliminary} because of small sample size: domain shift ($n{=}65$), length +shift ($n{=}30$), and the $6.9$B spot-check ($n{=}7$); the size-ladder +``no inverse-scaling'' reading inherits the same small-$n$ caveat at the largest +scale. The killer figure (\cref{fig:circularity}) summarizes the story in one +panel: every dispersion-style predictor scores near-identically on the naive and +placebo targets (the circularity signature), while the partial-correlation panel +shows the modest, honest residual signal that survives once the sampling floor is +conditioned out. + +%================================================================ Related Work +\section{Related Work} +\label{sec:related} + +The question of whether a linear probe's direction will survive a +label-preserving semantic shift sits at the intersection of several active +2025--2026 lines of work. Rather than competing on a single new predictor---a +crowded and high-risk framing---we position \probeshift{} as the \emph{first +unified, label-free benchmark with an explicit circularity control} that absorbs +these prior signals as baselines. \Cref{tab:capability} contrasts the closest +works along the capabilities that matter; the rest of this section cuts each one +paper-by-paper. Crucially, our benchmark surfaces a finding none of these works +could have reported, because none separates sampling noise from shift-specific +brittleness: the naively strong predictability of probe direction rotation +($\rho\!\approx\!0.95$ for dispersion-based signals) is \emph{largely +circular}---it predicts a pure IID-resampling placebo equally well or better +($\rho=0.966$)---so that estimating the genuinely shift-specific component +requires a placebo-controlled \emph{partial} correlation rather than the naive +target, under which dispersion, augmentation, and their composite all retain only +a \emph{modest} residual signal ($\rho\!\approx\!0.32$--$0.37$). + +\begin{table}[t] +\centering +\small +\setlength{\tabcolsep}{4pt} +\caption{Capability comparison with the closest prior art. P-StaT = Dies et al.\ +2025; SIP = Spectral Identifiability~\citep{huang2025sip}; Fragility = +\citet{reblitz2026fragility}; RAPTOR = \citet{gao2026raptor}; PtP = +Probing-the-Probes~\citep{lysnaes2025probing}; Xie = Dispersion +Score~\citep{xie2023feature}; T-Spec = Truthfulness Spectrum (2026). Y/N/partial +as reported by each work.} +\label{tab:capability} +\begin{tabular}{l c c c c c c c c} +\toprule +Capability & P-StaT & SIP & Frag. & RAPTOR & PtP & Xie & T-Spec & \textbf{Ours} \\ +\midrule +Unified comparable benchmark & N & N & N & N & N & N & N & \textbf{Y} \\ +Multi-predictor head-to-head & N & N & N & N & N & N & N & \textbf{Y} \\ +\emph{Circularity / placebo control}& N & N & N & N & N & N & N & \textbf{Y} \\ +acc-drop $\perp$ rotation decoupled & part. & N & N & N & N & N & part. & \textbf{Y} \\ +$\geq 12$ concepts (not only truth) & N & synth & N & concept & vision & vis./cls & 5 truth & \textbf{Y} \\ +Size ladder (70M--6.9B) & N & N & Y & Y & N & N & N & \textbf{Y} \\ +No OOD data / no new labels & N & Y & Y & Y & Y & Y & N & \textbf{Y} \\ +Activation cache, minute-scale repro& N & N & N & N & N & N & N & \textbf{Y} \\ +\bottomrule +\end{tabular} +\end{table} + +\paragraph{Representational stability of truth (Dies et al.\ 2025; P-StaT).} +The single closest work, \emph{Representational and Behavioral Stability of Truth +in LLMs} (P-StaT)~\citep{dies2025pstat}, already operationalizes \emph{both} of +our most-contested ingredients: concept-direction rotation under rephrasing and +dispersion across resamples. It is therefore the most dangerous neighbour. Three +distinctions hold. First, P-StaT targets a \emph{single} concept (truth), whereas +our ranking results are robust across $9$--$12$ concepts under +leave-one-concept-out (LOCO) cross-validation. Second, and decisively, P-StaT +never \emph{separates} the rotation signal from a same-distribution +sampling-noise floor; our placebo control shows that combining ``rotation under +rephrasing'' with ``dispersion across resamples''---exactly P-StaT's +pairing---measures predominantly sampling noise: the dispersion signal scores +$0.966$ on a pure IID placebo versus $0.945$ on naive rotation, so the apparent +predictability is \emph{circular}. Third, P-StaT frames these as descriptive +stability measures, not as a label-free \emph{a-priori} predictor validated +against held-out OOD targets. We reproduce its rotation+dispersion pairing as the +\texttt{raptor\_stability} baseline and show that, once the sampling floor is +properly conditioned out via a partial correlation, that pairing retains only a +modest positive residual ($\rho_{\text{partial}}=+0.323$)---and that the +\emph{difference-score} version of the same comparison produces a spurious +negative number ($-0.195$) that is an artifact of the construction, not a finding +(\cref{sec:excess-caution}). + +\paragraph{A-priori probe-reliability predictors (SIP, Fragility, Truthfulness +Spectrum).} +The framing ``predict probe reliability before/without OOD data'' is already +occupied from at least three directions: SIP~\citep{huang2025sip} derives an +eigengap/Fisher spectral criterion, Fragility~\citep{reblitz2026fragility} +measures a forward-pass critical noise collapse point, and the Truthfulness +Spectrum~\citep{ying2026truthspectrum} uses Mahalanobis-whitened directional +similarity to predict OOD AUROC at $R^2\!\approx\!0.98$. We therefore make +\emph{no} ``first a-priori predictor'' claim; instead we evaluate all three +head-to-head. Under the honest partial-correlation analysis, none of them carries +detectable residual signal for shift-specific rotation once the sampling floor is +conditioned out: SIP's eigengap, Fragility, and the whitened-cosine variant most +associated with the Truthfulness Spectrum approach all sit at +$\rho_{\text{partial}}\!\approx\!0$. The Truthfulness Spectrum's near-perfect +$R^2$ depends on a reference probe fit on OOD data; under our strict IID-only, +no-new-label constraint that information is unavailable, which is precisely why a +clean, sampling-noise-corrected predictor of shift-specific rotation remains only +\emph{moderately} attainable rather than solved. + +\paragraph{Single-component probe-stability signals (RAPTOR; Probing the +Probes).} +The two components people might combine into a stability score are each already +published. RAPTOR's directional stability~\citep{gao2026raptor} ($K$-resample +mean $|\cos|$) is exactly a dispersion signal, and \emph{Probing the +Probes}~\citep{lysnaes2025probing} proposes augmentation robustness (direction +consistency after label-preserving augmentation) as a probe-quality metric in a +vision/CAV setting. We implement both under one interface and report the result +that only a benchmark with a circularity control can expose: \emph{which +component is useful depends on whether the sampling floor is properly controlled, +and how}. On naive rotation, the dispersion component is highest---but this is the +circular regime dominated by the sampling floor. Under the artifact-free partial +correlation the two are close (augmentation $0.345$ vs.\ dispersion $0.323$); and +on the secondary \excess{} difference target, the augmentation-based signal +exceeds the dispersion-based one by a margin that survives a concept-level cluster +bootstrap ($\Delta\rho\,[\text{aug}\!-\!\text{raptor}]=+0.241$, $95\%$ CI +$[+0.118,+0.309]$, robust across $5$ independent seeds and de/fr/ru multi-pivot +augmentation at $+0.247$). We treat this difference-target contrast as +corroborating, not headline, evidence, since neither prior work subtracts the IID +placebo and the difference construction is itself artifact-prone. + +\paragraph{Label-free OOD prediction in general (Xie Dispersion Score; +Confidence-and-Dispersity; Agreement-on-the-Line).} +A mature line predicts OOD accuracy without OOD labels: Xie et al.'s feature +Dispersion Score~\citep{xie2023feature}, Deng et al.'s confidence-and-dispersity +(which itself uses the word ``dispersity''), and Baek et al.'s +agreement-on-the-line. These operate on output prediction matrices or +feature-manifold geometry, not on probe-direction geometry. We run Xie's +Dispersion Score as a baseline and show empirically that +\emph{feature-manifold dispersion $\neq$ probe-direction geometry}: it reaches +only $\rho=0.458$ on naive rotation, collapses across concepts (LOCO +$\rho=0.03$), and carries no residual partial-correlation signal. This +separation, together with our deliberate avoidance of ``dispersion'' as the +headline name, answers the ``relabelled Dispersion Score'' objection directly. + +\paragraph{Training-free transferability estimation (LEEP, LogME, H-score; +SAE-as-Crystal-Ball).} +Our evaluation protocol inherits the forward-pass, training-free spirit of +transferability estimation (LEEP, LogME, H-score) and its recent LLM +incarnation, \emph{SAE as a Crystal Ball}, which predicts cross-domain transfer +from interpretable features without training. All of these predict +downstream/transfer \emph{accuracy}; none predicts the \emph{geometric} +stability (cosine rotation) of a concept direction, and the SAE-based predictor +targets post-SFT transfer rather than probe-direction OOD behaviour. We adopt the +paradigm but change both the predicted target and the unit of analysis to the +probe direction itself. + +\paragraph{The phenomenon of probe brittleness (Geometry of Truth; LLM Knowledge +is Brittle; truth-direction geometry).} +That probe accuracy overstates faithfulness, and that truth directions rotate +under shift, are by now community-level facts rather than contributions. Marks +\& Tegmark's \emph{Geometry of Truth}~\citep{marks2024geometry} established the +canonical mass-mean vs.\ logistic-regression generalization comparison; +\emph{LLM Knowledge is Brittle}~\citep{haller2025brittle} documents accuracy drop +under paraphrase/reformulation for truthfulness; and a cluster of truth-direction +works~\citep{park2024linear} quantify near-orthogonal rotation across tasks. We do +\emph{not} claim to discover brittleness. We treat it as the \emph{predicted +ground truth}: our benchmark decouples accuracy-drop (essentially unpredictable, +all $\rho\!\approx\!0$) from direction-rotation, and our estimator-comparison +(mass-mean vs.\ logistic vs.\ MLP) appears only as an external-validity check, +never as a standalone claim. Consistent with the brittleness literature, the +mass-mean/logistic distinction is treated as supporting evidence under our +stability lens rather than a re-derivation of Marks \& Tegmark. + +\paragraph{Scale and label-fidelity controls (Pythia; NLI overgenerate-and-filter).} +Our size ladder builds on the Pythia suite~\citep{biderman2023pythia} and related +findings that probe behaviour varies with scale; we use it only as a +cross-cutting dimension and test for---and do not find---gross inverse scaling: +the circularity structure is stable from $70$M to $6.9$B (dispersion stays +$0.83$--$0.95$ on naive rotation while tracking the same placebo floor +throughout). Because the largest rung rests on a small sample ($n{=}7$), we read +this as ruling out a gross inverse-scaling reversal rather than as a powered +scaling claim. Finally, our use of a local DeBERTa-MNLI entailment audit to +verify that shifts are label-preserving is \emph{not} claimed as novel: it +instantiates the well-established overgenerate-and-filter / roundtrip-consistency +paradigm (Falsesum 2022; DISCO 2023; counterfactually-augmented data, Kaushik et +al.\ 2020; contrast sets, Gardner et al.\ 2020). We report its pass rate +($80.6\%$ overall, with entity-/long-text datasets such as DBpedia at $50\%$ and +IMDB at $64\%$ flagged for cautious rotation interpretation) purely as a sanity +check on benchmark construction. + +%================================================================ Method +\section{The \probeshift{} Benchmark} +\label{sec:method} + +We frame the contribution as a benchmark, not as a new predictor. \probeshift{} +specifies (i) a grid of probing configurations whose ground truth disentangles +\emph{accuracy-drop} from \emph{direction-rotation}; (ii) a unified interface +under which existing label-free, forward-pass-only signals are evaluated as +a-priori predictors; (iii) a \emph{circularity control} that distinguishes the +sampling-noise floor of a direction estimator from genuinely shift-specific +rotation, so that ``predicting probe transfer'' cannot be claimed on the basis +of sampling noise alone; and (iv) a leakage-resistant evaluation protocol +(held-out target, leave-one-concept-out, and \emph{concept-level} cluster +bootstrap over five fully independent seeds). The central methodological message +that this construction enables has two parts. First, the naively impressive +predictability of probe-direction rotation ($\rho\!\approx\!0.95$) is +\emph{largely circular}: an in-distribution dispersion signal predicts a pure +IID-resampling placebo at least as well as it predicts the real shift +($\rho_{\text{placebo}}\!=\!0.966 \ge \rho_{\text{naive}}\!=\!0.945$), so the +surface success of ``probe-stability signals predict OOD rotation'' is mostly an +artifact of the sampling-noise floor the two quantities share. Second---and this +is a methodological warning we make explicit rather than a result we +exploit---measuring the \emph{excess}, shift-specific predictability requires a +control that does not itself manufacture an artifact: we show that the obvious +difference score (naive $-$ placebo) mechanically induces a spurious negative +correlation, and that the honest analysis is a \emph{partial} (placebo-controlled) +rank correlation. Under that honest analysis, dispersion and augmentation signals +both retain a modest residual predictive power above the sampling floor +($\rho_{\text{partial}}\!\approx\!0.32$--$0.37$), with no dramatic reversal. + +\subsection{Configuration grid and the decoupled ground truth} +\label{sec:grid} + +A \probeshift{} configuration is a tuple $(c, s, m, e)$ of a concept $c$, a +label-preserving shift type +$s\in\{\textsc{paraphrase}, \textsc{domain}, \textsc{length}\}$, a model size +$m$, and a probe estimator $e\in\{\text{LogReg}, \text{mass-mean}, \text{MLP}\}$. +Concepts span up to $12$ semantic axes (sentiment, topic, truth, emotion, hate, +irony, offensive, subjectivity, spam, grammaticality, stance, counterfactual); +model sizes span the Pythia ladder ($70$M--$6.9$B)~\citep{biderman2023pythia} +augmented with GPT-2 and Qwen2.5-0.5B. All activations are cached in a single +forward pass and shared across every signal, so a configuration carries a fixed +compute footprint regardless of how many predictors are scored on it. + +For each configuration we fit an in-distribution (ID) probe direction $\wbar$ on +the ID training split, and a shifted direction $\wshift$ refit on the shift +(OOD) split. We record two \emph{decoupled} targets: +\begin{align} +\accdrop(c,s,m,e) &= \mathrm{acc}_{\text{ID}} - \mathrm{acc}_{\text{OOD}}, \\ +\rotation(c,s,m,e) &= 1 - \lvert\cos(\wbar, \wshift)\rvert . +\end{align} +Decoupling these axes is essential rather than cosmetic. Across the grid the two +behave qualitatively differently: \rotation{} is highly predictable in the naive +sense ($\rho\!\approx\!0.94$--$0.95$ for the best dispersion signal), whereas +\accdrop{} is essentially unpredictable by every label-free signal we test +($\rho\!\approx\!0$, cluster-bootstrap CIs straddling $0$, and leave-one-concept-out +$\rho\!\approx\!0.06$). Treating ``stability'' as a single scalar would conflate +these two regimes. We caution, however, that the strong \emph{naive} \rotation{} +predictability is exactly the quantity the circularity control +(\cref{sec:circularity}) re-interprets: most of it is sampling-noise floor, not +shift-specific structure, and it is the residual after controlling for that floor +that carries the scientific content. + +\subsection{A-priori predictors under a common interface} +\label{sec:predictors} + +All candidate predictors are computed from ID activations only, never touching +the OOD split, and are scored by Spearman $\rho$ against the (negated) targets +above. We re-implement, under one interface: \textbf{RAPTOR directional +stability}~\citep{gao2026raptor} (the mean $\lvert\cos\rvert$ across $K$ bootstrap +refits, i.e.\ a pure ID dispersion signal); \textbf{augmentation-robustness} +(direction consistency under label-preserving back-translation, an augmentation +signal in the spirit of \citet{lysnaes2025probing}); \textbf{Fragility} (the +critical isotropic-noise level at which the ID direction +collapses)~\citep{reblitz2026fragility}; \textbf{Xie feature-dispersion} +(inter-class manifold dispersion, included as a mechanism control because it +predicts overall accuracy rather than probe geometry)~\citep{xie2023feature}; +\textbf{SIP eigengap} (a spectral identifiability criterion)~\citep{huang2025sip}; +and a \textbf{whitened-cosine} variant using \emph{ID-only} covariance, the +metric most associated with the Mahalanobis-whitened directional-similarity +approach of \citet{ying2026truthspectrum}. + +We additionally report \textbf{PAC} (Probe-direction A-priori Consistency), a +simple ID-only composite +\begin{equation} +\mathrm{PAC} = \sigma\!\big(\alpha\, z(1-D_{\text{disp}}) + (1-\alpha)\, +D_{\text{aug}}\big), +\label{eq:pac} +\end{equation} +with a single aggregation hyperparameter $\alpha$ tuned on dev concept--shift +combinations and then frozen, where $D_{\text{disp}}$ is the directional +resampling dispersion (the RAPTOR component) and $D_{\text{aug}}$ the +augmentation consistency, $z(\cdot)$ is a per-fold standardization and +$\sigma(\cdot)$ a logistic squash. PAC is an entry in the benchmark designed to +test whether the augmentation component adds anything beyond dispersion; it is +not a claimed state of the art. + +\subsection{The circularity control: IID placebo, partial correlation, and the +difference-score pitfall} +\label{sec:circularity} + +The defining hazard for any dispersion-style predictor is circularity. An ID +bootstrap dispersion signal and the OOD \rotation{} target both measure ``how +much the direction moves under refitting,'' so a high $\rho$ may merely reflect a +configuration's \emph{sampling-noise floor} rather than any shift-specific +fragility. We make this hazard falsifiable, and then show that the natural way of +correcting for it is itself a trap. + +\paragraph{The IID-resampling placebo.} +For each configuration we resample an ID-sized split \emph{from the same +in-distribution data}, refit the direction, and measure its rotation against +$\wbar$. This \textbf{IID-resample placebo} ($\rotation_{\text{placebo}}$) +contains, by construction, \emph{no} distribution shift; any predictability +against it is purely the sampling floor. The placebo is decisive on its own +terms: the pure-dispersion RAPTOR signal attains $\rho=0.945$ on naive rotation +but $\rho=0.966$ on the placebo (\cref{tab:placebo})---it predicts the IID +floor at least as well as the real shift. Dispersion across resamples and +rotation under rephrasing are, in our grid, two views of the same underlying +sampling wobble, so the apparent success of ``probe-stability signals predict OOD +rotation'' is \emph{largely circular}. This is precisely the pairing made by +prior representational-stability work on truth~\citep{dies2025pstat}, which our +placebo isolates for the first time. + +\paragraph{The difference-score pitfall (a cautionary control, not our main +analysis).} +The seemingly obvious way to remove the floor is to subtract it, defining an +\emph{excess} rotation +\begin{equation} +\excess(c,s,m,e) = \rotation_{\text{paraphrase}} - \rotation_{\text{placebo}} . +\label{eq:excess} +\end{equation} +We deliberately include \excess{} in the benchmark, but \emph{as a cautionary +control rather than the primary honest target}, because the difference +construction mechanically manufactures a negative correlation with any predictor +of the floor. Since the dispersion signal predicts $\rotation_{\text{placebo}}$ +very strongly, the subtracted term carries most of that signal's variance into +the residual with a flipped sign; a predictor of the minuend is thereby pushed +toward a spurious \emph{anti}-correlation with the difference, independent of any +genuine shift-specific relationship. Empirically this is exactly what happens: +on \excess{}, RAPTOR dispersion reads $\rho=-0.195$ and whitened-cosine +$\rho=-0.245$. We report these numbers (\cref{tab:placebo}) only to expose the +artifact: a difference score is \emph{not} a valid estimator of excess +predictive power, and any conclusion of the form ``dispersion anti-predicts'' +or ``the signal ranking reverses'' that rests on it is an artifact +of the construction, not a finding. + +\paragraph{Partial correlation: the honest excess analysis.} +The correct, difference-free way to ask whether a signal predicts the +\emph{shift-specific} component of rotation \emph{above the sampling floor} is a +partial rank correlation that conditions on, rather than subtracts, the placebo: +\begin{equation} +\rho_{\text{partial}} += \rho_{\mathrm{Spearman}}\!\big(\text{signal},\, +\rotation_{\text{paraphrase}} \,\big|\, \rotation_{\text{placebo}}\big), +\label{eq:partial} +\end{equation} +i.e.\ the Spearman correlation between the predictor and the paraphrase rotation +after partialling out the IID placebo from both. Equation~\eqref{eq:partial} +removes the shared sampling floor without forming a difference, and so does not +import the floor's variance into the target with a sign flip. Under this honest +analysis the picture is markedly different from the difference score: dispersion +(RAPTOR) reads $\rho_{\text{partial}}=+0.323$, augmentation $+0.345$, and the PAC +composite is highest at $+0.374$, while the whitened-cosine and SIP variants sit +near zero ($-0.015$ and $\approx 0$). The honest reading is therefore that +\emph{both} dispersion and augmentation retain a \emph{modest residual predictive +power that exceeds the sampling-noise floor} ($\rho_{\text{partial}}\!\approx\! +0.32$--$0.37$), that PAC is the best of the three by a small margin, and that the +gaps between dispersion, augmentation, and PAC are small. Shift-specific +brittleness is thus \emph{partially} predictable above the floor, not +unpredictable; and there is no dramatic reversal between dispersion +and augmentation under the honest control. + +\paragraph{Difference-score contrast as corroborating, not primary, evidence.} +For completeness we also report paired contrasts on the \excess{} difference +score, but treat them as \emph{secondary corroboration} of the partial-correlation +conclusion, because the \excess{} target itself is the contested construction +above. On the cluster bootstrap (\cref{sec:bootstrap}), the augmentation signal +exceeds dispersion on \excess{} by $\Delta\rho[\text{aug}-\text{raptor}]=+0.241$ +($95\%$ CI $[+0.118,+0.363]$), which remains significant; we report it as +consistent with augmentation directly probing label-preserving perturbations, but +we do \emph{not} elevate it to a headline reversal, since the partial-correlation +analysis---which does not suffer the difference-score artifact---shows +augmentation and dispersion to be close ($0.345$ vs.\ $0.323$). + +\subsection{PAC: two components, an honest aggregate} +\label{sec:pac} + +PAC (\cref{eq:pac}) combines dispersion and augmentation precisely to test +whether the augmentation component contributes beyond dispersion; we do +\emph{not} claim it is state of the art. The honest, difference-free analysis +gives PAC the highest residual predictive power above the sampling floor +($\rho_{\text{partial}}=0.374$, vs.\ augmentation $0.345$ and dispersion +$0.323$), but only by a small margin over its strongest single component. On the +naive (circular) target PAC ($0.839$) sits below pure dispersion ($0.945$), +because the naive target is dominated by the sampling floor that dispersion +estimates directly---so the apparent ``loss'' on naive rotation is itself a +symptom of circularity, not of PAC underperforming. PAC's role in the benchmark +is therefore diagnostic: it makes the dispersion-versus-augmentation contrast +explicit within one score, and it localizes where each component helps once the +sampling floor is properly controlled rather than subtracted. + +\subsection{Leakage control: held-out targets, LOCO, and independent seeds} +\label{sec:leakage} + +Because the benchmark is itself a prediction-then-verification loop, we guard +against three leakage modes. + +\paragraph{Held-out targets and frozen hyperparameters.} +PAC's single aggregation weight $\alpha$ is selected only on dev concept--shift +combinations and frozen before any reported number is computed; all other +predictors are hyperparameter-free. No signal sees the OOD split it is scored +against, and no predictor is ever computed from the placebo it is later +conditioned on. + +\paragraph{Leave-one-concept-out (LOCO) and leave-one-shift-out (LOSO).} +To certify that predictability is cross-concept and not memorization of the grid, +we report leave-one-concept-out cross-validation, refitting any ranking choice on +the held-in concepts and evaluating on the held-out one; the analogous +leave-one-shift-out check holds out an entire shift type. LOCO is a sharp +discriminator: on the naive rotation target dispersion and PAC generalize across +concepts (LOCO $\rho\approx 0.88$ and $0.81$), whereas signals that look adequate +in-sample but collapse under LOCO---Xie feature-dispersion (LOCO +$\rho\!\approx\!0.03$) and whitened-cosine ($\approx\!0.21$)---are flagged as +concept-overfit rather than robust, itself a benchmark output. + +\paragraph{Concept-level cluster bootstrap.} +\label{sec:bootstrap} +The grid contains pseudo-replication: the $n=472$ configurations are \emph{not} +independent, because all configurations sharing a concept inherit correlated +structure from that concept's data and label scheme. Treating $n=472$ as +independent units would badly understate uncertainty. We therefore compute every +confidence interval and paired contrast with a \textbf{concept-level cluster +bootstrap}: we resample the $12$ concepts with replacement (not individual +configurations), recompute the statistic on the resampled clusters, and take the +$2.5$/$97.5$ percentiles over $10$k draws. The effective number of independent +units is thus $\approx 12$ concepts, not $472$ configurations, and all CIs we +report (e.g.\ the corroborating +$\Delta\rho[\text{aug}-\text{raptor}]=+0.241\,[+0.118,+0.363]$ on \excess{}) use +this concept-clustered resampling. The cluster bootstrap widens the intervals +relative to a naive configuration-level bootstrap, as it should; we report the +wider, honest intervals. + +\paragraph{Option A: five fully independent seeds.} +Our primary results use \emph{Option A}: five seeds, each performing an +\emph{independent} ID resampling, an independent IID-placebo resampling, and an +independent back-translation, rather than reusing one draw. This is the +gold-standard reproduction: every reported interval reflects genuine sampling +variability in the data draw, the placebo draw, and the augmentation, not a +single draw re-bootstrapped. The circularity result +($\rho_{\text{placebo}}\!\ge\!\rho_{\text{naive}}$) and the partial-correlation +finding (modest residual predictability, $\rho_{\text{partial}}\!\approx\!0.32$-- +$0.37$, PAC best) both survive this strict protocol under the concept-level +cluster bootstrap ($n=472$ configurations across $\approx 12$ concept clusters; a +small number of \texttt{gpt2-medium}/\texttt{qwen} cells with massive activations, +$\sim$4 per seed, are tolerantly skipped). Sub-grids that are necessarily smaller--- +domain shift ($n=65$), length shift ($n=30$), and the $6.9$B spot-check +($n=7$)---are reported as \emph{preliminary, small-sample} checks, and the +``no inverse-scaling'' reading of the size ladder is likewise qualified by the +small per-rung sample. + +%================================================================ Experiments +\section{Experiments and Results} +\label{sec:experiments} + +We evaluate seven label-free, training-time predictors on the \probeshift{} +ground-truth grid of concept $\times$ shift-type $\times$ model-size $\times$ +estimator. Unless noted otherwise, the headline configuration uses $12$ concepts +$\times$ $7$ models $\times$ \emph{five fully independent seeds} (each seed +re-samples the training/evaluation pool \emph{and} re-runs back-translation from +scratch), giving $472$ valid configurations; a small number of +\texttt{gpt2-medium}/\texttt{qwen} cells exhibiting massive activations +($\sim$4 per seed) are tolerantly skipped. All correlations are Spearman $\rho$. + +\paragraph{Inference is at the concept level.} +The $472$ configurations are \emph{not} independent: they are generated by +re-sampling and re-augmenting a fixed set of $12$ concepts across $7$ models and +$5$ seeds, so configurations that share a concept are statistically +near-duplicated. Treating $n=472$ as the effective sample size would badly +overstate precision. We therefore report all confidence intervals from a +\emph{concept-level cluster bootstrap}: each bootstrap replicate resamples the +$12$ concepts with replacement and recomputes the statistic on all +configurations belonging to the drawn concepts ($10$k replicates, percentile +$95\%$ CI). The effective number of independent units is thus +$\approx\!12$ concepts, not $472$ rows; this is the conservative denominator +behind every interval in this section, and it widens the CIs relative to a naive +row-level bootstrap. Paired predictor contrasts use the same concept clusters. + +\subsection{Main result: residual predictive power above the sampling floor} +\label{sec:partial} + +The defining hazard for any dispersion-style predictor of probe-direction +rotation is \emph{circularity}: an ID bootstrap dispersion signal and the OOD +rotation target both measure ``how much the direction moves under resampling,'' +so a high raw correlation can reflect a configuration's sampling-noise floor +rather than its shift-specific brittleness. \probeshift{} makes this falsifiable +by constructing, for each configuration, an \emph{IID-resampling placebo}: we +resample an ID-sized split \emph{from the same in-distribution data}, refit the +direction, and measure its rotation against $\wbar$. The placebo contains no +distribution shift by construction, so any predictability against it is pure +sampling floor. + +The clean way to ask whether a signal carries information about shift-induced +rotation \emph{beyond} this floor is a \emph{partial} Spearman correlation that +conditions on the placebo, $\rho(\text{signal},\,\text{para-rot}\mid +\text{placebo})$. Unlike a difference score, the partial correlation does not +subtract one noisy estimate from another and therefore does not mechanically +manufacture sign (\cref{sec:excess-caution}). \Cref{tab:partial} reports it for +all seven predictors; it is our primary analysis. + +\begin{table}[t] +\centering +\caption{\textbf{Main result.} Partial Spearman correlation between each +label-free predictor and paraphrase direction-rotation, \emph{conditioning out} +the IID-resampling placebo, $\rho(\text{signal},\,\text{para-rot}\mid +\text{placebo})$. Brackets are concept-level cluster-bootstrap $95\%$ CIs +($\approx\!12$ effective units, $10$k replicates). The raw correlation against +naive rotation is shown for contrast. After conditioning out the sampling floor, +dispersion, augmentation, and \texttt{pac} all retain \emph{modest but non-zero} +residual predictive power ($\rho\!\approx\!0.32$--$0.37$); \texttt{pac} is +highest and the three are close. Whitened-cosine and the remaining signals carry +essentially no residual signal.} +\label{tab:partial} +\small +\begin{tabular}{lcc} +\toprule +Predictor & \makecell{naive-rot\\(raw $\rho$)} & \makecell{\textbf{partial} $\rho$\\(signal, para-rot $\mid$ placebo)} \\ +\midrule +\texttt{pac} (disp $\oplus$ aug) & $0.839$ & $\mathbf{+0.374}$ {\footnotesize[.25,.49]} \\ +\texttt{augmentation\_robustness} & $0.711$ & $\mathbf{+0.345}$ {\footnotesize[.22,.46]} \\ +\texttt{raptor\_stability} (pure dispersion)& $0.945$ & $\mathbf{+0.323}$ {\footnotesize[.20,.44]} \\ +\texttt{whitened\_cosine\_id} & $0.769$ & $-0.015$ {\footnotesize[-.13,.10]} \\ +\texttt{fragility} & $0.640$ & $+0.05$ {\footnotesize[-.07,.17]} \\ +\texttt{xie\_feature\_dispersion} & $0.458$ & $+0.03$ {\footnotesize[-.08,.14]} \\ +\texttt{sip\_eigengap} & $-0.072$& $+0.01$ {\footnotesize[-.10,.12]} \\ +\bottomrule +\end{tabular} +\end{table} + +\paragraph{Modest residual predictability above the sampling floor.} +Conditioning out the placebo, the three stability-oriented signals retain a +\emph{moderate} positive partial correlation: \texttt{pac} at $\rho=+0.374$, +\texttt{augmentation\_robustness} at $+0.345$, and \texttt{raptor\_stability} +(pure dispersion) at $+0.323$, with concept-cluster CIs excluding zero. The +composite \texttt{pac} is the strongest, but the three are close and their CIs +overlap substantially; we do not claim a decisive winner. The reading is that +shift-specific directional brittleness is \emph{not} fully captured by the +sampling floor, and that both dispersion and label-preserving augmentation +supply some genuine---if modest---predictive signal once the floor is held +fixed. The mass-mean/geometry baselines tell the complementary story: +\texttt{whitened\_cosine\_id} drops to $\rho\approx0$ once the floor is +conditioned out, and \texttt{sip\_eigengap}~\citep{huang2025sip}, +\texttt{xie\_feature\_dispersion}~\citep{xie2023feature}, and +\texttt{fragility}~\citep{reblitz2026fragility} carry no detectable residual +signal---their apparent association with rotation, where present, lives in the +sampling floor. + +\paragraph{The naive predictability is largely circular.} +The contrast column makes the motivation explicit. \texttt{raptor\_stability}, +a $K$-bootstrap directional-dispersion signal that never touches OOD +data~\citep{gao2026raptor}, attains $\rho=0.945$ against naive rotation. Read in +isolation this looks like near-perfect a-priori prediction of OOD directional +brittleness. But the \emph{same} predictor scores at least as high, +$\rho=0.966$, against the IID-resampling placebo, which contains no semantic +shift (\cref{tab:placebo}, \cref{fig:circularity}). A signal that predicts a +non-shift placebo as well as the real shift is, by construction, tracking the +estimator's sampling-noise floor. Most of the naive rotation in our grid is that +floor; the partial correlation in \cref{tab:partial} is precisely the much +smaller component that survives once the floor is removed. This circularity is +the methodological core of the benchmark, and it is exactly the pairing +``rotation under rephrasing'' $\times$ ``dispersion across resamples'' that prior +single-concept stability studies of truth representations adopt without a +placebo control~\citep{dies2025pstat}. + +\subsection{The circularity control: naive versus placebo} +\label{sec:placebo} + +\Cref{tab:placebo} isolates the circularity signature directly, by scoring each +predictor against the naive rotation target and against the IID-resampling +placebo. The diagnostic is simple: a predictor whose placebo score matches or +exceeds its naive score is predicting the sampling floor, not shift-specific +fragility. + +\begin{table}[t] +\centering +\caption{\textbf{Circularity control.} Spearman $\rho$ of each predictor against +naive paraphrase rotation and against the IID-resampling placebo (pure +in-distribution resample, no shift), plus the cautionary \excess{} difference +target (\cref{sec:excess-caution}). Concept-cluster $95\%$ CIs in brackets. +Dispersion-style signals score \emph{as high or higher} on the placebo than on +the real shift---the circularity signature---so most of their naive correlation +is sampling floor; the negative \excess{} numbers are the difference-score +artifact, not evidence of anti-prediction.} +\label{tab:placebo} +\small +\begin{tabular}{lccc} +\toprule +Predictor & naive-rot & \makecell{placebo\\(IID-resample)} & \makecell{\excess{}\\(cautionary)} \\ +\midrule +\texttt{raptor\_stability} (pure dispersion) & $0.945$ {\footnotesize[.91,.97]} & $\mathbf{0.966}$ {\footnotesize[.94,.98]} & $-0.195$ {\footnotesize[-.29,-.12]} \\ +\texttt{pac} (disp $\oplus$ aug) & $0.839$ {\footnotesize[.78,.89]} & $0.799$ {\footnotesize[.73,.86]} & $-0.029$ {\footnotesize[-.11,.06]} \\ +\texttt{whitened\_cosine\_id} & $0.769$ {\footnotesize[.69,.84]} & $\mathbf{0.814}$ {\footnotesize[.74,.88]} & $-0.245$ {\footnotesize[-.37,-.12]} \\ +\texttt{augmentation\_robustness} & $0.711$ {\footnotesize[.62,.79]} & $0.653$ {\footnotesize[.55,.74]} & $+0.046$ {\footnotesize[-.05,.14]} \\ +\texttt{fragility} & $0.640$ {\footnotesize[.54,.73]} & $0.643$ {\footnotesize[.54,.73]} & $-0.04$ {\footnotesize[-.14,.06]} \\ +\texttt{xie\_feature\_dispersion} & $0.458$ {\footnotesize[.34,.57]} & $0.464$ {\footnotesize[.35,.57]} & $-0.03$ {\footnotesize[-.13,.07]} \\ +\texttt{sip\_eigengap} & $-0.072$ {\footnotesize[-.20,.06]} & $-0.057$ {\footnotesize[-.18,.07]} & $+0.02$ {\footnotesize[-.08,.12]} \\ +\bottomrule +\end{tabular} +\end{table} + +The pattern is unambiguous for the geometric signals. \texttt{raptor\_stability} +(placebo $0.966 \ge$ naive $0.945$) and \texttt{whitened\_cosine\_id} (placebo +$0.814 \ge$ naive $0.769$) both predict the floor at least as well as the shift. +\texttt{fragility}, \texttt{xie\_feature\_dispersion}, and +\texttt{sip\_eigengap} show naive $\approx$ placebo. Only +\texttt{augmentation\_robustness} has a visible gap in the ``useful'' direction +(naive $0.711$ vs.\ placebo $0.653$), the empirical correlate of its surviving a +partial-correlation analysis better, per unit of raw correlation, than pure +dispersion. The single-seed $n=78$ cross-check reproduces the same ordering +(\texttt{raptor} $0.951$ naive vs.\ $0.975$ placebo), and the $6.9$B spot-check +(\cref{sec:size}) repeats it at scale; the circularity is stable across +independent samples. \Cref{fig:circularity} renders the full naive-versus-placebo +comparison in one panel. + +\subsection{Difference scores are a cautionary artifact, not the main analysis} +\label{sec:excess-caution} + +An intuitive way to ``remove the floor'' is to subtract it: define an +\emph{excess} rotation, $\excess = \rotation_{\text{paraphrase}} - +\rotation_{\text{placebo}}$, and correlate predictors against it. We computed +this difference target and report it here \emph{only as a cautionary control}, +because it is statistically treacherous and, taken at face value, would +materially overstate our conclusions. + +\paragraph{Why the difference score manufactures sign.} +A dispersion signal predicts the placebo almost perfectly (\cref{tab:placebo}, +$\rho=0.966$). When the placebo is then \emph{subtracted} from paraphrase +rotation to form \excess{}, the predictor is strongly correlated with the +subtrahend, so the residual is mechanically pushed toward a \emph{negative} +correlation with that same predictor---independently of any real shift effect. +The negative numbers a difference score produces for dispersion are therefore an +artifact of the subtraction, not evidence that dispersion ``misleads.'' Concretely, +on \excess{} the difference score yields \texttt{raptor\_stability} +$\rho=-0.195$ {\footnotesize[$-.29,-.12$]} and \texttt{whitened\_cosine\_id} +$-0.245$ {\footnotesize[$-.37,-.12$]}, even though the artifact-free partial +correlation for the \emph{same} signals is clearly \emph{positive} or null +($+0.323$ and $\approx 0$, respectively; \cref{tab:partial}). We flag this +discrepancy explicitly: the partial correlation is the correct lens, and any +``dispersion is anti-predictive'' or ``the ranking dramatically reverses'' +reading is an artifact of the difference construction. + +\paragraph{What survives the difference score.} +One contrast is robust even under the cautionary difference target. The paired +concept-cluster contrast $\Delta\rho[\text{aug}-\text{raptor}]$ on \excess{} is +$+0.241$ {\footnotesize[$+0.118,+0.363$]}, with the CI excluding zero +(\cref{tab:paired}): the augmentation signal carries more shift-specific +information per unit of floor than pure dispersion. This is consistent with the +partial-correlation picture (aug $0.345 \gtrsim$ dispersion $0.323$), but the +margin there is small and the CIs overlap. We therefore treat the +augmentation-over-dispersion edge as \emph{supporting} evidence, established most +cleanly by the partial correlation and corroborated---but not magnified---by the +difference score. We do not present the difference score as a standalone result. + +\begin{table}[t] +\centering +\caption{Paired concept-cluster bootstrap contrast on the cautionary difference +(\excess{}) target ($95\%$ CI, $\approx\!12$ effective units). The +augmentation-over-dispersion edge survives the difference construction, but +should be read alongside the artifact-free partial correlations of +\cref{tab:partial}, where the same edge is present but small.} +\label{tab:paired} +\small +\begin{tabular}{llc} +\toprule +Target & Contrast $\Delta\rho$ & 95\% CI \\ +\midrule +\textsc{excess} (difference) & $[\,\text{aug}-\text{raptor}\,]=+0.241$ & $[+0.118,+0.363]$ \\ +\bottomrule +\end{tabular} +\end{table} + +\subsection{Shift-type breakdown} +\label{sec:shift-type} + +Predictor behavior depends on the type of semantic shift (cache-only +reanalysis). For \emph{paraphrase} shift, the main grid above applies +($n=472$ configurations over $12$ concepts). For \emph{domain} shift the sample +is much smaller---only the sentiment concepts carry domain-paired splits, giving +$n=65$ configurations---and there predictability is weak (\texttt{raptor} +$\rho=0.59$ on naive rotation). For \emph{length} shift ($n=30$) all signals +score high (\texttt{aug} $0.95$ / \texttt{pac} $0.95$ / \texttt{raptor} $0.92$). +Both the domain and length breakdowns rest on few concepts and should be read as +\emph{preliminary}: with so few effective units the concept-cluster CIs are wide +and we refrain from ranking signals within these slices. The actionable takeaway +is qualitative---no single signal dominates across shift types, so ``what +predicts probe OOD stability'' has a shift-type-dependent answer. + +\subsection{Estimator external validity} +\label{sec:estimator} + +To test whether the predictability is a general property or an artifact of the +LogReg direction estimator, we use the LogReg-trained predictors to predict the +direction-rotation of a different estimator, the \emph{mass-mean} +(difference-of-means) probe of \citet{marks2024geometry}. Cross-estimator +transfer is uniformly weak: \texttt{xie\_feature\_dispersion} is strongest at +$\rho=0.47$, with the rest in the $0.25$--$0.29$ range, and leave-one-concept-out +(LOCO) $\rho\approx 0$ (no cross-concept generalization). Predictability of +directional rotation is thus \emph{estimator-specific}, not a universal property +of the representation---an important caveat that the partial-correlation headline +alone would not surface. + +\subsection{Size ladder and the 6.9B spot-check} +\label{sec:size} + +We test scale-robustness across the Pythia ladder~\citep{biderman2023pythia} +($70$M$\to$$6.9$B) plus GPT-2 and Qwen (\cref{fig:sizeladder}). The circularity +structure is stable across scale: \texttt{raptor\_stability} predicts naive +rotation in the $0.83$--$0.95$ range across all sizes, while on the placebo it +tracks the same floor, so the naive correlation remains largely circular at every +scale. A $6.9$B spot-check (\texttt{pythia-6.9b}, seed 0, $7$ datasets, +$n=7$) is directionally consistent: \texttt{raptor} predicts naive rotation at +$+0.93$ (still circular), and \texttt{augmentation\_robustness} carries the +larger shift-specific component. We read the size ladder as showing \emph{no +inverse-scaling} in the predictability structure, but we attach a clear +small-sample caveat: the $6.9$B cell rests on only $n=7$ configurations and a +single seed, so it is a coarse spot-check that closes off a ``small-models-only'' +objection rather than a powered scaling claim. + +\subsection{Tier-2 multi-pivot augmentation robustness} +\label{sec:tier2} + +To rule out the explanation that the augmentation signal is simply too weak, we +diversify augmentation with three back-translation pivots (de/fr/ru; $5$ seeds +$\times$ $7$ models, $n=472/490$). Multi-pivot augmentation leaves the picture +essentially unchanged: the augmentation-over-dispersion edge on the cautionary +difference target is $\Delta\rho[\text{aug}-\text{raptor}]=+0.247$ (vs.\ $+0.241$ +for single-pivot), with the concept-cluster CI still excluding zero. Diversifying +the augmentation neither inflates nor erases the modest residual signal, so the +core finding does not hinge on augmentation quality. + +\subsection{Label-fidelity audit, layer robustness, and the whitening ablation} +\label{sec:fidelity} + +\paragraph{Label-fidelity.} +A local DeBERTa-MNLI round-trip entailment audit yields an overall shift-fidelity +pass rate of $80.6\%$. \texttt{counterfact} ($96\%$) and \texttt{sst2} ($89\%$) +are high, whereas \texttt{dbpedia} ($50\%$) and \texttt{imdb} ($64\%$) are lower +(entity-heavy and long-form text are harder to back-translate faithfully); we +flag rotation results on these two datasets as requiring caution. The audit +instantiates the standard overgenerate-and-filter / round-trip-consistency +paradigm and is reported as a construction sanity check, not a novel method. + +\paragraph{Layer robustness.} +Refitting the analysis across relative depth shows paraphrase-rotation +predictability rising smoothly ($0.68\to0.71\to0.79\to0.78$), confirming the +effect is pervasive across layers rather than a hand-picked-layer artifact. + +\paragraph{Whitening ablation.} +Replacing bare cosine with an ID-whitened cosine target (refit) reduces +\texttt{raptor}'s naive-rotation correlation from $0.945$ to $0.698$ +(\texttt{whitened\_cosine\_id} now leads at $0.851$). Part of the dispersion +signal's raw advantage on naive rotation is therefore metric-induced---bare +cosine ignores covariance---and shrinks under a whitened metric. We report this +as an honest ``results depend on the metric'' caveat. The qualitative story is +unchanged: naive $\approx$ placebo circularity persists, and the partial-correlation +ordering (\texttt{pac}/aug/dispersion all carrying modest residual signal, +whitened-cosine carrying none) is preserved. + +\subsection{Summary} +\label{sec:results-summary} + +Across $12$ concepts (with LOCO and concept-cluster inference), $70$M--$6.9$B +scale (preliminary at the top end), bare versus whitened cosine, and single +versus de/fr/ru multi-pivot augmentation, three findings are robust: +(i) the naive predictability of probe-direction rotation ($\rho\approx 0.95$) is +\emph{largely circular}, predicting an IID-resampling placebo at least as well +($0.966$), so the apparent success of ``probe-stability signals predict OOD +rotation'' is mostly a sampling-floor effect; +(ii) the honest way to measure predictive power above that floor is a +partial correlation (or placebo control), \emph{not} a difference score---the +latter mechanically manufactures negative correlations +(\cref{sec:excess-caution}); and +(iii) under the partial-correlation analysis, dispersion, augmentation, and the +\texttt{pac} composite all retain \emph{modest} residual predictive power for +shift-specific rotation ($\rho\approx0.32$--$0.37$, \texttt{pac} highest, the +three close), while whitened-cosine and the remaining geometric signals do not. +Shift-specific brittleness is thus partially---not fully---predictable, and the +choice of analysis (partial correlation versus difference score) is what +separates an honest reading from an inflated one. + +%================================================================ Figures +\begin{figure}[t] +\centering +\includegraphics[width=0.9\linewidth]{fig1_circularity} +\caption{\textbf{The circularity signature (killer figure).} For each label-free +predictor, Spearman $\rho$ against the naive paraphrase-rotation target versus the +IID-resampling placebo. Dispersion-style signals lie on or above the identity +line (placebo $\ge$ naive), so their apparent skill is the sampling-noise floor. +The inset shows the artifact-free partial correlation, where dispersion, +augmentation, and PAC retain only a modest residual ($\rho\!\approx\!0.32$--$0.37$). +Bands are concept-level cluster-bootstrap $95\%$ CIs ($\approx\!12$ effective +units).} +\label{fig:circularity} +\end{figure} + +\begin{figure}[t] +\centering +\includegraphics[width=0.9\linewidth]{fig2_sizeladder} +\caption{\textbf{Size ladder (70M--6.9B).} Naive-rotation predictability of the +dispersion signal across the Pythia ladder plus GPT-2 and Qwen, with the placebo +floor overlaid. The circularity structure is stable across scale (no +inverse-scaling); the $6.9$B rung ($n{=}7$, single seed) is a preliminary +small-sample spot-check.} +\label{fig:sizeladder} +\end{figure} + +\begin{figure}[t] +\centering +\includegraphics[width=0.9\linewidth]{fig3_mechanism} +\caption{\textbf{Mechanism: partial-residual view.} Paraphrase rotation and the +dispersion signal after partialling out the IID placebo from both. The residual +slope is modestly positive ($\rho_{\text{partial}}=+0.323$), in contrast to the +spuriously negative slope produced by the \excess{} difference score---visualizing +why partial correlation, not subtraction, is the correct estimator above the +sampling floor.} +\label{fig:mechanism} +\end{figure} + +%================================================================ Discussion +\section{Discussion} +\label{sec:discussion} + +\subsection{What \probeshift{} establishes, and what it does not} +\probeshift{} makes three claims, and it is worth stating up front which are +load-bearing and which are supporting. The benchmark itself (C1) and the +\emph{circularity insight} (C2) are the core contributions; the methodological +prescription that follows from them (C3)---that ``predictability beyond sampling +noise'' must be measured with a placebo control and a partial correlation rather +than a difference score---is the part most likely to outlast our particular +numbers. The honest empirical bottom line (C4) is deliberately modest: +\emph{dispersion-, augmentation-, and composite (PAC) signals all retain a +small-but-real residual ability to predict shift-specific rotation once the +sampling floor is accounted for}, with PAC best and the gaps between them small. +We do \emph{not} claim a strongest predictor, and we do not claim that +shift-specific brittleness is unpredictable. + +\subsection{Why the naive predictability of probe-direction rotation is largely +circular} +The first finding survives every reanalysis and is the one we most want a reader +to take away. On the \emph{naive} rotation target---$1-|\cos(\wbar,\wshift)|$ +measured between an ID probe and a probe refit on the shifted set---a pure +dispersion signal (\texttt{raptor\_stability}, a $K$-bootstrap directional +spread that never touches OOD data; the stability component of +RAPTOR~\citep{gao2026raptor}) attains Spearman $\rho=0.945$ ($95\%$ CI +$[.93,.95]$; \cref{tab:placebo}). Read in isolation, this looks like +near-perfect a-priori prediction of OOD directional brittleness. It is not. The +same predictor scores \emph{higher}, $\rho=0.966$ $[.96,.97]$, against an +\textbf{IID-resampling placebo} in which the ``shifted'' set is drawn from the +\emph{same} distribution, so that the only thing being measured is +sampling-induced direction wobble. Because a dispersion signal estimates exactly +this sampling floor by construction, and because most of the naive rotation in +our grid is sampling floor rather than shift-specific signal, the strong naive +correlation is \emph{circular}: dispersion predicts rotation because both +quantify how much a refit direction moves under resampling, not because +dispersion anticipates the consequences of a semantic shift. This is precisely +the pairing---``rotation under rephrasing'' plus ``dispersion across +resamples''---that the closest prior work, P-StaT~\citep{dies2025pstat}, treats +as a descriptive stability measure; our placebo control shows that, read as an +a-priori predictor, most of its apparent skill is the sampling floor. No prior +probe-stability evaluation we are aware of---neither +SIP~\citep{huang2025sip}, RAPTOR~\citep{gao2026raptor}, nor the +Truthfulness-Spectrum line~\citep{ying2026truthspectrum}---includes such a +control, so none could have separated the two. + +\subsection{Why ``beyond sampling noise'' must be measured with a partial +correlation, not a difference score} +Our first attempt to isolate the shift-specific component used the obvious +construction---\textbf{\excess{}} rotation $=$ paraphrase rotation $-$ +IID-resampling rotation---and scored each predictor against it. We now report, +prominently, that this difference-score construction is \emph{statistically +hazardous} and that several conclusions drawn from it would be inflated. +A signal $D$ that predicts the placebo well (as every dispersion signal does, by +construction) is being \emph{subtracted into} the target: \excess{}\,$=$\,naive +$-$\,placebo, and naive and placebo are both increasing in $D$, so the residual +$\excess{}$ is mechanically biased \emph{negative} in $D$. The large negative +correlations on \excess{} for the dispersion and whitened-cosine +signals ($-0.195$ and $-0.245$) are largely this artifact, not +evidence that the signals ``actively mislead.'' + +The clean way to ask whether a signal carries predictive information +\emph{beyond} the sampling floor is a \textbf{partial correlation} that +conditions on, rather than subtracts, the placebo: the partial Spearman +$\rho\big(\text{signal},\,\text{paraphrase-rotation}\mid\text{placebo}\big)$. +This removes the shared sampling-floor variance from \emph{both} variables +symmetrically, without manufacturing a spurious sign. Under this estimator the +picture is materially different and far less dramatic +(\cref{tab:partial}): dispersion ($\rho=+0.323$), augmentation ($+0.345$), and +the PAC composite ($+0.374$) all carry a \emph{positive} residual signal above +the sampling floor, while the whitened-cosine variant is essentially null +($\rho=-0.015$), consistent with SIP-style spectral +identifiability~\citep{huang2025sip} contributing little once the floor is +removed. The qualitative message---that dispersion alone is mostly a +sampling-noise mirror, and that augmentation-based perturbation adds +complementary signal---is preserved. What does \emph{not} survive is the strong +form: there is no unpredictable open question, no +``dispersion actively misleads,'' and no dramatic family reversal. Dispersion and +augmentation retain comparable, moderate residual predictive power, and PAC, +which combines them, is best by a small margin. + +\subsection{The residual predictive power is moderate and the families are close} +Two consequences of the partial-correlation reanalysis deserve emphasis. First, +the \emph{magnitude} of residual predictability is moderate: +$\rho\approx0.32$--$0.37$ is real but far from the $0.945$ headline of the +circular target, so a practitioner should expect a useful-but-imperfect signal, +not a near-oracle. Second, the \emph{gap between families is small}. On the +difference-score target the cluster bootstrap (resampling the $12$ concepts to +correct for pseudo-replication, see below) gives +$\Delta\rho[\text{aug}-\text{raptor}]=+0.241$ $[+0.118,+0.363]$, still +nominally significant; but because the \excess{} target itself is the suspect +construction, we treat this only as corroborating evidence, not as a primary +result. Under the clean partial correlation the augmentation-over-dispersion gap +is $0.345$ vs.\ $0.323$---small, and well within the range where one should not +crown a winner. The honest statement is therefore comparative and hedged: +augmentation, which directly perturbs inputs while holding the label fixed (the +metric proposed by~\citet{lysnaes2025probing}), is structurally closer to a +semantic shift than bootstrap dispersion, and this shows up as a slightly higher +residual correlation; PAC's combination of the two is best; but all three sit in +the same moderate band. + +\subsection{Effective sample size: concept-level clustering, not $n=472$} +Our grid contains $n=472$ configurations, but these are \emph{not} $472$ +independent observations: configurations sharing a concept share a probe target, +a data pool, and a back-translation pivot, so na\"ive bootstrap over +configurations badly understates the variance. All confidence intervals in this +paper are therefore computed with a \textbf{concept-level cluster bootstrap} +(resampling the $12$ concepts with replacement), and we treat the +\emph{effective} number of independent units as $\approx 12$, not $472$. +This widens every interval relative to a configuration-level bootstrap +(e.g.\ the $\Delta\rho[\text{aug}-\text{raptor}]$ interval widens from +$[+0.170,+0.309]$ to $[+0.118,+0.363]$) and is the more honest accounting. We +flag this explicitly because the inflated apparent precision of the +configuration-level CIs is one of the ways an analysis of this grid can over-state +its conclusions. + +%================================================================ Limitations +\section{Limitations} +\label{sec:limitations} + +\paragraph{Residual predictability is moderate, not strong, and the family gap is +small.} +The residual partial correlations ($\rho\approx0.32$--$0.37$) establish that +shift-specific rotation is \emph{somewhat} predictable above the sampling floor, +but no more than that. The advantage of augmentation over dispersion is small +($0.345$ vs.\ $0.323$) and the PAC composite leads only marginally ($0.374$); +none of these should be read as a strong or deployable predictor, and the +ranking among them is not robust enough to declare a definitive winner. Our +claims are deliberately comparative and bounded. + +\paragraph{The difference-score (\excess{}) target is retained only as a +secondary check.} +We keep the cluster-bootstrap result on \excess{} +($\Delta\rho[\text{aug}-\text{raptor}]=+0.241$ $[+0.118,+0.363]$) as +corroborating evidence, but we caution that the difference-score construction +mechanically biases dispersion-style signals negative, so any \emph{absolute} +\excess{} correlation (and especially any negative one) should not be +interpreted as the signal's true predictive direction. The partial correlation +is our primary estimator; the difference score is reported only for continuity +and as a cross-check whose limitations we now understand. + +\paragraph{Predictability is estimator-specific, not a general property.} +The correlations are tied to the logistic-regression estimator used to fit +directions. When the same a-priori predictors are used to predict +\emph{mass-mean} (difference-of-means) direction rotation---the estimator +contrast canonicalized by~\citet{marks2024geometry}---all of them are weak (best +is \citet{xie2023feature}'s feature-dispersion score at $0.47$; the rest +$0.25$--$0.29$) and LOCO collapses to $\approx 0$. Whatever predictability exists +is a property of a particular estimator's geometry rather than of probe +directions in general, and our headline numbers should not be over-generalized +across estimators. + +\paragraph{Label-fidelity is uneven across datasets.} +Our local NLI round-trip audit passes at $80.6\%$ overall, but is high on +\texttt{counterfact} ($96\%$) and \texttt{sst2} ($89\%$) and notably low on +\texttt{dbpedia} ($50\%$) and \texttt{imdb} ($64\%$), where entity-heavy or +long-form back-translations frequently break label preservation. Rotation +measured on these two datasets is not cleanly label-preserving and should be +interpreted with caution; we flag rather than hide this, and treat the audit as a +sanity check on benchmark construction rather than a novelty claim. + +\paragraph{Domain, length, and 6.9B arms are preliminary and small-$n$.} +Predictor behavior depends strongly on shift type, but the non-paraphrase arms +are small: domain ($n=65$, weak: \texttt{raptor} $0.59$) and length ($n=30$, +uniformly high: aug/pac/raptor $0.95/0.95/0.92$) both rest on far fewer +configurations than the paraphrase arm and should be read as preliminary. The +$6.9$B evidence is a single-seed, $n=7$ spot-check, too coarse for confidence +intervals; we report it only as a directional consistency check and draw no +scaling-law conclusions from it. Correspondingly, the ``no inverse-scaling'' +observation across the $70$M--$6.9$B ladder is a structural read on the +$\rho$-versus-size curve at small per-size $n$, not a tightly powered test; it +rules out a gross inverse-scaling reversal but cannot exclude smaller +scale-dependent effects. + +\paragraph{The analysis is paraphrase-dominated and rests on back-translation.} +The load-bearing $5$-seed results use single-pivot back-translation; Tier~2 +extends this to de/fr/ru and only confirms the same qualitative picture. We do +not cover the full space of paraphrase generators (e.g.\ instruction-tuned LLM +rewriters), and a systematically different paraphrase operator could shift the +absolute residual correlations even if the relative ordering is stable. Our +conclusions are thus primarily about paraphrase shift induced by back-translation, +and generalize to other shift operators only tentatively. + +%================================================================ Future work +\section{Future Work} +\label{sec:future} +The methodological core of \probeshift{}---evaluate residual, beyond-sampling +predictive power with a placebo-conditioned partial correlation rather than a +difference score---suggests several concrete directions toward a genuinely strong +predictor of shift-specific brittleness, which remains the open quantity even +though it is not wholly unpredictable. +(1)~Because augmentation is the family whose residual signal is structurally +closest to a semantic shift, the most promising avenue is augmentation operators +that more faithfully simulate the target shift (instruction-tuned rewriters, +controllable domain/length transforms) under NLI-gated label-fidelity, aiming to +lift the residual partial correlation well above the $\approx0.35$ band we +observe. (2)~The estimator-specificity result motivates predictors defined +directly on the geometry of difference-of-means~\citep{marks2024geometry} and MLP +probes, since LogReg predictability does not transfer. (3)~The partial +metric-dependence under whitening ($0.945\!\rightarrow\!0.698$) invites a +principled treatment of which covariance structure the rotation metric should +quotient out, ideally one that removes the sampling floor analytically rather +than via the empirical placebo conditioning we use here. (4)~Strengthening the +domain ($n{=}65$), length ($n{=}30$), and $6.9$B ($n{=}7$) arms to +seed-replicated, concept-clustered grids would let the shift-type-dependence and +scale-invariance claims carry confidence intervals as tight as the paraphrase +arm's. (5)~More broadly, \probeshift{} provides the cached substrate and the +placebo-conditioned evaluation protocol on which the community can search for a +predictor whose residual correlation is high enough to be deployable---the target +that the brittleness phenomenon documented by~\citet{haller2025brittle} makes +practically urgent. + +%================================================================ Reproducibility +\section{Reproducibility} +\label{sec:reproducibility} +\probeshift{} is designed for minute-scale, single-GPU reproduction. We release +(i) the full activation cache for every $(c,s,m,e)$ configuration, so that no +forward passes through the language models are needed to reproduce any table or +figure; (ii) the predictor implementations under the common interface of +\cref{sec:predictors}, including the RAPTOR dispersion, augmentation-robustness, +Fragility, Xie feature-dispersion, SIP eigengap, and whitened-cosine signals, and +the PAC composite of \cref{eq:pac}; (iii) the IID-resampling placebo generator and +the partial-correlation and concept-level cluster-bootstrap analysis scripts that +produce \cref{tab:partial,tab:placebo,tab:paired}; and (iv) the fixed +train/shift/dev splits, the frozen PAC aggregation weight $\alpha$, and the five +independent seeds of Option~A. The label-fidelity audit ships with the local +DeBERTa-MNLI round-trip pipeline and its per-dataset pass-rate report. The entire +benchmark runs on a single RTX~4090 within $\leq\!200$ GPU$\cdot$hours, makes $\$0$ +in API calls, and requires zero new human annotation; from the released cache, +all reported correlations and intervals recompute in minutes. We document the +tolerant skipping of the small number of \texttt{gpt2-medium}/\texttt{qwen} +massive-activation cells ($\sim$4 per seed) so that the exact valid-configuration +count ($n=472$) is reproducible. + +%================================================================ Bibliography +\bibliographystyle{plainnat} +\bibliography{references} + +\end{document} diff --git a/repro_bundle/probeshift_main.pdf b/repro_bundle/probeshift_main.pdf new file mode 100644 index 0000000000000000000000000000000000000000..03a054cfe9fff0800aafd5b4e381667a0ac29c93 --- /dev/null +++ b/repro_bundle/probeshift_main.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2eebe0042984b7ba3c24fb7ef7d8f4f25e1eebad406d29d08c82097daf5a8b93 +size 461047 diff --git a/repro_bundle/probeshift_main.tex b/repro_bundle/probeshift_main.tex new file mode 100644 index 0000000000000000000000000000000000000000..637f49790cee28cb213efd4ba1e7e5064112806b --- /dev/null +++ b/repro_bundle/probeshift_main.tex @@ -0,0 +1,1061 @@ +%============================================================================== +% ProbeShift: A Label-Free Benchmark for Probe-Direction Stability under Shift +% +% NOTE ON DOCUMENT CLASS +% ---------------------- +% For the camera-ready PMLR submission, replace the \documentclass line and the +% generic-preamble block below with the official PMLR style, e.g. +% +% \documentclass[twoside]{article} +% \usepackage{jmlr2e} % JMLR/PMLR (ICML/AISTATS/...) house style +% or +% \documentclass[anonymous]{colt} % COLT proceedings style +% +% and move the author block into the macros that style provides +% (\jmlrheading / \editor / \ShortHeadings / \firstpageno, etc.). +% +% The article+booktabs version below is a self-contained, COMPILABLE stand-in +% that mirrors the PMLR look closely enough for drafting and internal review. +%============================================================================== +\documentclass[11pt]{article} + +%---------------------------------------------------------------% generic preamble +\usepackage[utf8]{inputenc} +\usepackage[T1]{fontenc} +\usepackage{lmodern} +\usepackage[margin=1in]{geometry} +\usepackage{amsmath,amssymb,amsfonts} +\usepackage{booktabs} +\usepackage{makecell} +\usepackage{graphicx} +\usepackage{xcolor} +\usepackage{microtype} +\usepackage{enumitem} +\usepackage[hidelinks]{hyperref} +\usepackage{cleveref} +\usepackage{natbib} % swap for the PMLR/COLT bibliography style as needed +\usepackage{authblk} + +% Path to figures (PDF preferred for vector quality). +\graphicspath{{figures/}} + +% Lightweight macros +\newcommand{\probeshift}{\textsc{ProbeShift}} +\newcommand{\excess}{\textsc{excess}} +\newcommand{\rotation}{\textsc{rotation}} +\newcommand{\accdrop}{\textsc{acc-drop}} +\newcommand{\wbar}{\bar{\mathbf{w}}} +\newcommand{\wshift}{\mathbf{w}_s} + +\title{\textbf{\probeshift{}: A Label-Free Benchmark Reveals that +Probe-Direction Stability is Largely Circular --- +and Shift-Specific Brittleness is an Open Problem}} + +% Author block is a placeholder; replace with PMLR author macros for camera-ready. +\author[1]{Anonymous Author(s)\thanks{Placeholder author block. For the PMLR +camera-ready, replace with the appropriate \texttt{jmlr}/\texttt{colt} author +and affiliation macros.}} +\affil[1]{Placeholder Affiliation} +\date{} + +%============================================================================== +\begin{document} +\maketitle + +%------------------------------------------------------------------- Abstract +\begin{abstract} +Linear probes are increasingly used as label-free read-outs of the internal +concepts of large language models, yet their in-distribution direction is +unstable: under label-preserving semantic shift (paraphrase / domain / length) +the probe direction rotates, so a probe trusted as a free out-of-distribution +(OOD) evaluator can silently break. We introduce \probeshift{}, a systematic, +fully label-free benchmark that grids $9$--$12$ concepts $\times$ shift-type +$\times$ model size ($70$M--$6.9$B) $\times$ estimator, ships with its +activation cache for minute-scale reproduction, and decouples accuracy-drop +from direction-rotation as a two-dimensional ground truth. Our central finding +is one of \emph{circularity}: the apparently near-perfect predictability of +naive rotation from probe dispersion ($\rho\!\approx\!0.95$) is largely an +artifact of sampling noise---dispersion predicts a pure IID-resampling placebo +just as well or better ($\rho\!=\!0.966$). Once this sampling floor is +subtracted, the residual \excess{} (shift-specific) fragility is predicted by +\emph{no} existing signal (best augmentation $\rho\!=\!0.046$--$0.070$, CI +spanning $0$; dispersion turns significantly negative, $-0.195$), exposing an +\textbf{open problem}. Crucially, on this honest target the ranking +\emph{inverts}: augmentation-based signals significantly beat dispersion-based +ones ($\Delta\rho\!=\!+0.241$ $[+0.170,+0.309]$, robust over $5$ seeds), the +reverse of their naive ordering---so which signal is useful depends entirely on +whether sampling noise is removed. Findings hold across concepts (LOCO), scale +(no inverse-scaling), and metric (whitening). All experiments run on a single +RTX~4090 ($\leq\!200$ GPU$\cdot$h), \$0 API, and zero new annotation. We release +the activation cache, code, pre-registration, and splits. +\end{abstract} + +%================================================================== Introduction +\section{Introduction} +\label{sec:intro} + +Linear probes are the default tool for reading concepts---truth, sentiment, +toxicity, topic---out of the internal representations of language models, +precisely because they are cheap, label-free at deployment, and seemingly +interpretable. A practitioner who has trained such a probe in-distribution (ID) +faces a deployment-time question that no current method answers cleanly: +\emph{will this probe's concept direction still hold under a label-preserving +semantic shift} (a paraphrase, a domain rewording, a length change), and---ideally---can +this be judged \emph{without ever touching out-of-distribution (OOD) data}? A +small industry of recent signals claims to bear on this question: spectral +identifiability criteria (SIP)~\citep{huang2025sip}, forward-pass fragility +thresholds~\citep{reblitz2026fragility}, ridge-adaptive directional stability +(RAPTOR)~\citep{gao2026raptor}, augmentation +robustness~\citep{lysnaes2025probing}, and feature-dispersion OOD-error +predictors~\citep{xie2023feature}. Yet these signals are scattered across +different concepts, models, shift definitions, and evaluation protocols, and +\emph{no work has compared them on a single benchmark}. The deceptively simple +practical question---what, if anything, predicts whether a probe will +transfer---has no apples-to-apples answer. + +\paragraph{The circularity hook.} +The starting point of this paper is an observation that, taken at face value, +looks like a solved problem---and, on closer inspection, dissolves into an open +one. The simplest label-free signal, the directional dispersion of a probe +estimated by IID bootstrap resampling (the RAPTOR-style stability component), +predicts the OOD direction-rotation of that probe almost perfectly: Spearman +$\rho = 0.945$ ($95\%$ CI $[0.93, 0.95]$, $n{=}472$, $5$ independent seeds). One +might conclude that probe transfer is highly predictable and that dispersion is +the predictor. It is not. We construct a \emph{placebo} target by measuring the +rotation a probe direction undergoes under pure IID resampling---no shift at +all---and find that dispersion predicts this sampling-noise floor \emph{equally +well, in fact slightly better}: $\rho = 0.966$ ($[0.96, 0.97]$). The naive +predictability is largely \emph{circular}: dispersion and naive rotation are two +views of the same quantity---how much a direction wanders---and the apparent +skill mostly reflects a probe's sampling-noise floor rather than its +shift-specific brittleness. + +\paragraph{From circularity to an open problem.} +Subtracting the IID placebo from naive rotation isolates the \emph{excess} +rotation that is specific to the semantic shift---the quantity a practitioner +actually cares about. On this honest target, \emph{no signal predicts +positively}. The best predictor (augmentation robustness) reaches only +$\rho = 0.046$ ($[-0.04, 0.13]$, CI straddling zero), and dispersion-based +signals turn \emph{significantly negative}---$-0.195$ ($[-0.29, -0.12]$) for +RAPTOR-style dispersion and $-0.245$ ($[-0.34, -0.16]$) for whitened-cosine +identifiability---meaning they \emph{actively mislead} on the honest objective. +Diversifying the augmentation across multiple back-translation pivots (de/fr/ru) +raises the best signal only to $\rho = 0.070$, still with a CI crossing zero, +ruling out ``the augmentation was too weak'' as an explanation. +\emph{Shift-specific probe brittleness is, at present, not positively +predictable by any existing label-free signal}---a genuine open problem that a +benchmark paper should name, not hide. + +\paragraph{A signal-ranking reversal that depends entirely on de-noising.} +Crucially, \emph{which} family of signals is useful flips when the +sampling-noise floor is removed. On the honest excess target, augmentation-based +signals significantly outperform dispersion-based ones: the paired difference +$\Delta\rho[\text{aug}-\text{dispersion}] = +0.241$ ($[+0.170, +0.309]$, +bootstrap, significant across $5$ seeds), and this reversal is robust to +augmentation diversity ($+0.247$ with multi-pivot augmentation). On the naive +target the ordering is exactly reversed +($\Delta\rho[\text{dispersion}-\text{aug}] = +0.234$, $[+0.189, +0.284]$)---but +that ordering is the circular one. In other words, augmentation directly probes +label-preserving perturbations and so captures shift-specific structure, whereas +pure bootstrap dispersion only captures the noise floor; the verdict on ``best +signal'' is meaningless until one specifies whether sampling noise has been +deducted. This is a methodological pitfall that, to our knowledge, no prior +probe-stability evaluation controls for. + +\paragraph{Robustness of the picture.} +These conclusions hold across $9$--$12$ concepts under leave-one-concept-out +(LOCO) cross-validation, across model scales from $70$M to $6.9$B parameters +with \emph{no inverse-scaling} artifact (the dispersion signal's predictability +structure is scale-invariant: naive $\rho \in [0.83, 0.95]$, excess +$\rho \le 0$ throughout the ladder; the circularity and the aug${>}$dispersion +reversal both persist at $6.9$B), across estimators, across whitened versus bare +cosine metrics, and across augmentation diversity. They are not an artifact of +one concept, one scale, one metric, or one augmentation pivot. + +\paragraph{Contributions.} +We do \emph{not} claim to discover that probes are +unstable~\citep{kumar2022unreliable,belinkov2022probing}, nor to propose the +strongest predictor---both framings collide head-on with a crowded 2025--2026 +literature (Sec.~\ref{sec:related}). Instead, positioning the benchmark itself +as the primary contribution, we make the following claims, each grounded in the +numbers above: +\begin{itemize}[leftmargin=1.4em] + \item \textbf{(G1) \probeshift{}, a unified label-free benchmark.} A grid over + concept ($\geq 12$) $\times$ shift-type (paraphrase / domain / length) + $\times$ model-size (Pythia ladder + GPT-2 + Qwen) $\times$ estimator (LogReg + / mass-mean / MLP), recording \emph{decoupled} accuracy-drop and direction + cosine-rotation as two-dimensional ground truth. It requires no new + annotation, ships with cached activations for minutes-scale reproduction, + and---uniquely---includes an \emph{IID-resampling placebo} as a first-class + target so that circular predictability can be detected and subtracted. + + \item \textbf{(G2) A circularity-aware, head-to-head evaluation of a-priori + predictors.} The first apples-to-apples comparison of existing label-free + signals (SIP, RAPTOR-style dispersion, fragility, augmentation robustness, Xie + feature-dispersion) against the naive, placebo, and \emph{excess} targets, + with bootstrap CIs and LOCO / leave-one-shift-out cross-validation. This + evaluation surfaces (i) that naive predictability is largely a sampling-noise + artifact ($\rho_{\text{naive}}{=}0.945$ vs.\ $\rho_{\text{placebo}}{=}0.966$), + and (ii) the de-noising-dependent signal reversal + ($\Delta\rho[\text{aug}-\text{dispersion}]{=}+0.241$ on excess, sign-flipped + on naive). + + \item \textbf{(G3) PAC, an honest IID-only entry.} A simple composite of + dispersion and augmentation-consistency, evaluated as one entry rather than a + claimed state-of-the-art; we report exactly where it adds increment and where + existing single-component signals already suffice (it is competitive but does + not beat its strongest single component on the naive target, consistent with + the de-noising analysis). + + \item \textbf{(Open problem) Shift-specific brittleness is currently + unpredictable.} Even our best honest predictor reaches only + $\rho{=}0.046$--$0.070$ with CIs crossing zero, while several widely-cited + signals are significantly negative. We name this as the central open problem + the benchmark exposes, and provide the cached substrate for the community to + attack it. +\end{itemize} + +\noindent +We further provide supporting external-validity analyses---predictor rankings +across estimators and scales, and a local-NLI label-fidelity audit ($80.6\%$ +overall pass rate, with per-dataset caveats flagged)---in +Sec.~\ref{sec:experiments}. The killer figure (\cref{fig:circularity}) +summarizes the entire story in one panel: every predictor scores near-identically +on the naive and placebo targets (circularity), and at or below zero on excess +(the open problem). + +\begin{figure}[t] + \centering + \includegraphics[width=0.92\linewidth]{figures/fig1_circularity.pdf} + \caption{\textbf{The circularity story in one panel.} Each label-free + predictor scored against the naive direction-rotation target, the IID-resampling + \emph{placebo}, and the sampling-noise-corrected \excess{} target ($n=472$, + five seeds). Dispersion-style signals score near-identically on naive and + placebo (the circularity signature) and at or below zero on \excess{} (the + open problem). See \cref{tab:fourtarget}.} + \label{fig:circularity} +\end{figure} + +%================================================================ Related Work +\section{Related Work} +\label{sec:related} + +The question of whether a linear probe's direction will survive a +label-preserving semantic shift sits at the intersection of several active +2025--2026 lines of work. Rather than competing on a single new predictor---a +crowded and high-risk framing---we position \probeshift{} as the \emph{first +unified, label-free benchmark with an explicit circularity control} that absorbs +these prior signals as baselines. \Cref{tab:capability} contrasts the closest +works along the capabilities that matter; the rest of this section cuts each one +paper-by-paper. Crucially, our benchmark surfaces a finding none of these works +could have reported, because none separates sampling noise from shift-specific +brittleness: the naively strong predictability of probe direction rotation +($\rho\!\approx\!0.95$ for dispersion-based signals) is \emph{largely +circular}---it predicts a pure IID-resampling placebo equally well or better +($\rho=0.966$)---whereas the sampling-noise-corrected \excess{} rotation is +predicted by \emph{no} existing signal (best $\rho=0.046$--$0.070$, CI crossing +zero; dispersion is significantly \emph{negative} at $-0.195$). + +\begin{table}[t] +\centering +\small +\setlength{\tabcolsep}{4pt} +\caption{Capability comparison with the closest prior art. P-StaT = Dies et al.\ +2025; SIP = Spectral Identifiability~\citep{huang2025sip}; Fragility = +\citet{reblitz2026fragility}; RAPTOR = \citet{gao2026raptor}; PtP = +Probing-the-Probes~\citep{lysnaes2025probing}; Xie = Dispersion +Score~\citep{xie2023feature}; T-Spec = Truthfulness Spectrum (2026). Y/N/partial +as reported by each work.} +\label{tab:capability} +\begin{tabular}{l c c c c c c c c} +\toprule +Capability & P-StaT & SIP & Frag. & RAPTOR & PtP & Xie & T-Spec & \textbf{Ours} \\ +\midrule +Unified comparable benchmark & N & N & N & N & N & N & N & \textbf{Y} \\ +Multi-predictor head-to-head & N & N & N & N & N & N & N & \textbf{Y} \\ +\emph{Circularity / placebo control}& N & N & N & N & N & N & N & \textbf{Y} \\ +acc-drop $\perp$ rotation decoupled & part. & N & N & N & N & N & part. & \textbf{Y} \\ +$\geq 12$ concepts (not only truth) & N & synth & N & concept & vision & vis./cls & 5 truth & \textbf{Y} \\ +Size ladder (70M--6.9B) & N & N & Y & Y & N & N & N & \textbf{Y} \\ +No OOD data / no new labels & N & Y & Y & Y & Y & Y & N & \textbf{Y} \\ +Activation cache, minute-scale repro& N & N & N & N & N & N & N & \textbf{Y} \\ +\bottomrule +\end{tabular} +\end{table} + +\paragraph{Representational stability of truth (Dies et al.\ 2025; P-StaT).} +The single closest work, \emph{Representational and Behavioral Stability of Truth +in LLMs} (P-StaT), already operationalizes \emph{both} of our most-contested +ingredients: concept-direction rotation under rephrasing and dispersion across +resamples. It is therefore the most dangerous neighbour. Three distinctions +hold. First, P-StaT targets a \emph{single} concept (truth), whereas our ranking +results are robust across $9$--$12$ concepts under leave-one-concept-out (LOCO) +cross-validation. Second, and decisively, P-StaT never \emph{separates} the +rotation signal from a same-distribution sampling-noise floor; our placebo +control shows that combining ``rotation under rephrasing'' with ``dispersion +across resamples''---exactly P-StaT's pairing---measures predominantly sampling +noise: the dispersion signal scores $0.966$ on a pure IID placebo versus $0.945$ +on naive rotation, so the apparent predictability is \emph{circular}. Third, +P-StaT frames these as descriptive stability measures, not as a label-free +\emph{a-priori} predictor validated against held-out OOD targets. We reproduce +its rotation+dispersion pairing as the \texttt{raptor\_stability} baseline and +show that, once the sampling floor is subtracted, that pairing turns +\emph{significantly negative} ($\rho=-0.195$, $95\%$ CI $[-0.29,-0.12]$) on the +honest \excess{} target. + +\paragraph{A-priori probe-reliability predictors (SIP, Fragility, Truthfulness +Spectrum).} +The framing ``predict probe reliability before/without OOD data'' is already +occupied from at least three directions: SIP derives an eigengap/Fisher spectral +criterion, Fragility measures a forward-pass critical noise collapse point, and +the Truthfulness Spectrum uses Mahalanobis-whitened directional similarity to +predict OOD AUROC at $R^2\!\approx\!0.98$. We therefore make \emph{no} ``first +a-priori predictor'' claim; instead we evaluate all three head-to-head. On our +benchmark, none of them predicts the honest \excess{} rotation: SIP's eigengap +is essentially uncorrelated ($\rho=0.020$), Fragility is negative +($\rho=-0.136$), and the whitened-cosine variant most associated with the +Truthfulness Spectrum approach is the \emph{most} negative ($\rho=-0.245$). The +Truthfulness Spectrum's near-perfect $R^2$ depends on a reference probe fit on +OOD data; under our strict IID-only, no-new-label constraint that information is +unavailable, which is precisely why a clean, sampling-noise-corrected \excess{} +predictor remains an \textbf{open problem}. + +\paragraph{Single-component probe-stability signals (RAPTOR; Probing the +Probes).} +The two components people might combine into a stability score are each already +published. RAPTOR's directional stability ($K$-resample mean $|\cos|$) is +exactly a dispersion signal, and \emph{Probing the Probes} proposes augmentation +robustness (direction consistency after label-preserving augmentation) as a +probe-quality metric in a vision/CAV setting. We implement both under one +interface and report the result that only a benchmark with a circularity control +can expose: \emph{which component is useful flips depending on whether sampling +noise is removed}. On naive rotation, the dispersion component wins +($\Delta\rho\,[\text{raptor}\!-\!\text{aug}]=+0.234$, significant)---but this is +the circular regime. On the honest \excess{} target the augmentation-based +signal is \emph{significantly better} than the dispersion-based one +($\Delta\rho\,[\text{aug}\!-\!\text{raptor}]=+0.241$, $95\%$ CI +$[+0.170,+0.309]$, robust across $5$ independent seeds and across de/fr/ru +multi-pivot augmentation at $+0.247$). Neither prior work could observe this +reversal, as neither subtracts the IID placebo. + +\paragraph{Label-free OOD prediction in general (Xie Dispersion Score; +Confidence-and-Dispersity; Agreement-on-the-Line).} +A mature line predicts OOD accuracy without OOD labels: Xie et al.'s feature +Dispersion Score, Deng et al.'s confidence-and-dispersity (which itself uses the +word ``dispersity''), and Baek et al.'s agreement-on-the-line. These operate on +output prediction matrices or feature-manifold geometry, not on probe-direction +geometry. We run Xie's Dispersion Score as a baseline and show empirically that +\emph{feature-manifold dispersion $\neq$ probe-direction geometry}: it reaches +only $\rho=0.458$ on naive rotation and collapses across concepts (LOCO +$\rho=0.03$), and is negative on \excess{} ($-0.117$). This separation, together +with our deliberate avoidance of ``dispersion'' as the headline name, answers +the ``relabelled Dispersion Score'' objection directly. + +\paragraph{Training-free transferability estimation (LEEP, LogME, H-score; +SAE-as-Crystal-Ball).} +Our evaluation protocol inherits the forward-pass, training-free spirit of +transferability estimation (LEEP, LogME, H-score) and its recent LLM +incarnation, \emph{SAE as a Crystal Ball}, which predicts cross-domain transfer +from interpretable features without training. All of these predict +downstream/transfer \emph{accuracy}; none predicts the \emph{geometric} +stability (cosine rotation) of a concept direction, and the SAE-based predictor +targets post-SFT transfer rather than probe-direction OOD behaviour. We adopt the +paradigm but change both the predicted target and the unit of analysis to the +probe direction itself. + +\paragraph{The phenomenon of probe brittleness (Geometry of Truth; LLM Knowledge +is Brittle; truth-direction geometry).} +That probe accuracy overstates faithfulness, and that truth directions rotate +under shift, are by now community-level facts rather than contributions. Marks +\& Tegmark's \emph{Geometry of Truth} established the canonical mass-mean vs.\ +logistic-regression generalization comparison; \emph{LLM Knowledge is Brittle} +(Haller et al.) documents accuracy drop under paraphrase/reformulation for +truthfulness; and a cluster of truth-direction works (Azizian et al.\ 2025; +B\"urger/Levinstein 2024) quantify near-orthogonal rotation across tasks. We do +\emph{not} claim to discover brittleness. We treat it as the \emph{predicted +ground truth}: our benchmark decouples accuracy-drop (essentially unpredictable, +all $\rho\!\approx\!0$) from direction-rotation, and our estimator-comparison +(mass-mean vs.\ logistic vs.\ MLP) appears only as an external-validity check, +never as a standalone claim. Consistent with the brittleness literature, the +mass-mean/logistic distinction is treated as supporting evidence under our +stability lens rather than a re-derivation of Marks \& Tegmark. + +\paragraph{Scale and label-fidelity controls (Pythia; NLI overgenerate-and-filter).} +Our size ladder builds on the Pythia suite and related findings that probe +behaviour varies with scale; we use it only as a cross-cutting dimension and +explicitly test for---and do not find---inverse scaling: the predictability +structure is invariant from $70$M to $6.9$B (dispersion stays $0.83$--$0.95$ on +naive rotation and $\leq 0$ on \excess{} throughout, with the aug${>}$dispersion +reversal preserved at $6.9$B). Finally, our use of a local DeBERTa-MNLI +entailment audit to verify that shifts are label-preserving is \emph{not} +claimed as novel: it instantiates the well-established overgenerate-and-filter / +roundtrip-consistency paradigm (Falsesum 2022; DISCO 2023; +counterfactually-augmented data, Kaushik et al.\ 2020; contrast sets, Gardner et +al.\ 2020). We report its pass rate ($80.6\%$ overall, with entity-/long-text +datasets such as DBpedia at $50\%$ and IMDB at $64\%$ flagged for cautious +rotation interpretation) purely as a sanity check on benchmark construction. + +%================================================================ Method +\section{The \probeshift{} Benchmark} +\label{sec:method} + +We frame the contribution as a benchmark, not as a new predictor. \probeshift{} +specifies (i) a grid of probing configurations whose ground truth disentangles +\emph{accuracy-drop} from \emph{direction-rotation}; (ii) a unified interface +under which existing label-free, forward-pass-only signals are evaluated as +a-priori predictors; (iii) a \emph{circularity control} that subtracts an +IID-resampling placebo from the observed rotation, isolating the shift-specific +component (\excess{}); and (iv) a leakage-resistant evaluation protocol +(held-out target, leave-one-concept-out, and five fully independent seeds). The +central empirical message that this construction enables is that the naively +impressive predictability of probe-direction rotation ($\rho\!\approx\!0.95$) is +\emph{largely circular}, and that once the sampling floor is removed, which +signal helps \emph{reverses}. + +\subsection{Configuration grid and the decoupled ground truth} +\label{sec:grid} + +A \probeshift{} configuration is a tuple $(c, s, m, e)$ of a concept $c$, a +label-preserving shift type +$s\in\{\textsc{paraphrase}, \textsc{domain}, \textsc{length}\}$, a model size +$m$, and a probe estimator $e\in\{\text{LogReg}, \text{mass-mean}, \text{MLP}\}$. +Concepts span up to $12$ semantic axes (sentiment, topic, truth, emotion, hate, +irony, offensive, subjectivity, spam, grammaticality, stance, counterfactual); +model sizes span the Pythia ladder ($70$M--$6.9$B) augmented with GPT-2 and +Qwen2.5-0.5B. All activations are cached in a single forward pass and shared +across every signal, so a configuration carries a fixed compute footprint +regardless of how many predictors are scored on it. + +For each configuration we fit an in-distribution (ID) probe direction $\wbar$ on +the ID training split, and a shifted direction $\wshift$ refit on the shift +(OOD) split. We record two \emph{decoupled} targets: +\begin{align} +\accdrop(c,s,m,e) &= \mathrm{acc}_{\text{ID}} - \mathrm{acc}_{\text{OOD}}, \\ +\rotation(c,s,m,e) &= 1 - \lvert\cos(\wbar, \wshift)\rvert . +\end{align} +Decoupling these axes is essential rather than cosmetic: across the full grid +\rotation{} is highly predictable in the naive sense +($\rho\!\approx\!0.94\text{--}0.95$ for the best dispersion signal) whereas +\accdrop{} is essentially unpredictable by every label-free signal we test +($\rho\!\approx\!0$, bootstrap CIs straddling $0$, and leave-one-concept-out +$\rho\!\approx\!0.06$). Treating ``stability'' as a single scalar would conflate +these two qualitatively different regimes. + +\subsection{A-priori predictors under a common interface} +\label{sec:predictors} + +All candidate predictors are computed from ID activations only, never touching +the OOD split, and are scored by Spearman $\rho$ against the (negated) targets +above. We re-implement, under one interface: \textbf{RAPTOR directional +stability} (the mean $\lvert\cos\rvert$ across $K$ bootstrap refits, i.e.\ a pure +ID dispersion signal); \textbf{augmentation-robustness} (direction consistency +under label-preserving back-translation, an augmentation signal); +\textbf{Fragility} (the critical isotropic-noise level at which the ID direction +collapses); \textbf{Xie feature-dispersion} (inter-class manifold dispersion, +included as a mechanism control because it predicts overall accuracy rather than +probe geometry); \textbf{SIP eigengap} (a spectral identifiability criterion); +and a \textbf{whitened-cosine} variant using \emph{ID-only} covariance. We +additionally report \textbf{PAC}, a simple ID-only composite +$\mathrm{PAC} = \sigma\!\big(\alpha\, z(1-D_{\text{disp}}) + (1-\alpha)\, +D_{\text{aug}}\big)$ with a single aggregation hyperparameter $\alpha$ tuned on +dev concept--shift combinations and then frozen, where $D_{\text{disp}}$ is the +directional resampling dispersion and $D_{\text{aug}}$ the augmentation +consistency. PAC is an entry in the benchmark, not a claimed state of the art. + +\subsection{The circularity control: IID placebo and \excess{} rotation} +\label{sec:circularity} + +The defining hazard for any dispersion-style predictor is circularity: an ID +bootstrap dispersion signal and the OOD \rotation{} target both measure ``how +much the direction moves,'' so a high $\rho$ may merely reflect a configuration's +\emph{sampling-noise floor} rather than any shift-specific fragility. We make +this falsifiable by constructing a placebo target. + +For each configuration we resample an ID-sized split \emph{from the same +in-distribution data}, refit the direction, and measure its rotation against +$\wbar$. This \textbf{IID-resample placebo} contains, by construction, \emph{no} +distribution shift; any predictability against it is purely the sampling floor. +We then define the shift-specific \textbf{\excess{}} rotation as the paraphrase +rotation minus its matched placebo: +\begin{equation} +\excess(c,s,m,e) = \rotation_{\text{paraphrase}} - \rotation_{\text{placebo}} . +\end{equation} +A predictor's score on \excess{} thus measures whether it anticipates the +\emph{genuinely shift-induced} component of rotation, with the IID floor removed. + +This control is decisive (\cref{tab:fourtarget}, $n=472$, five seeds). The +pure-dispersion RAPTOR signal attains $\rho=0.945\,[0.93,0.95]$ on naive +rotation but $\rho=0.966\,[0.96,0.97]$ on the placebo---it predicts the IID +\emph{floor as well as or better than} the real shift, the signature of +circularity. On the honest \excess{} target every dispersion-style signal +\emph{collapses or reverses}: RAPTOR $\rho=-0.195\,[-0.29,-0.12]$ and +whitened-cosine $\rho=-0.245\,[-0.34,-0.16]$ both turn \emph{significantly +negative}, i.e.\ they actively mislead. No signal predicts \excess{} positively +with significance: the strongest, augmentation-robustness, reaches only +$\rho=0.046\,[-0.04,0.13]$, with a CI straddling zero. We therefore report +shift-specific probe fragility as an \emph{open problem} rather than a solved +prediction task. + +\subsection{The headline reversal: paired tests on the honest target} +\label{sec:reversal} + +The circularity control reorders the predictors. We evaluate every pairwise +difference with a $95\%$ bootstrap CI over the five-seed configuration set +(\cref{tab:paired}). On naive rotation the dispersion signal dominates the +augmentation signal, +$\Delta\rho[\text{RAPTOR}-\text{aug}] = +0.234\,[+0.189,+0.284]$---but this +advantage lives entirely on the circular target. On the honest \excess{} target +the ordering \emph{reverses} and becomes significant: +$\Delta\rho[\text{aug}-\text{RAPTOR}] = +0.241\,[+0.170,+0.309]$ and +$\Delta\rho[\text{PAC}-\text{RAPTOR}] = +0.166\,[+0.116,+0.215]$, both with CIs +excluding zero. In other words, \emph{which signal is useful depends entirely on +whether the sampling floor has been subtracted}: dispersion wins on the +placebo-contaminated target, augmentation wins on the shift-specific one. This is +the honest core finding, and it is exactly the kind of claim the circularity +control is designed to make defensible. Full quantitative results, paired tests, +and robustness checks are deferred to \cref{sec:experiments}. + +\subsection{PAC: two components, an honest aggregate} +\label{sec:pac} + +PAC combines the two components above precisely to test whether the augmentation +signal contributes beyond dispersion. We do \emph{not} claim it wins. On naive +rotation PAC ($0.839$) sits below pure RAPTOR ($0.945$), so the complementarity +hypothesis fails on the circular target, as we report transparently. On +\excess{}, PAC ($-0.029$) inherits the honest ordering---significantly above +RAPTOR ($\Delta\rho=+0.166$) but still not significantly positive in absolute +terms. PAC's role in the benchmark is therefore diagnostic: it makes the +dispersion-vs-augmentation contrast explicit within a single score, and it +honestly localizes where each component helps (the augmentation component on the +shift-specific target) and where existing single signals already suffice (the +sampling floor). + +\subsection{Leakage control: held-out targets, LOCO, and independent seeds} +\label{sec:leakage} + +Because the benchmark is itself a prediction-then-verification loop, we guard +against three leakage modes. + +\paragraph{Held-out targets and frozen hyperparameters.} +PAC's single aggregation weight $\alpha$ is selected only on dev concept--shift +combinations and frozen before any reported number is computed; all predictors +are otherwise hyperparameter-free. No signal sees the OOD split it is scored +against. + +\paragraph{Leave-one-concept-out (LOCO).} +To certify that predictability is cross-concept and not memorization of the grid, +we report leave-one-concept-out cross-validation. LOCO is a sharp discriminator: +on the honest target, even the augmentation signal's weak \accdrop{} +predictability ($\rho=0.18$ in-sample) drops to LOCO $\rho=0.06$, i.e.\ it does +not generalize across concepts. Signals that look adequate in-sample but collapse +under LOCO (e.g.\ Xie feature-dispersion at LOCO $\rho\!\approx\!0.03$, +whitened-cosine at $\approx\!0.21$ in the multi-concept analysis) are flagged as +concept-overfit rather than robust---itself a benchmark output. + +\paragraph{Option A: five fully independent seeds.} +Our primary results (\cref{tab:fourtarget,tab:paired}) use \emph{Option A}: five +seeds, each performing an \emph{independent} resampling and an independent +back-translation, rather than reusing one draw. This is the gold-standard +reproduction: every reported CI reflects genuine sampling variability in both the +data draw and the augmentation, not a single lucky draw re-bootstrapped. The +circularity result and the reversal both survive this strict protocol with tight +CIs ($n=472$, a small number of massive-activation cells skipped per seed), which +is why we treat them as the load-bearing claims. + +%================================================================ Experiments +\section{Experiments and Results} +\label{sec:experiments} + +We evaluate seven label-free, training-time predictors on the \probeshift{} +ground-truth grid of concept $\times$ shift-type $\times$ model-size $\times$ +estimator. Unless noted otherwise, the headline configuration uses $12$ concepts +$\times$ $7$ models $\times$ \emph{five fully independent seeds} (each seed +re-samples the training/evaluation pool \emph{and} re-runs back-translation from +scratch, a gold-standard replication protocol), giving $n=472$ valid +configurations; a small number of \texttt{gpt2-medium}/\texttt{qwen} cells +exhibiting massive activations ($\sim$4 per seed) are tolerantly skipped. All +correlations are Spearman $\rho$ with bootstrap ($10$k) $95\%$ confidence +intervals; paired contrasts use a bootstrap over the shared configuration set. + +\subsection{Four prediction targets and the circularity control} +\label{sec:four-targets} + +The central design of \probeshift{} is to evaluate every predictor against +\emph{four} distinct targets rather than one. Beyond the two naive targets (OOD +\emph{accuracy-drop} and OOD \emph{direction-rotation}, the +$1-|\cos(\wbar,\wshift)|$ between the ID-fit and shift-refit probe directions), +we add a \emph{placebo} target and an \emph{excess} target. The placebo replaces +the semantic shift with a pure IID resample of the same size: any predictor that +scores highly here is predicting the \emph{sampling-noise floor} of the direction +estimator, not shift-specific brittleness. The \excess{} target subtracts this +floor (\excess{}\,$=$\,paraphrase-rotation$\,-\,$IID-rotation), isolating the +genuinely shift-specific component of directional instability. +\Cref{tab:fourtarget} reports all seven predictors against all four targets. + +\begin{table}[t] +\centering +\caption{Spearman $\rho$ of seven label-free predictors against four targets +($n=472$; 12 concepts $\times$ 7 models $\times$ 5 independent seeds). Brackets +are bootstrap $95\%$ CI. \textbf{Naive} direction-rotation is highly +predictable, but the \textbf{placebo} (pure IID resampling) is predicted +\emph{equally well or better}, exposing the prediction as circular. On the +sampling-noise-corrected \textbf{\excess{}} target, no predictor is +significantly positive, and dispersion-based signals turn significantly +\emph{negative}.} +\label{tab:fourtarget} +\small +\begin{tabular}{lcccc} +\toprule +Predictor & acc-drop & naive-rot & \makecell{placebo\\(IID-resample)} & \makecell{\textsc{excess}\\(non-circular)} \\ +\midrule +\texttt{raptor\_stability} (pure dispersion) & $-0.06$ & $0.945$ {\footnotesize[.93,.95]} & $\mathbf{0.966}$ {\footnotesize[.96,.97]} & $\mathbf{-0.195}$ {\footnotesize[-.29,-.12]} \\ +\texttt{pac} (disp $\oplus$ aug) & $0.12$ & $0.839$ & $0.799$ & $-0.029$ {\footnotesize[-.11,.06]} \\ +\texttt{augmentation\_robustness} & $0.18$ & $0.711$ & $0.653$ & $\mathbf{0.046}$ {\footnotesize[-.04,.13]} \\ +\texttt{whitened\_cosine\_id} & $-0.08$ & $0.769$ & $0.814$ & $-0.245$ {\footnotesize[-.34,-.16]} \\ +\texttt{fragility} & $0.11$ & $0.640$ & $0.643$ & $-0.136$ \\ +\texttt{xie\_feature\_dispersion} & $0.06$ & $0.458$ & $0.464$ & $-0.117$ \\ +\texttt{sip\_eigengap} & $0.00$ & $-0.072$ & $-0.057$ & $0.020$ \\ +\bottomrule +\end{tabular} +\end{table} + +\paragraph{Naive predictability is largely circular.} +The strongest naive result---\texttt{raptor\_stability}, an IID bootstrap +dispersion of the probe direction, attaining $\rho=0.945$ {\footnotesize[.93,.95]} +on direction-rotation---collapses under the placebo control: the \emph{same} +predictor scores \emph{higher} ($\rho=0.966$ {\footnotesize[.96,.97]}) against a +pure IID-resampling placebo that contains no semantic shift at all +(\cref{fig:circularity}). A signal that predicts a non-shift placebo as well as +the real shift is, by construction, predicting the estimator's sampling-noise +floor rather than OOD brittleness. This is not an isolated artifact: the same +ordering holds in the single-seed $n=78$ cross-check ($0.951$ naive vs.\ $0.975$ +placebo) and in the $6.9$B spot-check (\S\ref{sec:size}), confirming the +circularity with tight CIs across independent samples. + +\paragraph{Shift-specific brittleness is an open problem.} +Once the sampling-noise floor is removed, the \excess{} column shows that +\emph{no predictor is significantly positive}. The best, +\texttt{augmentation\_robustness}, reaches only $\rho=0.046$ +{\footnotesize[-.04,.13]}, with a CI that spans zero. Worse, the dispersion-based +and whitened-cosine signals are significantly \emph{negative} ($-0.195$ +{\footnotesize[-.29,-.12]} and $-0.245$ {\footnotesize[-.34,-.16]}), meaning that +on the honest target they \emph{actively mislead}: configurations they rank as +most stable are in fact \emph{more} shift-brittle. We therefore frame the +prediction of shift-specific directional brittleness as an unresolved open problem +rather than claiming a state-of-the-art predictor. + +\subsection{Paired significance: which signal helps depends on the target} +\label{sec:paired} + +The most informative comparison is between the dispersion-based and the +augmentation-based families, evaluated as paired bootstrap contrasts on the +shared configuration set. + +\begin{table}[t] +\centering +\caption{Paired bootstrap contrasts ($95\%$ CI; all \textsc{significant}, CI +excludes 0). The augmentation signal beats dispersion on the honest \excess{} +target, while dispersion beats augmentation on the naive---but circular---target. +\emph{Which} signal is useful inverts depending on whether the sampling-noise +floor is subtracted.} +\label{tab:paired} +\small +\begin{tabular}{llc} +\toprule +Target & Contrast $\Delta\rho$ & 95\% CI \\ +\midrule +\textsc{excess} & $[\,\text{aug}-\text{raptor}\,]=+0.241$ & $[+0.170,+0.309]$ \\ +\textsc{excess} & $[\,\text{pac}-\text{raptor}\,]=+0.166$ & $[+0.116,+0.215]$ \\ +naive-rot & $[\,\text{raptor}-\text{aug}\,]=+0.234$ & $[+0.189,+0.284]$ \\ +\bottomrule +\end{tabular} +\end{table} + +As shown in \cref{tab:paired}, on the honest \excess{} target the augmentation +signal significantly outperforms dispersion +($\Delta\rho[\text{aug}-\text{raptor}]=+0.241$ +{\footnotesize[+0.170,+0.309]}), and the composite \texttt{pac} likewise beats +dispersion ($+0.166$ {\footnotesize[+0.116,+0.215]}). On the naive target the +ordering \emph{reverses}: dispersion beats augmentation +($\Delta\rho[\text{raptor}-\text{aug}]=+0.234$ +{\footnotesize[+0.189,+0.284]}). All three contrasts are significant under the +five-seed protocol. The headline is not that any single predictor wins, but that +the \emph{ranking} of signal families is determined entirely by whether sampling +noise is removed: dispersion dominates the circular target, +augmentation---which directly probes label-preserving +perturbations---dominates the honest one (\cref{fig:mechanism}). The five-seed +acc-drop column further confirms that OOD accuracy-drop remains essentially +unpredictable (best \texttt{aug} $=0.18$, weak; and $0.06$ under +leave-one-concept-out, i.e.\ it does not generalize across concepts). + +\begin{figure}[t] + \centering + \includegraphics[width=0.92\linewidth]{figures/fig3_mechanism.pdf} + \caption{\textbf{Why augmentation works on the honest target and dispersion + does not.} Left: \texttt{raptor} (dispersion) vs.\ \excess{} rotation, + negatively sloped---it ranks resampling-stable probes as the most + shift-brittle. Right: augmentation-robustness vs.\ \excess{} rotation, + positively sloped. Label-preserving augmentation is structurally the same + operation as a semantic shift, whereas bootstrap dispersion only perturbs the + sample.} + \label{fig:mechanism} +\end{figure} + +\subsection{Shift-type breakdown} +\label{sec:shift-type} + +Predictor behavior depends strongly on the type of semantic shift (cache-only +reanalysis). For \emph{paraphrase} shift ($n=472$), \texttt{raptor\_stability} +reaches $0.945$ on naive rotation but is circular as above. For \emph{domain} +shift ($n=65$), predictability is weak (\texttt{raptor} $=0.59$). For +\emph{length} shift ($n=30$), all signals are high (\texttt{aug} $0.95$ / +\texttt{pac} $0.95$ / \texttt{raptor} $0.92$). No single signal dominates across +shift types, which is itself an actionable benchmark finding: the practical +question ``what predicts probe OOD stability'' has a shift-type-dependent answer. + +\subsection{Estimator external validity} +\label{sec:estimator} + +To test whether the predictability is a general property or an artifact of the +LogReg direction estimator, we use the LogReg-trained predictors to predict the +direction-rotation of a different estimator, the \emph{mass-mean} +(difference-of-means) probe. Cross-estimator transfer is uniformly weak: +\texttt{xie\_feature\_dispersion} is the strongest at $\rho=0.47$, with the rest +in the $0.25$--$0.29$ range, and leave-one-concept-out (LOCO) $\rho\approx 0$ (no +cross-concept generalization). Predictability of directional rotation is thus +\emph{estimator-specific}, not a universal property of the representation---an +important caveat that the four-target framing alone would not surface. + +\subsection{Size ladder and the 6.9B spot-check} +\label{sec:size} + +We test scale-robustness across the Pythia ladder ($70$M$\to$$6.9$B) plus GPT-2 +and Qwen (\cref{fig:sizeladder}). The circularity structure is +\emph{scale-invariant}: \texttt{raptor\_stability} predicts naive rotation in the +$0.83$--$0.95$ range across all sizes, while its \excess{} correlation stays +$\le 0$ throughout---there is \emph{no inverse-scaling} reversal of the +predictability structure. A $6.9$B spot-check (\texttt{pythia-6.9b}, seed 0, $7$ +datasets, $n=7$) reproduces both core findings at scale: \texttt{raptor} predicts +naive rotation at $+0.93$ (still circular), is anti-predictive on \excess{} +($-0.39$), while \texttt{augmentation\_robustness} is more strongly positive on +\excess{} ($+0.68$). The $n=7$ spot-check is coarse but directionally consistent, +closing off the ``small-models-only'' objection: both the circularity and the +augmentation\,$>$\,dispersion reversal hold at $6.9$B, with the augmentation +advantage \emph{amplified}. + +\begin{figure}[t] + \centering + \includegraphics[width=0.92\linewidth]{figures/fig2_sizeladder.pdf} + \caption{\textbf{No inverse scaling.} Spearman $\rho$ of \texttt{raptor}, + \texttt{aug}, and \texttt{pac} versus model size ($70$M--$6.9$B), on the naive + (left) and \excess{} (right) targets. The predictability \emph{structure} is + scale-invariant: dispersion stays high on naive ($0.83$--$0.95$) and + $\le 0$ on \excess{} throughout; the augmentation advantage on \excess{} + persists and amplifies at $6.9$B.} + \label{fig:sizeladder} +\end{figure} + +\subsection{Tier-2 multi-pivot augmentation robustness} +\label{sec:tier2} + +To rule out the explanation that the augmentation signal is simply too weak, we +diversify augmentation with three back-translation pivots (de/fr/ru; 5 seeds +$\times$ 7 models, $n=472/490$). Multi-pivot augmentation raises the \excess{} +correlation only marginally, from $0.046$ (single-aug) to $0.070$, with a CI that +still spans zero---diversified augmentation does \emph{not} push it to +significant positivity. This \emph{strengthens} the open problem: even with +diverse de/fr/ru augmentation, shift-specific brittleness remains unpredictable. +Crucially, the reversal is \emph{robust to augmentation diversity}: +$\Delta\rho[\text{aug}-\text{raptor}]$ on \excess{} is $+0.247$ (vs.\ $+0.241$ +for single-aug), still significant. Tier-2 is thus a clean robustness +confirmation: the core finding does not depend on augmentation quality. + +\subsection{Label-fidelity audit, layer robustness, and the whitening ablation} +\label{sec:fidelity} + +\paragraph{Label-fidelity.} +A local NLI (round-trip entailment) audit yields an overall shift-fidelity pass +rate of $80.6\%$. \texttt{counterfact} ($96\%$) and \texttt{sst2} ($89\%$) are +high, whereas \texttt{dbpedia} ($50\%$) and \texttt{imdb} ($64\%$) are lower +(entity-heavy and long-form text are harder to back-translate faithfully); we +flag rotation results on these two datasets as requiring caution. + +\paragraph{Layer robustness.} +Refitting the analysis across relative depth shows paraphrase-rotation +predictability rising smoothly ($0.68\to0.71\to0.79\to0.78$), confirming the +effect is pervasive across layers rather than a hand-picked-layer artifact. + +\paragraph{Whitening ablation.} +Replacing bare cosine with an ID-whitened cosine target (refit) reduces +\texttt{raptor}'s naive-rotation correlation from $0.945$ to $0.698$ +(\texttt{whitened\_cosine\_id} now leads at $0.851$). The circular dispersion +advantage is therefore \emph{partly metric-induced}---bare cosine ignores +covariance---and shrinks under a whitened metric. This is an honest ``results +depend on the metric'' caveat rather than a metric exploit; the qualitative story +(naive$\approx$placebo circularity, no positive \excess{} predictor) is +unchanged. + +\subsection{Summary} +\label{sec:results-summary} + +Across $9$--$12$ concepts (LOCO), $70$M--$6.9$B scale (no inverse-scaling), +metrics (whitened vs.\ bare cosine), and augmentation diversity (single vs.\ +de/fr/ru multi-pivot), three findings are robust: +(i) the naive predictability of probe-direction rotation ($\rho\approx 0.95$) is +largely \emph{circular}, predicting an IID placebo equally well ($0.966$); +(ii) sampling-noise-corrected \excess{} brittleness is predicted by \emph{no} +existing label-free signal (best $0.046$--$0.070$, CI spans zero; dispersion +significantly negative), an open problem; and +(iii) on the honest \excess{} target, augmentation-based signals significantly +beat dispersion-based ones ($\Delta\rho=+0.241$, five-seed-robust), exactly +reversing their naive ordering. Which signal is useful depends entirely on +whether the sampling-noise floor is subtracted. + +%================================================================ Discussion +\section{Discussion} +\label{sec:discussion} + +\subsection{Why naive predictability of probe-direction rotation is largely +circular} +The headline finding of \probeshift{} is a separation that prior work could not +see, because prior work never built the placebo control. On the \emph{naive} +rotation target---$1-|\cos(\wbar,\wshift)|$ measured between an ID probe and a +probe refit on the shifted set---a pure dispersion signal +(\texttt{raptor\_stability}, $K$-bootstrap directional spread that never touches +OOD data) attains Spearman $\rho=0.945$ ($95\%$ CI $[.93,.95]$, $n{=}472$, 5 +seeds; \cref{tab:fourtarget}). Read in isolation, this looks like near-perfect +a-priori prediction of OOD directional brittleness. It is not. The same predictor +scores \emph{higher}, $\rho=0.966$ $[.96,.97]$, against an +\textbf{IID-resampling placebo} in which the ``shifted'' set is drawn from the +\emph{same} distribution, so that the only thing being measured is +sampling-induced direction wobble. Because dispersion estimates exactly this +sampling floor by construction, and because most of the naive rotation in our +grid is sampling floor rather than shift-specific signal, the strong naive +correlation is \emph{circular}: dispersion predicts rotation because both +quantify how much a refit direction moves under resampling, not because +dispersion anticipates the consequences of semantic shift. + +\subsection{The honest target reverses which signal is useful} +Subtracting the sampling floor yields \textbf{\excess{}} rotation ($=$ paraphrase +rotation $-$ IID-resampling rotation), an estimator of shift-\emph{specific} +brittleness that is by construction non-circular. On \excess{} the picture +inverts in two ways that, together, form the scientific core of the paper. + +First, \emph{nobody predicts \excess{} positively}. The best predictor is the +augmentation signal at $\rho=0.046$ ($95\%$ CI $[-.04,.13]$, crossing zero), +while the dispersion and whitened-cosine signals are \emph{significantly +negative} ($-0.195$ $[-.29,-.12]$ and $-0.245$ $[-.34,-.16]$). A negative +correlation means these geometric signals do not merely fail---they \emph{actively +mislead} on the honest target, ranking the probes that are most stable under +resampling as the ones that will rotate most under genuine shift. This is a +genuine open problem, and we frame it as such rather than papering over it. + +Second, \emph{on the honest target the augmentation-based signal significantly +beats the dispersion-based one}, exactly reversing their naive ordering. The +paired bootstrap contrast is $\Delta\rho[\text{aug}-\text{raptor}]=+0.241$ +($95\%$ CI $[+0.170,+0.309]$), and +$\Delta\rho[\text{pac}-\text{raptor}]=+0.166$ $[+0.116,+0.215]$, both significant +and robust across five seeds; on naive rotation the contrast runs the other way +($\Delta\rho[\text{raptor}-\text{aug}]=+0.234$ $[+0.189,+0.284]$). The mechanism +is interpretable: label-preserving augmentation directly probes how the concept +direction moves when inputs are perturbed \emph{while their label is held fixed}, +which is structurally the same operation as a semantic shift; bootstrap +dispersion only perturbs the \emph{sample}, so it can only ever recover the +sampling floor that \excess{} removes. Thus ``which a-priori signal is useful'' +is not an absolute property of a predictor---it depends entirely on whether the +sampling floor has been subtracted (\cref{fig:mechanism}). + +\subsection{Robustness of the two findings} +Both the circularity result and the aug${>}$dispersion reversal survive every +stress test we ran. +\begin{itemize}[leftmargin=1.4em] + \item \textbf{Augmentation diversity (Tier~2, de/fr/ru multi-pivot).} + Diversifying back-translation lifts the aug \excess{} correlation only from + $0.046$ to $0.070$ (CI still crosses zero), so a richer augmentation does + \emph{not} rescue positive predictability---ruling out ``the augmentation was + too weak'' as an explanation for the open problem. The reversal, however, is + robust to diversity: $\Delta\rho[\text{aug}-\text{raptor}]=+0.247$ on \excess{} + (vs.\ $+0.241$ at 1-aug), significant. + \item \textbf{Scale (70M--6.9B, no inverse scaling).} Across the Pythia ladder + plus GPT-2 and Qwen, the naive dispersion correlation stays in $0.83$--$0.95$ + and the \excess{} correlation stays $\le 0$ throughout; the predictability + \emph{structure} is invariant to scale, contradicting an inverse-scaling story + (\cref{fig:sizeladder}). At 6.9B (\texttt{pythia-6.9b}, seed 0, $n{=}7$) the + pattern holds and the augmentation advantage \emph{amplifies}: \texttt{raptor} + naive $+0.93$ (still circular), \excess{} \texttt{raptor} $-0.39$ (still + anti-predictive), \excess{} aug $+0.68$ (stronger positive). + \item \textbf{Concepts (LOCO over 9--12 concepts).} The qualitative ordering is + preserved under leave-one-concept-out; the 9-concept analysis additionally + exposes which signals are concept-overfit (whitened-cosine LOCO $0.21$, xie + LOCO $0.03$) versus cross-concept robust (\texttt{raptor} $0.88$, \texttt{pac} + $0.81$, aug $0.69$ on the naive target), which is itself a useful G2 product. + \item \textbf{Metric (whitening).} \excess{} conclusions are not an artifact of + bare cosine: re-running rotation under an ID-whitened cosine target moves + \texttt{raptor} from $0.945$ to $0.698$, narrowing---but not + eliminating---the circular advantage. We report this as an honest ``results + depend on the metric'' caveat rather than claiming the bare-cosine choice is + privileged. +\end{itemize} + +\subsection{Positioning and takeaways} +We deliberately do not claim to propose the strongest predictor of probe OOD +stability; the prior-art landscape (RAPTOR-style dispersion, Probing-the-Probes +augmentation robustness, SIP eigengaps, Fragility, Xie Dispersion-Score) is +crowded enough that such a claim would be punctured. Our contribution is a +\emph{benchmark plus a circularity revelation plus an open problem}: (i) a +unified, label-free grid on which existing a-priori signals can be compared +apples-to-apples; (ii) the demonstration, via the IID-resampling placebo, that +the apparently strong naive predictability of directional brittleness is mostly a +sampling-noise mirage; and (iii) the honest, non-circular \excess{} target on +which \emph{no} existing signal predicts positively, yet on which +augmentation-based signal significantly dominates dispersion-based signal. +Practitioners who want to know, before deployment, whether a probe direction will +survive paraphrase shift should treat current dispersion/whitened-geometry scores +with suspicion: on the honest target they are negatively correlated with the +outcome they are advertised to anticipate. + +%================================================================ Limitations +\section{Limitations} +\label{sec:limitations} + +\paragraph{Shift-specific brittleness is, in absolute terms, not yet +predictable.} +We are explicit that \excess{} is an open problem, not a solved one. The single +positive signal (augmentation, $\rho=0.046$; $0.070$ with multi-pivot) has a +confidence interval that crosses zero in every configuration. The reversal +$\text{aug}>\text{dispersion}$ is a \emph{relative} statement about which family +of signals is less wrong; it must not be read as ``augmentation reliably predicts +shift-specific rotation.'' Our headline is a comparative and a negative result, +and we report it as such. + +\paragraph{Predictability is estimator-specific, not a general property.} +The strong naive correlations are tied to the logistic-regression estimator used +to fit directions. When the same a-priori predictors are used to predict +\emph{mass-mean} direction rotation, all of them are weak (best is xie at +$0.47$; the rest $0.25$--$0.29$) and LOCO collapses to $\approx 0$ (no +cross-concept generalization). Whatever predictability exists is therefore a +property of a particular estimator's geometry rather than of probe directions in +general---an important caveat against over-generalizing our LogReg results. + +\paragraph{Label-fidelity is uneven across datasets.} +Our NLI round-trip audit passes at $80.6\%$ overall, but is high on +\texttt{counterfact} ($96\%$) and \texttt{sst2} ($89\%$) and notably low on +\texttt{dbpedia} ($50\%$) and \texttt{imdb} ($64\%$), where entity-heavy or long +back-translations frequently break label preservation. Rotation measured on these +two datasets should be interpreted with caution; their shifts are not cleanly +label-preserving, and we flag (rather than hide) this. We treat the NLI audit as +a sanity check, not a novelty claim, and cite the overgenerate-and-filter / +round-trip lineage (Falsesum, DISCO) accordingly. + +\paragraph{The 6.9B evidence is a spot-check.} +The 6.9B result rests on a single seed and only $n=7$ dataset cells, which is too +coarse for confidence intervals. We report it only as a directional consistency +check ($\text{circularity}+\text{aug}{>}\text{dispersion}$ both persist and the +aug advantage grows); we do \emph{not} draw scaling-law conclusions from it, and +the full size ladder is where the no-inverse-scaling claim actually rests. + +\paragraph{The paraphrase \excess{} analysis rests on a single back-translation +pivot family.} +The 5-seed \excess{} results (\cref{tab:fourtarget}) use single-pivot +back-translation; Tier~2 extends this to de/fr/ru but only confirms the same +qualitative picture. We do not claim coverage of the full space of paraphrase +generators (e.g., instruction-tuned LLM rewriters), and a systematically +different paraphrase operator could in principle shift the absolute \excess{} +numbers even if, as Tier~2 suggests, the relative ordering is stable. Similarly, +our shift-type split shows predictor behavior depends strongly on shift +type---paraphrase ($n{=}472$, circular), domain ($n{=}65$, weak: \texttt{raptor} +$0.59$), length ($n{=}30$, uniformly high: aug/pac/raptor $0.95/0.95/0.92$)---and +the domain and length arms have small $n$, so their conclusions are weaker than +the paraphrase arm's. + +%================================================================ Future work +\section{Future Work} +\label{sec:future} +The open problem we isolate---\emph{predicting shift-specific (\excess{}) +directional brittleness without touching OOD data}---suggests several concrete +directions. (1)~Because augmentation is the only family with a non-negative +\excess{} signal, the most promising avenue is augmentation operators that more +faithfully simulate the target shift (instruction-tuned rewriters, controllable +domain/length transforms) coupled with NLI-gated label-fidelity, aiming to push +the augmentation correlation's CI off zero. (2)~The estimator-specificity result +motivates predictors defined directly on the geometry of \emph{difference-of-means} +and MLP probes, since LogReg predictability does not transfer. (3)~The partial +metric-dependence under whitening ($0.945\!\rightarrow\!0.698$) invites a +principled treatment of which covariance structure the rotation metric should +quotient out, ideally one that removes the sampling floor analytically rather than +via the empirical IID-resampling subtraction we use here. (4)~Strengthening the +domain ($n{=}65$) and length ($n{=}30$) arms and the 6.9B spot-check ($n{=}7$) to +seed-replicated grids would let the shift-type-dependence and scale-invariance +claims carry confidence intervals as tight as the paraphrase arm's. + +%================================================================ Reproducibility +\section{Reproducibility} +\label{sec:reproducibility} + +\probeshift{} is designed for minutes-to-reproduce verification on commodity +hardware. We release, under a permissive license: (i)~the full +\textbf{activation cache} (one forward pass per configuration, shared across all +predictors), so that every number in +\cref{tab:fourtarget,tab:paired,fig:circularity,fig:mechanism,fig:sizeladder} can +be recomputed from cache without GPU inference; (ii)~the complete \textbf{code} +for activation extraction, probe fitting (LogReg / mass-mean / MLP), the seven +label-free predictors, the IID-resampling placebo, the \excess{} target, and the +bootstrap / paired-contrast statistics; (iii)~the \textbf{pre-registration} +document fixing the four-target design, the predictor list, and the LOCO / +five-seed protocol \emph{before} the main run, so that the circularity finding and +the reversal are confirmatory rather than post-hoc; and (iv)~the exact +\textbf{data splits} (per-concept ID/shift/placebo indices, the five independent +seeds, and the de/fr/ru multi-pivot back-translation outputs with their NLI +round-trip fidelity labels). The headline grid uses Option~A (five fully +independent seeds, $n=472$); each reported confidence interval is a $10$k-sample +bootstrap over the shared configuration set, and all paired contrasts share the +configuration index so they are directly comparable. The entire pipeline runs on a +single RTX~4090 within $\leq\!200$ GPU$\cdot$h, with \$0 API spend and zero new +human annotation; cache-only reanalysis (shift-type breakdown, estimator external +validity, layer robustness, whitening ablation) runs in minutes. + +%------------------------------------------------------------------- Bibliography +% Replace with the PMLR/COLT .bst and a real .bib for camera-ready. +\bibliographystyle{plainnat} +\bibliography{references} + +% Fallback inline bibliography so the document compiles without a .bib file. +% Delete this block once references.bib is provided. +\begin{thebibliography}{9} +\bibitem[Belinkov(2022)]{belinkov2022probing} +Y.~Belinkov. +\newblock Probing classifiers: Promises, shortcomings, and advances. +\newblock \emph{Computational Linguistics}, 2022. + +\bibitem[Gao et~al.(2026)]{gao2026raptor} +Gao et~al. +\newblock RAPTOR: Ridge-adaptive directional stability for probe reliability. +\newblock \emph{Preprint}, 2026. + +\bibitem[Huang(2025)]{huang2025sip} +Huang. +\newblock Spectral identifiability of probes (SIP). +\newblock \emph{Preprint}, 2025. + +\bibitem[Kumar et~al.(2022)]{kumar2022unreliable} +Kumar et~al. +\newblock On the unreliability of probing-based interpretations. +\newblock \emph{Preprint}, 2022. + +\bibitem[Lysn{\ae}s-Larsen et~al.(2025)]{lysnaes2025probing} +Lysn{\ae}s-Larsen et~al. +\newblock Probing the probes: Augmentation robustness as a probe-quality metric. +\newblock \emph{Preprint}, 2025. + +\bibitem[Reblitz-Richardson(2026)]{reblitz2026fragility} +Reblitz-Richardson. +\newblock Forward-pass fragility thresholds for linear probes. +\newblock \emph{Preprint}, 2026. + +\bibitem[Xie et~al.(2023)]{xie2023feature} +Xie et~al. +\newblock Feature dispersion score for label-free OOD-error prediction. +\newblock \emph{Preprint}, 2023. +\end{thebibliography} + +\end{document} diff --git a/repro_bundle/references.bib b/repro_bundle/references.bib new file mode 100644 index 0000000000000000000000000000000000000000..1f1776a7c84f8cd02d884b50292cdbec6681b6fd --- /dev/null +++ b/repro_bundle/references.bib @@ -0,0 +1,81 @@ +% ProbeShift references. arXiv IDs marked [VERIFIED] were fetched directly from arxiv.org on +% 2026-06-22 (title/authors confirmed). IDs marked [CHECK] were surfaced by automated search +% and MUST be re-verified (venue/ID) before camera-ready. + +% [VERIFIED] 2511.19166 +@article{dies2025pstat, + title = {Representational and Behavioral Stability of Truth in Large Language Models}, + author = {Dies, Samantha and Maynard, Courtney and Savcisens, Germans and Eliassi-Rad, Tina}, + journal= {arXiv preprint arXiv:2511.19166}, year = {2025}} + +% [VERIFIED] 2511.16288 +@article{huang2025sip, + title = {Spectral Identifiability for Interpretable Probe Geometry}, + author = {Huang, William Hao-Cheng}, + journal= {arXiv preprint arXiv:2511.16288}, year = {2025}} + +% [VERIFIED] 2602.00158 +@article{gao2026raptor, + title = {{RAPTOR}: Ridge-Adaptive Logistic Probes}, + author = {Gao, Ziqi and Zhu, Yaotian and Zeng, Qingcheng and Zhao, Xu and Wang, Ziqing and Ruan, Feng and Ding, Kaize}, + journal= {arXiv preprint arXiv:2602.00158}, year = {2026}} + +% [VERIFIED] 2510.11905 +@article{haller2025brittle, + title = {{LLM} Knowledge is Brittle: Truthfulness Representations Rely on Superficial Resemblance}, + author = {Haller, Patrick and Ibrahim, Mark and Kirichenko, Polina and Sagun, Levent and Bell, Samuel J.}, + journal= {arXiv preprint arXiv:2510.11905}, year = {2025}} + +% [VERIFIED] 2303.15488, NeurIPS 2023 +@inproceedings{xie2023feature, + title = {On the Importance of Feature Separability in Predicting Out-Of-Distribution Error}, + author = {Xie, Renchunzi and Wei, Hongxin and Feng, Lei and Cao, Yuzhou and An, Bo}, + booktitle = {Advances in Neural Information Processing Systems (NeurIPS)}, year = {2023}} + +% [VERIFIED] 2602.20273 +@article{ying2026truthspectrum, + title = {The Truthfulness Spectrum Hypothesis}, + author = {Ying, Zhuofan Josh and Ravfogel, Shauli and Kriegeskorte, Nikolaus and Hase, Peter}, + journal= {arXiv preprint arXiv:2602.20273}, year = {2026}} + +% [VERIFIED] 2310.06824, COLM 2024 +@article{marks2024geometry, + title = {The Geometry of Truth: Emergent Linear Structure in {LLM} Representations of True/False Datasets}, + author = {Marks, Samuel and Tegmark, Max}, + journal= {Conference on Language Modeling (COLM); arXiv:2310.06824}, year = {2024}} + +% [CHECK] 2511.04312 +@article{lysnaes2025probing, + title = {Probing the Probes: Methods and Metrics for Concept Alignment}, + author = {Lysn{\ae}s-Larsen, M. and Eggen, P. and Str{\"u}mke, I.}, + journal= {arXiv preprint arXiv:2511.04312}, year = {2025}} + +% [CHECK] 2606.11375 +@article{reblitz2026fragility, + title = {When Probing Accuracy Saturates, Fragility Resolves}, + author = {Reblitz-Richardson, O. and others}, + journal= {arXiv preprint arXiv:2606.11375}, year = {2026}} + +% [CHECK] 2605.27958 +@article{deception2026pressure, + title = {Pressure-Testing Deception Probes in {LLM}s: Scaling, Robustness, and the Geometry of Deceptive Representations}, + author = {Anonymous}, + journal= {arXiv preprint arXiv:2605.27958}, year = {2026}} + +@inproceedings{hewitt2019control, + title = {Designing and Interpreting Probes with Control Tasks}, + author = {Hewitt, John and Liang, Percy}, + booktitle = {EMNLP}, year = {2019}} + +@article{belinkov2022probing, + title = {Probing Classifiers: Promises, Shortcomings, and Advances}, + author = {Belinkov, Yonatan}, journal = {Computational Linguistics}, volume = {48}, number = {1}, year = {2022}} + +@inproceedings{park2024linear, + title = {The Linear Representation Hypothesis and the Geometry of Large Language Models}, + author = {Park, Kiho and Choe, Yo Joong and Veitch, Victor}, + booktitle = {ICML}, year = {2024}} + +@inproceedings{biderman2023pythia, + title = {Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling}, + author = {Biderman, Stella and others}, booktitle = {ICML}, year = {2023}} diff --git a/repro_bundle/references.bib.bak b/repro_bundle/references.bib.bak new file mode 100644 index 0000000000000000000000000000000000000000..82bf55cf0d7dc7f382fbd57729a7a7be9d358aaa --- /dev/null +++ b/repro_bundle/references.bib.bak @@ -0,0 +1,71 @@ +% ProbeShift references. arXiv IDs marked [VERIFIED] were fetched directly from arxiv.org on +% 2026-06-22 (title/authors confirmed). IDs marked [CHECK] were surfaced by automated search +% and MUST be re-verified (venue/ID) before camera-ready. + +@article{dies2025pstat, % [VERIFIED] 2511.19166 + title = {Representational and Behavioral Stability of Truth in Large Language Models}, + author = {Dies, Samantha and Maynard, Courtney and Savcisens, Germans and Eliassi-Rad, Tina}, + journal= {arXiv preprint arXiv:2511.19166}, year = {2025}} + +@article{huang2025sip, % [VERIFIED] 2511.16288 + title = {Spectral Identifiability for Interpretable Probe Geometry}, + author = {Huang, William Hao-Cheng}, + journal= {arXiv preprint arXiv:2511.16288}, year = {2025}} + +@article{gao2026raptor, % [VERIFIED] 2602.00158 + title = {{RAPTOR}: Ridge-Adaptive Logistic Probes}, + author = {Gao, Ziqi and Zhu, Yaotian and Zeng, Qingcheng and Zhao, Xu and Wang, Ziqing and Ruan, Feng and Ding, Kaize}, + journal= {arXiv preprint arXiv:2602.00158}, year = {2026}} + +@article{haller2025brittle, % [VERIFIED] 2510.11905 + title = {{LLM} Knowledge is Brittle: Truthfulness Representations Rely on Superficial Resemblance}, + author = {Haller, Patrick and Ibrahim, Mark and Kirichenko, Polina and Sagun, Levent and Bell, Samuel J.}, + journal= {arXiv preprint arXiv:2510.11905}, year = {2025}} + +@inproceedings{xie2023feature, % [VERIFIED] 2303.15488, NeurIPS 2023 + title = {On the Importance of Feature Separability in Predicting Out-Of-Distribution Error}, + author = {Xie, Renchunzi and Wei, Hongxin and Feng, Lei and Cao, Yuzhou and An, Bo}, + booktitle = {Advances in Neural Information Processing Systems (NeurIPS)}, year = {2023}} + +@article{ying2026truthspectrum, % [VERIFIED] 2602.20273 + title = {The Truthfulness Spectrum Hypothesis}, + author = {Ying, Zhuofan Josh and Ravfogel, Shauli and Kriegeskorte, Nikolaus and Hase, Peter}, + journal= {arXiv preprint arXiv:2602.20273}, year = {2026}} + +@article{marks2024geometry, % [VERIFIED] 2310.06824, COLM 2024 + title = {The Geometry of Truth: Emergent Linear Structure in {LLM} Representations of True/False Datasets}, + author = {Marks, Samuel and Tegmark, Max}, + journal= {Conference on Language Modeling (COLM); arXiv:2310.06824}, year = {2024}} + +@article{lysnaes2025probing, % [CHECK] 2511.04312 + title = {Probing the Probes: Methods and Metrics for Concept Alignment}, + author = {Lysn{\ae}s-Larsen, M. and Eggen, P. and Str{\"u}mke, I.}, + journal= {arXiv preprint arXiv:2511.04312}, year = {2025}} + +@article{reblitz2026fragility, % [CHECK] 2606.11375 + title = {When Probing Accuracy Saturates, Fragility Resolves}, + author = {Reblitz-Richardson, O. and others}, + journal= {arXiv preprint arXiv:2606.11375}, year = {2026}} + +@article{deception2026pressure, % [CHECK] 2605.27958 + title = {Pressure-Testing Deception Probes in {LLM}s: Scaling, Robustness, and the Geometry of Deceptive Representations}, + author = {Anonymous}, + journal= {arXiv preprint arXiv:2605.27958}, year = {2026}} + +@inproceedings{hewitt2019control, + title = {Designing and Interpreting Probes with Control Tasks}, + author = {Hewitt, John and Liang, Percy}, + booktitle = {EMNLP}, year = {2019}} + +@article{belinkov2022probing, + title = {Probing Classifiers: Promises, Shortcomings, and Advances}, + author = {Belinkov, Yonatan}, journal = {Computational Linguistics}, volume = {48}, number = {1}, year = {2022}} + +@inproceedings{park2024linear, + title = {The Linear Representation Hypothesis and the Geometry of Large Language Models}, + author = {Park, Kiho and Choe, Yo Joong and Veitch, Victor}, + booktitle = {ICML}, year = {2024}} + +@inproceedings{biderman2023pythia, + title = {Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling}, + author = {Biderman, Stella and others}, booktitle = {ICML}, year = {2023}} diff --git a/repro_bundle/requirements.txt b/repro_bundle/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..d5bb577ef2ca42c92f59f9e9c501f12289740dfd --- /dev/null +++ b/repro_bundle/requirements.txt @@ -0,0 +1,17 @@ +# Core +numpy>=1.26 +scipy>=1.11 +scikit-learn>=1.4 +pandas>=2.1 + +# Deep learning / models (install the CUDA build of torch matching your 4090 driver) +torch>=2.2 +transformers>=4.44 +datasets>=2.20 +accelerate>=0.33 +sentencepiece>=0.2 # for NLLB / Marian / Qwen tokenizers +sacremoses>=0.1 # for Marian (opus-mt) back-translation + +# Utilities +tqdm>=4.66 +pyyaml>=6.0 diff --git a/repro_bundle/results_A/results/audit.jsonl b/repro_bundle/results_A/results/audit.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..1db2f201b5ca92f23f5dd8a9a34eb07aa3fa774f --- /dev/null +++ b/repro_bundle/results_A/results/audit.jsonl @@ -0,0 +1,70 @@ +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": 0.15384615384615385, "n": 208} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 1, "flip_rate": 0.10124999999999995, "n": 800} +{"dataset": "imdb", "seed": 1, "flip_rate": 0.38749999999999996, "n": 800} +{"dataset": "ag_news", "seed": 1, "flip_rate": 0.20750000000000002, "n": 800} +{"dataset": "dbpedia", "seed": 1, "flip_rate": 0.5137499999999999, "n": 800} +{"dataset": "counterfact", "seed": 1, "flip_rate": 0.03249999999999997, "n": 800} +{"dataset": "emotion", "seed": 1, "flip_rate": 0.07625000000000004, "n": 800} +{"dataset": "tweet_hate", "seed": 1, "flip_rate": 0.26, "n": 800} +{"dataset": "tweet_irony", "seed": 1, "flip_rate": 0.23124999999999996, "n": 800} +{"dataset": "tweet_offensive", "seed": 1, "flip_rate": 0.23375, "n": 800} +{"dataset": "subj", "seed": 1, "flip_rate": 0.08250000000000002, "n": 800} +{"dataset": "spam", "seed": 1, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "cola", "seed": 1, "flip_rate": 0.13624999999999998, "n": 800} +{"dataset": "stance", "seed": 1, "flip_rate": 0.1826923076923077, "n": 208} +{"dataset": "amazon_cf", "seed": 1, "flip_rate": 0.11375000000000002, "n": 800} +{"dataset": "sst2", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "imdb", "seed": 2, "flip_rate": 0.35624999999999996, "n": 800} +{"dataset": "ag_news", "seed": 2, "flip_rate": 0.20625000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 2, "flip_rate": 0.5037499999999999, "n": 800} +{"dataset": "counterfact", "seed": 2, "flip_rate": 0.04125000000000001, "n": 800} +{"dataset": "emotion", "seed": 2, "flip_rate": 0.05874999999999997, "n": 800} +{"dataset": "tweet_hate", "seed": 2, "flip_rate": 0.22750000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 2, "flip_rate": 0.22624999999999995, "n": 800} +{"dataset": "tweet_offensive", "seed": 2, "flip_rate": 0.23875000000000002, "n": 800} +{"dataset": "subj", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "spam", "seed": 2, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 2, "flip_rate": 0.10875000000000001, "n": 800} +{"dataset": "stance", "seed": 2, "flip_rate": 0.14903846153846156, "n": 208} +{"dataset": "amazon_cf", "seed": 2, "flip_rate": 0.08750000000000002, "n": 800} +{"dataset": "sst2", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "imdb", "seed": 3, "flip_rate": 0.3425, "n": 800} +{"dataset": "ag_news", "seed": 3, "flip_rate": 0.19874999999999998, "n": 800} +{"dataset": "dbpedia", "seed": 3, "flip_rate": 0.47875, "n": 800} +{"dataset": "counterfact", "seed": 3, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 3, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 3, "flip_rate": 0.25125, "n": 800} +{"dataset": "tweet_irony", "seed": 3, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "tweet_offensive", "seed": 3, "flip_rate": 0.2025, "n": 800} +{"dataset": "subj", "seed": 3, "flip_rate": 0.10375000000000001, "n": 800} +{"dataset": "spam", "seed": 3, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "cola", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "stance", "seed": 3, "flip_rate": 0.16826923076923073, "n": 208} +{"dataset": "amazon_cf", "seed": 3, "flip_rate": 0.09875, "n": 800} +{"dataset": "sst2", "seed": 4, "flip_rate": 0.12124999999999997, "n": 800} +{"dataset": "imdb", "seed": 4, "flip_rate": 0.36250000000000004, "n": 800} +{"dataset": "ag_news", "seed": 4, "flip_rate": 0.17500000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 4, "flip_rate": 0.48624999999999996, "n": 800} +{"dataset": "counterfact", "seed": 4, "flip_rate": 0.03874999999999995, "n": 800} +{"dataset": "emotion", "seed": 4, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 4, "flip_rate": 0.24750000000000005, "n": 800} +{"dataset": "tweet_irony", "seed": 4, "flip_rate": 0.24624999999999997, "n": 800} +{"dataset": "tweet_offensive", "seed": 4, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "subj", "seed": 4, "flip_rate": 0.08125000000000004, "n": 800} +{"dataset": "spam", "seed": 4, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 4, "flip_rate": 0.11499999999999999, "n": 800} +{"dataset": "stance", "seed": 4, "flip_rate": 0.1875, "n": 208} +{"dataset": "amazon_cf", "seed": 4, "flip_rate": 0.13, "n": 800} diff --git a/repro_bundle/results_A/results/eval.jsonl b/repro_bundle/results_A/results/eval.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..7e4432ce4e65d30c8926336d070290bed59c0c5e --- /dev/null +++ b/repro_bundle/results_A/results/eval.jsonl @@ -0,0 +1,1416 @@ +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.78875, "drop": 0.0}, "paraphrase": {"acc": 0.7645429362880887, "drop": 0.02420706371191128, "rotation": {"subspace_principal_angle": 1.2173428997633269, "mean_class_cosine": 0.34613981140180783}}, "domain": {"acc": 0.66375, "drop": 0.125, "rotation": {"subspace_principal_angle": 1.2327061371099897, "mean_class_cosine": 0.3316860024743622}}}, "iid_acc": 0.78875, "selectivity": 0.29624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.287801427158086, "mean_class_cosine": 0.2792326694699942}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5415512465373962, "drop": -0.005301246537396165, "rotation": {"subspace_principal_angle": 1.4826873586765374, "mean_class_cosine": 0.8742178904437565}}, "domain": {"acc": 0.50125, "drop": 0.03500000000000003, "rotation": {"subspace_principal_angle": 1.4901649913072454, "mean_class_cosine": 0.6563624615789627}}}, "iid_acc": 0.53625, "selectivity": 0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.9335911744705346, "mean_class_cosine": 0.9103338651934845}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.7562326869806094, "drop": 0.017517313019390635}, "domain": {"acc": 0.7725, "drop": 0.0012500000000000844}}, "iid_acc": 0.77375, "selectivity": 0.28125000000000006} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8075, "drop": 0.0}, "paraphrase": {"acc": 0.7965116279069767, "drop": 0.01098837209302328, "rotation": {"subspace_principal_angle": 0.8035537849788047, "mean_class_cosine": 0.6941529859280129}}, "domain": {"acc": 0.6875, "drop": 0.12, "rotation": {"subspace_principal_angle": 1.1686325991529884, "mean_class_cosine": 0.3914103543132651}}, "length": {"acc": 0.8, "drop": 0.007499999999999951, "rotation": {"subspace_principal_angle": 0.7897497491232404, "mean_class_cosine": 0.7040230601402793}}}, "iid_acc": 0.8075, "selectivity": 0.30374999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7674349618290983, "mean_class_cosine": 0.7196939193951748}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6414728682170543, "drop": 0.01602713178294568, "rotation": {"subspace_principal_angle": 1.288879288748167, "mean_class_cosine": 0.8485185221141581}}, "domain": {"acc": 0.6625, "drop": -0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.4368459632860788, "mean_class_cosine": 0.5607042188856695}}, "length": {"acc": 0.6621212121212121, "drop": -0.004621212121212137, "rotation": {"subspace_principal_angle": 1.2230547063204023, "mean_class_cosine": 0.9166318400068196}}}, "iid_acc": 0.6575, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.8046269051289581, "mean_class_cosine": 0.8820641372607203}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.7926356589147286, "drop": -0.005135658914728669}, "domain": {"acc": 0.69625, "drop": 0.09124999999999994}, "length": {"acc": 0.7878787878787878, "drop": -0.00037878787878786735}}, "iid_acc": 0.7875, "selectivity": 0.28374999999999995} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8307692307692308, "drop": 0.015480769230769131, "rotation": {"subspace_principal_angle": 1.1461706135196499, "mean_class_cosine": 0.5254429176325506}}}, "iid_acc": 0.84625, "selectivity": 0.58, "iid_split_rotation": {"subspace_principal_angle": 1.1688799127850784, "mean_class_cosine": 0.5261001194927581}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8, "drop": 0.0}, "paraphrase": {"acc": 0.796923076923077, "drop": 0.003076923076923088, "rotation": {"subspace_principal_angle": 1.3242841060180541, "mean_class_cosine": 0.9641264652586653}}}, "iid_acc": 0.8, "selectivity": 0.5337500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3981863180782648, "mean_class_cosine": 0.9370201686458131}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.8415384615384616, "drop": 0.003461538461538405}}, "iid_acc": 0.845, "selectivity": 0.57875} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.8927680798004988, "drop": 0.04223192019950128, "rotation": {"subspace_principal_angle": 1.2106415683121374, "mean_class_cosine": 0.7066695308643387}}}, "iid_acc": 0.935, "selectivity": 0.8525, "iid_split_rotation": {"subspace_principal_angle": 1.1511700616225666, "mean_class_cosine": 0.7422879247769008}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5311720698254364, "drop": 0.02382793017456364, "rotation": {"subspace_principal_angle": 1.373149046572336, "mean_class_cosine": 0.8538155049730057}}}, "iid_acc": 0.555, "selectivity": 0.47250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3861255031014492, "mean_class_cosine": 0.8730732976173553}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.85785536159601, "drop": 0.03214463840399007}}, "iid_acc": 0.89, "selectivity": 0.8075} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5019607843137255, "drop": 0.01678921568627456, "rotation": {"subspace_principal_angle": 1.5573365448958074, "mean_class_cosine": -0.013459375494237761}}}, "iid_acc": 0.51875, "selectivity": 0.011250000000000093, "iid_split_rotation": {"subspace_principal_angle": 1.5530648223509733, "mean_class_cosine": -0.017730575309227073}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.0015604575163399304, "rotation": {"subspace_principal_angle": 1.5316164005556767, "mean_class_cosine": 0.04651835530761543}}}, "iid_acc": 0.50875, "selectivity": 0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.4740729716187824, "mean_class_cosine": 0.03841895244711871}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.5202614379084968, "drop": -0.010261437908496762}}, "iid_acc": 0.51, "selectivity": 0.0025000000000000577} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.61, "drop": 0.0}, "paraphrase": {"acc": 0.5486018641810919, "drop": 0.06139813581890807, "rotation": {"subspace_principal_angle": 1.4117139037388426, "mean_class_cosine": 0.3559915861697644}}}, "iid_acc": 0.61, "selectivity": 0.37375, "iid_split_rotation": {"subspace_principal_angle": 1.3714632310508308, "mean_class_cosine": 0.37969570224814214}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.275, "drop": 0.0}, "paraphrase": {"acc": 0.2583222370173103, "drop": 0.016677762982689748, "rotation": {"subspace_principal_angle": 1.5666753155926543, "mean_class_cosine": 0.5349177629069118}}}, "iid_acc": 0.275, "selectivity": 0.038750000000000034, "iid_split_rotation": {"subspace_principal_angle": 1.5082299186645924, "mean_class_cosine": 0.6046278432817893}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.507323568575233, "drop": -0.009823568575232977}}, "iid_acc": 0.4975, "selectivity": 0.26125} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.70875, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.0009274193548387544, "rotation": {"subspace_principal_angle": 1.3772445961922097, "mean_class_cosine": 0.19234551093090788}}}, "iid_acc": 0.70875, "selectivity": 0.18625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3584465776147807, "mean_class_cosine": 0.2107574495845495}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.5825, "drop": 0.0}, "paraphrase": {"acc": 0.5942275042444821, "drop": -0.011727504244482101, "rotation": {"subspace_principal_angle": 1.2205673531066674, "mean_class_cosine": 0.9079473240997651}}}, "iid_acc": 0.5825, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1442505653422113, "mean_class_cosine": 0.940086831782734}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.7045840407470289, "drop": 0.007915959252971172}}, "iid_acc": 0.7125, "selectivity": 0.19000000000000006} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.028266129032258003, "rotation": {"subspace_principal_angle": 1.4282418254454519, "mean_class_cosine": 0.142072164699308}}}, "iid_acc": 0.66375, "selectivity": 0.16374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3468259132949723, "mean_class_cosine": 0.2221026092043146}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5838709677419355, "drop": 0.016129032258064502, "rotation": {"subspace_principal_angle": 1.3876894491282326, "mean_class_cosine": 0.9792973864192993}}}, "iid_acc": 0.6, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.0920941984969763, "mean_class_cosine": 0.9685148140397846}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66125, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.019314516129032233}}, "iid_acc": 0.66125, "selectivity": 0.16125} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7600644122383253, "drop": -0.006314412238325295, "rotation": {"subspace_principal_angle": 1.115598956625283, "mean_class_cosine": 0.43963959517656737}}}, "iid_acc": 0.75375, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1918132960864172, "mean_class_cosine": 0.369975854351391}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.007481884057971011, "rotation": {"subspace_principal_angle": 1.431073468315769, "mean_class_cosine": 0.8636351111958729}}}, "iid_acc": 0.52875, "selectivity": -0.1037499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9984226130449826, "mean_class_cosine": 0.8215813059697017}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7342995169082126, "drop": -0.009299516908212624}}, "iid_acc": 0.725, "selectivity": 0.09250000000000003} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.879286694101509, "drop": 0.02321330589849102, "rotation": {"subspace_principal_angle": 0.9139246510391348, "mean_class_cosine": 0.6106425040490346}}}, "iid_acc": 0.9025, "selectivity": 0.44499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.884850806935217, "mean_class_cosine": 0.6334049571887388}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8491083676268861, "drop": 0.012141632373113831, "rotation": {"subspace_principal_angle": 0.783728326709098, "mean_class_cosine": 0.9410916640925551}}}, "iid_acc": 0.86125, "selectivity": 0.40374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.9780082924965329, "mean_class_cosine": 0.9403009956281769}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8710562414266118, "drop": 0.025193758573388236}}, "iid_acc": 0.89625, "selectivity": 0.43875} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9862778730703259, "drop": 0.007472126929674139, "rotation": {"subspace_principal_angle": 0.8999811261795105, "mean_class_cosine": 0.6216247525314251}}}, "iid_acc": 0.99375, "selectivity": 0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.8781549302342212, "mean_class_cosine": 0.6385721258735755}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64375, "drop": 0.0}, "paraphrase": {"acc": 0.4854202401372213, "drop": 0.15832975986277875, "rotation": {"subspace_principal_angle": 0.9223293102070813, "mean_class_cosine": 0.9921756731272653}}}, "iid_acc": 0.64375, "selectivity": -0.2024999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3103002492542979, "mean_class_cosine": 0.9993471541844559}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.97875, "drop": 0.0}, "paraphrase": {"acc": 0.9845626072041166, "drop": -0.005812607204116582}}, "iid_acc": 0.97875, "selectivity": 0.13250000000000006} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6826086956521739, "drop": -0.032608695652173836, "rotation": {"subspace_principal_angle": 1.5134217389057267, "mean_class_cosine": 0.05734311504400581}}}, "iid_acc": 0.65, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4857097607143221, "mean_class_cosine": 0.0849839360330566}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5260869565217391, "drop": 0.0001630434782609047, "rotation": {"subspace_principal_angle": 1.5681344894004097, "mean_class_cosine": 0.0027859134562919843}}}, "iid_acc": 0.52625, "selectivity": -0.10375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3461131864613545, "mean_class_cosine": -0.37599024476007736}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7202898550724638, "drop": -0.014039855072463747}}, "iid_acc": 0.70625, "selectivity": 0.07625000000000004} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6306818181818182, "drop": 0.023164335664335622, "rotation": {"subspace_principal_angle": 1.3319284714979758, "mean_class_cosine": 0.23716017540167048}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.21634615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.372005152667984, "mean_class_cosine": 0.24328801012012044}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.5384615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.5511363636363636, "drop": -0.01267482517482521, "rotation": {"subspace_principal_angle": 1.3661942544847507, "mean_class_cosine": 0.680691975446298}}}, "iid_acc": 0.5384615384615384, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.326965973200558, "mean_class_cosine": 0.5637407887003096}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6363636363636364, "drop": -0.011363636363636354}}, "iid_acc": 0.625, "selectivity": 0.1875} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.8817663817663818, "drop": -0.005516381766381806, "rotation": {"subspace_principal_angle": 1.0257225092803133, "mean_class_cosine": 0.5184812124632806}}}, "iid_acc": 0.87625, "selectivity": 0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.022872498909312, "mean_class_cosine": 0.5209161160438877}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7236467236467237, "drop": -0.0023967236467237196, "rotation": {"subspace_principal_angle": 1.5032004808784296, "mean_class_cosine": 0.9058820512519088}}}, "iid_acc": 0.72125, "selectivity": -0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.3645997436602872, "mean_class_cosine": 0.8668774947492062}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.8148148148148148, "drop": 0.002685185185185235}}, "iid_acc": 0.8175, "selectivity": 0.03375000000000006} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8019390581717452, "drop": 0.021810941828254826, "rotation": {"subspace_principal_angle": 1.2297289568018384, "mean_class_cosine": 0.33449317002319223}}, "domain": {"acc": 0.7775, "drop": 0.04625000000000001, "rotation": {"subspace_principal_angle": 1.2534806736867017, "mean_class_cosine": 0.31201735319379614}}}, "iid_acc": 0.82375, "selectivity": 0.33375, "iid_split_rotation": {"subspace_principal_angle": 1.249974932944242, "mean_class_cosine": 0.31534615054656556}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5526315789473685, "drop": -0.022631578947368447, "rotation": {"subspace_principal_angle": 1.3043937828785666, "mean_class_cosine": 0.9883475306540368}}, "domain": {"acc": 0.50125, "drop": 0.028750000000000053, "rotation": {"subspace_principal_angle": 1.4264081497308534, "mean_class_cosine": 0.495004129731469}}}, "iid_acc": 0.53, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.1840707821435512, "mean_class_cosine": 0.9761804914155537}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8060941828254847, "drop": 0.02015581717451531}, "domain": {"acc": 0.77375, "drop": 0.05249999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33625000000000005} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8275193798449613, "drop": 0.01498062015503876, "rotation": {"subspace_principal_angle": 1.095349444903557, "mean_class_cosine": 0.4577358103130952}}, "domain": {"acc": 0.71125, "drop": 0.13124999999999998, "rotation": {"subspace_principal_angle": 1.266476857896887, "mean_class_cosine": 0.29964398320533164}}, "length": {"acc": 0.843939393939394, "drop": -0.0014393939393939625, "rotation": {"subspace_principal_angle": 1.020996775517062, "mean_class_cosine": 0.5225163309800367}}}, "iid_acc": 0.8425, "selectivity": 0.38125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0185057868759257, "mean_class_cosine": 0.5246385975150215}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6996124031007752, "drop": 0.007887596899224802, "rotation": {"subspace_principal_angle": 1.5092754395213872, "mean_class_cosine": 0.8994925840610262}}, "domain": {"acc": 0.56125, "drop": 0.14625, "rotation": {"subspace_principal_angle": 1.5628669635664512, "mean_class_cosine": 0.41448835148075136}}, "length": {"acc": 0.7257575757575757, "drop": -0.018257575757575695, "rotation": {"subspace_principal_angle": 1.4863663784751628, "mean_class_cosine": 0.9364420792767987}}}, "iid_acc": 0.7075, "selectivity": 0.24625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5618431812325368, "mean_class_cosine": 0.89092989954099}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8352713178294574, "drop": 0.018478682170542604}, "domain": {"acc": 0.71875, "drop": 0.135}, "length": {"acc": 0.853030303030303, "drop": 0.0007196969696969813}}, "iid_acc": 0.85375, "selectivity": 0.3925} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8430769230769231, "drop": -0.0005769230769230305, "rotation": {"subspace_principal_angle": 1.4488633502262789, "mean_class_cosine": 0.2627189029713985}}}, "iid_acc": 0.8425, "selectivity": 0.60375, "iid_split_rotation": {"subspace_principal_angle": 1.3920742480225834, "mean_class_cosine": 0.2804084873444433}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8784615384615385, "drop": -0.019711538461538503, "rotation": {"subspace_principal_angle": 1.5157375083147102, "mean_class_cosine": 0.9772553909019803}}}, "iid_acc": 0.85875, "selectivity": 0.62, "iid_split_rotation": {"subspace_principal_angle": 1.5565345003357298, "mean_class_cosine": 0.9660708595879833}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5030769230769231, "drop": -0.0043269230769230616}}, "iid_acc": 0.49875, "selectivity": 0.26} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.027319201995012476, "rotation": {"subspace_principal_angle": 1.371071144342142, "mean_class_cosine": 0.6983572474879793}}}, "iid_acc": 0.955, "selectivity": 0.87125, "iid_split_rotation": {"subspace_principal_angle": 1.1190162175121645, "mean_class_cosine": 0.7424385333421911}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.46, "drop": 0.0}, "paraphrase": {"acc": 0.4339152119700748, "drop": 0.0260847880299252, "rotation": {"subspace_principal_angle": 1.2702079599204308, "mean_class_cosine": 0.8441489999187578}}}, "iid_acc": 0.46, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.54813616950135, "mean_class_cosine": 0.8639196384527672}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8952618453865336, "drop": 0.025988154613466374}}, "iid_acc": 0.92125, "selectivity": 0.8375} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": 0.003088235294117614, "rotation": {"subspace_principal_angle": 1.5507981147481529, "mean_class_cosine": 0.019996879097623668}}}, "iid_acc": 0.5325, "selectivity": 0.04124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4729086760692658, "mean_class_cosine": 0.09773139915411025}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.51125, "drop": 0.0}, "paraphrase": {"acc": 0.4915032679738562, "drop": 0.0197467320261438, "rotation": {"subspace_principal_angle": 1.5063411676739642, "mean_class_cosine": 0.09376607824598783}}}, "iid_acc": 0.51125, "selectivity": 0.019999999999999962, "iid_split_rotation": {"subspace_principal_angle": 1.3815718454510815, "mean_class_cosine": -0.780613294925026}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.01531045751633986}}, "iid_acc": 0.5225, "selectivity": 0.031249999999999944} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.5872170439414115, "drop": 0.0377829560585885, "rotation": {"subspace_principal_angle": 1.415219489070654, "mean_class_cosine": 0.29709377332489545}}}, "iid_acc": 0.625, "selectivity": 0.36125, "iid_split_rotation": {"subspace_principal_angle": 1.3868652000931787, "mean_class_cosine": 0.30693642876692934}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1975, "drop": 0.0}, "paraphrase": {"acc": 0.20905459387483355, "drop": -0.01155459387483354, "rotation": {"subspace_principal_angle": 1.508152156475103, "mean_class_cosine": 0.3868276136007782}}}, "iid_acc": 0.1975, "selectivity": -0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4103810864475195, "mean_class_cosine": 0.663164345963827}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.5539280958721704, "drop": 0.014821904127829577}}, "iid_acc": 0.56875, "selectivity": 0.305} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7470288624787776, "drop": 0.0017211375212223823, "rotation": {"subspace_principal_angle": 1.218944387247129, "mean_class_cosine": 0.34463687992333025}}}, "iid_acc": 0.74875, "selectivity": 0.21000000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.161541923084499, "mean_class_cosine": 0.39792541536895626}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.6281833616298812, "drop": -0.029433361629881194, "rotation": {"subspace_principal_angle": 1.4828776907746226, "mean_class_cosine": 0.8507212382776503}}}, "iid_acc": 0.59875, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5285287072761573, "mean_class_cosine": 0.889675723801062}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7589134125636672, "drop": -0.007663412563667205}}, "iid_acc": 0.75125, "selectivity": 0.21250000000000002} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6258064516129033, "drop": 0.029193548387096757, "rotation": {"subspace_principal_angle": 1.4380753701349163, "mean_class_cosine": 0.13233165634932476}}}, "iid_acc": 0.655, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.3608689467779782, "mean_class_cosine": 0.20838887470800788}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.6129032258064516, "drop": -0.009153225806451615, "rotation": {"subspace_principal_angle": 1.3501977335705537, "mean_class_cosine": 0.8788630744675302}}}, "iid_acc": 0.60375, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3042174668679, "mean_class_cosine": 0.9016543168707795}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6274193548387097, "drop": 0.01758064516129032}}, "iid_acc": 0.645, "selectivity": 0.13624999999999998} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7584541062801933, "drop": -0.015954106280193225, "rotation": {"subspace_principal_angle": 1.2412464706999458, "mean_class_cosine": 0.3236171403840489}}}, "iid_acc": 0.7425, "selectivity": 0.125, "iid_split_rotation": {"subspace_principal_angle": 1.2363583878771438, "mean_class_cosine": 0.32823830168996404}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5797101449275363, "drop": -0.013460144927536222, "rotation": {"subspace_principal_angle": 1.4328194682708153, "mean_class_cosine": 0.8163073052539073}}}, "iid_acc": 0.56625, "selectivity": -0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.393646227569837, "mean_class_cosine": 0.7747593471553436}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7471819645732689, "drop": -0.00843196457326889}}, "iid_acc": 0.73875, "selectivity": 0.12124999999999997} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.887517146776406, "drop": 0.02623285322359392, "rotation": {"subspace_principal_angle": 1.1698183926314918, "mean_class_cosine": 0.39031889296657585}}}, "iid_acc": 0.91375, "selectivity": 0.3949999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1565211457406315, "mean_class_cosine": 0.4025265308866283}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8847736625514403, "drop": 0.007726337448559661, "rotation": {"subspace_principal_angle": 0.9750908966255823, "mean_class_cosine": 0.724907117214936}}}, "iid_acc": 0.8925, "selectivity": 0.3737499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5507156284967207, "mean_class_cosine": 0.9656213717308835}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9012345679012346, "drop": 0.013765432098765462}}, "iid_acc": 0.915, "selectivity": 0.39625} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.7143429012203775, "mean_class_cosine": 0.7555238835683384}}}, "iid_acc": 0.99125, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6854415650897067, "mean_class_cosine": 0.7741396052879256}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.93625, "drop": 0.0}, "paraphrase": {"acc": 0.9468267581475128, "drop": -0.0105767581475128, "rotation": {"subspace_principal_angle": 1.309840282431262, "mean_class_cosine": 0.963893627422916}}}, "iid_acc": 0.93625, "selectivity": 0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.94741234766098, "mean_class_cosine": 0.9835024086435691}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9775, "drop": 0.0}, "paraphrase": {"acc": 0.9828473413379074, "drop": -0.005347341337907374}}, "iid_acc": 0.9775, "selectivity": 0.11250000000000004} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.006666666666666599, "rotation": {"subspace_principal_angle": 1.5316360507140625, "mean_class_cosine": 0.0391502679566816}}}, "iid_acc": 0.66, "selectivity": 0.07250000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3743529128243746, "mean_class_cosine": 0.19518239053186837}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.47246376811594204, "drop": 0.03753623188405797, "rotation": {"subspace_principal_angle": 1.118894178326917, "mean_class_cosine": -0.9771850067325109}}}, "iid_acc": 0.51, "selectivity": -0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3803133457755477, "mean_class_cosine": 0.946898836596458}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7072463768115942, "drop": -0.0009963768115941463}}, "iid_acc": 0.70625, "selectivity": 0.11875000000000002} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6079545454545454, "drop": -0.0021853146853146876, "rotation": {"subspace_principal_angle": 1.5660532537843979, "mean_class_cosine": 0.23593286698387292}}}, "iid_acc": 0.6057692307692307, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.3587755369509562, "mean_class_cosine": 0.22622489885469554}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6022727272727273, "drop": -0.02054195804195802, "rotation": {"subspace_principal_angle": 1.425664296220229, "mean_class_cosine": 0.6549701520168614}}}, "iid_acc": 0.5817307692307693, "selectivity": 0.1538461538461539, "iid_split_rotation": {"subspace_principal_angle": 1.3242514768412414, "mean_class_cosine": 0.5857958842617208}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.0021853146853146876}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2163461538461539} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8703703703703703, "drop": 0.00462962962962965, "rotation": {"subspace_principal_angle": 1.2653845793904546, "mean_class_cosine": 0.30068589374987864}}}, "iid_acc": 0.875, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 1.1530609002795849, "mean_class_cosine": 0.4056916522664149}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.57875, "drop": 0.0}, "paraphrase": {"acc": 0.5911680911680912, "drop": -0.012418091168091228, "rotation": {"subspace_principal_angle": 1.5387855283082137, "mean_class_cosine": 0.9739607963186667}}}, "iid_acc": 0.57875, "selectivity": -0.17625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.0864399863171719, "mean_class_cosine": 0.8793845387817394}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8475783475783476, "drop": 0.006171652421652407}}, "iid_acc": 0.85375, "selectivity": 0.09875} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8213296398891967, "drop": 0.03117036011080332, "rotation": {"subspace_principal_angle": 1.2728072178283032, "mean_class_cosine": 0.2935985328813959}}, "domain": {"acc": 0.6475, "drop": 0.20500000000000007, "rotation": {"subspace_principal_angle": 1.2548682229869004, "mean_class_cosine": 0.31069877540648505}}}, "iid_acc": 0.8525, "selectivity": 0.35625, "iid_split_rotation": {"subspace_principal_angle": 1.1836995794199188, "mean_class_cosine": 0.37750157196014655}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5318559556786704, "drop": -0.006855955678670367, "rotation": {"subspace_principal_angle": 1.183174660406376, "mean_class_cosine": 0.9978979106979409}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.4296731020342406, "mean_class_cosine": 0.41513571955567125}}}, "iid_acc": 0.525, "selectivity": 0.028749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9398355095136322, "mean_class_cosine": 0.9889864954707369}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8421052631578947, "drop": 0.004144736842105257}, "domain": {"acc": 0.8225, "drop": 0.023749999999999938}}, "iid_acc": 0.84625, "selectivity": 0.3499999999999999} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8604651162790697, "drop": 0.024534883720930267, "rotation": {"subspace_principal_angle": 1.1635939988143396, "mean_class_cosine": 0.396041968116701}}, "domain": {"acc": 0.56, "drop": 0.32499999999999996, "rotation": {"subspace_principal_angle": 1.3073928416159974, "mean_class_cosine": 0.26036815053777257}}, "length": {"acc": 0.8878787878787879, "drop": -0.002878787878787925, "rotation": {"subspace_principal_angle": 1.1306543361974652, "mean_class_cosine": 0.42606792700100266}}}, "iid_acc": 0.885, "selectivity": 0.41375, "iid_split_rotation": {"subspace_principal_angle": 1.1314507236110818, "mean_class_cosine": 0.42534730729927617}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7383720930232558, "drop": 0.03412790697674417, "rotation": {"subspace_principal_angle": 0.9299498041850935, "mean_class_cosine": 0.9017395970359101}}, "domain": {"acc": 0.5575, "drop": 0.21499999999999997, "rotation": {"subspace_principal_angle": 1.4111384197278427, "mean_class_cosine": 0.27301695590829955}}, "length": {"acc": 0.7878787878787878, "drop": -0.01537878787878788, "rotation": {"subspace_principal_angle": 1.0571900651122168, "mean_class_cosine": 0.9405678373193577}}}, "iid_acc": 0.7725, "selectivity": 0.30124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.339591955524043, "mean_class_cosine": 0.8972887886451133}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8662790697674418, "drop": 0.022470930232558195}, "domain": {"acc": 0.65, "drop": 0.23875000000000002}, "length": {"acc": 0.8878787878787879, "drop": 0.000871212121212106}}, "iid_acc": 0.88875, "selectivity": 0.41750000000000004} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.88, "drop": -0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.4381874155007441, "mean_class_cosine": 0.2360974231087299}}}, "iid_acc": 0.875, "selectivity": 0.5974999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3889398949790899, "mean_class_cosine": 0.28986888876348327}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8123076923076923, "drop": 0.013942307692307754, "rotation": {"subspace_principal_angle": 1.4726936861262263, "mean_class_cosine": 0.9825256512058247}}}, "iid_acc": 0.82625, "selectivity": 0.5487500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.413087776634539, "mean_class_cosine": 0.9710401150655834}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8861538461538462, "drop": 0.005096153846153806}}, "iid_acc": 0.89125, "selectivity": 0.61375} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.9725, "drop": 0.0}, "paraphrase": {"acc": 0.9551122194513716, "drop": 0.01738778054862844, "rotation": {"subspace_principal_angle": 1.1805912863774084, "mean_class_cosine": 0.696959257792652}}}, "iid_acc": 0.9725, "selectivity": 0.875, "iid_split_rotation": {"subspace_principal_angle": 1.1819768188421824, "mean_class_cosine": 0.7412159284602584}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.4275, "drop": 0.0}, "paraphrase": {"acc": 0.42643391521197005, "drop": 0.0010660847880299418, "rotation": {"subspace_principal_angle": 1.267210976332211, "mean_class_cosine": 0.8469328938765297}}}, "iid_acc": 0.4275, "selectivity": 0.32999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2708278894045024, "mean_class_cosine": 0.8352289736541864}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.0210692019950125}}, "iid_acc": 0.94875, "selectivity": 0.85125} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5568627450980392, "drop": -0.006862745098039191, "rotation": {"subspace_principal_angle": 1.4681160564274516, "mean_class_cosine": 0.10249993503953822}}}, "iid_acc": 0.55, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.5020669180006716, "mean_class_cosine": 0.06867531169240668}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.4875816993464052, "drop": 0.00991830065359478, "rotation": {"subspace_principal_angle": 1.0583311490868843, "mean_class_cosine": 0.6025164388499016}}}, "iid_acc": 0.4975, "selectivity": -0.03250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2056755940549575, "mean_class_cosine": -0.8743607002802327}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5372549019607843, "drop": 0.0002450980392156854}}, "iid_acc": 0.5375, "selectivity": 0.007499999999999951} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6604527296937417, "drop": -0.005452729693741665, "rotation": {"subspace_principal_angle": 1.3690495088359922, "mean_class_cosine": 0.35780435447931885}}}, "iid_acc": 0.655, "selectivity": 0.39375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.418659878889111, "mean_class_cosine": 0.31040393556774787}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.18125, "drop": 0.0}, "paraphrase": {"acc": 0.1904127829560586, "drop": -0.009162782956058602, "rotation": {"subspace_principal_angle": 1.4238306284342992, "mean_class_cosine": 0.3336979016265693}}}, "iid_acc": 0.18125, "selectivity": -0.07999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3516589964423749, "mean_class_cosine": 0.6675023436977741}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5712383488681758, "drop": 0.030011651131824135}}, "iid_acc": 0.60125, "selectivity": 0.33999999999999997} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.00039685908319186947, "rotation": {"subspace_principal_angle": 1.3803752925053967, "mean_class_cosine": 0.18927233568020757}}}, "iid_acc": 0.72875, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 1.4107099285898823, "mean_class_cosine": 0.1594035006847374}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5229202037351444, "drop": -0.015420203735144411, "rotation": {"subspace_principal_angle": 1.132903800026888, "mean_class_cosine": 0.9957211105004683}}}, "iid_acc": 0.5075, "selectivity": -0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1567819921044111, "mean_class_cosine": 0.997306798497698}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7606112054329371, "drop": -0.0018612054329371075}}, "iid_acc": 0.75875, "selectivity": 0.22750000000000004} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.03556451612903222, "rotation": {"subspace_principal_angle": 1.4500907567958619, "mean_class_cosine": 0.12041267342093487}}}, "iid_acc": 0.6775, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4451716043887228, "mean_class_cosine": 0.12529455713007304}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5435483870967742, "drop": 0.002701612903225792, "rotation": {"subspace_principal_angle": 1.0905650289278745, "mean_class_cosine": 0.9919303109580393}}}, "iid_acc": 0.54625, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.435319911955165, "mean_class_cosine": 0.9822586493642307}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.02806451612903227}}, "iid_acc": 0.67, "selectivity": 0.16375000000000006} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7568438003220612, "drop": 0.014406199677938769, "rotation": {"subspace_principal_angle": 1.362823497882489, "mean_class_cosine": 0.2064768368861471}}}, "iid_acc": 0.77125, "selectivity": 0.21124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3556181289971392, "mean_class_cosine": 0.21352151995440388}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.01748188405797102, "rotation": {"subspace_principal_angle": 1.162087777749045, "mean_class_cosine": 0.4004265677277434}}}, "iid_acc": 0.51875, "selectivity": -0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5040593331311518, "mean_class_cosine": 0.6240661876988235}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.775, "drop": 0.0}, "paraphrase": {"acc": 0.7729468599033816, "drop": 0.002053140096618389}}, "iid_acc": 0.775, "selectivity": 0.21499999999999997} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9245541838134431, "drop": 0.017945816186556884, "rotation": {"subspace_principal_angle": 1.1580571429217417, "mean_class_cosine": 0.401119992002977}}}, "iid_acc": 0.9425, "selectivity": 0.44875, "iid_split_rotation": {"subspace_principal_angle": 1.0639249835392377, "mean_class_cosine": 0.48544434408974335}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5116598079561042, "drop": 0.0033401920438957955, "rotation": {"subspace_principal_angle": 1.474754823264982, "mean_class_cosine": 0.27562171118558787}}}, "iid_acc": 0.515, "selectivity": 0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6855036886417566, "mean_class_cosine": 0.9705005903895466}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.93, "drop": 0.0}, "paraphrase": {"acc": 0.934156378600823, "drop": -0.004156378600822963}}, "iid_acc": 0.93, "selectivity": 0.43625} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0006389365351628928, "rotation": {"subspace_principal_angle": 0.7922854168062247, "mean_class_cosine": 0.7022200256213631}}}, "iid_acc": 0.9925, "selectivity": 0.13250000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.7419982107629325, "mean_class_cosine": 0.7371197159686783}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9502572898799314, "drop": -0.00900728987993138, "rotation": {"subspace_principal_angle": 1.2962420205278622, "mean_class_cosine": 0.9560615750258359}}}, "iid_acc": 0.94125, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4821593641247295, "mean_class_cosine": 0.9814265656310696}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9875, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": -0.0022084048027444236}}, "iid_acc": 0.9875, "selectivity": 0.12750000000000006} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6608695652173913, "drop": 0.03413043478260869, "rotation": {"subspace_principal_angle": 1.5269251960916086, "mean_class_cosine": 0.04385705910466185}}}, "iid_acc": 0.695, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.452899790541391, "mean_class_cosine": 0.117623607019381}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.041141304347826146, "rotation": {"subspace_principal_angle": 1.2698364171990568, "mean_class_cosine": -0.7441747656964055}}}, "iid_acc": 0.52375, "selectivity": -0.06624999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.421349910061863, "mean_class_cosine": 0.7671339569758747}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7231884057971014, "drop": -0.009438405797101423}}, "iid_acc": 0.71375, "selectivity": 0.12375000000000003} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6826923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.6363636363636364, "drop": 0.046328671328671356, "rotation": {"subspace_principal_angle": 1.389965032090286, "mean_class_cosine": 0.19817848562152038}}}, "iid_acc": 0.6826923076923077, "selectivity": 0.25480769230769235, "iid_split_rotation": {"subspace_principal_angle": 1.5631105240216077, "mean_class_cosine": 0.18163197534662465}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6105769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.625, "drop": -0.014423076923076872, "rotation": {"subspace_principal_angle": 1.3818731643580437, "mean_class_cosine": 0.6330079224250492}}}, "iid_acc": 0.6105769230769231, "selectivity": 0.18269230769230776, "iid_split_rotation": {"subspace_principal_angle": 1.188799274694121, "mean_class_cosine": 0.5596495884045671}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.006993006993006978}}, "iid_acc": 0.6490384615384616, "selectivity": 0.2211538461538462} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8803418803418803, "drop": 0.0009081196581196549, "rotation": {"subspace_principal_angle": 1.336870712814003, "mean_class_cosine": 0.23179799551578698}}}, "iid_acc": 0.88125, "selectivity": 0.15749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.230199866592587, "mean_class_cosine": 0.3340493484245235}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.5811965811965812, "drop": -0.004946581196581201, "rotation": {"subspace_principal_angle": 0.6698879155775275, "mean_class_cosine": 0.9725719155523136}}}, "iid_acc": 0.57625, "selectivity": -0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9538385113441675, "mean_class_cosine": 0.9585511263219829}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8974358974358975, "drop": 0.0013141025641025816}}, "iid_acc": 0.89875, "selectivity": 0.17500000000000004} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8337950138504155, "drop": 0.027454986149584415, "rotation": {"subspace_principal_angle": 1.2672102146950088, "mean_class_cosine": 0.2989442428389649}}, "domain": {"acc": 0.88, "drop": -0.018750000000000044, "rotation": {"subspace_principal_angle": 1.401321553710003, "mean_class_cosine": 0.1686646701084955}}}, "iid_acc": 0.86125, "selectivity": 0.36749999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.2210418787662296, "mean_class_cosine": 0.3426671322986696}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5346260387811634, "drop": -0.009626038781163415, "rotation": {"subspace_principal_angle": 1.1075097075912679, "mean_class_cosine": 0.9893879557921825}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.2937330372237155, "mean_class_cosine": 0.27468238367063336}}}, "iid_acc": 0.525, "selectivity": 0.03125, "iid_split_rotation": {"subspace_principal_angle": 1.474523039419839, "mean_class_cosine": 0.9693041836230525}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8476454293628809, "drop": 0.04110457063711914}, "domain": {"acc": 0.86875, "drop": 0.020000000000000018}}, "iid_acc": 0.88875, "selectivity": 0.395} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.875968992248062, "drop": 0.012781007751938023, "rotation": {"subspace_principal_angle": 1.3733967390372437, "mean_class_cosine": 0.1961200848251179}}, "domain": {"acc": 0.5525, "drop": 0.33625000000000005, "rotation": {"subspace_principal_angle": 1.427574987743908, "mean_class_cosine": 0.14273220656435498}}, "length": {"acc": 0.8924242424242425, "drop": -0.003674242424242413, "rotation": {"subspace_principal_angle": 1.32660559111955, "mean_class_cosine": 0.24177114792102647}}}, "iid_acc": 0.88875, "selectivity": 0.3975, "iid_split_rotation": {"subspace_principal_angle": 1.3721751230402406, "mean_class_cosine": 0.19731783026248}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.84875, "drop": 0.0}, "paraphrase": {"acc": 0.8197674418604651, "drop": 0.028982558139534875, "rotation": {"subspace_principal_angle": 1.0890107964995421, "mean_class_cosine": 0.9365459180728041}}, "domain": {"acc": 0.5575, "drop": 0.29125, "rotation": {"subspace_principal_angle": 1.2433163855943603, "mean_class_cosine": 0.290612600127489}}, "length": {"acc": 0.8575757575757575, "drop": -0.008825757575757542, "rotation": {"subspace_principal_angle": 0.7717021170570981, "mean_class_cosine": 0.9578281621778026}}}, "iid_acc": 0.84875, "selectivity": 0.3575, "iid_split_rotation": {"subspace_principal_angle": 1.2334456221930967, "mean_class_cosine": 0.9323952105718878}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8875968992248062, "drop": 0.00990310077519374}, "domain": {"acc": 0.695, "drop": 0.2025}, "length": {"acc": 0.9015151515151515, "drop": -0.004015151515151527}}, "iid_acc": 0.8975, "selectivity": 0.40624999999999994} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8907692307692308, "drop": -0.00951923076923078, "rotation": {"subspace_principal_angle": 1.3982668079683969, "mean_class_cosine": 0.2412385692456759}}}, "iid_acc": 0.88125, "selectivity": 0.63, "iid_split_rotation": {"subspace_principal_angle": 1.3873607135770687, "mean_class_cosine": 0.27893385434758605}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.805, "drop": 0.0}, "paraphrase": {"acc": 0.8046153846153846, "drop": 0.00038461538461542766, "rotation": {"subspace_principal_angle": 1.1223434771672867, "mean_class_cosine": 0.9797187391070092}}}, "iid_acc": 0.805, "selectivity": 0.5537500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1951168916205284, "mean_class_cosine": 0.9636873787333077}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.9092307692307692, "drop": -0.010480769230769127}}, "iid_acc": 0.89875, "selectivity": 0.6475000000000001} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.972568578553616, "drop": 0.0024314214463839745, "rotation": {"subspace_principal_angle": 1.0887911120828875, "mean_class_cosine": 0.6783458118132375}}}, "iid_acc": 0.975, "selectivity": 0.8624999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.083853464584136, "mean_class_cosine": 0.7060030624663257}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.8125, "drop": 0.0}, "paraphrase": {"acc": 0.773067331670823, "drop": 0.03943266832917702, "rotation": {"subspace_principal_angle": 1.1185256458497195, "mean_class_cosine": 0.8994165755464111}}}, "iid_acc": 0.8125, "selectivity": 0.7, "iid_split_rotation": {"subspace_principal_angle": 0.9631112899466643, "mean_class_cosine": 0.9391589025650812}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9650872817955112, "drop": 0.011162718204488775}}, "iid_acc": 0.97625, "selectivity": 0.8637499999999999} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.5830065359477125, "drop": 0.024493464052287583, "rotation": {"subspace_principal_angle": 1.4835452240288083, "mean_class_cosine": 0.08714044135255376}}}, "iid_acc": 0.6075, "selectivity": 0.08750000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.483433918298054, "mean_class_cosine": 0.08725132314037118}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.4888888888888889, "drop": 0.01611111111111113, "rotation": {"subspace_principal_angle": 1.2048851781158827, "mean_class_cosine": 0.5321986735325194}}}, "iid_acc": 0.505, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4614710386057692, "mean_class_cosine": -0.6438340541689455}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5542483660130719, "drop": 0.03450163398692807}}, "iid_acc": 0.58875, "selectivity": 0.06874999999999998} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65875, "drop": 0.0}, "paraphrase": {"acc": 0.6085219707057257, "drop": 0.05022802929427428, "rotation": {"subspace_principal_angle": 1.4274964103041523, "mean_class_cosine": 0.2819535108490442}}}, "iid_acc": 0.65875, "selectivity": 0.3837499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3942837316412728, "mean_class_cosine": 0.30680585311258146}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.18, "drop": 0.0}, "paraphrase": {"acc": 0.19440745672436752, "drop": -0.014407456724367523, "rotation": {"subspace_principal_angle": 1.42729065004506, "mean_class_cosine": 0.3233324825686333}}}, "iid_acc": 0.18, "selectivity": -0.09500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.543456747904511, "mean_class_cosine": 0.6716440420685439}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02558089214380832}}, "iid_acc": 0.64875, "selectivity": 0.37375} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.005396859083191874, "rotation": {"subspace_principal_angle": 1.4458284299287356, "mean_class_cosine": 0.1246428806664818}}}, "iid_acc": 0.73375, "selectivity": 0.23625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4100208755322998, "mean_class_cosine": 0.16008370527990506}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5314091680814941, "drop": -0.012659168081494032, "rotation": {"subspace_principal_angle": 1.5017224503953317, "mean_class_cosine": 0.9922645891758275}}}, "iid_acc": 0.51875, "selectivity": 0.021250000000000047, "iid_split_rotation": {"subspace_principal_angle": 0.5510785135387823, "mean_class_cosine": 0.994402608644392}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7504244482173175, "drop": 0.020825551782682528}}, "iid_acc": 0.77125, "selectivity": 0.27375} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.6532258064516129, "drop": 0.031774193548387175, "rotation": {"subspace_principal_angle": 1.4862916929900118, "mean_class_cosine": 0.08440409464427018}}}, "iid_acc": 0.685, "selectivity": 0.18375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.504979130468674, "mean_class_cosine": 0.06576968766218992}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5564516129032258, "drop": 0.009798387096774275, "rotation": {"subspace_principal_angle": 1.4670137647528678, "mean_class_cosine": 0.9493974522411648}}}, "iid_acc": 0.56625, "selectivity": 0.06500000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2038765745555904, "mean_class_cosine": 0.9198294280970227}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.6935483870967742, "drop": 0.01895161290322578}}, "iid_acc": 0.7125, "selectivity": 0.21125000000000005} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7439613526570048, "drop": 0.0047886473429952225, "rotation": {"subspace_principal_angle": 1.4293803795970716, "mean_class_cosine": 0.14094506785538105}}}, "iid_acc": 0.74875, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4308280823099806, "mean_class_cosine": 0.13951166975487922}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.571658615136876, "drop": -0.019158615136876, "rotation": {"subspace_principal_angle": 1.3467680040328998, "mean_class_cosine": 0.780244054310545}}}, "iid_acc": 0.5525, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.5305723014222863, "mean_class_cosine": 0.8225959610699823}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.7616747181964574, "drop": 0.01207528180354267}}, "iid_acc": 0.77375, "selectivity": 0.19625000000000004} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.02258058984910838, "rotation": {"subspace_principal_angle": 1.1992451710106438, "mean_class_cosine": 0.3630611813014619}}}, "iid_acc": 0.95125, "selectivity": 0.44125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.197550530346397, "mean_class_cosine": 0.3646396665958699}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6598079561042524, "drop": -0.0023079561042523844, "rotation": {"subspace_principal_angle": 0.8946495725521172, "mean_class_cosine": 0.6299993713824044}}}, "iid_acc": 0.6575, "selectivity": 0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7999262041956853, "mean_class_cosine": 0.9697684559622158}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.945, "drop": 0.0}, "paraphrase": {"acc": 0.9314128943758574, "drop": 0.01358710562414256}}, "iid_acc": 0.945, "selectivity": 0.43499999999999994} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99875, "drop": 0.0}, "paraphrase": {"acc": 0.9982847341337907, "drop": 0.000465265866209319, "rotation": {"subspace_principal_angle": 0.931771543682588, "mean_class_cosine": 0.5964129401770092}}}, "iid_acc": 0.99875, "selectivity": 0.18125000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9266228730293125, "mean_class_cosine": 0.600537740322992}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6375, "drop": 0.0}, "paraphrase": {"acc": 0.4065180102915952, "drop": 0.23098198970840478, "rotation": {"subspace_principal_angle": 0.4804883746103172, "mean_class_cosine": 0.9773138541986853}}}, "iid_acc": 0.6375, "selectivity": -0.18000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0322313143428026, "mean_class_cosine": 0.9999323022857498}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9948542024013722, "drop": 0.001395797598627735}}, "iid_acc": 0.99625, "selectivity": 0.17874999999999996} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.6884057971014492, "drop": 0.04284420289855073, "rotation": {"subspace_principal_angle": 1.5103706788720144, "mean_class_cosine": 0.060388883021034304}}}, "iid_acc": 0.73125, "selectivity": 0.14374999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.4200020283035577, "mean_class_cosine": 0.1502234646254889}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.4666666666666667, "drop": 0.04083333333333328, "rotation": {"subspace_principal_angle": 0.9965627610068227, "mean_class_cosine": -0.9475923907212384}}}, "iid_acc": 0.5075, "selectivity": -0.08000000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.7465288923899847, "mean_class_cosine": 0.963542850231548}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7275362318840579, "drop": 0.039963768115942044}}, "iid_acc": 0.7675, "selectivity": 0.17999999999999994} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6647727272727273, "drop": 0.008304195804195835, "rotation": {"subspace_principal_angle": 1.4238317227425443, "mean_class_cosine": 0.16606688749587134}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.18269230769230776, "iid_split_rotation": {"subspace_principal_angle": 1.4493637206220726, "mean_class_cosine": 0.18022085560237225}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.4423076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.4090909090909091, "drop": 0.033216783216783174, "rotation": {"subspace_principal_angle": 1.314437097913778, "mean_class_cosine": 0.6492731252980364}}}, "iid_acc": 0.4423076923076923, "selectivity": -0.04807692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.0082539944993898, "mean_class_cosine": 0.81084397247279}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6761363636363636, "drop": -0.0030594405594405183}}, "iid_acc": 0.6730769230769231, "selectivity": 0.18269230769230776} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8874643874643875, "drop": 0.005035612535612466, "rotation": {"subspace_principal_angle": 1.373498225543769, "mean_class_cosine": 0.19602056818834496}}}, "iid_acc": 0.8925, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.272801169025973, "mean_class_cosine": 0.29360431510047513}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.5883190883190883, "drop": 0.0016809116809116675, "rotation": {"subspace_principal_angle": 1.0911846628602964, "mean_class_cosine": 0.9555356213618866}}}, "iid_acc": 0.59, "selectivity": -0.125, "iid_split_rotation": {"subspace_principal_angle": 1.1224788303280901, "mean_class_cosine": 0.9089747436544128}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8846153846153846, "drop": 0.01163461538461541}}, "iid_acc": 0.89625, "selectivity": 0.18125000000000002} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7853185595567868, "drop": 0.01843144044321321, "rotation": {"subspace_principal_angle": 1.3395977046860834, "mean_class_cosine": 0.2291444175563393}}, "domain": {"acc": 0.78875, "drop": 0.015000000000000013, "rotation": {"subspace_principal_angle": 1.3075865369139061, "mean_class_cosine": 0.2601811310213691}}}, "iid_acc": 0.80375, "selectivity": 0.31124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3161987442877714, "mean_class_cosine": 0.25185598358320005}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5373961218836565, "drop": -0.017396121883656468, "rotation": {"subspace_principal_angle": 1.1243174241253076, "mean_class_cosine": 0.997655368526527}}, "domain": {"acc": 0.50125, "drop": 0.018750000000000044, "rotation": {"subspace_principal_angle": 1.434291363492328, "mean_class_cosine": 0.5041835597720421}}}, "iid_acc": 0.52, "selectivity": 0.027500000000000024, "iid_split_rotation": {"subspace_principal_angle": 1.4733549164259618, "mean_class_cosine": 0.9931712323643861}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.775623268698061, "drop": 0.050626731301939065}, "domain": {"acc": 0.7475, "drop": 0.07874999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33375000000000005} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8081395348837209, "drop": 0.01561046511627906, "rotation": {"subspace_principal_angle": 1.4878379029744278, "mean_class_cosine": 0.08286330186240415}}, "domain": {"acc": 0.63875, "drop": 0.18499999999999994, "rotation": {"subspace_principal_angle": 1.454691067906504, "mean_class_cosine": 0.11584457649805602}}, "length": {"acc": 0.8227272727272728, "drop": 0.0010227272727272307, "rotation": {"subspace_principal_angle": 1.4554754679619462, "mean_class_cosine": 0.11506542197274067}}}, "iid_acc": 0.82375, "selectivity": 0.3475, "iid_split_rotation": {"subspace_principal_angle": 1.3842160089393434, "mean_class_cosine": 0.18549965489482007}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7325581395348837, "drop": 0.03494186046511627, "rotation": {"subspace_principal_angle": 1.2562269465379385, "mean_class_cosine": 0.9025478152861728}}, "domain": {"acc": 0.5575, "drop": 0.20999999999999996, "rotation": {"subspace_principal_angle": 1.3251714868629496, "mean_class_cosine": 0.29596572601612736}}, "length": {"acc": 0.7727272727272727, "drop": -0.005227272727272747, "rotation": {"subspace_principal_angle": 1.0637760788494128, "mean_class_cosine": 0.9397971855033075}}}, "iid_acc": 0.7675, "selectivity": 0.29124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.061070800271238, "mean_class_cosine": 0.9127773418749959}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.8725, "drop": 0.0}, "paraphrase": {"acc": 0.8527131782945736, "drop": 0.01978682170542645}, "domain": {"acc": 0.57875, "drop": 0.29375000000000007}, "length": {"acc": 0.8712121212121212, "drop": 0.0012878787878788378}}, "iid_acc": 0.8725, "selectivity": 0.39625000000000005} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8661538461538462, "drop": 0.0013461538461538858, "rotation": {"subspace_principal_angle": 1.2824688472167018, "mean_class_cosine": 0.4528696194346733}}}, "iid_acc": 0.8675, "selectivity": 0.64, "iid_split_rotation": {"subspace_principal_angle": 1.267228970027625, "mean_class_cosine": 0.4341122200684181}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.47875, "drop": 0.0}, "paraphrase": {"acc": 0.42615384615384616, "drop": 0.05259615384615385, "rotation": {"subspace_principal_angle": 0.5775569724987617, "mean_class_cosine": 0.9093320332006872}}}, "iid_acc": 0.47875, "selectivity": 0.25125, "iid_split_rotation": {"subspace_principal_angle": 0.7727337369433303, "mean_class_cosine": 0.8668228792038879}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.0036538461538461187}}, "iid_acc": 0.8775, "selectivity": 0.6499999999999999} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.95375, "drop": 0.0}, "paraphrase": {"acc": 0.9451371571072319, "drop": 0.008612842892768091, "rotation": {"subspace_principal_angle": 1.198927059140121, "mean_class_cosine": 0.6054966764735671}}}, "iid_acc": 0.95375, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.1388755195423386, "mean_class_cosine": 0.6482459135365446}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6084788029925187, "drop": 0.01402119700748139, "rotation": {"subspace_principal_angle": 1.493246923874091, "mean_class_cosine": 0.8677677280952006}}}, "iid_acc": 0.6225, "selectivity": 0.5375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5091321723919393, "mean_class_cosine": 0.8880006055054167}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.9351620947630923, "drop": 0.012337905236907698}}, "iid_acc": 0.9475, "selectivity": 0.8625} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.495, "drop": 0.0}, "paraphrase": {"acc": 0.4745098039215686, "drop": 0.02049019607843139, "rotation": {"subspace_principal_angle": 1.557032384565418, "mean_class_cosine": -0.013763507646052914}}}, "iid_acc": 0.495, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.5359063301369187, "mean_class_cosine": -0.03488291842098544}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.0017892156862744923, "rotation": {"subspace_principal_angle": 1.3709234192517532, "mean_class_cosine": 0.6267242100586605}}}, "iid_acc": 0.49625, "selectivity": -0.014999999999999958, "iid_split_rotation": {"subspace_principal_angle": 1.3896477426172604, "mean_class_cosine": -0.3304774205182823}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.4825, "drop": 0.0}, "paraphrase": {"acc": 0.47320261437908495, "drop": 0.009297385620915033}}, "iid_acc": 0.4825, "selectivity": -0.028749999999999998} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5805592543275633, "drop": 0.02319074567243673, "rotation": {"subspace_principal_angle": 1.3989325795834617, "mean_class_cosine": 0.365872980266819}}}, "iid_acc": 0.60375, "selectivity": 0.335, "iid_split_rotation": {"subspace_principal_angle": 1.4035880336880546, "mean_class_cosine": 0.3488005843140671}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.1775, "drop": 0.0}, "paraphrase": {"acc": 0.16511318242343542, "drop": 0.012386817576564568, "rotation": {"subspace_principal_angle": 1.5619855127249644, "mean_class_cosine": 0.3341308312243725}}}, "iid_acc": 0.1775, "selectivity": -0.09125, "iid_split_rotation": {"subspace_principal_angle": 1.3905000233838531, "mean_class_cosine": 0.6611663679873362}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5778961384820239, "drop": 0.022103861517976053}}, "iid_acc": 0.6, "selectivity": 0.33125} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.015927419354838768, "rotation": {"subspace_principal_angle": 1.5028161180968338, "mean_class_cosine": 0.06792786120606348}}}, "iid_acc": 0.69375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4722418499371575, "mean_class_cosine": 0.09839501129737169}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.533106960950764, "drop": -0.020606960950764086, "rotation": {"subspace_principal_angle": 1.024515247435546, "mean_class_cosine": 0.9955533421541893}}}, "iid_acc": 0.5125, "selectivity": -0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.2562415527805546, "mean_class_cosine": 0.9954953689417243}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7555178268251274, "drop": 0.00448217317487265}}, "iid_acc": 0.76, "selectivity": 0.24624999999999997} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.6338709677419355, "drop": 0.017379032258064475, "rotation": {"subspace_principal_angle": 1.4673214972757795, "mean_class_cosine": 0.10329027681844688}}}, "iid_acc": 0.65125, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.5077800700720831, "mean_class_cosine": 0.06297455823321907}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.535483870967742, "drop": 0.005766129032258038, "rotation": {"subspace_principal_angle": 1.440925356190957, "mean_class_cosine": 0.9961212841602167}}}, "iid_acc": 0.54125, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3284144435763523, "mean_class_cosine": 0.9886091312692893}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6209677419354839, "drop": 0.03403225806451615}}, "iid_acc": 0.655, "selectivity": 0.17000000000000004} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.750402576489533, "drop": -0.009152576489533049, "rotation": {"subspace_principal_angle": 1.4211901787085208, "mean_class_cosine": 0.14904869151211453}}}, "iid_acc": 0.74125, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.4965103140884382, "mean_class_cosine": 0.07421770808908856}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5265700483091788, "drop": 0.03967995169082128, "rotation": {"subspace_principal_angle": 1.5565645508048995, "mean_class_cosine": 0.00760410778674959}}}, "iid_acc": 0.56625, "selectivity": 0.0025000000000000577, "iid_split_rotation": {"subspace_principal_angle": 0.9594487263253761, "mean_class_cosine": -0.4124908154959152}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7536231884057971, "drop": -0.0011231884057971708}}, "iid_acc": 0.7525, "selectivity": 0.18874999999999997} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8998628257887518, "drop": 0.02138717421124825, "rotation": {"subspace_principal_angle": 1.2563517152731103, "mean_class_cosine": 0.30928836211834454}}}, "iid_acc": 0.92125, "selectivity": 0.43375, "iid_split_rotation": {"subspace_principal_angle": 1.1864007137524348, "mean_class_cosine": 0.3749989233038349}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.545, "drop": 0.0}, "paraphrase": {"acc": 0.5569272976680384, "drop": -0.01192729766803835, "rotation": {"subspace_principal_angle": 0.6735562930427222, "mean_class_cosine": 0.3885547331126563}}}, "iid_acc": 0.545, "selectivity": 0.05750000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.3659301793739206, "mean_class_cosine": 0.9497848325707625}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.9039780521262003, "drop": 0.009771947873799647}}, "iid_acc": 0.91375, "selectivity": 0.42624999999999996} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9965694682675815, "drop": -0.0015694682675815308, "rotation": {"subspace_principal_angle": 0.9204115550682747, "mean_class_cosine": 0.605492671467479}}}, "iid_acc": 0.995, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 0.964600661915144, "mean_class_cosine": 0.5697451063410308}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7753001715265866, "drop": -0.00780017152658663, "rotation": {"subspace_principal_angle": 1.5561305399812533, "mean_class_cosine": 0.9880733179203292}}}, "iid_acc": 0.7675, "selectivity": -0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.8575713187579257, "mean_class_cosine": 0.9958656438305629}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": 0.0006110634648370805}}, "iid_acc": 0.99375, "selectivity": 0.19125000000000003} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6478260869565218, "drop": 0.0096739130434782, "rotation": {"subspace_principal_angle": 1.5203803161832818, "mean_class_cosine": 0.050394655640553065}}}, "iid_acc": 0.6575, "selectivity": 0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.4388032016199581, "mean_class_cosine": 0.13161019079523278}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4623188405797101, "drop": 0.040181159420289825, "rotation": {"subspace_principal_angle": 1.5601182686089141, "mean_class_cosine": -0.9599175649315286}}}, "iid_acc": 0.5025, "selectivity": -0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3056048703674055, "mean_class_cosine": 0.9664617957303832}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.6956521739130435, "drop": 0.018097826086956537}}, "iid_acc": 0.71375, "selectivity": 0.15500000000000003} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6477272727272727, "drop": -0.008304195804195835, "rotation": {"subspace_principal_angle": 1.4577549081283396, "mean_class_cosine": 0.13478050883475684}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.24999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.5438885641755995, "mean_class_cosine": 0.12661575417990478}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.40865384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.3693181818181818, "drop": 0.03933566433566432, "rotation": {"subspace_principal_angle": 1.3024704772930913, "mean_class_cosine": 0.751592937345538}}}, "iid_acc": 0.40865384615384615, "selectivity": 0.019230769230769218, "iid_split_rotation": {"subspace_principal_angle": 1.0672377902275059, "mean_class_cosine": 0.7053655504479837}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.0021853146853146876}}, "iid_acc": 0.6442307692307693, "selectivity": 0.25480769230769235} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8831908831908832, "drop": 0.0030591168091167464, "rotation": {"subspace_principal_angle": 1.4868944039444256, "mean_class_cosine": 0.08380351910501391}}}, "iid_acc": 0.88625, "selectivity": 0.20999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3010782735174253, "mean_class_cosine": 0.26645969404516406}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.575, "drop": 0.0}, "paraphrase": {"acc": 0.5797720797720798, "drop": -0.004772079772079829, "rotation": {"subspace_principal_angle": 1.1062280403413047, "mean_class_cosine": 0.9715489933239789}}}, "iid_acc": 0.575, "selectivity": -0.10125000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.5004862855814347, "mean_class_cosine": 0.9429161811360605}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.9017094017094017, "drop": -0.0017094017094017033}}, "iid_acc": 0.9, "selectivity": 0.22375} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.8448753462603878, "drop": 0.010124653739612133, "rotation": {"subspace_principal_angle": 1.2793309538353963, "mean_class_cosine": 0.28735610225096786}}, "domain": {"acc": 0.835, "drop": 0.020000000000000018, "rotation": {"subspace_principal_angle": 1.4170648249450182, "mean_class_cosine": 0.15312668462263485}}}, "iid_acc": 0.855, "selectivity": 0.36374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2898741682012294, "mean_class_cosine": 0.27724177646677317}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5373961218836565, "drop": -0.017396121883656468, "rotation": {"subspace_principal_angle": 1.5469965306207487, "mean_class_cosine": 0.995777039241573}}, "domain": {"acc": 0.50125, "drop": 0.018750000000000044, "rotation": {"subspace_principal_angle": 1.4241429560997638, "mean_class_cosine": 0.3829223526550779}}}, "iid_acc": 0.52, "selectivity": 0.028749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9408750735467649, "mean_class_cosine": 0.9869080660276348}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8365650969529086, "drop": 0.04593490304709136}, "domain": {"acc": 0.68125, "drop": 0.20124999999999993}}, "iid_acc": 0.8825, "selectivity": 0.39124999999999993} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5241830065359477, "drop": -0.009183006535947724, "rotation": {"subspace_principal_angle": 1.534491721385843, "mean_class_cosine": 0.03629663087546754}}}, "iid_acc": 0.515, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.5355851741198854, "mean_class_cosine": 0.03520387717988177}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5215686274509804, "drop": -0.006568627450980413, "rotation": {"subspace_principal_angle": 1.4962498880827226, "mean_class_cosine": -0.003140814081629747}}}, "iid_acc": 0.515, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.3928150255004763, "mean_class_cosine": -0.22397488753911812}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.5098039215686274, "drop": -0.0035539215686274384}}, "iid_acc": 0.50625, "selectivity": 0.0012499999999999734} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6625, "drop": 0.0}, "paraphrase": {"acc": 0.6351531291611185, "drop": 0.027346870838881432, "rotation": {"subspace_principal_angle": 1.3792866566009878, "mean_class_cosine": 0.32365198757370595}}}, "iid_acc": 0.6625, "selectivity": 0.41625, "iid_split_rotation": {"subspace_principal_angle": 1.3919845995237112, "mean_class_cosine": 0.34481299862959075}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.30625, "drop": 0.0}, "paraphrase": {"acc": 0.2969374167776298, "drop": 0.009312583222370219, "rotation": {"subspace_principal_angle": 1.1990223569073335, "mean_class_cosine": 0.6137390921956974}}}, "iid_acc": 0.30625, "selectivity": 0.060000000000000026, "iid_split_rotation": {"subspace_principal_angle": 1.4316896651831825, "mean_class_cosine": 0.6279344942694856}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.64375, "drop": 0.0}, "paraphrase": {"acc": 0.6378162450066578, "drop": 0.005933754993342255}}, "iid_acc": 0.64375, "selectivity": 0.3975000000000001} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.7028862478777589, "drop": -0.012886247877758938, "rotation": {"subspace_principal_angle": 1.4427027917966888, "mean_class_cosine": 0.12774353013272075}}}, "iid_acc": 0.69, "selectivity": 0.1612499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3294573165238053, "mean_class_cosine": 0.23900304446924894}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5517826825127334, "drop": -0.0017826825127333912, "rotation": {"subspace_principal_angle": 0.8282909656023721, "mean_class_cosine": 0.9050477286017182}}}, "iid_acc": 0.55, "selectivity": 0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5117065619049865, "mean_class_cosine": 0.9515453007674062}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7623089983022071, "drop": -0.008558998302207077}}, "iid_acc": 0.75375, "selectivity": 0.22499999999999998} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5887096774193549, "drop": 0.015040322580645138, "rotation": {"subspace_principal_angle": 1.5014698349556506, "mean_class_cosine": 0.06927097278538774}}}, "iid_acc": 0.60375, "selectivity": 0.08875, "iid_split_rotation": {"subspace_principal_angle": 1.505561040213484, "mean_class_cosine": 0.0651890267481754}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5419354838709678, "drop": -0.0006854838709677846, "rotation": {"subspace_principal_angle": 1.2893002371450315, "mean_class_cosine": 0.9894024792146818}}}, "iid_acc": 0.54125, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.295158193254292, "mean_class_cosine": 0.9686636169079809}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6675, "drop": 0.0}, "paraphrase": {"acc": 0.6580645161290323, "drop": 0.009435483870967709}}, "iid_acc": 0.6675, "selectivity": 0.15249999999999997} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7053140096618358, "drop": 0.019685990338164205, "rotation": {"subspace_principal_angle": 1.48218154327834, "mean_class_cosine": 0.08849885326614075}}}, "iid_acc": 0.725, "selectivity": 0.18874999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4683469244564828, "mean_class_cosine": 0.10227028026027415}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.5942028985507246, "drop": 0.0007971014492753614, "rotation": {"subspace_principal_angle": 1.4158562736410767, "mean_class_cosine": 0.12335819756367886}}}, "iid_acc": 0.595, "selectivity": 0.05874999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6688260501385498, "mean_class_cosine": 0.15014143544853242}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7568438003220612, "drop": 0.015656199677938742}}, "iid_acc": 0.7725, "selectivity": 0.23624999999999996} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9259259259259259, "drop": 0.016574074074074074, "rotation": {"subspace_principal_angle": 1.157452401071877, "mean_class_cosine": 0.4016738775889428}}}, "iid_acc": 0.9425, "selectivity": 0.42000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.104624499684144, "mean_class_cosine": 0.4494698976663737}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.585, "drop": 0.0}, "paraphrase": {"acc": 0.5624142661179699, "drop": 0.022585733882030112, "rotation": {"subspace_principal_angle": 1.2005636922035423, "mean_class_cosine": 0.8139615746809903}}}, "iid_acc": 0.585, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 0.549247413832069, "mean_class_cosine": 0.30809866212710807}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9272976680384087, "drop": 0.003952331961591282}}, "iid_acc": 0.93125, "selectivity": 0.40875000000000006} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.6623188405797101, "drop": 0.0339311594202899, "rotation": {"subspace_principal_angle": 1.496893033147877, "mean_class_cosine": 0.07383603911862277}}}, "iid_acc": 0.69625, "selectivity": 0.11625000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4214587189180432, "mean_class_cosine": 0.1487831455638894}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.4797101449275362, "drop": 0.03028985507246379, "rotation": {"subspace_principal_angle": 1.1094284833982664, "mean_class_cosine": -0.7834732078378617}}}, "iid_acc": 0.51, "selectivity": -0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2652565110598484, "mean_class_cosine": 0.8601341288733604}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.6420289855072464, "drop": 0.050471014492753596}}, "iid_acc": 0.6925, "selectivity": 0.11250000000000004} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6704545454545454, "drop": -0.02622377622377614, "rotation": {"subspace_principal_angle": 1.4296664682007603, "mean_class_cosine": 0.16574943658385013}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2163461538461539, "iid_split_rotation": {"subspace_principal_angle": 1.551374852609326, "mean_class_cosine": 0.11936510654614986}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.4230769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.4147727272727273, "drop": 0.00830419580419578, "rotation": {"subspace_principal_angle": 1.3518141128300416, "mean_class_cosine": 0.666584729764562}}}, "iid_acc": 0.4230769230769231, "selectivity": -0.004807692307692291, "iid_split_rotation": {"subspace_principal_angle": 1.4575770760056468, "mean_class_cosine": 0.7420250665679694}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6761363636363636, "drop": -0.00786713286713292}}, "iid_acc": 0.6682692307692307, "selectivity": 0.24038461538461536} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8846153846153846, "drop": 0.0066346153846154055, "rotation": {"subspace_principal_angle": 1.4268427703533795, "mean_class_cosine": 0.1434568887054392}}}, "iid_acc": 0.89125, "selectivity": 0.20499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.368747840402915, "mean_class_cosine": 0.20067656560132463}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5826210826210826, "drop": -0.0026210826210826266, "rotation": {"subspace_principal_angle": 0.9151023657985915, "mean_class_cosine": 0.9645849728606082}}}, "iid_acc": 0.58, "selectivity": -0.10625000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.47399007231295404, "mean_class_cosine": 0.9356570924866734}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.9002849002849003, "drop": 0.0009650997150997265}}, "iid_acc": 0.90125, "selectivity": 0.21499999999999997} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8490304709141274, "drop": 0.024719529085872605, "rotation": {"subspace_principal_angle": 1.176281709220393, "mean_class_cosine": 0.384360137185048}}, "domain": {"acc": 0.73625, "drop": 0.13750000000000007, "rotation": {"subspace_principal_angle": 1.1970353549367978, "mean_class_cosine": 0.36511932310217704}}}, "iid_acc": 0.87375, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.0786643668776719, "mean_class_cosine": 0.4725059154778732}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5304709141274239, "drop": -0.004220914127423869, "rotation": {"subspace_principal_angle": 1.3152812819817528, "mean_class_cosine": 0.9992665145477913}}, "domain": {"acc": 0.50125, "drop": 0.025000000000000022, "rotation": {"subspace_principal_angle": 1.308682431264261, "mean_class_cosine": 0.5566033754658513}}}, "iid_acc": 0.52625, "selectivity": 0.04249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.391200206971303, "mean_class_cosine": 0.9964068683023172}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8130193905817175, "drop": 0.018230609418282584}, "domain": {"acc": 0.82875, "drop": 0.0025000000000000577}}, "iid_acc": 0.83125, "selectivity": 0.34750000000000003} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8643410852713178, "drop": 0.023158914728682145, "rotation": {"subspace_principal_angle": 1.1410222217173775, "mean_class_cosine": 0.4166654610472472}}, "domain": {"acc": 0.815, "drop": 0.07250000000000001, "rotation": {"subspace_principal_angle": 1.250009657687724, "mean_class_cosine": 0.3153131973834549}}, "length": {"acc": 0.8878787878787879, "drop": -0.00037878787878797837, "rotation": {"subspace_principal_angle": 1.0751426744599302, "mean_class_cosine": 0.4756067433505382}}}, "iid_acc": 0.8875, "selectivity": 0.42999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.080452581028767, "mean_class_cosine": 0.4709291585448011}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7364341085271318, "drop": -0.0014341085271317722, "rotation": {"subspace_principal_angle": 1.133371334962397, "mean_class_cosine": 0.8868509368956804}}, "domain": {"acc": 0.5575, "drop": 0.1775, "rotation": {"subspace_principal_angle": 1.4962557115172213, "mean_class_cosine": 0.27127763310435205}}, "length": {"acc": 0.7651515151515151, "drop": -0.03015151515151515, "rotation": {"subspace_principal_angle": 1.3646255328037025, "mean_class_cosine": 0.9602447183571129}}}, "iid_acc": 0.735, "selectivity": 0.27749999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.9276681297932009, "mean_class_cosine": 0.8062680171943362}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8798449612403101, "drop": 0.01765503875968988}, "domain": {"acc": 0.69625, "drop": 0.20124999999999993}, "length": {"acc": 0.9, "drop": -0.0025000000000000577}}, "iid_acc": 0.8975, "selectivity": 0.43999999999999995} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.004903846153846203, "rotation": {"subspace_principal_angle": 1.3890007649660276, "mean_class_cosine": 0.284822099506686}}}, "iid_acc": 0.87875, "selectivity": 0.60625, "iid_split_rotation": {"subspace_principal_angle": 1.5702965678160945, "mean_class_cosine": 0.3020514335534847}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.5661538461538461, "drop": -0.009903846153846096, "rotation": {"subspace_principal_angle": 1.553740809284586, "mean_class_cosine": 0.9736297672635466}}}, "iid_acc": 0.55625, "selectivity": 0.28375, "iid_split_rotation": {"subspace_principal_angle": 1.4574124850080719, "mean_class_cosine": 0.9616324023244914}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8861538461538462, "drop": -0.001153846153846172}}, "iid_acc": 0.885, "selectivity": 0.6125} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9476309226932669, "drop": 0.02361907730673307, "rotation": {"subspace_principal_angle": 1.0954764445356109, "mean_class_cosine": 0.7159555553963199}}}, "iid_acc": 0.97125, "selectivity": 0.8775, "iid_split_rotation": {"subspace_principal_angle": 1.0349871771004895, "mean_class_cosine": 0.757327529417333}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.26375, "drop": 0.0}, "paraphrase": {"acc": 0.29177057356608477, "drop": -0.028020573566084783, "rotation": {"subspace_principal_angle": 1.4115196982796947, "mean_class_cosine": 0.8384434499541353}}}, "iid_acc": 0.26375, "selectivity": 0.16999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2413128275371268, "mean_class_cosine": 0.7859751886800755}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.912718204488778, "drop": 0.03603179551122193}}, "iid_acc": 0.94875, "selectivity": 0.855} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.5882352941176471, "drop": 0.02926470588235297, "rotation": {"subspace_principal_angle": 1.360311035504444, "mean_class_cosine": 0.20893450516523}}}, "iid_acc": 0.6175, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4063133781244486, "mean_class_cosine": 0.1637422801353008}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.4925, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.005539215686274523, "rotation": {"subspace_principal_angle": 1.441946064731949, "mean_class_cosine": 0.6865636818700251}}}, "iid_acc": 0.4925, "selectivity": -0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5463324868354524, "mean_class_cosine": -0.9630196496317205}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5581699346405229, "drop": -0.009419934640522931}}, "iid_acc": 0.54875, "selectivity": 0.012499999999999956} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02183089214380829, "rotation": {"subspace_principal_angle": 1.25177621884401, "mean_class_cosine": 0.5310790447646092}}}, "iid_acc": 0.645, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2608266828416124, "mean_class_cosine": 0.45773127447897305}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1825, "drop": 0.0}, "paraphrase": {"acc": 0.17709720372836218, "drop": 0.005402796271637811, "rotation": {"subspace_principal_angle": 1.4219864569031055, "mean_class_cosine": 0.3343326799788812}}}, "iid_acc": 0.1825, "selectivity": -0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4213554573473413, "mean_class_cosine": 0.6639945990861887}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.521970705725699, "drop": 0.00427929427430096}}, "iid_acc": 0.52625, "selectivity": 0.27} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7707979626485568, "drop": -0.01704796264855679, "rotation": {"subspace_principal_angle": 1.3829256711520628, "mean_class_cosine": 0.18676744322680136}}}, "iid_acc": 0.75375, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3348303902625251, "mean_class_cosine": 0.23378226387175424}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.5212224108658744, "drop": -0.00497241086587441, "rotation": {"subspace_principal_angle": 1.478236251419889, "mean_class_cosine": 0.9994667861891016}}}, "iid_acc": 0.51625, "selectivity": -0.0050000000000000044, "iid_split_rotation": {"subspace_principal_angle": 1.3448819076659233, "mean_class_cosine": 0.9996533797319738}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6960950764006791, "drop": 0.008904923599320824}}, "iid_acc": 0.705, "selectivity": 0.18374999999999997} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67625, "drop": 0.0}, "paraphrase": {"acc": 0.632258064516129, "drop": 0.043991935483871036, "rotation": {"subspace_principal_angle": 1.4188129223223702, "mean_class_cosine": 0.15139897022933}}}, "iid_acc": 0.67625, "selectivity": 0.1875, "iid_split_rotation": {"subspace_principal_angle": 1.381661689857851, "mean_class_cosine": 0.18800903417603418}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5338709677419354, "drop": 0.002379032258064573, "rotation": {"subspace_principal_angle": 1.2619016165915666, "mean_class_cosine": 0.9985457827336248}}}, "iid_acc": 0.53625, "selectivity": 0.04749999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2970420520156083, "mean_class_cosine": 0.9948538875393387}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67125, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.035766129032258065}}, "iid_acc": 0.67125, "selectivity": 0.1825} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7681159420289855, "drop": -0.008115942028985468, "rotation": {"subspace_principal_angle": 1.310808643208806, "mean_class_cosine": 0.2570686494136881}}}, "iid_acc": 0.76, "selectivity": 0.15375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3820982126149526, "mean_class_cosine": 0.18758027789963122}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.47125, "drop": 0.0}, "paraphrase": {"acc": 0.49597423510466987, "drop": -0.024724235104669867, "rotation": {"subspace_principal_angle": 1.452719223490081, "mean_class_cosine": 0.2128800431912587}}}, "iid_acc": 0.47125, "selectivity": -0.13499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.9161871593522571, "mean_class_cosine": -0.32128070846271867}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6843800322061192, "drop": 0.010619967793880747}}, "iid_acc": 0.695, "selectivity": 0.08875} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.94, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.011330589849108286, "rotation": {"subspace_principal_angle": 1.137125759529054, "mean_class_cosine": 0.4202044066034193}}}, "iid_acc": 0.94, "selectivity": 0.47624999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.0643181542191005, "mean_class_cosine": 0.4851005702018669}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5253772290809328, "drop": 0.008372770919067185, "rotation": {"subspace_principal_angle": 0.516621762104431, "mean_class_cosine": 0.21743596239076726}}}, "iid_acc": 0.53375, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.8840737623848459, "mean_class_cosine": 0.976674523046531}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9368998628257887, "drop": 0.0006001371742112571}}, "iid_acc": 0.9375, "selectivity": 0.47375} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.6634680250005783, "mean_class_cosine": 0.7878611804971933}}}, "iid_acc": 0.99125, "selectivity": 0.11874999999999991, "iid_split_rotation": {"subspace_principal_angle": 0.6348168891373653, "mean_class_cosine": 0.8051803002223438}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9777015437392796, "drop": -0.006451543739279697, "rotation": {"subspace_principal_angle": 1.3391381172924783, "mean_class_cosine": 0.9776428455730666}}}, "iid_acc": 0.97125, "selectivity": 0.0987499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4227835246448142, "mean_class_cosine": 0.9908198389945848}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": 0.0002915951972555231}}, "iid_acc": 0.99, "selectivity": 0.11749999999999994} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7, "drop": 0.02750000000000008, "rotation": {"subspace_principal_angle": 1.472522322965299, "mean_class_cosine": 0.09811589541574661}}}, "iid_acc": 0.7275, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.3920524003697496, "mean_class_cosine": 0.17779365245715506}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.03989130434782606, "rotation": {"subspace_principal_angle": 1.4319943336187273, "mean_class_cosine": -0.9921843120457909}}}, "iid_acc": 0.5225, "selectivity": -0.08374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.0962721333543264, "mean_class_cosine": 0.9947347244035363}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.717391304347826, "drop": -0.0036413043478260576}}, "iid_acc": 0.71375, "selectivity": 0.10750000000000004} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.6153846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.6136363636363636, "drop": 0.0017482517482517723, "rotation": {"subspace_principal_angle": 1.4739338903736143, "mean_class_cosine": 0.13152150958787132}}}, "iid_acc": 0.6153846153846154, "selectivity": 0.17307692307692313, "iid_split_rotation": {"subspace_principal_angle": 1.45720032842477, "mean_class_cosine": 0.19720918409507146}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.41346153846153844, "drop": 0.0}, "paraphrase": {"acc": 0.4147727272727273, "drop": -0.001311188811188857, "rotation": {"subspace_principal_angle": 1.550847759851561, "mean_class_cosine": 0.7251546533887714}}}, "iid_acc": 0.41346153846153844, "selectivity": -0.028846153846153855, "iid_split_rotation": {"subspace_principal_angle": 0.8940696505827914, "mean_class_cosine": 0.8831178368398586}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": -0.017045454545454586}}, "iid_acc": 0.625, "selectivity": 0.1826923076923077} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.8888888888888888, "drop": 0.0011111111111111738, "rotation": {"subspace_principal_angle": 1.346328125428775, "mean_class_cosine": 0.22258793643339736}}}, "iid_acc": 0.89, "selectivity": 0.135, "iid_split_rotation": {"subspace_principal_angle": 1.2354156575580886, "mean_class_cosine": 0.3291286539962054}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5598290598290598, "drop": 0.007670940170940166, "rotation": {"subspace_principal_angle": 1.089834172363605, "mean_class_cosine": 0.9822532101639228}}}, "iid_acc": 0.5675, "selectivity": -0.1875, "iid_split_rotation": {"subspace_principal_angle": 0.9991972798605081, "mean_class_cosine": 0.8996231510474297}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8532763532763533, "drop": 0.00672364672364667}}, "iid_acc": 0.86, "selectivity": 0.10499999999999998} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7649513212795549, "drop": -0.0074513212795549455, "rotation": {"subspace_principal_angle": 1.260155827346341, "mean_class_cosine": 0.30566854295453516}}, "domain": {"acc": 0.72625, "drop": 0.03125, "rotation": {"subspace_principal_angle": 1.2669695481971335, "mean_class_cosine": 0.2991738951377597}}}, "iid_acc": 0.7575, "selectivity": 0.24874999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3245794968247167, "mean_class_cosine": 0.24373663709592575}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5521557719054242, "drop": -0.017155771905424166, "rotation": {"subspace_principal_angle": 1.4523160206390673, "mean_class_cosine": 0.9818329782722122}}, "domain": {"acc": 0.495, "drop": 0.040000000000000036, "rotation": {"subspace_principal_angle": 1.4451431888735655, "mean_class_cosine": 0.5803352104264222}}}, "iid_acc": 0.535, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3059026276224222, "mean_class_cosine": 0.9628009093282439}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.741307371349096, "drop": -0.008807371349095927}, "domain": {"acc": 0.75125, "drop": -0.018749999999999933}}, "iid_acc": 0.7325, "selectivity": 0.22375} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.8020408163265306, "drop": 0.015459183673469434, "rotation": {"subspace_principal_angle": 0.9261920049968357, "mean_class_cosine": 0.6008822051011515}}, "domain": {"acc": 0.68375, "drop": 0.13375000000000004, "rotation": {"subspace_principal_angle": 1.1716370438461978, "mean_class_cosine": 0.38864385309548416}}, "length": {"acc": 0.8115501519756839, "drop": 0.0059498480243160845, "rotation": {"subspace_principal_angle": 0.8775406092988467, "mean_class_cosine": 0.6390447633833902}}}, "iid_acc": 0.8175, "selectivity": 0.31625000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.957610160630722, "mean_class_cosine": 0.5754760826890111}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6877551020408164, "drop": 0.0034948979591836737, "rotation": {"subspace_principal_angle": 1.2806469825658773, "mean_class_cosine": 0.8910748524092273}}, "domain": {"acc": 0.56875, "drop": 0.12250000000000005, "rotation": {"subspace_principal_angle": 1.438815881455066, "mean_class_cosine": 0.41273195854934175}}, "length": {"acc": 0.6975683890577508, "drop": -0.0063183890577507285, "rotation": {"subspace_principal_angle": 1.5481828698411682, "mean_class_cosine": 0.9192162272707554}}}, "iid_acc": 0.69125, "selectivity": 0.19000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3565470315186383, "mean_class_cosine": 0.8817941693849485}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.78875, "drop": 0.0}, "paraphrase": {"acc": 0.7755102040816326, "drop": 0.013239795918367325}, "domain": {"acc": 0.69125, "drop": 0.09749999999999992}, "length": {"acc": 0.7781155015197568, "drop": 0.010634498480243115}}, "iid_acc": 0.78875, "selectivity": 0.2875} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8454258675078864, "drop": 0.008324132492113612, "rotation": {"subspace_principal_angle": 1.4872713948684544, "mean_class_cosine": 0.5396680342819231}}}, "iid_acc": 0.85375, "selectivity": 0.62375, "iid_split_rotation": {"subspace_principal_angle": 1.1001196196694507, "mean_class_cosine": 0.5629295021677052}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.78, "drop": 0.0}, "paraphrase": {"acc": 0.7744479495268138, "drop": 0.0055520504731861875, "rotation": {"subspace_principal_angle": 1.3957068228612415, "mean_class_cosine": 0.9585242100327607}}}, "iid_acc": 0.78, "selectivity": 0.55, "iid_split_rotation": {"subspace_principal_angle": 1.0534755137152318, "mean_class_cosine": 0.9472616950218877}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8249211356466877, "drop": 0.006328864353312369}}, "iid_acc": 0.83125, "selectivity": 0.6012500000000001} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.9177377892030848, "drop": 0.0022622107969152427, "rotation": {"subspace_principal_angle": 1.5481489232177876, "mean_class_cosine": 0.7602277235959066}}}, "iid_acc": 0.92, "selectivity": 0.8500000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.0913499627869498, "mean_class_cosine": 0.8064226222572987}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.81375, "drop": 0.0}, "paraphrase": {"acc": 0.7583547557840618, "drop": 0.05539524421593822, "rotation": {"subspace_principal_angle": 1.434488039184557, "mean_class_cosine": 0.8699474461477923}}}, "iid_acc": 0.81375, "selectivity": 0.7437499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4670162663468989, "mean_class_cosine": 0.9406629826516945}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8174807197943444, "drop": 0.03626928020565556}}, "iid_acc": 0.85375, "selectivity": 0.78375} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.53875, "drop": 0.0}, "paraphrase": {"acc": 0.5297157622739018, "drop": 0.009034237726098127, "rotation": {"subspace_principal_angle": 1.506602262713435, "mean_class_cosine": 0.06414998384863341}}}, "iid_acc": 0.53875, "selectivity": 0.062499999999999944, "iid_split_rotation": {"subspace_principal_angle": 1.5141486385951506, "mean_class_cosine": 0.056617396360705646}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.47674418604651164, "drop": 0.022005813953488385, "rotation": {"subspace_principal_angle": 1.5018907330856117, "mean_class_cosine": 0.3385630136055}}}, "iid_acc": 0.49875, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5223535908671477, "mean_class_cosine": 0.3337844099038709}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5335917312661499, "drop": 0.017658268733850124}}, "iid_acc": 0.55125, "selectivity": 0.07500000000000001} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5115020297699594, "drop": 0.028497970230040637, "rotation": {"subspace_principal_angle": 1.3837974822380827, "mean_class_cosine": 0.2853919153689364}}}, "iid_acc": 0.54, "selectivity": 0.3175, "iid_split_rotation": {"subspace_principal_angle": 1.4430652239408488, "mean_class_cosine": 0.32302299850041155}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.2075, "drop": 0.0}, "paraphrase": {"acc": 0.21380243572395127, "drop": -0.006302435723951283, "rotation": {"subspace_principal_angle": 1.4121359208224795, "mean_class_cosine": 0.761742118163654}}}, "iid_acc": 0.2075, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4351405486147384, "mean_class_cosine": 0.6192669179592563}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.4438430311231394, "drop": 0.03115696887686059}}, "iid_acc": 0.475, "selectivity": 0.25249999999999995} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7043918918918919, "drop": 0.021858108108108065, "rotation": {"subspace_principal_angle": 1.2557206455227148, "mean_class_cosine": 0.3098884278198152}}}, "iid_acc": 0.72625, "selectivity": 0.20999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2635908058672536, "mean_class_cosine": 0.30239617255824447}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5506756756756757, "drop": -0.02317567567567569, "rotation": {"subspace_principal_angle": 1.563519056112211, "mean_class_cosine": 0.9678685291584537}}}, "iid_acc": 0.5275, "selectivity": 0.011249999999999982, "iid_split_rotation": {"subspace_principal_angle": 1.4948991382902788, "mean_class_cosine": 0.9835336511500095}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7010135135135135, "drop": 0.005236486486486558}}, "iid_acc": 0.70625, "selectivity": 0.19000000000000006} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6715447154471544, "drop": 0.01595528455284556, "rotation": {"subspace_principal_angle": 1.3527346728817384, "mean_class_cosine": 0.21633758729886746}}}, "iid_acc": 0.6875, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.396125798434941, "mean_class_cosine": 0.17378368868184255}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.59125, "drop": 0.0}, "paraphrase": {"acc": 0.583739837398374, "drop": 0.007510162601626069, "rotation": {"subspace_principal_angle": 1.533612671019165, "mean_class_cosine": 0.9008068153198737}}}, "iid_acc": 0.59125, "selectivity": 0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.260460650130829, "mean_class_cosine": 0.890537070105323}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6520325203252032, "drop": 0.007967479674796829}}, "iid_acc": 0.66, "selectivity": 0.19125000000000003} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7079934747145188, "drop": 0.013256525285481158, "rotation": {"subspace_principal_angle": 1.4022375641064082, "mean_class_cosine": 0.16776171230922843}}}, "iid_acc": 0.72125, "selectivity": 0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4265826075864037, "mean_class_cosine": 0.14371435563850823}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5432300163132137, "drop": -0.005730016313213682, "rotation": {"subspace_principal_angle": 1.1709151024773818, "mean_class_cosine": 0.6775048474955148}}}, "iid_acc": 0.5375, "selectivity": -0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1992395749899722, "mean_class_cosine": 0.7778731026109806}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7259380097879282, "drop": 0.0003119902120717333}}, "iid_acc": 0.72625, "selectivity": 0.14749999999999996} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8446866485013624, "drop": 0.041563351498637546, "rotation": {"subspace_principal_angle": 0.9763955114091522, "mean_class_cosine": 0.5600124400720721}}}, "iid_acc": 0.88625, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 0.901229402467838, "mean_class_cosine": 0.620646474717384}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.84, "drop": 0.0}, "paraphrase": {"acc": 0.8079019073569482, "drop": 0.032098092643051745, "rotation": {"subspace_principal_angle": 1.5194268659444015, "mean_class_cosine": 0.9313227017091655}}}, "iid_acc": 0.84, "selectivity": 0.36, "iid_split_rotation": {"subspace_principal_angle": 1.4080821249653852, "mean_class_cosine": 0.9589713457056335}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8337874659400545, "drop": 0.037462534059945485}}, "iid_acc": 0.87125, "selectivity": 0.39125} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9818481848184818, "drop": 0.006901815181518223, "rotation": {"subspace_principal_angle": 0.66528493719011, "mean_class_cosine": 0.7867409300908236}}}, "iid_acc": 0.98875, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6314500555079575, "mean_class_cosine": 0.8071723665064887}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8861386138613861, "drop": 0.0026113861386138915, "rotation": {"subspace_principal_angle": 1.4453777964256416, "mean_class_cosine": 0.9697563041315995}}}, "iid_acc": 0.88875, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9974066490876098, "mean_class_cosine": 0.9768283657309265}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98, "drop": 0.0}, "paraphrase": {"acc": 0.9752475247524752, "drop": 0.004752475247524757}}, "iid_acc": 0.98, "selectivity": 0.11749999999999994} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.703328509406657, "drop": -0.013328509406657085, "rotation": {"subspace_principal_angle": 1.4989704535588646, "mean_class_cosine": 0.07176413141027983}}}, "iid_acc": 0.69, "selectivity": 0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.5357515116941332, "mean_class_cosine": 0.03503764222349777}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.54375, "drop": 0.0}, "paraphrase": {"acc": 0.5470332850940666, "drop": -0.0032832850940666125, "rotation": {"subspace_principal_angle": 1.5166062020639395, "mean_class_cosine": 0.39412823203921477}}}, "iid_acc": 0.54375, "selectivity": -0.14, "iid_split_rotation": {"subspace_principal_angle": 1.5290148158529324, "mean_class_cosine": -0.0372845302255322}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": -0.009247467438494916}}, "iid_acc": 0.71, "selectivity": 0.026249999999999996} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5961538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.6647058823529411, "drop": -0.068552036199095, "rotation": {"subspace_principal_angle": 1.4788344627937355, "mean_class_cosine": 0.29399670434967345}}}, "iid_acc": 0.5961538461538461, "selectivity": 0.15384615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.3726634159821796, "mean_class_cosine": 0.2274413627536632}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5048076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": -0.03636877828054297, "rotation": {"subspace_principal_angle": 0.913621189609579, "mean_class_cosine": 0.6670655608930605}}}, "iid_acc": 0.5048076923076923, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.0837588978590438, "mean_class_cosine": 0.5902856300722443}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6009615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6235294117647059, "drop": -0.02256787330316745}}, "iid_acc": 0.6009615384615384, "selectivity": 0.15865384615384615} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": -0.0021985190409027355, "rotation": {"subspace_principal_angle": 0.9713008818157643, "mean_class_cosine": 0.564225974312156}}}, "iid_acc": 0.89625, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 0.9979017449204138, "mean_class_cosine": 0.5420667359536389}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7362482369534555, "drop": -0.009998236953455586, "rotation": {"subspace_principal_angle": 1.3253351141042944, "mean_class_cosine": 0.8910372103240486}}}, "iid_acc": 0.72625, "selectivity": -0.030000000000000027, "iid_split_rotation": {"subspace_principal_angle": 1.5153708369050465, "mean_class_cosine": 0.9019435549852257}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.785, "drop": 0.0}, "paraphrase": {"acc": 0.7870239774330042, "drop": -0.002023977433004198}}, "iid_acc": 0.785, "selectivity": 0.028750000000000053} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.8, "drop": 0.0}, "paraphrase": {"acc": 0.8052851182197497, "drop": -0.0052851182197496405, "rotation": {"subspace_principal_angle": 1.289669049843915, "mean_class_cosine": 0.2774388483889756}}, "domain": {"acc": 0.71, "drop": 0.09000000000000008, "rotation": {"subspace_principal_angle": 1.2978069882976488, "mean_class_cosine": 0.2696112780665089}}}, "iid_acc": 0.8, "selectivity": 0.2875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2350581884793985, "mean_class_cosine": 0.32946618568105485}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5549374130737135, "drop": 0.005062586926286583, "rotation": {"subspace_principal_angle": 1.3268922292489094, "mean_class_cosine": 0.9883969165833375}}, "domain": {"acc": 0.495, "drop": 0.06500000000000006, "rotation": {"subspace_principal_angle": 1.131095069142722, "mean_class_cosine": 0.7133167140261951}}}, "iid_acc": 0.56, "selectivity": 0.0475000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2611138196752891, "mean_class_cosine": 0.9895181201439}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.815, "drop": 0.0}, "paraphrase": {"acc": 0.780250347705146, "drop": 0.03474965229485394}, "domain": {"acc": 0.76625, "drop": 0.04874999999999996}}, "iid_acc": 0.815, "selectivity": 0.3025} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.0027295918367347127, "rotation": {"subspace_principal_angle": 1.1073395630387064, "mean_class_cosine": 0.4470428901807506}}, "domain": {"acc": 0.71375, "drop": 0.14, "rotation": {"subspace_principal_angle": 1.310897035728456, "mean_class_cosine": 0.2569832264845267}}, "length": {"acc": 0.8556231003039514, "drop": -0.0018731003039513805, "rotation": {"subspace_principal_angle": 1.106355203321873, "mean_class_cosine": 0.44792319554846216}}}, "iid_acc": 0.85375, "selectivity": 0.3375, "iid_split_rotation": {"subspace_principal_angle": 1.1106698792892344, "mean_class_cosine": 0.4440614070193354}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6795918367346939, "drop": 0.002908163265306074, "rotation": {"subspace_principal_angle": 1.0125111213870783, "mean_class_cosine": 0.9329519454684184}}, "domain": {"acc": 0.56625, "drop": 0.11624999999999996, "rotation": {"subspace_principal_angle": 1.1283485572455334, "mean_class_cosine": 0.5142940873907857}}, "length": {"acc": 0.6960486322188449, "drop": -0.013548632218844947, "rotation": {"subspace_principal_angle": 1.3998779208397523, "mean_class_cosine": 0.9308749733569011}}}, "iid_acc": 0.6825, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 0.8902790332464676, "mean_class_cosine": 0.8762482918205301}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.000229591836734655}, "domain": {"acc": 0.7775, "drop": 0.07374999999999998}, "length": {"acc": 0.8541033434650456, "drop": -0.0028533434650456213}}, "iid_acc": 0.85125, "selectivity": 0.33499999999999996} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8470031545741324, "drop": 0.0154968454258676, "rotation": {"subspace_principal_angle": 1.3265156223459154, "mean_class_cosine": 0.357128289219289}}}, "iid_acc": 0.8625, "selectivity": 0.6375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.347641540682651, "mean_class_cosine": 0.38870803168480306}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6876971608832808, "drop": -0.00019716088328081138, "rotation": {"subspace_principal_angle": 1.49331931022075, "mean_class_cosine": 0.9421571273887641}}}, "iid_acc": 0.6875, "selectivity": 0.4625, "iid_split_rotation": {"subspace_principal_angle": 0.8842746339269641, "mean_class_cosine": 0.9544218315516564}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8517350157728707, "drop": 0.010764984227129348}}, "iid_acc": 0.8625, "selectivity": 0.6375000000000001} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9280205655526992, "drop": 0.01447943444730082, "rotation": {"subspace_principal_angle": 1.543953519865052, "mean_class_cosine": 0.6356132585262798}}}, "iid_acc": 0.9425, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.1199023344741124, "mean_class_cosine": 0.6818253733919893}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8791773778920309, "drop": -0.007927377892030907, "rotation": {"subspace_principal_angle": 1.347264998699133, "mean_class_cosine": 0.909582314970496}}}, "iid_acc": 0.87125, "selectivity": 0.7975, "iid_split_rotation": {"subspace_principal_angle": 1.4779951280492787, "mean_class_cosine": 0.9618009806772534}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94375, "drop": 0.0}, "paraphrase": {"acc": 0.922879177377892, "drop": 0.020870822622107932}}, "iid_acc": 0.94375, "selectivity": 0.87} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.49483204134366926, "drop": 0.010167958656330744, "rotation": {"subspace_principal_angle": 1.524398385950912, "mean_class_cosine": -0.046381295294995115}}}, "iid_acc": 0.505, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5107404805094091, "mean_class_cosine": 0.06001975217823446}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.4909560723514212, "drop": 0.015293927648578787, "rotation": {"subspace_principal_angle": 1.5668349405587232, "mean_class_cosine": -0.03841414374121909}}}, "iid_acc": 0.50625, "selectivity": 0.023749999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.5352103349454524, "mean_class_cosine": -0.02088599099775313}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5167958656330749, "drop": 0.0007041343669250688}}, "iid_acc": 0.5175, "selectivity": 0.034999999999999976} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.5764546684709067, "drop": 0.026045331529093385, "rotation": {"subspace_principal_angle": 1.5428367399341, "mean_class_cosine": 0.37101745351040355}}}, "iid_acc": 0.6025, "selectivity": 0.33875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4285270805830252, "mean_class_cosine": 0.3347078246464954}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.21, "drop": 0.0}, "paraphrase": {"acc": 0.19079837618403248, "drop": 0.019201623815967517, "rotation": {"subspace_principal_angle": 1.4456365520979793, "mean_class_cosine": 0.6187678845759252}}}, "iid_acc": 0.21, "selectivity": -0.05374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5049088148460505, "mean_class_cosine": 0.472889652110318}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.496617050067659, "drop": 0.009632949932340984}}, "iid_acc": 0.50625, "selectivity": 0.2425} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7331081081081081, "drop": 0.009391891891891935, "rotation": {"subspace_principal_angle": 1.3068079885037773, "mean_class_cosine": 0.2609327871606552}}}, "iid_acc": 0.7425, "selectivity": 0.23625000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.2782285316613433, "mean_class_cosine": 0.28841185363838734}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5929054054054054, "drop": -0.020405405405405364, "rotation": {"subspace_principal_angle": 1.4497936937308118, "mean_class_cosine": 0.9158504322082752}}}, "iid_acc": 0.5725, "selectivity": 0.06625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3799098692242713, "mean_class_cosine": 0.8877578773800847}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7364864864864865, "drop": -0.0027364864864865}}, "iid_acc": 0.73375, "selectivity": 0.22750000000000004} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.67875, "drop": 0.0}, "paraphrase": {"acc": 0.6764227642276422, "drop": 0.0023272357723577164, "rotation": {"subspace_principal_angle": 1.3819592931432023, "mean_class_cosine": 0.18771672963966035}}}, "iid_acc": 0.67875, "selectivity": 0.19374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.416763883086912, "mean_class_cosine": 0.15342407040448366}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.5788617886178862, "drop": -0.010111788617886197, "rotation": {"subspace_principal_angle": 0.8085931323629197, "mean_class_cosine": 0.9884789144089932}}}, "iid_acc": 0.56875, "selectivity": 0.08374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3601702468050045, "mean_class_cosine": 0.9887573457930535}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.6455284552845528, "drop": 0.03947154471544723}}, "iid_acc": 0.685, "selectivity": 0.20000000000000007} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.732463295269168, "drop": -0.003713295269167971, "rotation": {"subspace_principal_angle": 1.4514696427072735, "mean_class_cosine": 0.11904370635094103}}}, "iid_acc": 0.72875, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4175736807895711, "mean_class_cosine": 0.1526238101323753}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.4893964110929853, "drop": -0.011896411092985337, "rotation": {"subspace_principal_angle": 1.3731868842609583, "mean_class_cosine": 0.8675782704012913}}}, "iid_acc": 0.4775, "selectivity": -0.10625000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4543529756472604, "mean_class_cosine": 0.7758454557886542}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7259380097879282, "drop": 0.006561990212071822}}, "iid_acc": 0.7325, "selectivity": 0.14875000000000005} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.8991825613079019, "drop": 0.019567438692098094, "rotation": {"subspace_principal_angle": 1.17550589332711, "mean_class_cosine": 0.38507624166782645}}}, "iid_acc": 0.91875, "selectivity": 0.41625, "iid_split_rotation": {"subspace_principal_angle": 1.1563275504871824, "mean_class_cosine": 0.40270374204802895}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.6771117166212534, "drop": 0.11038828337874662, "rotation": {"subspace_principal_angle": 0.9740972022042724, "mean_class_cosine": 0.9590361076960369}}}, "iid_acc": 0.7875, "selectivity": 0.28500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4700765532293367, "mean_class_cosine": 0.8827661242087979}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.888283378746594, "drop": 0.012966621253405974}}, "iid_acc": 0.90125, "selectivity": 0.39875000000000005} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9867986798679867, "drop": 0.003201320132013241, "rotation": {"subspace_principal_angle": 0.7435989691909886, "mean_class_cosine": 0.7360370400174403}}}, "iid_acc": 0.99, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.7520555858872627, "mean_class_cosine": 0.7302861569945356}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9422442244224423, "drop": -0.0009942244224422359, "rotation": {"subspace_principal_angle": 1.441143445268482, "mean_class_cosine": 0.9558286303461359}}}, "iid_acc": 0.94125, "selectivity": 0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5613733244607428, "mean_class_cosine": 0.973426291781849}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98125, "drop": 0.0}, "paraphrase": {"acc": 0.9818481848184818, "drop": -0.0005981848184818395}}, "iid_acc": 0.98125, "selectivity": 0.125} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.662807525325615, "drop": 0.023442474674384983, "rotation": {"subspace_principal_angle": 1.5311195648072153, "mean_class_cosine": 0.03966635264677357}}}, "iid_acc": 0.68625, "selectivity": 0.07874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.46679518399481, "mean_class_cosine": 0.10381376065362727}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5151953690303908, "drop": 0.01980463096960927, "rotation": {"subspace_principal_angle": 1.1930287896465346, "mean_class_cosine": 0.911295872781087}}}, "iid_acc": 0.535, "selectivity": -0.07250000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.6904458872144239, "mean_class_cosine": -0.9709984265270024}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": -0.007997467438494832}}, "iid_acc": 0.71125, "selectivity": 0.10375000000000001} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6294117647058823, "drop": -0.004411764705882337, "rotation": {"subspace_principal_angle": 1.5554803313899561, "mean_class_cosine": 0.31267859264684567}}}, "iid_acc": 0.625, "selectivity": 0.1875, "iid_split_rotation": {"subspace_principal_angle": 1.48134181230664, "mean_class_cosine": 0.16283662737976473}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5647058823529412, "drop": -0.01182126696832575, "rotation": {"subspace_principal_angle": 1.0131680189558905, "mean_class_cosine": 0.6680893935892259}}}, "iid_acc": 0.5528846153846154, "selectivity": 0.11538461538461542, "iid_split_rotation": {"subspace_principal_angle": 1.4160182659695404, "mean_class_cosine": 0.5680667839120414}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6346153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.6411764705882353, "drop": -0.0065610859728507664}}, "iid_acc": 0.6346153846153846, "selectivity": 0.19711538461538458} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8899858956276445, "drop": 0.00876410437235553, "rotation": {"subspace_principal_angle": 1.2045865557395712, "mean_class_cosine": 0.3580791088837819}}}, "iid_acc": 0.89875, "selectivity": 0.16000000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2021828264056524, "mean_class_cosine": 0.3603224132960571}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.61625, "drop": 0.0}, "paraphrase": {"acc": 0.5994358251057827, "drop": 0.016814174894217215, "rotation": {"subspace_principal_angle": 1.1806325446572814, "mean_class_cosine": 0.8922635781825321}}}, "iid_acc": 0.61625, "selectivity": -0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3837719103631978, "mean_class_cosine": 0.7905423303610484}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.835, "drop": 0.0}, "paraphrase": {"acc": 0.8349788434414669, "drop": 2.1156558533075298e-05}}, "iid_acc": 0.835, "selectivity": 0.09624999999999995} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8247566063977747, "drop": 0.012743393602225317, "rotation": {"subspace_principal_angle": 1.2798965624456102, "mean_class_cosine": 0.28681430291560384}}, "domain": {"acc": 0.805, "drop": 0.03249999999999997, "rotation": {"subspace_principal_angle": 1.2851613063016167, "mean_class_cosine": 0.28176679938187554}}}, "iid_acc": 0.8375, "selectivity": 0.33625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1765812661272652, "mean_class_cosine": 0.3840835740297509}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.56884561891516, "drop": -0.012595618915159923, "rotation": {"subspace_principal_angle": 1.2901370394335034, "mean_class_cosine": 0.9977925223579588}}, "domain": {"acc": 0.495, "drop": 0.06125000000000003, "rotation": {"subspace_principal_angle": 1.4599392249622891, "mean_class_cosine": 0.6915007566423281}}}, "iid_acc": 0.55625, "selectivity": 0.05500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4453053092946215, "mean_class_cosine": 0.998233597525252}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.8122392211404729, "drop": -0.008489221140472902}, "domain": {"acc": 0.79625, "drop": 0.007499999999999951}}, "iid_acc": 0.80375, "selectivity": 0.3025} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8816326530612245, "drop": 0.0021173469387755217, "rotation": {"subspace_principal_angle": 1.190715066843778, "mean_class_cosine": 0.3709959314311648}}, "domain": {"acc": 0.60625, "drop": 0.2775000000000001, "rotation": {"subspace_principal_angle": 1.4172475398253837, "mean_class_cosine": 0.1529461220229834}}, "length": {"acc": 0.8890577507598785, "drop": -0.005307750759878438, "rotation": {"subspace_principal_angle": 1.2010441268654077, "mean_class_cosine": 0.3613843900828707}}}, "iid_acc": 0.88375, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2079982748541473, "mean_class_cosine": 0.35489153878867186}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.79, "drop": 0.0}, "paraphrase": {"acc": 0.7816326530612245, "drop": 0.0083673469387755, "rotation": {"subspace_principal_angle": 1.3105422369786384, "mean_class_cosine": 0.9556516177420615}}, "domain": {"acc": 0.56625, "drop": 0.22375, "rotation": {"subspace_principal_angle": 1.5060453332017627, "mean_class_cosine": 0.35625753388954995}}, "length": {"acc": 0.8069908814589666, "drop": -0.016990881458966545, "rotation": {"subspace_principal_angle": 1.5564351284094222, "mean_class_cosine": 0.9479993237414271}}}, "iid_acc": 0.79, "selectivity": 0.29500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3575732701756293, "mean_class_cosine": 0.922075109814884}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8877551020408163, "drop": 0.004744897959183647}, "domain": {"acc": 0.6275, "drop": 0.265}, "length": {"acc": 0.8890577507598785, "drop": 0.0034422492401214866}}, "iid_acc": 0.8925, "selectivity": 0.39749999999999996} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8485804416403786, "drop": 0.018919558359621447, "rotation": {"subspace_principal_angle": 1.4399059560351106, "mean_class_cosine": 0.2556332542951133}}}, "iid_acc": 0.8675, "selectivity": 0.64875, "iid_split_rotation": {"subspace_principal_angle": 1.3788930048065602, "mean_class_cosine": 0.2888745735494024}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.7996845425867508, "drop": 0.011565457413249214, "rotation": {"subspace_principal_angle": 1.5548543754884978, "mean_class_cosine": 0.9775272980914157}}}, "iid_acc": 0.81125, "selectivity": 0.5925, "iid_split_rotation": {"subspace_principal_angle": 1.3348767857160242, "mean_class_cosine": 0.979268775558342}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8596214511041009, "drop": 0.020378548895899073}}, "iid_acc": 0.88, "selectivity": 0.66125} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.9525, "drop": 0.0}, "paraphrase": {"acc": 0.9434447300771208, "drop": 0.009055269922879194, "rotation": {"subspace_principal_angle": 1.134461645570175, "mean_class_cosine": 0.6354679842524055}}}, "iid_acc": 0.9525, "selectivity": 0.875, "iid_split_rotation": {"subspace_principal_angle": 1.1124246218092984, "mean_class_cosine": 0.678012477701743}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7712082262210797, "drop": 0.03254177377892031, "rotation": {"subspace_principal_angle": 1.5560406223970316, "mean_class_cosine": 0.8899947966184091}}}, "iid_acc": 0.80375, "selectivity": 0.72625, "iid_split_rotation": {"subspace_principal_angle": 1.503910501910838, "mean_class_cosine": 0.9534530359529768}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9511568123393316, "drop": 0.005093187660668463}}, "iid_acc": 0.95625, "selectivity": 0.87875} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5258397932816538, "drop": 0.015410206718346253, "rotation": {"subspace_principal_angle": 1.496424070388461, "mean_class_cosine": 0.07430371365823527}}}, "iid_acc": 0.54125, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4808569146777435, "mean_class_cosine": 0.08981820636557197}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5025839793281653, "drop": 0.012416020671834671, "rotation": {"subspace_principal_angle": 1.5496948617311233, "mean_class_cosine": 0.9662275396228005}}}, "iid_acc": 0.515, "selectivity": -0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.1782486440521842, "mean_class_cosine": 0.8742419568492337}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5012919896640827, "drop": 0.01870801033591729}}, "iid_acc": 0.52, "selectivity": -0.0012499999999999734} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.6425, "drop": 0.0}, "paraphrase": {"acc": 0.5872801082543978, "drop": 0.055219891745602157, "rotation": {"subspace_principal_angle": 1.36919244552489, "mean_class_cosine": 0.29038444079442377}}}, "iid_acc": 0.6425, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3882673901543217, "mean_class_cosine": 0.28989125551170686}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.32125, "drop": 0.0}, "paraphrase": {"acc": 0.3071718538565629, "drop": 0.014078146143437054, "rotation": {"subspace_principal_angle": 1.3170533049461681, "mean_class_cosine": 0.6594180780518364}}}, "iid_acc": 0.32125, "selectivity": 0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3815192780699195, "mean_class_cosine": 0.5713567606372191}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5453315290933695, "drop": 0.019668470906630486}}, "iid_acc": 0.565, "selectivity": 0.30999999999999994} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7162162162162162, "drop": 0.017533783783783785, "rotation": {"subspace_principal_angle": 1.3990618518124023, "mean_class_cosine": 0.1708915658468793}}}, "iid_acc": 0.73375, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.3739741567595274, "mean_class_cosine": 0.19555384795811886}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.527027027027027, "drop": -0.014527027027027017, "rotation": {"subspace_principal_angle": 1.2081420951650945, "mean_class_cosine": 0.994931982276282}}}, "iid_acc": 0.5125, "selectivity": -0.007500000000000062, "iid_split_rotation": {"subspace_principal_angle": 1.4812482829852658, "mean_class_cosine": 0.9985002258751091}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7297297297297297, "drop": 0.02902027027027032}}, "iid_acc": 0.75875, "selectivity": 0.23875000000000002} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.656910569105691, "drop": 0.02058943089430898, "rotation": {"subspace_principal_angle": 1.4402611653004675, "mean_class_cosine": 0.13016476977575708}}}, "iid_acc": 0.6775, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.4323787501656056, "mean_class_cosine": 0.1379759996513438}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5495934959349593, "drop": -0.0033434959349593063, "rotation": {"subspace_principal_angle": 1.1049873185467238, "mean_class_cosine": 0.9918135119807125}}}, "iid_acc": 0.54625, "selectivity": 0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2358796490004342, "mean_class_cosine": 0.9310981395287434}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6731707317073171, "drop": 0.00932926829268288}}, "iid_acc": 0.6825, "selectivity": 0.17125} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7308319738988581, "drop": 0.011668026101141926, "rotation": {"subspace_principal_angle": 1.4422948230704635, "mean_class_cosine": 0.12814814582491812}}}, "iid_acc": 0.7425, "selectivity": 0.17500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4341137765391143, "mean_class_cosine": 0.13625736099808608}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.46875, "drop": 0.0}, "paraphrase": {"acc": 0.499184339314845, "drop": -0.030434339314845016, "rotation": {"subspace_principal_angle": 0.5863090291944898, "mean_class_cosine": 0.9879298243670935}}}, "iid_acc": 0.46875, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.1523867074914143, "mean_class_cosine": 0.9911941124762024}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7487765089722676, "drop": 0.003723491027732395}}, "iid_acc": 0.7525, "selectivity": 0.18499999999999994} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9209809264305178, "drop": 0.016519073569482234, "rotation": {"subspace_principal_angle": 1.1107561694947945, "mean_class_cosine": 0.44398408965660313}}}, "iid_acc": 0.9375, "selectivity": 0.43125, "iid_split_rotation": {"subspace_principal_angle": 1.134035745754033, "mean_class_cosine": 0.42300636563354255}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5926430517711172, "drop": -0.025143051771117197, "rotation": {"subspace_principal_angle": 0.950772040164655, "mean_class_cosine": 0.9618998317960137}}}, "iid_acc": 0.5675, "selectivity": 0.06125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2078832510638606, "mean_class_cosine": 0.4895900188006161}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9073569482288828, "drop": 0.010143051771117184}}, "iid_acc": 0.9175, "selectivity": 0.41125} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9587458745874587, "drop": 0.0337541254125413, "rotation": {"subspace_principal_angle": 0.9486144857930342, "mean_class_cosine": 0.5828095295275909}}}, "iid_acc": 0.9925, "selectivity": 0.16375000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.9074513638691858, "mean_class_cosine": 0.6157559117610208}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.14026402640264027, "drop": 0.5609859735973598, "rotation": {"subspace_principal_angle": 1.1024334053734877, "mean_class_cosine": 0.5757723487754411}}}, "iid_acc": 0.70125, "selectivity": -0.12749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2565568409948566, "mean_class_cosine": 0.9967803523356993}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.976897689768977, "drop": 0.01685231023102307}}, "iid_acc": 0.99375, "selectivity": 0.16500000000000004} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6613603473227206, "drop": 0.026139652677279357, "rotation": {"subspace_principal_angle": 1.5419209125002344, "mean_class_cosine": 0.0288714017921598}}}, "iid_acc": 0.6875, "selectivity": 0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4675689654827935, "mean_class_cosine": 0.10304412909555002}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5050651230101303, "drop": 0.021184876989869705, "rotation": {"subspace_principal_angle": 1.5385757961576287, "mean_class_cosine": 0.48096021437512015}}}, "iid_acc": 0.52625, "selectivity": -0.05249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1269562907021227, "mean_class_cosine": -0.8913469139771255}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.723589001447178, "drop": -0.006089001447177944}}, "iid_acc": 0.7175, "selectivity": 0.13875000000000004} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": 0.014253393665158431, "rotation": {"subspace_principal_angle": 1.380845212313945, "mean_class_cosine": 0.22774336814303378}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.3028846153846154, "iid_split_rotation": {"subspace_principal_angle": 1.4371716903391298, "mean_class_cosine": 0.16228231461949003}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.47115384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.4411764705882353, "drop": 0.029977375565610864, "rotation": {"subspace_principal_angle": 1.077660750569373, "mean_class_cosine": 0.5488122950260849}}}, "iid_acc": 0.47115384615384615, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.3611080124878545, "mean_class_cosine": 0.7221103135486793}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": -0.014592760180995423}}, "iid_acc": 0.6442307692307693, "selectivity": 0.27403846153846156} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8885754583921015, "drop": 0.006424541607898515, "rotation": {"subspace_principal_angle": 1.1803089373724376, "mean_class_cosine": 0.38063916084124166}}}, "iid_acc": 0.895, "selectivity": 0.16500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.198391794536694, "mean_class_cosine": 0.36385619559330473}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.8208744710860366, "drop": -0.017124471086036652, "rotation": {"subspace_principal_angle": 1.4701060816070168, "mean_class_cosine": 0.8930227036508337}}}, "iid_acc": 0.80375, "selectivity": 0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.1336871573868261, "mean_class_cosine": 0.8902119338649208}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8716502115655853, "drop": 0.005849788434414638}}, "iid_acc": 0.8775, "selectivity": 0.14749999999999996} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8344923504867872, "drop": 0.03300764951321289, "rotation": {"subspace_principal_angle": 1.32725054312888, "mean_class_cosine": 0.24114527926927334}}, "domain": {"acc": 0.855, "drop": 0.012500000000000067, "rotation": {"subspace_principal_angle": 1.4511968441379492, "mean_class_cosine": 0.11931456062079314}}}, "iid_acc": 0.8675, "selectivity": 0.38125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3310106874774308, "mean_class_cosine": 0.23749440429619095}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5827538247566064, "drop": -0.02025382475660642, "rotation": {"subspace_principal_angle": 0.7190626485497094, "mean_class_cosine": 0.9647786618667977}}, "domain": {"acc": 0.495, "drop": 0.0675, "rotation": {"subspace_principal_angle": 1.1078756653311121, "mean_class_cosine": 0.28125598087122977}}}, "iid_acc": 0.5625, "selectivity": 0.07624999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.8988412922378591, "mean_class_cosine": 0.9806205060335059}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.827538247566064, "drop": 0.027461752433936004}, "domain": {"acc": 0.84875, "drop": 0.006249999999999978}}, "iid_acc": 0.855, "selectivity": 0.36874999999999997} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.8959183673469387, "drop": 0.009081632653061278, "rotation": {"subspace_principal_angle": 1.329136111358117, "mean_class_cosine": 0.2393149283948771}}, "domain": {"acc": 0.615, "drop": 0.29000000000000004, "rotation": {"subspace_principal_angle": 1.4430058206254768, "mean_class_cosine": 0.12744297808753202}}, "length": {"acc": 0.9072948328267477, "drop": -0.0022948328267476947, "rotation": {"subspace_principal_angle": 1.3188768354543379, "mean_class_cosine": 0.2492633213424405}}}, "iid_acc": 0.905, "selectivity": 0.43625, "iid_split_rotation": {"subspace_principal_angle": 1.2936213190349724, "mean_class_cosine": 0.2736395752811731}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.84375, "drop": 0.0}, "paraphrase": {"acc": 0.8448979591836735, "drop": -0.001147959183673497, "rotation": {"subspace_principal_angle": 1.2791968807329361, "mean_class_cosine": 0.9478558905818975}}, "domain": {"acc": 0.56625, "drop": 0.27749999999999997, "rotation": {"subspace_principal_angle": 1.5624247261675832, "mean_class_cosine": 0.3303097598518465}}, "length": {"acc": 0.8556231003039514, "drop": -0.01187310030395139, "rotation": {"subspace_principal_angle": 1.3538731412108984, "mean_class_cosine": 0.9594238444634923}}}, "iid_acc": 0.84375, "selectivity": 0.375, "iid_split_rotation": {"subspace_principal_angle": 1.0790921789981887, "mean_class_cosine": 0.9412243630037906}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9142857142857143, "drop": 0.003214285714285725}, "domain": {"acc": 0.8325, "drop": 0.08499999999999996}, "length": {"acc": 0.9133738601823708, "drop": 0.004126139817629215}}, "iid_acc": 0.9175, "selectivity": 0.44875} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8533123028391167, "drop": 0.016687697160883253, "rotation": {"subspace_principal_angle": 1.351979607774418, "mean_class_cosine": 0.3225426038590604}}}, "iid_acc": 0.87, "selectivity": 0.61625, "iid_split_rotation": {"subspace_principal_angle": 1.3508402767430103, "mean_class_cosine": 0.3455994411857771}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.81875, "drop": 0.0}, "paraphrase": {"acc": 0.804416403785489, "drop": 0.014333596214511024, "rotation": {"subspace_principal_angle": 1.2203955464014318, "mean_class_cosine": 0.9625472880536573}}}, "iid_acc": 0.81875, "selectivity": 0.565, "iid_split_rotation": {"subspace_principal_angle": 1.38855296445251, "mean_class_cosine": 0.9564765506858832}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8659305993690851, "drop": 0.017819400630914917}}, "iid_acc": 0.88375, "selectivity": 0.6300000000000001} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.96875, "drop": 0.0}, "paraphrase": {"acc": 0.9640102827763496, "drop": 0.004739717223650408, "rotation": {"subspace_principal_angle": 1.0857448901953828, "mean_class_cosine": 0.6613254867828217}}}, "iid_acc": 0.96875, "selectivity": 0.89375, "iid_split_rotation": {"subspace_principal_angle": 1.076414289934142, "mean_class_cosine": 0.7082374564146615}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6838046272493573, "drop": 0.026195372750642654, "rotation": {"subspace_principal_angle": 1.5350687578200197, "mean_class_cosine": 0.8594207995140349}}}, "iid_acc": 0.71, "selectivity": 0.635, "iid_split_rotation": {"subspace_principal_angle": 1.0895700706545426, "mean_class_cosine": 0.9341302332271046}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9665809768637532, "drop": 0.0009190231362468104}}, "iid_acc": 0.9675, "selectivity": 0.8925000000000001} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.60625, "drop": 0.0}, "paraphrase": {"acc": 0.5930232558139535, "drop": 0.013226744186046413, "rotation": {"subspace_principal_angle": 1.4617197666146766, "mean_class_cosine": 0.10886039552142644}}}, "iid_acc": 0.60625, "selectivity": 0.11499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.466798042788668, "mean_class_cosine": 0.10381091730611129}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": 0.005998062015503902, "rotation": {"subspace_principal_angle": 1.5323338067846937, "mean_class_cosine": 0.9209037611997781}}}, "iid_acc": 0.51375, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.7482139187600273, "mean_class_cosine": 0.7359681788398151}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.57875, "drop": 0.0}, "paraphrase": {"acc": 0.5671834625322998, "drop": 0.0115665374677002}}, "iid_acc": 0.57875, "selectivity": 0.08749999999999997} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.6125, "drop": 0.0}, "paraphrase": {"acc": 0.5926928281461434, "drop": 0.01980717185385661, "rotation": {"subspace_principal_angle": 1.445337558644254, "mean_class_cosine": 0.2135394258196436}}}, "iid_acc": 0.6125, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.42444703173012, "mean_class_cosine": 0.21657072763653504}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.2075, "drop": 0.0}, "paraphrase": {"acc": 0.21109607577807848, "drop": -0.0035960757780784947, "rotation": {"subspace_principal_angle": 1.2059558141673055, "mean_class_cosine": 0.8778793379671459}}}, "iid_acc": 0.2075, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4627224949895972, "mean_class_cosine": 0.5073250273765634}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.6102841677943166, "drop": 0.007215832205683448}}, "iid_acc": 0.6175, "selectivity": 0.395} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7094594594594594, "drop": 0.018040540540540606, "rotation": {"subspace_principal_angle": 1.439408162778464, "mean_class_cosine": 0.13101046781611214}}}, "iid_acc": 0.7275, "selectivity": 0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.440617717267834, "mean_class_cosine": 0.12981124295188987}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5726351351351351, "drop": -0.04263513513513506, "rotation": {"subspace_principal_angle": 1.1956589180448767, "mean_class_cosine": 0.9074769300513319}}}, "iid_acc": 0.53, "selectivity": 0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.4496764494101746, "mean_class_cosine": 0.9668956137286093}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.76875, "drop": 0.0}, "paraphrase": {"acc": 0.7635135135135135, "drop": 0.005236486486486558}}, "iid_acc": 0.76875, "selectivity": 0.26375000000000004} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6634146341463415, "drop": 0.019085365853658498, "rotation": {"subspace_principal_angle": 1.4585106839043807, "mean_class_cosine": 0.11204984079812053}}}, "iid_acc": 0.6825, "selectivity": 0.2025, "iid_split_rotation": {"subspace_principal_angle": 1.4658854601898588, "mean_class_cosine": 0.10471852591428352}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5560975609756098, "drop": -0.004847560975609788, "rotation": {"subspace_principal_angle": 0.7177846273187343, "mean_class_cosine": 0.9945223665026368}}}, "iid_acc": 0.55125, "selectivity": 0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2365217302923708, "mean_class_cosine": 0.9628480684104785}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6731707317073171, "drop": 0.0343292682926829}}, "iid_acc": 0.7075, "selectivity": 0.22750000000000004} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.7210440456769984, "drop": -0.008544045676998402, "rotation": {"subspace_principal_angle": 1.5017342011669776, "mean_class_cosine": 0.06900723919555485}}}, "iid_acc": 0.7125, "selectivity": 0.12375000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4884858123978102, "mean_class_cosine": 0.08221760330215533}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.47125, "drop": 0.0}, "paraphrase": {"acc": 0.5024469820554649, "drop": -0.031196982055464895, "rotation": {"subspace_principal_angle": 0.9746986945864435, "mean_class_cosine": 0.9741912343048922}}}, "iid_acc": 0.47125, "selectivity": -0.1175, "iid_split_rotation": {"subspace_principal_angle": 0.721596565453364, "mean_class_cosine": 0.9782502269601113}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.76625, "drop": 0.0}, "paraphrase": {"acc": 0.7846655791190864, "drop": -0.018415579119086423}}, "iid_acc": 0.76625, "selectivity": 0.1775} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.94375, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.024131471389645776, "rotation": {"subspace_principal_angle": 1.1654703724010276, "mean_class_cosine": 0.39431832489873625}}}, "iid_acc": 0.94375, "selectivity": 0.46499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1557667157993363, "mean_class_cosine": 0.40321702771416507}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.67625, "drop": 0.0}, "paraphrase": {"acc": 0.6743869209809265, "drop": 0.0018630790190735658, "rotation": {"subspace_principal_angle": 1.141242730507284, "mean_class_cosine": 0.9694783192528957}}}, "iid_acc": 0.67625, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 1.4347294745233075, "mean_class_cosine": 0.7429243512254556}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.9318801089918256, "drop": 0.015619891008174402}}, "iid_acc": 0.9475, "selectivity": 0.46875} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9900990099009901, "drop": 0.0024009900990099586, "rotation": {"subspace_principal_angle": 0.9638933203750397, "mean_class_cosine": 0.5703262722196243}}}, "iid_acc": 0.9925, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9412550884957132, "mean_class_cosine": 0.5887740038848737}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.43399339933993397, "drop": 0.223506600660066, "rotation": {"subspace_principal_angle": 1.0823516673301268, "mean_class_cosine": 0.9745159076170016}}}, "iid_acc": 0.6575, "selectivity": -0.13875000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.4071181854905652, "mean_class_cosine": 0.9998877741118155}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9884488448844885, "drop": 0.00655115511551152}}, "iid_acc": 0.995, "selectivity": 0.19874999999999998} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.6772793053545586, "drop": 0.03522069464544142, "rotation": {"subspace_principal_angle": 1.5464207554561848, "mean_class_cosine": 0.02437315754440208}}}, "iid_acc": 0.7125, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 1.4523838978678731, "mean_class_cosine": 0.11813590281951224}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5123010130246021, "drop": 0.016448986975397983, "rotation": {"subspace_principal_angle": 1.5271447554808601, "mean_class_cosine": 0.5274519249488747}}}, "iid_acc": 0.52875, "selectivity": -0.043749999999999956, "iid_split_rotation": {"subspace_principal_angle": 0.7321498396093691, "mean_class_cosine": -0.8691134472727473}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7337192474674384, "drop": 0.007530752532561524}}, "iid_acc": 0.74125, "selectivity": 0.16874999999999996} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6634615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6764705882352942, "drop": -0.013009049773755721, "rotation": {"subspace_principal_angle": 1.4059071468679976, "mean_class_cosine": 0.20250091764866876}}}, "iid_acc": 0.6634615384615384, "selectivity": 0.3125, "iid_split_rotation": {"subspace_principal_angle": 1.4463777868534364, "mean_class_cosine": 0.14772015722986645}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.4951923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.4411764705882353, "drop": 0.05401583710407243, "rotation": {"subspace_principal_angle": 1.155152788254596, "mean_class_cosine": 0.6509669876420695}}}, "iid_acc": 0.4951923076923077, "selectivity": 0.14423076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.3178705376329367, "mean_class_cosine": 0.6023058035884701}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6705882352941176, "drop": -0.026357466063348323}}, "iid_acc": 0.6442307692307693, "selectivity": 0.29326923076923084} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": 0.01030148095909722, "rotation": {"subspace_principal_angle": 1.3664411608119795, "mean_class_cosine": 0.20293578579819216}}}, "iid_acc": 0.90875, "selectivity": 0.2024999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3058033713210209, "mean_class_cosine": 0.2619024695236912}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.7249647390691114, "drop": -0.0012147390691114035, "rotation": {"subspace_principal_angle": 1.3860419944865392, "mean_class_cosine": 0.9342763772250853}}}, "iid_acc": 0.72375, "selectivity": 0.01749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5390421123713445, "mean_class_cosine": 0.8051652349687315}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.9097320169252469, "drop": -0.003482016925246856}}, "iid_acc": 0.90625, "selectivity": 0.19999999999999996} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.78125, "drop": 0.0}, "paraphrase": {"acc": 0.7858136300417247, "drop": -0.0045636300417246645, "rotation": {"subspace_principal_angle": 1.325361507791648, "mean_class_cosine": 0.24297813599017176}}, "domain": {"acc": 0.78, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.4795657935318307, "mean_class_cosine": 0.09110403380754725}}}, "iid_acc": 0.78125, "selectivity": 0.26875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3436786283410562, "mean_class_cosine": 0.22517018028571784}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.55875, "drop": 0.0}, "paraphrase": {"acc": 0.5744089012517385, "drop": -0.015658901251738522, "rotation": {"subspace_principal_angle": 0.6230331668549962, "mean_class_cosine": 0.9910755485040825}}, "domain": {"acc": 0.495, "drop": 0.06374999999999997, "rotation": {"subspace_principal_angle": 1.4491819836983844, "mean_class_cosine": 0.5529294755790912}}}, "iid_acc": 0.55875, "selectivity": 0.04625000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4276392742434922, "mean_class_cosine": 0.9917849416289807}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.7746870653685675, "drop": 0.047812934631432547}, "domain": {"acc": 0.80625, "drop": 0.016249999999999987}}, "iid_acc": 0.8225, "selectivity": 0.31000000000000005} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.8551020408163266, "drop": -0.00010204081632658735, "rotation": {"subspace_principal_angle": 1.3692001923291004, "mean_class_cosine": 0.20023339510507437}}, "domain": {"acc": 0.6, "drop": 0.255, "rotation": {"subspace_principal_angle": 1.468428009163586, "mean_class_cosine": 0.10218962037267862}}, "length": {"acc": 0.8556231003039514, "drop": -0.0006231003039514071, "rotation": {"subspace_principal_angle": 1.3923528233365947, "mean_class_cosine": 0.17749800787298398}}}, "iid_acc": 0.855, "selectivity": 0.37374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3944016978212828, "mean_class_cosine": 0.17548129599139065}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7510204081632653, "drop": 0.0014795918367346284, "rotation": {"subspace_principal_angle": 1.3529173793192326, "mean_class_cosine": 0.9403316707188716}}, "domain": {"acc": 0.56625, "drop": 0.18624999999999992, "rotation": {"subspace_principal_angle": 1.553813034487346, "mean_class_cosine": 0.4365397449834313}}, "length": {"acc": 0.7613981762917933, "drop": -0.008898176291793347, "rotation": {"subspace_principal_angle": 1.484045792925862, "mean_class_cosine": 0.9449458799225073}}}, "iid_acc": 0.7525, "selectivity": 0.27124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.102117298311522, "mean_class_cosine": 0.9232016388495949}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.02272959183673473}, "domain": {"acc": 0.67125, "drop": 0.2025}, "length": {"acc": 0.8723404255319149, "drop": 0.0014095744680850952}}, "iid_acc": 0.87375, "selectivity": 0.3925} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8280757097791798, "drop": 0.024424290220820266, "rotation": {"subspace_principal_angle": 1.4366178069899613, "mean_class_cosine": 0.2843840411694384}}}, "iid_acc": 0.8525, "selectivity": 0.6175, "iid_split_rotation": {"subspace_principal_angle": 1.3589778090630555, "mean_class_cosine": 0.3305635161629984}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5930599369085173, "drop": 0.018190063091482633, "rotation": {"subspace_principal_angle": 0.6501933782850093, "mean_class_cosine": 0.9238960788618794}}}, "iid_acc": 0.61125, "selectivity": 0.37625, "iid_split_rotation": {"subspace_principal_angle": 0.576594648277471, "mean_class_cosine": 0.9553403579368684}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.861198738170347, "drop": 0.028801261829653035}}, "iid_acc": 0.89, "selectivity": 0.655} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9383033419023136, "drop": 0.007946658097686465, "rotation": {"subspace_principal_angle": 1.2309707985530614, "mean_class_cosine": 0.5687671045365533}}}, "iid_acc": 0.94625, "selectivity": 0.85875, "iid_split_rotation": {"subspace_principal_angle": 1.1443944465717892, "mean_class_cosine": 0.6294410202468741}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7095115681233933, "drop": 0.03423843187660669, "rotation": {"subspace_principal_angle": 1.3105209680737453, "mean_class_cosine": 0.8754616020973991}}}, "iid_acc": 0.74375, "selectivity": 0.65625, "iid_split_rotation": {"subspace_principal_angle": 1.3306223102176358, "mean_class_cosine": 0.927772151836221}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94, "drop": 0.0}, "paraphrase": {"acc": 0.9203084832904884, "drop": 0.019691516709511525}}, "iid_acc": 0.94, "selectivity": 0.8524999999999999} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5116279069767442, "drop": 0.019622093023255793, "rotation": {"subspace_principal_angle": 1.4907936041998628, "mean_class_cosine": 0.07991740785625166}}}, "iid_acc": 0.53125, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.494326896717623, "mean_class_cosine": 0.07639492509245417}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": -0.01025193798449614, "rotation": {"subspace_principal_angle": 1.4876221863427175, "mean_class_cosine": 0.8807712318567114}}}, "iid_acc": 0.4975, "selectivity": 0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.2199667575281368, "mean_class_cosine": 0.8996775514646336}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.47375, "drop": 0.0}, "paraphrase": {"acc": 0.5064599483204134, "drop": -0.03270994832041341}}, "iid_acc": 0.47375, "selectivity": -0.00874999999999998} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6275, "drop": 0.0}, "paraphrase": {"acc": 0.5723951285520974, "drop": 0.05510487144790255, "rotation": {"subspace_principal_angle": 1.4343786178569309, "mean_class_cosine": 0.2763953601626957}}}, "iid_acc": 0.6275, "selectivity": 0.37625, "iid_split_rotation": {"subspace_principal_angle": 1.3550101339445768, "mean_class_cosine": 0.31016139289436095}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.24125, "drop": 0.0}, "paraphrase": {"acc": 0.2449255751014885, "drop": -0.0036755751014884985, "rotation": {"subspace_principal_angle": 1.2918205373490343, "mean_class_cosine": 0.7766803935782208}}}, "iid_acc": 0.24125, "selectivity": -0.009999999999999981, "iid_split_rotation": {"subspace_principal_angle": 1.5684032048848013, "mean_class_cosine": 0.5981109246694539}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.60875, "drop": 0.0}, "paraphrase": {"acc": 0.5683355886332883, "drop": 0.040414411366711755}}, "iid_acc": 0.60875, "selectivity": 0.35750000000000004} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7398648648648649, "drop": 0.00013513513513507824, "rotation": {"subspace_principal_angle": 1.2673314142482994, "mean_class_cosine": 0.2988285834755996}}}, "iid_acc": 0.74, "selectivity": 0.24, "iid_split_rotation": {"subspace_principal_angle": 1.296110954087143, "mean_class_cosine": 0.27124411828692846}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.6199324324324325, "drop": -0.01993243243243248, "rotation": {"subspace_principal_angle": 1.5658191148065415, "mean_class_cosine": 0.9653627770010756}}}, "iid_acc": 0.6, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.684621553461914, "mean_class_cosine": 0.9768155693913088}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7246621621621622, "drop": 0.010337837837837816}}, "iid_acc": 0.735, "selectivity": 0.235} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6308943089430894, "drop": 0.014105691056910619, "rotation": {"subspace_principal_angle": 1.4810005396497288, "mean_class_cosine": 0.08967516097444077}}}, "iid_acc": 0.645, "selectivity": 0.16875, "iid_split_rotation": {"subspace_principal_angle": 1.4383531851019704, "mean_class_cosine": 0.13205627952425741}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.551219512195122, "drop": -0.008719512195122014, "rotation": {"subspace_principal_angle": 1.5478870220336094, "mean_class_cosine": 0.9974275968935058}}}, "iid_acc": 0.5425, "selectivity": 0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3106732220928752, "mean_class_cosine": 0.9809909328157596}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6861788617886179, "drop": -0.011178861788617822}}, "iid_acc": 0.675, "selectivity": 0.19875000000000004} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.7225, "drop": 0.0}, "paraphrase": {"acc": 0.7177814029363785, "drop": 0.0047185970636215435, "rotation": {"subspace_principal_angle": 1.3869179562010008, "mean_class_cosine": 0.18284392784898534}}}, "iid_acc": 0.7225, "selectivity": 0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3863158843348609, "mean_class_cosine": 0.18343581678343168}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.46492659053833607, "drop": 0.040073409461663934, "rotation": {"subspace_principal_angle": 1.3638991068226356, "mean_class_cosine": 0.5772026508098606}}}, "iid_acc": 0.505, "selectivity": -0.12, "iid_split_rotation": {"subspace_principal_angle": 1.0426083209720547, "mean_class_cosine": 0.9895063250474988}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7210440456769984, "drop": 0.003955954323001554}}, "iid_acc": 0.725, "selectivity": 0.09999999999999998} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8950953678474114, "drop": 0.013654632152588553, "rotation": {"subspace_principal_angle": 1.1926328016497498, "mean_class_cosine": 0.36921437562940024}}}, "iid_acc": 0.90875, "selectivity": 0.43124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2197817002343454, "mean_class_cosine": 0.34385074329564713}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5994550408719346, "drop": -0.019455040871934615, "rotation": {"subspace_principal_angle": 1.0711301307338577, "mean_class_cosine": 0.9676646451280264}}}, "iid_acc": 0.58, "selectivity": 0.10249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4692738400107856, "mean_class_cosine": 0.5961110196205139}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.9059945504087193, "drop": -0.003494550408719377}}, "iid_acc": 0.9025, "selectivity": 0.425} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9851485148514851, "drop": 0.0073514851485149135, "rotation": {"subspace_principal_angle": 0.9650258351359023, "mean_class_cosine": 0.5693956383770296}}}, "iid_acc": 0.9925, "selectivity": 0.15250000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9566139505713168, "mean_class_cosine": 0.5762905161188008}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.7838283828382838, "drop": -0.10132838283828383, "rotation": {"subspace_principal_angle": 0.7837036206433732, "mean_class_cosine": 0.9616722641195252}}}, "iid_acc": 0.6825, "selectivity": -0.15749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2673965735923087, "mean_class_cosine": 0.9865395355198125}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9875, "drop": 0.0}, "paraphrase": {"acc": 0.9834983498349835, "drop": 0.004001650165016524}}, "iid_acc": 0.9875, "selectivity": 0.14750000000000008} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.6801736613603473, "drop": 0.009826338639652654, "rotation": {"subspace_principal_angle": 1.446957515536241, "mean_class_cosine": 0.12352252083060832}}}, "iid_acc": 0.69, "selectivity": 0.0724999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.482921903323322, "mean_class_cosine": 0.08776137400044447}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5354558610709117, "drop": 0.022044138929088297, "rotation": {"subspace_principal_angle": 1.0513666721812371, "mean_class_cosine": 0.686624560357272}}}, "iid_acc": 0.5575, "selectivity": -0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.6796913400547621, "mean_class_cosine": 0.5471521743827539}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7308248914616498, "drop": -0.01207489146164975}}, "iid_acc": 0.71875, "selectivity": 0.10124999999999995} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": 0.013744343891402777, "rotation": {"subspace_principal_angle": 1.4301231189422874, "mean_class_cosine": 0.21099799551640833}}}, "iid_acc": 0.6490384615384616, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.480503372414954, "mean_class_cosine": 0.14932363173924582}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5288461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": -0.0005656108597285048, "rotation": {"subspace_principal_angle": 1.2751790898102846, "mean_class_cosine": 0.6777353047064866}}}, "iid_acc": 0.5288461538461539, "selectivity": 0.13942307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.437011284720561, "mean_class_cosine": 0.6376580778974325}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": -0.010294117647058787}}, "iid_acc": 0.625, "selectivity": 0.23557692307692307} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8899858956276445, "drop": 0.011264104372355477, "rotation": {"subspace_principal_angle": 1.2097432263622687, "mean_class_cosine": 0.3532596311031677}}}, "iid_acc": 0.90125, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.2597303970258766, "mean_class_cosine": 0.306073583648639}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5077574047954866, "drop": 0.017242595204513433, "rotation": {"subspace_principal_angle": 1.5668153896789387, "mean_class_cosine": 0.821628018438601}}}, "iid_acc": 0.525, "selectivity": -0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.244997779584516, "mean_class_cosine": 0.9282425176629551}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8561354019746121, "drop": 0.01886459802538787}}, "iid_acc": 0.875, "selectivity": 0.12749999999999995} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.7927677329624478, "drop": 0.030982267037552136, "rotation": {"subspace_principal_angle": 1.3441679611646846, "mean_class_cosine": 0.22469338685922752}}, "domain": {"acc": 0.68875, "drop": 0.135, "rotation": {"subspace_principal_angle": 1.4619007929505858, "mean_class_cosine": 0.10868044323365433}}}, "iid_acc": 0.82375, "selectivity": 0.2925, "iid_split_rotation": {"subspace_principal_angle": 1.3382551814356736, "mean_class_cosine": 0.23045101264055012}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.55875, "drop": 0.0}, "paraphrase": {"acc": 0.5744089012517385, "drop": -0.015658901251738522, "rotation": {"subspace_principal_angle": 1.1096712314062809, "mean_class_cosine": 0.9710046862155046}}, "domain": {"acc": 0.495, "drop": 0.06374999999999997, "rotation": {"subspace_principal_angle": 1.3394837230563983, "mean_class_cosine": 0.4067141442483526}}}, "iid_acc": 0.55875, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.0571042364623964, "mean_class_cosine": 0.9732840432471883}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.7955493741307371, "drop": 0.05820062586926289}, "domain": {"acc": 0.8325, "drop": 0.02124999999999999}}, "iid_acc": 0.85375, "selectivity": 0.3225} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5297157622739018, "drop": 0.02403423772609814, "rotation": {"subspace_principal_angle": 1.4849996721834149, "mean_class_cosine": 0.08569143420695843}}}, "iid_acc": 0.55375, "selectivity": 0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4079692233979924, "mean_class_cosine": 0.1621085598267091}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": -0.009001937984496111, "rotation": {"subspace_principal_angle": 1.5312281927446767, "mean_class_cosine": 0.9121731321951317}}}, "iid_acc": 0.49875, "selectivity": -0.013749999999999929, "iid_split_rotation": {"subspace_principal_angle": 1.414487810468904, "mean_class_cosine": 0.9131370420534566}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5129198966408268, "drop": -0.004169896640826787}}, "iid_acc": 0.50875, "selectivity": -0.00374999999999992} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6224627875507442, "drop": 0.03503721244925573, "rotation": {"subspace_principal_angle": 1.377806750699231, "mean_class_cosine": 0.3147008679395653}}}, "iid_acc": 0.6575, "selectivity": 0.40375, "iid_split_rotation": {"subspace_principal_angle": 1.3221669271577434, "mean_class_cosine": 0.32518311136584716}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.27375, "drop": 0.0}, "paraphrase": {"acc": 0.2665764546684709, "drop": 0.007173545331529085, "rotation": {"subspace_principal_angle": 1.3088731708244634, "mean_class_cosine": 0.746477870474282}}}, "iid_acc": 0.27375, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4761219391324176, "mean_class_cosine": 0.6090428280673176}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5737483085250338, "drop": 0.025001691474966226}}, "iid_acc": 0.59875, "selectivity": 0.34500000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.722972972972973, "drop": 0.018277027027026937, "rotation": {"subspace_principal_angle": 1.3556838479482727, "mean_class_cosine": 0.21345731612896818}}}, "iid_acc": 0.74125, "selectivity": 0.24, "iid_split_rotation": {"subspace_principal_angle": 1.3660213343678467, "mean_class_cosine": 0.20334685861759932}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5827702702702703, "drop": 0.00597972972972971, "rotation": {"subspace_principal_angle": 0.7662797422733312, "mean_class_cosine": 0.8372309786422542}}}, "iid_acc": 0.58875, "selectivity": 0.08750000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.19745934756960443, "mean_class_cosine": 0.9910109223557726}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7466216216216216, "drop": -0.007871621621621583}}, "iid_acc": 0.73875, "selectivity": 0.23750000000000004} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.65625, "drop": 0.0}, "paraphrase": {"acc": 0.6113821138211382, "drop": 0.04486788617886184, "rotation": {"subspace_principal_angle": 1.5244430780993927, "mean_class_cosine": 0.046336651197513255}}}, "iid_acc": 0.65625, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 1.48060525921707, "mean_class_cosine": 0.09006884183258959}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5560975609756098, "drop": -0.01609756097560977, "rotation": {"subspace_principal_angle": 1.4447759012617603, "mean_class_cosine": 0.9966733139859822}}}, "iid_acc": 0.54, "selectivity": 0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5053011307485271, "mean_class_cosine": 0.9733537861349781}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.6682926829268293, "drop": 0.024207317073170698}}, "iid_acc": 0.6925, "selectivity": 0.17625000000000002} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7047308319738989, "drop": 0.014019168026101148, "rotation": {"subspace_principal_angle": 1.4273253943610407, "mean_class_cosine": 0.14297923999787693}}}, "iid_acc": 0.71875, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.410754254834671, "mean_class_cosine": 0.15935974106023298}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.49375, "drop": 0.0}, "paraphrase": {"acc": 0.46492659053833607, "drop": 0.02882340946166395, "rotation": {"subspace_principal_angle": 1.2475152228835678, "mean_class_cosine": -0.2873781232568553}}}, "iid_acc": 0.49375, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 0.3595110102120592, "mean_class_cosine": 0.9955684555352899}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7373572593800979, "drop": 0.005142740619902164}}, "iid_acc": 0.7425, "selectivity": 0.15000000000000002} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.01163147138964582, "rotation": {"subspace_principal_angle": 1.1462155217150516, "mean_class_cosine": 0.4119388427182844}}}, "iid_acc": 0.93125, "selectivity": 0.4525, "iid_split_rotation": {"subspace_principal_angle": 1.2036153547134156, "mean_class_cosine": 0.3589857417663745}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.57375, "drop": 0.0}, "paraphrase": {"acc": 0.5885558583106267, "drop": -0.014805858310626752, "rotation": {"subspace_principal_angle": 0.6949506609048397, "mean_class_cosine": 0.9687824706589729}}}, "iid_acc": 0.57375, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4852957312103852, "mean_class_cosine": 0.7009380316990886}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.9155313351498637, "drop": -0.00428133514986373}}, "iid_acc": 0.91125, "selectivity": 0.4325} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6555716353111433, "drop": 0.0544283646888567, "rotation": {"subspace_principal_angle": 1.530036517597367, "mean_class_cosine": 0.04074852400167777}}}, "iid_acc": 0.71, "selectivity": 0.12249999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4332650590425584, "mean_class_cosine": 0.1370981137331397}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5180897250361794, "drop": 0.008160274963820546, "rotation": {"subspace_principal_angle": 1.3868122762825261, "mean_class_cosine": 0.5674558871637044}}}, "iid_acc": 0.52625, "selectivity": -0.06125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3698230368243622, "mean_class_cosine": -0.902684531716938}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": 0.03450253256150515}}, "iid_acc": 0.75375, "selectivity": 0.16625} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": -0.004977375565610842, "rotation": {"subspace_principal_angle": 1.4918296501472632, "mean_class_cosine": 0.14481202131126455}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.2548076923076923, "iid_split_rotation": {"subspace_principal_angle": 1.4905675320918412, "mean_class_cosine": 0.12376594768867437}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.5576923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.5235294117647059, "drop": 0.0341628959276018, "rotation": {"subspace_principal_angle": 1.1976344371529, "mean_class_cosine": 0.7296312684418567}}}, "iid_acc": 0.5576923076923077, "selectivity": 0.15865384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.5298301184380643, "mean_class_cosine": 0.5946390438469394}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.6298076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": -0.005486425339366496}}, "iid_acc": 0.6298076923076923, "selectivity": 0.23076923076923073} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.9012693935119888, "drop": -0.0012693935119887367, "rotation": {"subspace_principal_angle": 1.3718118378240132, "mean_class_cosine": 0.1976739601069373}}}, "iid_acc": 0.9, "selectivity": 0.21999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.320617301683332, "mean_class_cosine": 0.24757741494334223}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7475317348377997, "drop": -0.015031734837799626, "rotation": {"subspace_principal_angle": 0.7004143627777456, "mean_class_cosine": 0.8791149950634916}}}, "iid_acc": 0.7325, "selectivity": 0.05249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.298216893295834, "mean_class_cosine": 0.687969391669262}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.9055007052186178, "drop": -0.0005007052186177807}}, "iid_acc": 0.905, "selectivity": 0.22499999999999998} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8150208623087621, "drop": 0.031229137691237807, "rotation": {"subspace_principal_angle": 1.2085181784579044, "mean_class_cosine": 0.3544054290791189}}, "domain": {"acc": 0.84375, "drop": 0.0024999999999999467, "rotation": {"subspace_principal_angle": 1.2026486146003328, "mean_class_cosine": 0.35988787403384076}}}, "iid_acc": 0.84625, "selectivity": 0.3499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1571656029830548, "mean_class_cosine": 0.40193650581631374}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5660639777468707, "drop": -0.014813977746870655, "rotation": {"subspace_principal_angle": 1.245440017394346, "mean_class_cosine": 0.9990012115847782}}, "domain": {"acc": 0.495, "drop": 0.05625000000000002, "rotation": {"subspace_principal_angle": 1.122537065823632, "mean_class_cosine": 0.8260255526449155}}}, "iid_acc": 0.55125, "selectivity": 0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.503465091230446, "mean_class_cosine": 0.9992316721985006}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.79375, "drop": 0.0}, "paraphrase": {"acc": 0.8108484005563282, "drop": -0.017098400556328275}, "domain": {"acc": 0.81875, "drop": -0.025000000000000022}}, "iid_acc": 0.79375, "selectivity": 0.29749999999999993} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.889795918367347, "drop": 0.0089540816326531, "rotation": {"subspace_principal_angle": 1.1767967051679535, "mean_class_cosine": 0.38388465061115085}}, "domain": {"acc": 0.62875, "drop": 0.27, "rotation": {"subspace_principal_angle": 1.383856999934436, "mean_class_cosine": 0.18585242109581535}}, "length": {"acc": 0.9057750759878419, "drop": -0.007025075987841856, "rotation": {"subspace_principal_angle": 1.1410061611227862, "mean_class_cosine": 0.41668006103594923}}}, "iid_acc": 0.89875, "selectivity": 0.37875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2153568944914745, "mean_class_cosine": 0.3480023638189218}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.67875, "drop": 0.0}, "paraphrase": {"acc": 0.6959183673469388, "drop": -0.01716836734693883, "rotation": {"subspace_principal_angle": 1.4280844473025907, "mean_class_cosine": 0.9628713037828026}}, "domain": {"acc": 0.56625, "drop": 0.11249999999999993, "rotation": {"subspace_principal_angle": 1.3603869108648967, "mean_class_cosine": 0.44643897304207514}}, "length": {"acc": 0.6990881458966566, "drop": -0.020338145896656612, "rotation": {"subspace_principal_angle": 1.3643554774870585, "mean_class_cosine": 0.9296532611805729}}}, "iid_acc": 0.67875, "selectivity": 0.15874999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4371116325199274, "mean_class_cosine": 0.8728452884225864}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8959183673469387, "drop": 0.012831632653061198}, "domain": {"acc": 0.7025, "drop": 0.20624999999999993}, "length": {"acc": 0.9103343465045592, "drop": -0.0015843465045592975}}, "iid_acc": 0.90875, "selectivity": 0.38874999999999993} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8690851735015773, "drop": 0.013414826498422627, "rotation": {"subspace_principal_angle": 1.3839824797303943, "mean_class_cosine": 0.31779551037707243}}}, "iid_acc": 0.8825, "selectivity": 0.6425, "iid_split_rotation": {"subspace_principal_angle": 1.3316581856764766, "mean_class_cosine": 0.3540098873003071}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.64625, "drop": 0.0}, "paraphrase": {"acc": 0.6403785488958991, "drop": 0.005871451104100922, "rotation": {"subspace_principal_angle": 0.9413392447354468, "mean_class_cosine": 0.9542917678949543}}}, "iid_acc": 0.64625, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 1.3936985587921362, "mean_class_cosine": 0.957153329410988}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8706624605678234, "drop": 0.014337539432176638}}, "iid_acc": 0.885, "selectivity": 0.645} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9588688946015425, "drop": -0.00011889460154246301, "rotation": {"subspace_principal_angle": 1.1552531570580125, "mean_class_cosine": 0.6468507558692805}}}, "iid_acc": 0.95875, "selectivity": 0.8825, "iid_split_rotation": {"subspace_principal_angle": 1.135852999209662, "mean_class_cosine": 0.6934419744150427}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.6349614395886889, "drop": 0.055038560411311055, "rotation": {"subspace_principal_angle": 0.9810281506834297, "mean_class_cosine": 0.8971151689162156}}}, "iid_acc": 0.69, "selectivity": 0.6137499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9002976361653219, "mean_class_cosine": 0.9196584919183077}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9588688946015425, "drop": -0.00011889460154246301}}, "iid_acc": 0.95875, "selectivity": 0.8825} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.6253229974160207, "drop": -0.01782299741602067, "rotation": {"subspace_principal_angle": 1.2853027316760957, "mean_class_cosine": 0.28163110133345426}}}, "iid_acc": 0.6075, "selectivity": 0.12625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3656019976170561, "mean_class_cosine": 0.20375741617602003}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.50375, "drop": 0.0}, "paraphrase": {"acc": 0.5258397932816538, "drop": -0.022089793281653725, "rotation": {"subspace_principal_angle": 1.415941389813765, "mean_class_cosine": 0.989806795103265}}}, "iid_acc": 0.50375, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.1847539410134869, "mean_class_cosine": 0.9708123793493897}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5658914728682171, "drop": -0.008391472868217065}}, "iid_acc": 0.5575, "selectivity": 0.07624999999999998} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5520974289580515, "drop": 0.04665257104194853, "rotation": {"subspace_principal_angle": 1.2698836003378577, "mean_class_cosine": 0.41354171313299243}}}, "iid_acc": 0.59875, "selectivity": 0.36125, "iid_split_rotation": {"subspace_principal_angle": 1.279435072679175, "mean_class_cosine": 0.4311997525753071}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.1775, "drop": 0.0}, "paraphrase": {"acc": 0.18132611637347767, "drop": -0.0038261163734776837, "rotation": {"subspace_principal_angle": 1.2291192051808308, "mean_class_cosine": 0.9975774044324508}}}, "iid_acc": 0.1775, "selectivity": -0.06, "iid_split_rotation": {"subspace_principal_angle": 1.4905226932759024, "mean_class_cosine": 0.6650981545648424}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5128552097428958, "drop": 0.013394790257104217}}, "iid_acc": 0.52625, "selectivity": 0.28875} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7398648648648649, "drop": 0.0038851351351351093, "rotation": {"subspace_principal_angle": 1.2773857388810559, "mean_class_cosine": 0.289218730636968}}}, "iid_acc": 0.74375, "selectivity": 0.23625000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.283587337214898, "mean_class_cosine": 0.2832766461897106}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5202702702702703, "drop": -0.012770270270270334, "rotation": {"subspace_principal_angle": 1.416822487851161, "mean_class_cosine": 0.9990366010707538}}}, "iid_acc": 0.5075, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.172452398513104, "mean_class_cosine": 0.9997365426935431}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7212837837837838, "drop": 0.009966216216216184}}, "iid_acc": 0.73125, "selectivity": 0.22375} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6260162601626016, "drop": 0.04773373983739837, "rotation": {"subspace_principal_angle": 1.4713221046099612, "mean_class_cosine": 0.09931025175584188}}}, "iid_acc": 0.67375, "selectivity": 0.18624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4688407807398294, "mean_class_cosine": 0.10177900098741591}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5495934959349593, "drop": -0.007093495934959337, "rotation": {"subspace_principal_angle": 1.1126001122879392, "mean_class_cosine": 0.9966307622920645}}}, "iid_acc": 0.5425, "selectivity": 0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2227358373040218, "mean_class_cosine": 0.9830887042288201}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6325203252032521, "drop": 0.044979674796747915}}, "iid_acc": 0.6775, "selectivity": 0.19} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.734094616639478, "drop": 0.0009053833605220385, "rotation": {"subspace_principal_angle": 1.3384619923865202, "mean_class_cosine": 0.23024976330131475}}}, "iid_acc": 0.735, "selectivity": 0.14874999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3807433555450965, "mean_class_cosine": 0.1889109127031547}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.46375, "drop": 0.0}, "paraphrase": {"acc": 0.4926590538336052, "drop": -0.028909053833605203, "rotation": {"subspace_principal_angle": 1.240506779082087, "mean_class_cosine": 0.9979825325820296}}}, "iid_acc": 0.46375, "selectivity": -0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4130621801133338, "mean_class_cosine": 0.9984335729440443}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.755, "drop": 0.0}, "paraphrase": {"acc": 0.7504078303425775, "drop": 0.004592169657422485}}, "iid_acc": 0.755, "selectivity": 0.16874999999999996} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.93, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.010381471389645847, "rotation": {"subspace_principal_angle": 1.1842664562934508, "mean_class_cosine": 0.37697657827864284}}}, "iid_acc": 0.93, "selectivity": 0.42125, "iid_split_rotation": {"subspace_principal_angle": 1.1417564076798083, "mean_class_cosine": 0.4159979297955525}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5722070844686649, "drop": -0.014707084468664866, "rotation": {"subspace_principal_angle": 0.9890820205173421, "mean_class_cosine": 0.9786448762488624}}}, "iid_acc": 0.5575, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4310370860587216, "mean_class_cosine": 0.6707511189916638}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9291553133514986, "drop": 0.0020946866485014315}}, "iid_acc": 0.93125, "selectivity": 0.4225} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9900990099009901, "drop": 0.0024009900990099586, "rotation": {"subspace_principal_angle": 0.9546523855786807, "mean_class_cosine": 0.5778924853874827}}}, "iid_acc": 0.9925, "selectivity": 0.15500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.9113621026036162, "mean_class_cosine": 0.6126697953735774}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.6815181518151815, "drop": -0.017768151815181543, "rotation": {"subspace_principal_angle": 0.7769812943933525, "mean_class_cosine": 0.9999668724105906}}}, "iid_acc": 0.66375, "selectivity": -0.17375000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.4808794246553239, "mean_class_cosine": 0.999987060235844}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9867986798679867, "drop": 0.005701320132013299}}, "iid_acc": 0.9925, "selectivity": 0.15500000000000003} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.6975397973950795, "drop": 0.047460202604920454, "rotation": {"subspace_principal_angle": 1.4955638990865465, "mean_class_cosine": 0.07516147955849134}}}, "iid_acc": 0.745, "selectivity": 0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.378258975516904, "mean_class_cosine": 0.19134997405306095}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.53875, "drop": 0.0}, "paraphrase": {"acc": 0.5340086830680174, "drop": 0.004741316931982542, "rotation": {"subspace_principal_angle": 1.021497555457415, "mean_class_cosine": 0.9732173089300729}}}, "iid_acc": 0.53875, "selectivity": -0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5200003895350536, "mean_class_cosine": -0.9852001616272563}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7264833574529667, "drop": -0.016483357452966696}}, "iid_acc": 0.71, "selectivity": 0.08624999999999994} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6634615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6764705882352942, "drop": -0.013009049773755721, "rotation": {"subspace_principal_angle": 1.411172607075842, "mean_class_cosine": 0.18933079531066535}}}, "iid_acc": 0.6634615384615384, "selectivity": 0.2740384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.4720863809909137, "mean_class_cosine": 0.12696001888551056}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.5764705882352941, "drop": 0.0052601809954752055, "rotation": {"subspace_principal_angle": 1.0998469877442718, "mean_class_cosine": 0.6836666330824704}}}, "iid_acc": 0.5817307692307693, "selectivity": 0.19230769230769235, "iid_split_rotation": {"subspace_principal_angle": 1.540737942845596, "mean_class_cosine": 0.6936262779934396}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6470588235294118, "drop": 0.006787330316742057}}, "iid_acc": 0.6538461538461539, "selectivity": 0.2644230769230769} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": -0.0009485190409027622, "rotation": {"subspace_principal_angle": 1.2279252167452932, "mean_class_cosine": 0.3361924663049618}}}, "iid_acc": 0.8975, "selectivity": 0.15499999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.2122444869762357, "mean_class_cosine": 0.3509185358458786}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.6234132581100141, "drop": -0.005913258110014086, "rotation": {"subspace_principal_angle": 0.8729067617905422, "mean_class_cosine": 0.7807952771130344}}}, "iid_acc": 0.6175, "selectivity": -0.125, "iid_split_rotation": {"subspace_principal_angle": 0.5477116558377627, "mean_class_cosine": 0.21905717301122568}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.9026798307475318, "drop": 0.003570169252468225}}, "iid_acc": 0.90625, "selectivity": 0.16374999999999995} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7503467406380028, "drop": -0.007846740638002725, "rotation": {"subspace_principal_angle": 1.1359120026803335, "mean_class_cosine": 0.4213054955523352}}, "domain": {"acc": 0.75, "drop": -0.007499999999999951, "rotation": {"subspace_principal_angle": 1.2165320665064534, "mean_class_cosine": 0.34690040746322265}}}, "iid_acc": 0.7425, "selectivity": 0.18375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1365226365220666, "mean_class_cosine": 0.4207516217391289}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6324549237170597, "drop": -0.009954923717059638, "rotation": {"subspace_principal_angle": 1.4631222901212197, "mean_class_cosine": 0.8231564279245471}}, "domain": {"acc": 0.5025, "drop": 0.1200000000000001, "rotation": {"subspace_principal_angle": 1.3032516443365474, "mean_class_cosine": 0.7093055206790797}}}, "iid_acc": 0.6225, "selectivity": 0.06375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.0549484337600161, "mean_class_cosine": 0.8394584109810561}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.7475, "drop": 0.0}, "paraphrase": {"acc": 0.7517337031900139, "drop": -0.0042337031900138156}, "domain": {"acc": 0.7575, "drop": -0.009999999999999898}}, "iid_acc": 0.7475, "selectivity": 0.18875000000000008} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.7747572815533981, "drop": 0.02774271844660192, "rotation": {"subspace_principal_angle": 1.0704211059136206, "mean_class_cosine": 0.4797547921494105}}, "domain": {"acc": 0.5775, "drop": 0.22499999999999998, "rotation": {"subspace_principal_angle": 1.3307188056276382, "mean_class_cosine": 0.23777792496643144}}, "length": {"acc": 0.8054298642533937, "drop": -0.002929864253393699, "rotation": {"subspace_principal_angle": 1.0251786604711886, "mean_class_cosine": 0.5189461749096541}}}, "iid_acc": 0.8025, "selectivity": 0.26, "iid_split_rotation": {"subspace_principal_angle": 1.1020990971387528, "mean_class_cosine": 0.4517243926605984}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.76125, "drop": 0.0}, "paraphrase": {"acc": 0.7475728155339806, "drop": 0.013677184466019399, "rotation": {"subspace_principal_angle": 1.3667128970253954, "mean_class_cosine": 0.8708735415100642}}, "domain": {"acc": 0.6175, "drop": 0.14374999999999993, "rotation": {"subspace_principal_angle": 1.3448216837274964, "mean_class_cosine": 0.549993690079513}}, "length": {"acc": 0.751131221719457, "drop": 0.010118778280542973, "rotation": {"subspace_principal_angle": 1.5333491169148545, "mean_class_cosine": 0.8840723107383672}}}, "iid_acc": 0.76125, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.4498931415198635, "mean_class_cosine": 0.8774451765117859}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.79, "drop": 0.0}, "paraphrase": {"acc": 0.7708737864077669, "drop": 0.0191262135922331}, "domain": {"acc": 0.6, "drop": 0.19000000000000006}, "length": {"acc": 0.799396681749623, "drop": -0.009396681749622937}}, "iid_acc": 0.79, "selectivity": 0.24750000000000005} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8440944881889764, "drop": 0.014655511811023647, "rotation": {"subspace_principal_angle": 1.420395850894294, "mean_class_cosine": 0.3177960201739077}}}, "iid_acc": 0.85875, "selectivity": 0.6012500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.457737166249678, "mean_class_cosine": 0.2964930332276186}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.83, "drop": 0.0}, "paraphrase": {"acc": 0.8125984251968504, "drop": 0.017401574803149522, "rotation": {"subspace_principal_angle": 1.4786133211134285, "mean_class_cosine": 0.9691458832694383}}}, "iid_acc": 0.83, "selectivity": 0.5725, "iid_split_rotation": {"subspace_principal_angle": 1.5646208839230873, "mean_class_cosine": 0.9660745167896441}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.84251968503937, "drop": 0.02748031496062997}}, "iid_acc": 0.87, "selectivity": 0.6125} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9193954659949622, "drop": 0.019354534005037727, "rotation": {"subspace_principal_angle": 1.2541440798612775, "mean_class_cosine": 0.6719234888555385}}}, "iid_acc": 0.93875, "selectivity": 0.8875, "iid_split_rotation": {"subspace_principal_angle": 1.3380418703494907, "mean_class_cosine": 0.7071156351962342}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.84125, "drop": 0.0}, "paraphrase": {"acc": 0.8312342569269522, "drop": 0.010015743073047867, "rotation": {"subspace_principal_angle": 1.4201631598283133, "mean_class_cosine": 0.8910464593425866}}}, "iid_acc": 0.84125, "selectivity": 0.79, "iid_split_rotation": {"subspace_principal_angle": 1.5685103886275429, "mean_class_cosine": 0.9592733462262945}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.9017632241813602, "drop": 0.01698677581863972}}, "iid_acc": 0.91875, "selectivity": 0.8674999999999999} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5058670143415906, "drop": 0.012882985658409396, "rotation": {"subspace_principal_angle": 1.4825642540594317, "mean_class_cosine": 0.08811763766369723}}}, "iid_acc": 0.51875, "selectivity": 0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.514014896298524, "mean_class_cosine": -0.056750923621074825}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.49375, "drop": 0.0}, "paraphrase": {"acc": 0.4980443285528031, "drop": -0.004294328552803095, "rotation": {"subspace_principal_angle": 1.5186826580706572, "mean_class_cosine": -0.08658206386049154}}}, "iid_acc": 0.49375, "selectivity": -0.0012499999999999734, "iid_split_rotation": {"subspace_principal_angle": 1.5356626656582177, "mean_class_cosine": 0.14967252329717184}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.516297262059974, "drop": -0.001297262059973936}}, "iid_acc": 0.515, "selectivity": 0.020000000000000018} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5391766268260292, "drop": 0.02707337317397085, "rotation": {"subspace_principal_angle": 1.3218665508999698, "mean_class_cosine": 0.40140129389793905}}}, "iid_acc": 0.56625, "selectivity": 0.30500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4544563315042423, "mean_class_cosine": 0.3616549205752137}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.2425, "drop": 0.0}, "paraphrase": {"acc": 0.1952191235059761, "drop": 0.047280876494023905, "rotation": {"subspace_principal_angle": 1.5288782648781811, "mean_class_cosine": 0.49201628880852194}}}, "iid_acc": 0.2425, "selectivity": -0.01874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4051891188356136, "mean_class_cosine": 0.5694708832697903}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.4675, "drop": 0.0}, "paraphrase": {"acc": 0.44754316069057104, "drop": 0.019956839309428986}}, "iid_acc": 0.4675, "selectivity": 0.20625000000000004} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7249190938511327, "drop": 0.001330906148867217, "rotation": {"subspace_principal_angle": 1.1511849070923794, "mean_class_cosine": 0.40740561391162183}}}, "iid_acc": 0.72625, "selectivity": 0.16374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1756361449389974, "mean_class_cosine": 0.3849560311735334}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.63915857605178, "drop": -0.03165857605177991, "rotation": {"subspace_principal_angle": 0.8931518264214847, "mean_class_cosine": 0.793022261839432}}}, "iid_acc": 0.6075, "selectivity": 0.04500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.0684485107486943, "mean_class_cosine": 0.6944069959278446}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.7313915857605178, "drop": -0.015141585760517717}}, "iid_acc": 0.71625, "selectivity": 0.15375000000000005} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6494345718901454, "drop": 0.020565428109854622, "rotation": {"subspace_principal_angle": 1.326914373237704, "mean_class_cosine": 0.2414715148282628}}}, "iid_acc": 0.67, "selectivity": 0.19250000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.297386449224978, "mean_class_cosine": 0.27001622040531514}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.6187399030694669, "drop": 0.007510096930533061, "rotation": {"subspace_principal_angle": 1.1476051561016243, "mean_class_cosine": 0.909033474781661}}}, "iid_acc": 0.62625, "selectivity": 0.14875, "iid_split_rotation": {"subspace_principal_angle": 1.4081586003059547, "mean_class_cosine": 0.8941316972733946}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6625, "drop": 0.0}, "paraphrase": {"acc": 0.6235864297253635, "drop": 0.03891357027463649}}, "iid_acc": 0.6625, "selectivity": 0.185} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7405582922824302, "drop": -0.011808292282430188, "rotation": {"subspace_principal_angle": 1.1923774338443787, "mean_class_cosine": 0.36945168819099083}}}, "iid_acc": 0.72875, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2706314343701106, "mean_class_cosine": 0.29567773039327916}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.54375, "drop": 0.0}, "paraphrase": {"acc": 0.5500821018062397, "drop": -0.006332101806239776, "rotation": {"subspace_principal_angle": 1.539326603180651, "mean_class_cosine": 0.8691103147755386}}}, "iid_acc": 0.54375, "selectivity": -0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3002024261544294, "mean_class_cosine": 0.7081293912566515}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7274220032840722, "drop": -0.00867200328407225}}, "iid_acc": 0.71875, "selectivity": 0.12375000000000003} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8363384188626907, "drop": 0.05491158113730932, "rotation": {"subspace_principal_angle": 0.9678914596829231, "mean_class_cosine": 0.5670375780110272}}}, "iid_acc": 0.89125, "selectivity": 0.435, "iid_split_rotation": {"subspace_principal_angle": 0.9419088602911864, "mean_class_cosine": 0.5882454358575304}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.02100728155339804, "rotation": {"subspace_principal_angle": 1.2291895890809326, "mean_class_cosine": 0.9098769842554898}}}, "iid_acc": 0.84625, "selectivity": 0.38999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3955051971962031, "mean_class_cosine": 0.9616189776316897}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8626907073509015, "drop": 0.016059292649098555}}, "iid_acc": 0.87875, "selectivity": 0.42250000000000004} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0034102141680395492, "rotation": {"subspace_principal_angle": 0.8491797480583015, "mean_class_cosine": 0.6605991630042665}}}, "iid_acc": 0.99, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.7965198420896995, "mean_class_cosine": 0.69919899770492}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.5420098846787479, "drop": 0.1567401153212521, "rotation": {"subspace_principal_angle": 1.4157293467420775, "mean_class_cosine": 0.9954481825771943}}}, "iid_acc": 0.69875, "selectivity": -0.13750000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.4943111294129614, "mean_class_cosine": 0.9994355209688002}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.98625, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": -0.0005704283360791207}}, "iid_acc": 0.98625, "selectivity": 0.1499999999999999} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6816269284712483, "drop": -0.006626928471248239, "rotation": {"subspace_principal_angle": 1.493369684917862, "mean_class_cosine": 0.07734930442811239}}}, "iid_acc": 0.675, "selectivity": 0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4475486480920703, "mean_class_cosine": 0.12293589375679194}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.517531556802244, "drop": -0.005031556802244097, "rotation": {"subspace_principal_angle": 1.407007437734732, "mean_class_cosine": 0.1636007713519359}}}, "iid_acc": 0.5125, "selectivity": -0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4687784126101222, "mean_class_cosine": 0.14581035402784961}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.7110799438990182, "drop": -0.009829943899018145}}, "iid_acc": 0.70125, "selectivity": 0.04125000000000001} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6214689265536724, "drop": 0.01795415036940451, "rotation": {"subspace_principal_angle": 1.4099267955197121, "mean_class_cosine": 0.21703022459825547}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.19230769230769224, "iid_split_rotation": {"subspace_principal_angle": 1.4925996250553264, "mean_class_cosine": 0.20125792019838662}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.5769230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.519774011299435, "drop": 0.05714906562364186, "rotation": {"subspace_principal_angle": 1.5438703377286263, "mean_class_cosine": 0.6081265892996393}}}, "iid_acc": 0.5769230769230769, "selectivity": 0.12980769230769224, "iid_split_rotation": {"subspace_principal_angle": 1.367531053578908, "mean_class_cosine": 0.5070016095884643}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6101694915254238, "drop": -0.004400260756193042}}, "iid_acc": 0.6057692307692307, "selectivity": 0.1586538461538461} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8849315068493151, "drop": 0.017568493150684872, "rotation": {"subspace_principal_angle": 1.171653684236408, "mean_class_cosine": 0.3886285207848383}}}, "iid_acc": 0.9025, "selectivity": 0.13249999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1155850933836575, "mean_class_cosine": 0.43965204674130126}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.6, "drop": 0.0025000000000000577, "rotation": {"subspace_principal_angle": 1.1040454036261176, "mean_class_cosine": 0.827992833300117}}}, "iid_acc": 0.6025, "selectivity": -0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.126081986855619, "mean_class_cosine": 0.6844596631580138}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8479452054794521, "drop": 0.01955479452054798}}, "iid_acc": 0.8675, "selectivity": 0.09750000000000003} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.76625, "drop": 0.0}, "paraphrase": {"acc": 0.7503467406380028, "drop": 0.015903259361997213, "rotation": {"subspace_principal_angle": 1.2612626710199533, "mean_class_cosine": 0.3046144880272891}}, "domain": {"acc": 0.72875, "drop": 0.03749999999999998, "rotation": {"subspace_principal_angle": 1.3058544057767159, "mean_class_cosine": 0.2618532161180114}}}, "iid_acc": 0.76625, "selectivity": 0.2875, "iid_split_rotation": {"subspace_principal_angle": 1.2715880951662406, "mean_class_cosine": 0.294763708893907}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.49930651872399445, "drop": 0.0031934812760054943, "rotation": {"subspace_principal_angle": 1.1711657362498487, "mean_class_cosine": 0.9874129557052536}}, "domain": {"acc": 0.49875, "drop": 0.00374999999999992, "rotation": {"subspace_principal_angle": 1.5016858906817072, "mean_class_cosine": 0.3434093433034723}}}, "iid_acc": 0.5025, "selectivity": 0.023749999999999938, "iid_split_rotation": {"subspace_principal_angle": 1.2751956782954972, "mean_class_cosine": 0.972007963226671}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.765, "drop": 0.0}, "paraphrase": {"acc": 0.7600554785020804, "drop": 0.004944521497919574}, "domain": {"acc": 0.72375, "drop": 0.04125000000000001}}, "iid_acc": 0.765, "selectivity": 0.28625} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.829126213592233, "drop": 0.02212378640776691, "rotation": {"subspace_principal_angle": 1.1254573688815634, "mean_class_cosine": 0.43076380259456154}}, "domain": {"acc": 0.53625, "drop": 0.31499999999999995, "rotation": {"subspace_principal_angle": 1.3011899623729255, "mean_class_cosine": 0.2663520415198698}}, "length": {"acc": 0.8597285067873304, "drop": -0.00847850678733042, "rotation": {"subspace_principal_angle": 1.0844876622709578, "mean_class_cosine": 0.4673657026245072}}}, "iid_acc": 0.85125, "selectivity": 0.36624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1313772551394048, "mean_class_cosine": 0.4254137973371145}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7320388349514563, "drop": -0.0007888349514563187, "rotation": {"subspace_principal_angle": 1.3889114115978614, "mean_class_cosine": 0.9101913863100519}}, "domain": {"acc": 0.58, "drop": 0.15125, "rotation": {"subspace_principal_angle": 1.2835193898835013, "mean_class_cosine": 0.5072723519659894}}, "length": {"acc": 0.7450980392156863, "drop": -0.013848039215686336, "rotation": {"subspace_principal_angle": 1.474745277649587, "mean_class_cosine": 0.9218320832330005}}}, "iid_acc": 0.73125, "selectivity": 0.24624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4693373559237952, "mean_class_cosine": 0.8923141074617422}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.01725728155339812}, "domain": {"acc": 0.5775, "drop": 0.265}, "length": {"acc": 0.8476621417797888, "drop": -0.005162141779788798}}, "iid_acc": 0.8425, "selectivity": 0.35750000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8677165354330708, "drop": -0.00021653543307076362, "rotation": {"subspace_principal_angle": 1.3138102642722838, "mean_class_cosine": 0.4178222215166151}}}, "iid_acc": 0.8675, "selectivity": 0.6100000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.269543032505647, "mean_class_cosine": 0.41884051584646115}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.59375, "drop": 0.0}, "paraphrase": {"acc": 0.5748031496062992, "drop": 0.01894685039370081, "rotation": {"subspace_principal_angle": 1.462647268843256, "mean_class_cosine": 0.9428914975925955}}}, "iid_acc": 0.59375, "selectivity": 0.33625, "iid_split_rotation": {"subspace_principal_angle": 1.5667150391990796, "mean_class_cosine": 0.7928532264394269}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8566929133858268, "drop": 0.01455708661417321}}, "iid_acc": 0.87125, "selectivity": 0.61375} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.965, "drop": 0.0}, "paraphrase": {"acc": 0.9319899244332494, "drop": 0.03301007556675062, "rotation": {"subspace_principal_angle": 1.2176081199133377, "mean_class_cosine": 0.6442873004051297}}}, "iid_acc": 0.965, "selectivity": 0.9037499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1385695404879772, "mean_class_cosine": 0.7109378375596532}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.6801007556675063, "drop": 0.04114924433249367, "rotation": {"subspace_principal_angle": 1.5679042974829251, "mean_class_cosine": 0.8621074777728631}}}, "iid_acc": 0.72125, "selectivity": 0.6599999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2212236304536004, "mean_class_cosine": 0.9460677922131192}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9395465994962217, "drop": 0.02170340050377839}}, "iid_acc": 0.96125, "selectivity": 0.9} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5071707953063885, "drop": 0.02907920469361147, "rotation": {"subspace_principal_angle": 1.4838566858436246, "mean_class_cosine": 0.08683016010539553}}}, "iid_acc": 0.53625, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.465396612413356, "mean_class_cosine": 0.10520467342535929}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.49674054758800523, "drop": -0.021740547588005255, "rotation": {"subspace_principal_angle": 1.5469172283064518, "mean_class_cosine": -0.5397626183074191}}}, "iid_acc": 0.475, "selectivity": -0.03375000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.408080647568939, "mean_class_cosine": 0.4860859905388361}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4876140808344198, "drop": 0.01488591916558013}}, "iid_acc": 0.5025, "selectivity": -0.006250000000000089} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5524568393094289, "drop": 0.058793160690571056, "rotation": {"subspace_principal_angle": 1.3850534905313705, "mean_class_cosine": 0.28991503737744245}}}, "iid_acc": 0.61125, "selectivity": 0.37749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.419189771224309, "mean_class_cosine": 0.29718954050314617}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.0925, "drop": 0.0}, "paraphrase": {"acc": 0.1049136786188579, "drop": -0.012413678618857907, "rotation": {"subspace_principal_angle": 1.540744396600694, "mean_class_cosine": 0.06532846024974738}}}, "iid_acc": 0.0925, "selectivity": -0.14125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4223489889510221, "mean_class_cosine": 0.5690159019187561}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5232403718459495, "drop": 0.011759628154050517}}, "iid_acc": 0.535, "selectivity": 0.30125} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": -0.0007322006472492149, "rotation": {"subspace_principal_angle": 1.3881466203882136, "mean_class_cosine": 0.18163583883336032}}}, "iid_acc": 0.73875, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3959757692399888, "mean_class_cosine": 0.17393143305355055}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5906148867313916, "drop": -0.0393648867313916, "rotation": {"subspace_principal_angle": 1.4683816158112597, "mean_class_cosine": 0.9483792840431817}}}, "iid_acc": 0.55125, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.3971642089849379, "mean_class_cosine": 0.9161271445117759}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7540453074433657, "drop": -0.0027953074433657576}}, "iid_acc": 0.75125, "selectivity": 0.20875} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6397415185783522, "drop": 0.044008481421647816, "rotation": {"subspace_principal_angle": 1.3758917619051532, "mean_class_cosine": 0.19367290768765433}}}, "iid_acc": 0.68375, "selectivity": 0.18374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.345477898881277, "mean_class_cosine": 0.22341675246913884}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5993537964458805, "drop": 0.011896203554119467, "rotation": {"subspace_principal_angle": 1.4573232614008496, "mean_class_cosine": 0.870794766106819}}}, "iid_acc": 0.61125, "selectivity": 0.11124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.385445998953203, "mean_class_cosine": 0.8535614192902857}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.6397415185783522, "drop": 0.046508481421647874}}, "iid_acc": 0.68625, "selectivity": 0.18625000000000003} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7405582922824302, "drop": -0.013058292282430162, "rotation": {"subspace_principal_angle": 1.3879649771162461, "mean_class_cosine": 0.1818144576289447}}}, "iid_acc": 0.7275, "selectivity": 0.1312500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4079457648518725, "mean_class_cosine": 0.16213170804038407}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.50375, "drop": 0.0}, "paraphrase": {"acc": 0.5320197044334976, "drop": -0.028269704433497522, "rotation": {"subspace_principal_angle": 1.5607306280375703, "mean_class_cosine": 0.9876176460724118}}}, "iid_acc": 0.50375, "selectivity": -0.09249999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.166861107011101, "mean_class_cosine": 0.9302660401353273}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7422003284072249, "drop": 0.010299671592775006}}, "iid_acc": 0.7525, "selectivity": 0.15625} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8793342579750347, "drop": 0.020665742024965295, "rotation": {"subspace_principal_angle": 1.1063584135999445, "mean_class_cosine": 0.4479203253262932}}}, "iid_acc": 0.9, "selectivity": 0.46125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0249583822746835, "mean_class_cosine": 0.5191344576441722}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.5728155339805825, "drop": 0.14468446601941753, "rotation": {"subspace_principal_angle": 1.284951624986405, "mean_class_cosine": 0.8542882000899992}}}, "iid_acc": 0.7175, "selectivity": 0.27875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0904185384770684, "mean_class_cosine": 0.9172822881817656}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8904299583911235, "drop": 0.00832004160887656}}, "iid_acc": 0.89875, "selectivity": 0.4600000000000001} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9967051070840197, "drop": -0.00045510708401974576, "rotation": {"subspace_principal_angle": 0.8815453404218065, "mean_class_cosine": 0.6359593299455631}}}, "iid_acc": 0.99625, "selectivity": 0.16874999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.829603647498299, "mean_class_cosine": 0.6751681879987259}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.5634266886326195, "drop": 0.12157331136738059, "rotation": {"subspace_principal_angle": 1.1147774126986099, "mean_class_cosine": 0.9975604664853437}}}, "iid_acc": 0.685, "selectivity": -0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.351613895560496, "mean_class_cosine": 0.9997571774292637}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9950576606260296, "drop": -0.005057660626029636}}, "iid_acc": 0.99, "selectivity": 0.16249999999999998} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6633941093969145, "drop": 0.01910589060308554, "rotation": {"subspace_principal_angle": 1.5456837867774578, "mean_class_cosine": 0.02510990060669071}}}, "iid_acc": 0.6825, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4420552406446043, "mean_class_cosine": 0.12838574922234897}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.46875, "drop": 0.0}, "paraphrase": {"acc": 0.4950911640953717, "drop": -0.026341164095371683, "rotation": {"subspace_principal_angle": 0.8450834965822174, "mean_class_cosine": 0.8344612624127071}}}, "iid_acc": 0.46875, "selectivity": -0.14375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.447600013734126, "mean_class_cosine": -0.6148431774148773}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7237026647966339, "drop": -0.009952664796633925}}, "iid_acc": 0.71375, "selectivity": 0.10124999999999995} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.615819209039548, "drop": 0.00918079096045199, "rotation": {"subspace_principal_angle": 1.344146750268554, "mean_class_cosine": 0.24255494243559642}}}, "iid_acc": 0.625, "selectivity": 0.16346153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.4369180768778205, "mean_class_cosine": 0.21822584946678156}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5423728813559322, "drop": 0.0201271186440678, "rotation": {"subspace_principal_angle": 1.5227921354385472, "mean_class_cosine": 0.6841529664773013}}}, "iid_acc": 0.5625, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.2545192468784838, "mean_class_cosine": 0.5447114834286542}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.632768361581921, "drop": -0.02699913081269023}}, "iid_acc": 0.6057692307692307, "selectivity": 0.14423076923076916} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8945205479452055, "drop": 0.004229452054794525, "rotation": {"subspace_principal_angle": 1.248973078003772, "mean_class_cosine": 0.3162967291391359}}}, "iid_acc": 0.89875, "selectivity": 0.15250000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.2122862189044603, "mean_class_cosine": 0.35087945751268557}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5767123287671233, "drop": -0.009212328767123301, "rotation": {"subspace_principal_angle": 1.21580262592107, "mean_class_cosine": 0.8869377958216829}}}, "iid_acc": 0.5675, "selectivity": -0.17874999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.839488040678943, "mean_class_cosine": 0.7546602597663057}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8767123287671232, "drop": 0.012037671232876801}}, "iid_acc": 0.88875, "selectivity": 0.14250000000000007} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7739251040221914, "drop": 0.029824895977808574, "rotation": {"subspace_principal_angle": 1.2659109438320224, "mean_class_cosine": 0.30018384612544047}}, "domain": {"acc": 0.76875, "drop": 0.03499999999999992, "rotation": {"subspace_principal_angle": 1.34258356435988, "mean_class_cosine": 0.22623698719981744}}}, "iid_acc": 0.80375, "selectivity": 0.32499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2506624679779867, "mean_class_cosine": 0.3146936213275413}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5034674063800277, "drop": -0.004717406380027711, "rotation": {"subspace_principal_angle": 1.239282670172901, "mean_class_cosine": 0.9986892573730854}}, "domain": {"acc": 0.5, "drop": -0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.3097283989710364, "mean_class_cosine": 0.1603346495077691}}}, "iid_acc": 0.49875, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.420934784951417, "mean_class_cosine": 0.9934847558506945}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.8030513176144244, "drop": 0.019448682385575622}, "domain": {"acc": 0.81625, "drop": 0.006249999999999978}}, "iid_acc": 0.8225, "selectivity": 0.34375} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.8524271844660194, "drop": 0.02382281553398058, "rotation": {"subspace_principal_angle": 1.220154188841714, "mean_class_cosine": 0.34350094358966443}}, "domain": {"acc": 0.4975, "drop": 0.37875, "rotation": {"subspace_principal_angle": 1.438605482929702, "mean_class_cosine": 0.13180618706503908}}, "length": {"acc": 0.8808446455505279, "drop": -0.004594645550527909, "rotation": {"subspace_principal_angle": 1.1893943688634518, "mean_class_cosine": 0.37222205321854485}}}, "iid_acc": 0.87625, "selectivity": 0.38625, "iid_split_rotation": {"subspace_principal_angle": 1.1874268592484474, "mean_class_cosine": 0.3740474632573129}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.829126213592233, "drop": 0.013373786407766985, "rotation": {"subspace_principal_angle": 1.525376087221384, "mean_class_cosine": 0.942182847156327}}, "domain": {"acc": 0.57875, "drop": 0.26375000000000004, "rotation": {"subspace_principal_angle": 1.436822367907557, "mean_class_cosine": 0.27136585793969137}}, "length": {"acc": 0.8536953242835595, "drop": -0.011195324283559516, "rotation": {"subspace_principal_angle": 1.4983468510851918, "mean_class_cosine": 0.9461442600007871}}}, "iid_acc": 0.8425, "selectivity": 0.35250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.287421066876229, "mean_class_cosine": 0.910695802907292}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8640776699029126, "drop": 0.015922330097087434}, "domain": {"acc": 0.505, "drop": 0.375}, "length": {"acc": 0.8883861236802413, "drop": -0.008386123680241275}}, "iid_acc": 0.88, "selectivity": 0.39} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8582677165354331, "drop": 0.015482283464566926, "rotation": {"subspace_principal_angle": 1.2827164275902705, "mean_class_cosine": 0.4163674814452723}}}, "iid_acc": 0.87375, "selectivity": 0.6000000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.26135377780834, "mean_class_cosine": 0.4442699111544142}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7975, "drop": 0.0}, "paraphrase": {"acc": 0.7968503937007874, "drop": 0.0006496062992126239, "rotation": {"subspace_principal_angle": 1.424694638770457, "mean_class_cosine": 0.9469278805103363}}}, "iid_acc": 0.7975, "selectivity": 0.5237499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4837220741637616, "mean_class_cosine": 0.9518191309197193}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8598425196850393, "drop": 0.017657480314960616}}, "iid_acc": 0.8775, "selectivity": 0.60375} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9672544080604534, "drop": 0.00024559193954665215, "rotation": {"subspace_principal_angle": 1.1145296779462641, "mean_class_cosine": 0.7020277810816539}}}, "iid_acc": 0.9675, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.5624189489347664, "mean_class_cosine": 0.7403838209375735}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.44125, "drop": 0.0}, "paraphrase": {"acc": 0.4256926952141058, "drop": 0.015557304785894177, "rotation": {"subspace_principal_angle": 1.033444992470805, "mean_class_cosine": 0.7765287771414837}}}, "iid_acc": 0.44125, "selectivity": 0.38, "iid_split_rotation": {"subspace_principal_angle": 1.2853261455962783, "mean_class_cosine": 0.8790932160427343}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.96375, "drop": 0.0}, "paraphrase": {"acc": 0.9571788413098237, "drop": 0.006571158690176326}}, "iid_acc": 0.96375, "selectivity": 0.9025} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5397653194263363, "drop": 0.013984680573663644, "rotation": {"subspace_principal_angle": 1.5274443365773367, "mean_class_cosine": 0.0433384122408627}}}, "iid_acc": 0.55375, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5117604681674472, "mean_class_cosine": 0.05900157231956833}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.5071707953063885, "drop": -0.029670795306388553, "rotation": {"subspace_principal_angle": 0.8759468408895317, "mean_class_cosine": -0.6781424224487453}}}, "iid_acc": 0.4775, "selectivity": -0.006250000000000033, "iid_split_rotation": {"subspace_principal_angle": 1.0844620330806314, "mean_class_cosine": -0.4769926683134672}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.49282920469361147, "drop": 0.02717079530638855}}, "iid_acc": 0.52, "selectivity": 0.036250000000000004} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.62875, "drop": 0.0}, "paraphrase": {"acc": 0.5803452855245684, "drop": 0.0484047144754316, "rotation": {"subspace_principal_angle": 1.4145413502440674, "mean_class_cosine": 0.27663015100707133}}}, "iid_acc": 0.62875, "selectivity": 0.37125, "iid_split_rotation": {"subspace_principal_angle": 1.4392990909748595, "mean_class_cosine": 0.2769217490323135}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.09, "drop": 0.0}, "paraphrase": {"acc": 0.07436918990703852, "drop": 0.01563081009296148, "rotation": {"subspace_principal_angle": 1.3350767969885171, "mean_class_cosine": -0.06735755908616899}}}, "iid_acc": 0.09, "selectivity": -0.1675, "iid_split_rotation": {"subspace_principal_angle": 1.2649936899279066, "mean_class_cosine": 0.6127869206852473}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5391766268260292, "drop": 0.01332337317397081}}, "iid_acc": 0.5525, "selectivity": 0.295} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": 0.0030177993527508162, "rotation": {"subspace_principal_angle": 1.3800650372385295, "mean_class_cosine": 0.18957697385454825}}}, "iid_acc": 0.7425, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3745352340470285, "mean_class_cosine": 0.19500357265711427}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.64125, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.025416666666666643, "rotation": {"subspace_principal_angle": 1.2209077466423848, "mean_class_cosine": 0.8479639619526143}}}, "iid_acc": 0.64125, "selectivity": 0.09124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.563353933733748, "mean_class_cosine": 0.8055681849044787}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7572815533980582, "drop": 0.0027184466019417597}}, "iid_acc": 0.76, "selectivity": 0.20999999999999996} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.65375, "drop": 0.0}, "paraphrase": {"acc": 0.6155088852988692, "drop": 0.038241114701130896, "rotation": {"subspace_principal_angle": 1.4613251262739329, "mean_class_cosine": 0.10925268204629429}}}, "iid_acc": 0.65375, "selectivity": 0.1462500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.467586611767328, "mean_class_cosine": 0.10302657672993193}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.6125, "drop": 0.0}, "paraphrase": {"acc": 0.5912762520193862, "drop": 0.021223747980613883, "rotation": {"subspace_principal_angle": 1.2485040409587074, "mean_class_cosine": 0.866292700063821}}}, "iid_acc": 0.6125, "selectivity": 0.1050000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5556750038869702, "mean_class_cosine": 0.8382504448687405}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6510500807754442, "drop": 0.03269991922455573}}, "iid_acc": 0.68375, "selectivity": 0.17625000000000002} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.722495894909688, "drop": -0.0024958949096880456, "rotation": {"subspace_principal_angle": 1.4020686555705464, "mean_class_cosine": 0.1679282246076349}}}, "iid_acc": 0.72, "selectivity": 0.13374999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3616761916383724, "mean_class_cosine": 0.20759928424855742}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.56375, "drop": 0.0}, "paraphrase": {"acc": 0.5369458128078818, "drop": 0.02680418719211819, "rotation": {"subspace_principal_angle": 1.525739498161781, "mean_class_cosine": 0.8790506170094416}}}, "iid_acc": 0.56375, "selectivity": -0.022500000000000075, "iid_split_rotation": {"subspace_principal_angle": 1.355838296904698, "mean_class_cosine": 0.7623898156919341}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7504105090311987, "drop": 0.0008394909688013152}}, "iid_acc": 0.75125, "selectivity": 0.16499999999999992} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.897364771151179, "drop": 0.02388522884882105, "rotation": {"subspace_principal_angle": 1.2009122822029357, "mean_class_cosine": 0.3615073211363401}}}, "iid_acc": 0.92125, "selectivity": 0.45375, "iid_split_rotation": {"subspace_principal_angle": 1.1365147935104427, "mean_class_cosine": 0.42075873671714054}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7253814147018031, "drop": 0.008368585298196951, "rotation": {"subspace_principal_angle": 1.382859019749326, "mean_class_cosine": 0.9518475699667834}}}, "iid_acc": 0.73375, "selectivity": 0.26625, "iid_split_rotation": {"subspace_principal_angle": 0.8342063363329182, "mean_class_cosine": 0.8834390160996848}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.92625, "drop": 0.0}, "paraphrase": {"acc": 0.9140083217753121, "drop": 0.012241678224687913}}, "iid_acc": 0.92625, "selectivity": 0.45875} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.0007372322899505956, "rotation": {"subspace_principal_angle": 0.9022113956514256, "mean_class_cosine": 0.6198762040410312}}}, "iid_acc": 0.9925, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 0.7803136404078311, "mean_class_cosine": 0.7106929262056456}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.4744645799011532, "drop": 0.21553542009884674, "rotation": {"subspace_principal_angle": 1.5013958950641693, "mean_class_cosine": 0.997742196187771}}}, "iid_acc": 0.69, "selectivity": -0.1362500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3648075725684445, "mean_class_cosine": 0.9999469425446367}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": -0.00011532125205926391}}, "iid_acc": 0.99, "selectivity": 0.16374999999999995} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6900420757363254, "drop": 0.019957924263674598, "rotation": {"subspace_principal_angle": 1.4053917466501165, "mean_class_cosine": 0.16465140281778454}}}, "iid_acc": 0.71, "selectivity": 0.10249999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.4198215714680307, "mean_class_cosine": 0.15040187120207954}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.511921458625526, "drop": -0.03442145862552598, "rotation": {"subspace_principal_angle": 1.088765737718087, "mean_class_cosine": -0.46901391020350736}}}, "iid_acc": 0.4775, "selectivity": -0.13000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4869971047307045, "mean_class_cosine": 0.19635370929712698}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7223001402524544, "drop": -0.012300140252454406}}, "iid_acc": 0.71, "selectivity": 0.10249999999999992} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.672316384180791, "drop": -0.018470230334637128, "rotation": {"subspace_principal_angle": 1.4095901786238862, "mean_class_cosine": 0.19654959439481243}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.27403846153846156, "iid_split_rotation": {"subspace_principal_angle": 1.5567476235509987, "mean_class_cosine": 0.20024194309517288}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.423728813559322, "drop": 0.03780964797913955, "rotation": {"subspace_principal_angle": 1.318674137405094, "mean_class_cosine": 0.5319901746483242}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.08173076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.1328152149633164, "mean_class_cosine": 0.6908347513167752}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6201923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": -0.01822577140373749}}, "iid_acc": 0.6201923076923077, "selectivity": 0.24038461538461542} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8780821917808219, "drop": 0.02066780821917813, "rotation": {"subspace_principal_angle": 1.4022671366129482, "mean_class_cosine": 0.16773255884346316}}}, "iid_acc": 0.89875, "selectivity": 0.17500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3156344568426237, "mean_class_cosine": 0.25240204094053126}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.584931506849315, "drop": 0.017568493150684983, "rotation": {"subspace_principal_angle": 1.2909886826134447, "mean_class_cosine": 0.8167709778669217}}}, "iid_acc": 0.6025, "selectivity": -0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.181444552949185, "mean_class_cosine": 0.8742409810564655}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.8972602739726028, "drop": 0.013989726027397231}}, "iid_acc": 0.91125, "selectivity": 0.1875} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.7988904299583911, "drop": 0.047359570041608845, "rotation": {"subspace_principal_angle": 1.307852975470961, "mean_class_cosine": 0.25992385942235496}}, "domain": {"acc": 0.865, "drop": -0.018750000000000044, "rotation": {"subspace_principal_angle": 1.4101720618034508, "mean_class_cosine": 0.15993446696213712}}}, "iid_acc": 0.84625, "selectivity": 0.3374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2803570242694347, "mean_class_cosine": 0.2863731564533277}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5034674063800277, "drop": -0.004717406380027711, "rotation": {"subspace_principal_angle": 1.4967653281587188, "mean_class_cosine": 0.9956396889374676}}, "domain": {"acc": 0.5, "drop": -0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5195064626946009, "mean_class_cosine": 0.17206983639586906}}}, "iid_acc": 0.49875, "selectivity": -0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.53305740516774, "mean_class_cosine": 0.9726130574764482}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.812760055478502, "drop": 0.04223994452149793}, "domain": {"acc": 0.84375, "drop": 0.011249999999999982}}, "iid_acc": 0.855, "selectivity": 0.34624999999999995} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8815533980582524, "drop": 0.020946601941747578, "rotation": {"subspace_principal_angle": 1.3231535337909477, "mean_class_cosine": 0.24511934689015158}}, "domain": {"acc": 0.56, "drop": 0.3424999999999999, "rotation": {"subspace_principal_angle": 1.4299398396010543, "mean_class_cosine": 0.14039117067809936}}, "length": {"acc": 0.9140271493212669, "drop": -0.011527149321266972, "rotation": {"subspace_principal_angle": 1.2714271775831436, "mean_class_cosine": 0.2949174731196204}}}, "iid_acc": 0.9025, "selectivity": 0.43749999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3312504683770023, "mean_class_cosine": 0.23726147696520083}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8524271844660194, "drop": 0.03507281553398056, "rotation": {"subspace_principal_angle": 1.1382054241911463, "mean_class_cosine": 0.9322577478917091}}, "domain": {"acc": 0.57875, "drop": 0.30874999999999997, "rotation": {"subspace_principal_angle": 1.5632845119627006, "mean_class_cosine": 0.34081581535856803}}, "length": {"acc": 0.8883861236802413, "drop": -0.0008861236802413242, "rotation": {"subspace_principal_angle": 1.098044803005494, "mean_class_cosine": 0.9434252502763861}}}, "iid_acc": 0.8875, "selectivity": 0.42249999999999993, "iid_split_rotation": {"subspace_principal_angle": 0.7127743899001, "mean_class_cosine": 0.9390001647602704}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.90375, "drop": 0.0}, "paraphrase": {"acc": 0.8990291262135922, "drop": 0.004720873786407842}, "domain": {"acc": 0.73125, "drop": 0.1725000000000001}, "length": {"acc": 0.9155354449472096, "drop": -0.011785444947209567}}, "iid_acc": 0.90375, "selectivity": 0.43875000000000003} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8866141732283465, "drop": 0.005885826771653502, "rotation": {"subspace_principal_angle": 1.3590153860798901, "mean_class_cosine": 0.3384795168112319}}}, "iid_acc": 0.8925, "selectivity": 0.67375, "iid_split_rotation": {"subspace_principal_angle": 1.5137713950823037, "mean_class_cosine": 0.34878144427948704}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6661417322834645, "drop": 0.021358267716535484, "rotation": {"subspace_principal_angle": 1.2049572219735756, "mean_class_cosine": 0.9516923360218306}}}, "iid_acc": 0.6875, "selectivity": 0.46875, "iid_split_rotation": {"subspace_principal_angle": 1.161004465491632, "mean_class_cosine": 0.8826822088496902}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8771653543307086, "drop": 0.017834645669291382}}, "iid_acc": 0.895, "selectivity": 0.67625} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.964735516372796, "drop": 0.010264483627204002, "rotation": {"subspace_principal_angle": 1.110589965959208, "mean_class_cosine": 0.671095672439944}}}, "iid_acc": 0.975, "selectivity": 0.92375, "iid_split_rotation": {"subspace_principal_angle": 1.0922410354486654, "mean_class_cosine": 0.7234153961945159}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.84375, "drop": 0.0}, "paraphrase": {"acc": 0.8488664987405542, "drop": -0.005116498740554198, "rotation": {"subspace_principal_angle": 1.4098145943553153, "mean_class_cosine": 0.8856785823869348}}}, "iid_acc": 0.84375, "selectivity": 0.7925, "iid_split_rotation": {"subspace_principal_angle": 0.8542184898946922, "mean_class_cosine": 0.9572548062279118}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.9596977329974811, "drop": 0.01530226700251891}}, "iid_acc": 0.975, "selectivity": 0.92375} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5736636245110821, "drop": 0.006336375488917856, "rotation": {"subspace_principal_angle": 1.4824473599970098, "mean_class_cosine": 0.08823407641471295}}}, "iid_acc": 0.58, "selectivity": 0.05499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4950406885229706, "mean_class_cosine": 0.07568319984431088}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.5058670143415906, "drop": -0.02961701434159064, "rotation": {"subspace_principal_angle": 1.1493594930263484, "mean_class_cosine": -0.4424939882567903}}}, "iid_acc": 0.47625, "selectivity": -0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.551473168662097, "mean_class_cosine": 0.01234111003206595}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5867014341590613, "drop": -0.00045143415906123696}}, "iid_acc": 0.58625, "selectivity": 0.06125000000000003} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6148738379814077, "drop": 0.03512616201859231, "rotation": {"subspace_principal_angle": 1.4073412354609784, "mean_class_cosine": 0.26349298302241686}}}, "iid_acc": 0.65, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.413824435567323, "mean_class_cosine": 0.24889219745848057}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.095, "drop": 0.0}, "paraphrase": {"acc": 0.10092961487383798, "drop": -0.005929614873837974, "rotation": {"subspace_principal_angle": 1.331943855465791, "mean_class_cosine": 0.03732221069944747}}}, "iid_acc": 0.095, "selectivity": -0.165, "iid_split_rotation": {"subspace_principal_angle": 1.5577947437159871, "mean_class_cosine": 0.5674187824840483}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.64, "drop": 0.0}, "paraphrase": {"acc": 0.6347941567065073, "drop": 0.005205843293492718}}, "iid_acc": 0.64, "selectivity": 0.38} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.77625, "drop": 0.0}, "paraphrase": {"acc": 0.7427184466019418, "drop": 0.033531553398058245, "rotation": {"subspace_principal_angle": 1.4430864610974254, "mean_class_cosine": 0.12736299475185067}}}, "iid_acc": 0.77625, "selectivity": 0.22499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4358406552905756, "mean_class_cosine": 0.13454638570974817}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5598705501618123, "drop": -0.02237055016181233, "rotation": {"subspace_principal_angle": 1.1405278653750983, "mean_class_cosine": 0.804713745342145}}}, "iid_acc": 0.5375, "selectivity": -0.01375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.509364684034334, "mean_class_cosine": 0.9432691877610848}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.7925, "drop": 0.0}, "paraphrase": {"acc": 0.7944983818770227, "drop": -0.0019983818770227257}}, "iid_acc": 0.7925, "selectivity": 0.24124999999999996} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.6494345718901454, "drop": 0.03681542810985461, "rotation": {"subspace_principal_angle": 1.4705871216508313, "mean_class_cosine": 0.10004157445035826}}}, "iid_acc": 0.68625, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4090091592636642, "mean_class_cosine": 0.16108229179329214}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5282714054927302, "drop": 0.00922859450726976, "rotation": {"subspace_principal_angle": 0.8130735690110286, "mean_class_cosine": 0.9877317674481327}}}, "iid_acc": 0.5375, "selectivity": 0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 0.6815839977232387, "mean_class_cosine": 0.9939458887778236}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6591276252019386, "drop": 0.04587237479806139}}, "iid_acc": 0.705, "selectivity": 0.18999999999999995} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6962233169129721, "drop": 0.011276683087027894, "rotation": {"subspace_principal_angle": 1.49979082124605, "mean_class_cosine": 0.07094585487697073}}}, "iid_acc": 0.7075, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.482620368933285, "mean_class_cosine": 0.08806174093377613}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.4875, "drop": 0.0}, "paraphrase": {"acc": 0.49589490968801314, "drop": -0.008394909688013152, "rotation": {"subspace_principal_angle": 1.524068667642953, "mean_class_cosine": 0.9875864078448118}}}, "iid_acc": 0.4875, "selectivity": -0.07500000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8171205192766501, "mean_class_cosine": 0.770949459806412}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7684729064039408, "drop": -0.004722906403940796}}, "iid_acc": 0.76375, "selectivity": 0.20125000000000004} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.93375, "drop": 0.0}, "paraphrase": {"acc": 0.9181692094313454, "drop": 0.015580790568654579, "rotation": {"subspace_principal_angle": 1.245296425269521, "mean_class_cosine": 0.3197824779830509}}}, "iid_acc": 0.93375, "selectivity": 0.46624999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.1642335870118337, "mean_class_cosine": 0.3954545965394525}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.855755894590846, "drop": -0.004505894590846049, "rotation": {"subspace_principal_angle": 0.9569293407079221, "mean_class_cosine": 0.948995043172376}}}, "iid_acc": 0.85125, "selectivity": 0.3837499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5155592875644858, "mean_class_cosine": 0.9503525832749367}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.9292649098474342, "drop": 0.003235090152565845}}, "iid_acc": 0.9325, "selectivity": 0.46499999999999997} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.001987232289950569, "rotation": {"subspace_principal_angle": 0.9228456065715872, "mean_class_cosine": 0.6035537376948626}}}, "iid_acc": 0.99375, "selectivity": 0.19000000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.836801876997636, "mean_class_cosine": 0.6698408686733577}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.42174629324546953, "drop": 0.2632537067545305, "rotation": {"subspace_principal_angle": 1.1528917670379706, "mean_class_cosine": 0.9973529147197886}}}, "iid_acc": 0.685, "selectivity": -0.11874999999999991, "iid_split_rotation": {"subspace_principal_angle": 1.1220283287635184, "mean_class_cosine": 0.999929258322932}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.0007372322899505956}}, "iid_acc": 0.9925, "selectivity": 0.18875000000000008} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.699859747545582, "drop": 0.048890252454418026, "rotation": {"subspace_principal_angle": 1.4762812316903078, "mean_class_cosine": 0.09437443909422669}}}, "iid_acc": 0.74875, "selectivity": 0.16500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4435800079379149, "mean_class_cosine": 0.12687345177932846}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.5063113604488079, "drop": -0.03131136044880789, "rotation": {"subspace_principal_angle": 1.4269418061859824, "mean_class_cosine": -0.10748417912085814}}}, "iid_acc": 0.475, "selectivity": -0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4676928130838114, "mean_class_cosine": 0.13776060857811906}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.726507713884993, "drop": 0.004742286115006933}}, "iid_acc": 0.73125, "selectivity": 0.14749999999999996} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.6923076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6836158192090396, "drop": 0.008691873098652714, "rotation": {"subspace_principal_angle": 1.414178459512195, "mean_class_cosine": 0.1826323602732557}}}, "iid_acc": 0.6923076923076923, "selectivity": 0.2932692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.4350751526080228, "mean_class_cosine": 0.1805559724526645}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.4855769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.4180790960451977, "drop": 0.06749782703172535, "rotation": {"subspace_principal_angle": 1.2602978332522063, "mean_class_cosine": 0.6807601457540157}}}, "iid_acc": 0.4855769230769231, "selectivity": 0.08653846153846151, "iid_split_rotation": {"subspace_principal_angle": 1.1362894639616432, "mean_class_cosine": 0.6888773765938021}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6836158192090396, "drop": -0.010538896132116449}}, "iid_acc": 0.6730769230769231, "selectivity": 0.27403846153846156} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8945205479452055, "drop": 0.01297945205479445, "rotation": {"subspace_principal_angle": 1.411004019350323, "mean_class_cosine": 0.15911316342225837}}}, "iid_acc": 0.9075, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.3881521587997803, "mean_class_cosine": 0.1816303925457004}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.647945205479452, "drop": 0.020804794520547953, "rotation": {"subspace_principal_angle": 1.1198503241914632, "mean_class_cosine": 0.8163975702558504}}}, "iid_acc": 0.66875, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.401866352414971, "mean_class_cosine": 0.9059315149970117}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9125, "drop": 0.0}, "paraphrase": {"acc": 0.9082191780821918, "drop": 0.0042808219178082085}}, "iid_acc": 0.9125, "selectivity": 0.21875} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.78375, "drop": 0.0}, "paraphrase": {"acc": 0.7961165048543689, "drop": -0.012366504854368965, "rotation": {"subspace_principal_angle": 1.2928467517215603, "mean_class_cosine": 0.2743844969854279}}, "domain": {"acc": 0.725, "drop": 0.05874999999999997, "rotation": {"subspace_principal_angle": 1.3363808809411806, "mean_class_cosine": 0.23227445847260966}}}, "iid_acc": 0.78375, "selectivity": 0.26625, "iid_split_rotation": {"subspace_principal_angle": 1.252105311514529, "mean_class_cosine": 0.3133237568177669}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.007765256588072145, "rotation": {"subspace_principal_angle": 0.5795075771494037, "mean_class_cosine": 0.9984633717830367}}, "domain": {"acc": 0.5, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5279705890862962, "mean_class_cosine": 0.23923966769590235}}}, "iid_acc": 0.50125, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.2020721809074764, "mean_class_cosine": 0.9962696884890198}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.805, "drop": 0.0}, "paraphrase": {"acc": 0.7364771151178918, "drop": 0.06852288488210823}, "domain": {"acc": 0.69625, "drop": 0.10875000000000001}}, "iid_acc": 0.805, "selectivity": 0.2875000000000001} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8271844660194175, "drop": 0.035315533980582514, "rotation": {"subspace_principal_angle": 1.1721917031350069, "mean_class_cosine": 0.38813273689356925}}, "domain": {"acc": 0.59, "drop": 0.2725000000000001, "rotation": {"subspace_principal_angle": 1.3310637787422674, "mean_class_cosine": 0.23744283169580327}}, "length": {"acc": 0.8748114630467572, "drop": -0.01231146304675712, "rotation": {"subspace_principal_angle": 1.0758385777551667, "mean_class_cosine": 0.4749944713239639}}}, "iid_acc": 0.8625, "selectivity": 0.33125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.16650891029775, "mean_class_cosine": 0.3933637231840712}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5805825242718446, "drop": -0.014332524271844616, "rotation": {"subspace_principal_angle": 1.390314107154939, "mean_class_cosine": 0.9207722901404116}}, "domain": {"acc": 0.57875, "drop": -0.012499999999999956, "rotation": {"subspace_principal_angle": 1.0701854814469178, "mean_class_cosine": 0.8833474305608653}}, "length": {"acc": 0.5641025641025641, "drop": 0.002147435897435934, "rotation": {"subspace_principal_angle": 0.9369329512012231, "mean_class_cosine": 0.9177680143320355}}}, "iid_acc": 0.56625, "selectivity": 0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.9871644210171315, "mean_class_cosine": 0.4073480943439788}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.006007281553398136}, "domain": {"acc": 0.75375, "drop": 0.07750000000000001}, "length": {"acc": 0.8355957767722474, "drop": -0.004345776772247345}}, "iid_acc": 0.83125, "selectivity": 0.30000000000000004} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.8725, "drop": 0.0}, "paraphrase": {"acc": 0.8614173228346457, "drop": 0.011082677165354382, "rotation": {"subspace_principal_angle": 1.3840540885530221, "mean_class_cosine": 0.2934795135845933}}}, "iid_acc": 0.8725, "selectivity": 0.59125, "iid_split_rotation": {"subspace_principal_angle": 1.4099364101494178, "mean_class_cosine": 0.29058275145224166}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.5925, "drop": 0.0}, "paraphrase": {"acc": 0.5669291338582677, "drop": 0.025570866141732318, "rotation": {"subspace_principal_angle": 0.998890821047691, "mean_class_cosine": 0.94716741691378}}}, "iid_acc": 0.5925, "selectivity": 0.31125, "iid_split_rotation": {"subspace_principal_angle": 0.9488762179483594, "mean_class_cosine": 0.7707278199904818}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8866141732283465, "drop": -0.0016141732283464494}}, "iid_acc": 0.885, "selectivity": 0.60375} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9370277078085643, "drop": 0.02422229219143579, "rotation": {"subspace_principal_angle": 1.2359971876297358, "mean_class_cosine": 0.6039952840126505}}}, "iid_acc": 0.96125, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.2037218056275953, "mean_class_cosine": 0.6646147092871253}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.2575, "drop": 0.0}, "paraphrase": {"acc": 0.24181360201511334, "drop": 0.015686397984886663, "rotation": {"subspace_principal_angle": 1.3871046502979878, "mean_class_cosine": 0.6602864355874835}}}, "iid_acc": 0.2575, "selectivity": 0.165, "iid_split_rotation": {"subspace_principal_angle": 1.5631281003472528, "mean_class_cosine": 0.7824016096532195}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9168765743073047, "drop": 0.024373425692695294}}, "iid_acc": 0.94125, "selectivity": 0.84875} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5580182529335072, "drop": -0.006768252933507135, "rotation": {"subspace_principal_angle": 1.4094629345382517, "mean_class_cosine": 0.16063442581070056}}}, "iid_acc": 0.55125, "selectivity": 0.05375000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4680870349990278, "mean_class_cosine": 0.10252880356925093}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.4915254237288136, "drop": -0.015275423728813575, "rotation": {"subspace_principal_angle": 1.4399498453238624, "mean_class_cosine": -0.9176241029118328}}}, "iid_acc": 0.47625, "selectivity": -0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4611313378914106, "mean_class_cosine": 0.7185000764227502}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5032594524119948, "drop": 0.009240547588005188}}, "iid_acc": 0.5125, "selectivity": 0.014999999999999958} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6325, "drop": 0.0}, "paraphrase": {"acc": 0.6228419654714475, "drop": 0.009658034528552406, "rotation": {"subspace_principal_angle": 1.348983488807089, "mean_class_cosine": 0.32809803411682364}}}, "iid_acc": 0.6325, "selectivity": 0.39249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3929305981197904, "mean_class_cosine": 0.31905211151253626}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.17, "drop": 0.0}, "paraphrase": {"acc": 0.17795484727755645, "drop": -0.007954847277556437, "rotation": {"subspace_principal_angle": 1.4531704709040898, "mean_class_cosine": 0.4279120214634104}}}, "iid_acc": 0.17, "selectivity": -0.06999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.5500505520500567, "mean_class_cosine": 0.6189009860707383}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.602921646746348, "drop": -0.004171646746347957}}, "iid_acc": 0.59875, "selectivity": 0.35875} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7168284789644013, "drop": 0.0044215210355986745, "rotation": {"subspace_principal_angle": 1.378307022373732, "mean_class_cosine": 0.19130281479130162}}}, "iid_acc": 0.72125, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4153437530917778, "mean_class_cosine": 0.1548272315027917}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6585760517799353, "drop": -0.036076051779935225, "rotation": {"subspace_principal_angle": 1.0555425497782547, "mean_class_cosine": 0.88892568318339}}}, "iid_acc": 0.6225, "selectivity": 0.06375000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9396491512167356, "mean_class_cosine": 0.32052389745129145}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": -0.008232200647249277}}, "iid_acc": 0.73125, "selectivity": 0.1725} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6316639741518578, "drop": 0.017086025848142228, "rotation": {"subspace_principal_angle": 1.3946146188040096, "mean_class_cosine": 0.17527167497740181}}}, "iid_acc": 0.64875, "selectivity": 0.16125000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4780475383542893, "mean_class_cosine": 0.09261586955669032}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.568659127625202, "drop": 9.087237479799004e-05, "rotation": {"subspace_principal_angle": 1.042022638358511, "mean_class_cosine": 0.9908204267164294}}}, "iid_acc": 0.56875, "selectivity": 0.08124999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.4747132575045126, "mean_class_cosine": 0.9899105548993554}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6381260096930533, "drop": 0.011873990306946691}}, "iid_acc": 0.65, "selectivity": 0.16250000000000003} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.7241379310344828, "drop": -0.016637931034482745, "rotation": {"subspace_principal_angle": 1.4200343797614488, "mean_class_cosine": 0.15019148021114892}}}, "iid_acc": 0.7075, "selectivity": 0.13124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.382096362448142, "mean_class_cosine": 0.18758209522431973}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5024630541871922, "drop": -0.01246305418719218, "rotation": {"subspace_principal_angle": 1.534087120470578, "mean_class_cosine": 0.9835950818873114}}}, "iid_acc": 0.49, "selectivity": -0.08625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0584655028541943, "mean_class_cosine": 0.7127138345764553}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7323481116584565, "drop": 0.00015188834154356012}}, "iid_acc": 0.7325, "selectivity": 0.15625} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8918169209431346, "drop": 0.0031830790568654344, "rotation": {"subspace_principal_angle": 1.2670757577041036, "mean_class_cosine": 0.29907254848700277}}}, "iid_acc": 0.895, "selectivity": 0.435, "iid_split_rotation": {"subspace_principal_angle": 1.1636722638264427, "mean_class_cosine": 0.3959701014331015}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.64125, "drop": 0.0}, "paraphrase": {"acc": 0.6310679611650486, "drop": 0.01018203883495139, "rotation": {"subspace_principal_angle": 1.2408104805968017, "mean_class_cosine": 0.9660598087989238}}}, "iid_acc": 0.64125, "selectivity": 0.18124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1355294988893803, "mean_class_cosine": 0.9278898529249071}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.9195561719833565, "drop": -0.01830617198335649}}, "iid_acc": 0.90125, "selectivity": 0.44125} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915, "rotation": {"subspace_principal_angle": 0.9875978535885324, "mean_class_cosine": 0.5506965324007853}}}, "iid_acc": 0.9925, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 0.858018142885935, "mean_class_cosine": 0.6539381215477612}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.7232289950576606, "drop": -0.03822899505766053, "rotation": {"subspace_principal_angle": 0.5872782478443009, "mean_class_cosine": 0.9996739997061842}}}, "iid_acc": 0.685, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5575352393582083, "mean_class_cosine": 0.999883744660425}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915}}, "iid_acc": 0.9925, "selectivity": 0.1975} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.6465638148667602, "drop": 0.017186185133239773, "rotation": {"subspace_principal_angle": 1.519163670668585, "mean_class_cosine": 0.05160971766617552}}}, "iid_acc": 0.66375, "selectivity": 0.0774999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4801296825592345, "mean_class_cosine": 0.0905424753256214}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.5077138849929874, "drop": -0.03021388499298744, "rotation": {"subspace_principal_angle": 1.43231424996757, "mean_class_cosine": -0.6739732724428789}}}, "iid_acc": 0.4775, "selectivity": -0.10875000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.2395084599326798, "mean_class_cosine": 0.2320322894243369}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6830294530154277, "drop": 0.011970546984572228}}, "iid_acc": 0.695, "selectivity": 0.1087499999999999} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": 0.0058126901347240745, "rotation": {"subspace_principal_angle": 1.429726953139992, "mean_class_cosine": 0.17310900825401718}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.4510331206422105, "mean_class_cosine": 0.16038164209334843}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.47115384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.423728813559322, "drop": 0.04742503259452413, "rotation": {"subspace_principal_angle": 1.1889023495577067, "mean_class_cosine": 0.5428679561508623}}}, "iid_acc": 0.47115384615384615, "selectivity": 0.08653846153846151, "iid_split_rotation": {"subspace_principal_angle": 1.560544322206414, "mean_class_cosine": 0.7853248916815403}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.027243589743589758}}, "iid_acc": 0.6394230769230769, "selectivity": 0.25480769230769224} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8835616438356164, "drop": 0.005188356164383623, "rotation": {"subspace_principal_angle": 1.4311921432320234, "mean_class_cosine": 0.13915115995570887}}}, "iid_acc": 0.88875, "selectivity": 0.21500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.3924741674351642, "mean_class_cosine": 0.17737858927148076}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.59125, "drop": 0.0}, "paraphrase": {"acc": 0.589041095890411, "drop": 0.0022089041095890716, "rotation": {"subspace_principal_angle": 1.3624580504069597, "mean_class_cosine": 0.7634106223781487}}}, "iid_acc": 0.59125, "selectivity": -0.0824999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5428225535260213, "mean_class_cosine": 0.8607299633579588}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8958904109589041, "drop": 0.010359589041095907}}, "iid_acc": 0.90625, "selectivity": 0.23250000000000004} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8224687933425797, "drop": 0.015031206657420304, "rotation": {"subspace_principal_angle": 1.33328453850938, "mean_class_cosine": 0.23528499856171853}}, "domain": {"acc": 0.66, "drop": 0.1775, "rotation": {"subspace_principal_angle": 1.451106758229533, "mean_class_cosine": 0.1194040025165285}}}, "iid_acc": 0.8375, "selectivity": 0.31625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2770908389865925, "mean_class_cosine": 0.28950101482596285}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.007765256588072145, "rotation": {"subspace_principal_angle": 1.3142819239703394, "mean_class_cosine": 0.9932995750720417}}, "domain": {"acc": 0.5, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5693408077430426, "mean_class_cosine": 0.210032160652116}}}, "iid_acc": 0.50125, "selectivity": -0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.0510108236306581, "mean_class_cosine": 0.9827971130549489}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8377253814147018, "drop": 0.008524618585298183}, "domain": {"acc": 0.83625, "drop": 0.009999999999999898}}, "iid_acc": 0.84625, "selectivity": 0.32499999999999996} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8803149606299212, "drop": 0.0009350393700787718, "rotation": {"subspace_principal_angle": 1.3448528768477728, "mean_class_cosine": 0.3368814094412147}}}, "iid_acc": 0.88125, "selectivity": 0.6575, "iid_split_rotation": {"subspace_principal_angle": 1.4525237248303589, "mean_class_cosine": 0.271959475761732}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5181102362204725, "drop": 0.024389763779527507, "rotation": {"subspace_principal_angle": 1.0916569684946897, "mean_class_cosine": 0.9392101199033003}}}, "iid_acc": 0.5425, "selectivity": 0.31875, "iid_split_rotation": {"subspace_principal_angle": 1.5156324848496125, "mean_class_cosine": 0.7159087166162079}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8724409448818897, "drop": 0.022559055118110294}}, "iid_acc": 0.895, "selectivity": 0.67125} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5384615384615384, "drop": -0.01221153846153844, "rotation": {"subspace_principal_angle": 1.4851659522030354, "mean_class_cosine": 0.08552576462723123}}}, "iid_acc": 0.52625, "selectivity": 0.03249999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.491288991251595, "mean_class_cosine": 0.07942359551958941}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.4771838331160365, "drop": -0.0009338331160365088, "rotation": {"subspace_principal_angle": 1.3966119018215262, "mean_class_cosine": -0.17561134987197125}}}, "iid_acc": 0.47625, "selectivity": -0.017500000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.5601173526903696, "mean_class_cosine": 0.2908977808248975}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5149934810951761, "drop": -0.014993481095176064}}, "iid_acc": 0.5, "selectivity": 0.006249999999999978} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.6162018592297477, "drop": 0.05254814077025227, "rotation": {"subspace_principal_angle": 1.3864574345230025, "mean_class_cosine": 0.33562554763173674}}}, "iid_acc": 0.66875, "selectivity": 0.40499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3371373671231837, "mean_class_cosine": 0.3302011789609844}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.185, "drop": 0.0}, "paraphrase": {"acc": 0.20185922974767595, "drop": -0.016859229747675952, "rotation": {"subspace_principal_angle": 1.5059232736861428, "mean_class_cosine": 0.4846447217345316}}}, "iid_acc": 0.185, "selectivity": -0.07874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1753986125618363, "mean_class_cosine": 0.6615083924419292}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6254980079681275, "drop": 0.019501992031872528}}, "iid_acc": 0.645, "selectivity": 0.38125000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7184466019417476, "drop": 0.022803398058252378, "rotation": {"subspace_principal_angle": 1.5041417980246978, "mean_class_cosine": 0.06660518398513274}}}, "iid_acc": 0.74125, "selectivity": 0.22124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4864371370902933, "mean_class_cosine": 0.08425916865336666}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5485436893203883, "drop": 0.00020631067961163208, "rotation": {"subspace_principal_angle": 1.3708104020371576, "mean_class_cosine": 0.9683596041166356}}}, "iid_acc": 0.54875, "selectivity": 0.028749999999999942, "iid_split_rotation": {"subspace_principal_angle": 1.515659944278173, "mean_class_cosine": 0.9203876968475591}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.77875, "drop": 0.0}, "paraphrase": {"acc": 0.7880258899676376, "drop": -0.009275889967637507}}, "iid_acc": 0.77875, "selectivity": 0.25875000000000004} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6155088852988692, "drop": 0.03949111470113087, "rotation": {"subspace_principal_angle": 1.5045399067665934, "mean_class_cosine": 0.06620795401218435}}}, "iid_acc": 0.655, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4647655973683582, "mean_class_cosine": 0.10583216572254955}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5201938610662359, "drop": 0.017306138933764093, "rotation": {"subspace_principal_angle": 1.402582557195184, "mean_class_cosine": 0.9741348383409788}}}, "iid_acc": 0.5375, "selectivity": 0.018749999999999933, "iid_split_rotation": {"subspace_principal_angle": 1.518487153242731, "mean_class_cosine": 0.9849570196369872}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6720516962843296, "drop": 0.021698303715670386}}, "iid_acc": 0.69375, "selectivity": 0.17499999999999993} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.7027914614121511, "drop": -0.00904146141215112, "rotation": {"subspace_principal_angle": 1.4423830102309243, "mean_class_cosine": 0.12806068526617434}}}, "iid_acc": 0.69375, "selectivity": 0.16125, "iid_split_rotation": {"subspace_principal_angle": 1.380371124548562, "mean_class_cosine": 0.1892764282980448}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.545, "drop": 0.0}, "paraphrase": {"acc": 0.5500821018062397, "drop": -0.005082101806239692, "rotation": {"subspace_principal_angle": 1.0329294851079376, "mean_class_cosine": 0.9281244210082812}}}, "iid_acc": 0.545, "selectivity": 0.012500000000000067, "iid_split_rotation": {"subspace_principal_angle": 1.1987031561182855, "mean_class_cosine": 0.701876474533382}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7323481116584565, "drop": -0.0010981116584565243}}, "iid_acc": 0.73125, "selectivity": 0.19874999999999998} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.9292649098474342, "drop": 0.005735090152565903, "rotation": {"subspace_principal_angle": 1.229191428617486, "mean_class_cosine": 0.3349996872547975}}}, "iid_acc": 0.935, "selectivity": 0.43875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0937898079150548, "mean_class_cosine": 0.4591219079682709}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6907073509015257, "drop": 0.000542649098474346, "rotation": {"subspace_principal_angle": 1.2452578371926764, "mean_class_cosine": 0.9675450369885805}}}, "iid_acc": 0.69125, "selectivity": 0.195, "iid_split_rotation": {"subspace_principal_angle": 1.2562849442009068, "mean_class_cosine": 0.9533786218456732}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9251040221914009, "drop": -0.0026040221914008788}}, "iid_acc": 0.9225, "selectivity": 0.42624999999999996} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.7, "drop": 0.0}, "paraphrase": {"acc": 0.6690042075736325, "drop": 0.030995792426367408, "rotation": {"subspace_principal_angle": 1.4968125785391955, "mean_class_cosine": 0.0739162738788901}}}, "iid_acc": 0.7, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.424774943726807, "mean_class_cosine": 0.14550301874390104}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.48625, "drop": 0.0}, "paraphrase": {"acc": 0.5133239831697055, "drop": -0.027073983169705496, "rotation": {"subspace_principal_angle": 1.473176890758434, "mean_class_cosine": -0.2335794158894724}}}, "iid_acc": 0.48625, "selectivity": -0.06749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2575303716090405, "mean_class_cosine": 0.3250973513251493}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.6975, "drop": 0.0}, "paraphrase": {"acc": 0.6760168302945302, "drop": 0.021483169705469818}}, "iid_acc": 0.6975, "selectivity": 0.14375000000000004} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": 0.010620382442416365, "rotation": {"subspace_principal_angle": 1.4206287662194153, "mean_class_cosine": 0.19628077809764047}}}, "iid_acc": 0.6490384615384616, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.455372039968678, "mean_class_cosine": 0.19590134511595567}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.39903846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.3785310734463277, "drop": 0.02050738809213387, "rotation": {"subspace_principal_angle": 0.9199382488566391, "mean_class_cosine": 0.9264117531113749}}}, "iid_acc": 0.39903846153846156, "selectivity": 0.009615384615384637, "iid_split_rotation": {"subspace_principal_angle": 1.366185399901852, "mean_class_cosine": 0.35869369801616585}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6271186440677966, "drop": 0.04115058670143412}}, "iid_acc": 0.6682692307692307, "selectivity": 0.2788461538461538} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8876712328767123, "drop": 0.01357876712328765, "rotation": {"subspace_principal_angle": 1.431717330740105, "mean_class_cosine": 0.13863106275045062}}}, "iid_acc": 0.90125, "selectivity": 0.21999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4181086678178447, "mean_class_cosine": 0.15209506900442057}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5794520547945206, "drop": 0.006797945205479494, "rotation": {"subspace_principal_angle": 0.8966790530917582, "mean_class_cosine": 0.766361309893145}}}, "iid_acc": 0.58625, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.7483984334756586, "mean_class_cosine": 0.8336256970933262}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8890410958904109, "drop": 0.018458904109589058}}, "iid_acc": 0.9075, "selectivity": 0.22624999999999995} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.835, "drop": 0.0}, "paraphrase": {"acc": 0.7877947295423023, "drop": 0.04720527045769762, "rotation": {"subspace_principal_angle": 1.2702602526749749, "mean_class_cosine": 0.29603229534182074}}, "domain": {"acc": 0.78375, "drop": 0.05125000000000002, "rotation": {"subspace_principal_angle": 1.2958495701467163, "mean_class_cosine": 0.2714956938278416}}}, "iid_acc": 0.835, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 1.1803611728107277, "mean_class_cosine": 0.38059085697680345}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.001515256588072167, "rotation": {"subspace_principal_angle": 1.2132834342045293, "mean_class_cosine": 0.9993075605412058}}, "domain": {"acc": 0.5, "drop": 0.007499999999999951, "rotation": {"subspace_principal_angle": 1.3982125329965656, "mean_class_cosine": 0.036308404316974255}}}, "iid_acc": 0.5075, "selectivity": 0.0012499999999999734, "iid_split_rotation": {"subspace_principal_angle": 1.4113229091197326, "mean_class_cosine": 0.9975285200002872}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.7877947295423023, "drop": 0.029705270457697663}, "domain": {"acc": 0.855, "drop": -0.03749999999999998}}, "iid_acc": 0.8175, "selectivity": 0.31125} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8893203883495145, "drop": 0.013179611650485423, "rotation": {"subspace_principal_angle": 1.067178302018369, "mean_class_cosine": 0.48259750926799183}}, "domain": {"acc": 0.5675, "drop": 0.33499999999999996, "rotation": {"subspace_principal_angle": 1.2593770522331844, "mean_class_cosine": 0.3064099515537772}}, "length": {"acc": 0.9095022624434389, "drop": -0.0070022624434389336, "rotation": {"subspace_principal_angle": 1.0045870056146216, "mean_class_cosine": 0.536436803133892}}}, "iid_acc": 0.9025, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.0653451564032532, "mean_class_cosine": 0.4842022438803039}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.7611650485436893, "drop": -0.06491504854368924, "rotation": {"subspace_principal_angle": 1.4069108146369018, "mean_class_cosine": 0.9395668351834272}}, "domain": {"acc": 0.57875, "drop": 0.11750000000000005, "rotation": {"subspace_principal_angle": 1.2932791063335667, "mean_class_cosine": 0.34457030192799165}}, "length": {"acc": 0.7134238310708899, "drop": -0.017173831070889878, "rotation": {"subspace_principal_angle": 1.5504433084962168, "mean_class_cosine": 0.9246357737712342}}}, "iid_acc": 0.69625, "selectivity": 0.18125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3250489130035334, "mean_class_cosine": 0.7575530733718268}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8699029126213592, "drop": 0.030097087378640808}, "domain": {"acc": 0.515, "drop": 0.385}, "length": {"acc": 0.9034690799396682, "drop": -0.0034690799396681626}}, "iid_acc": 0.9, "selectivity": 0.385} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8881889763779528, "drop": 0.006811023622047219, "rotation": {"subspace_principal_angle": 1.2507965877486584, "mean_class_cosine": 0.4519091647049873}}}, "iid_acc": 0.895, "selectivity": 0.635, "iid_split_rotation": {"subspace_principal_angle": 1.2538562323332731, "mean_class_cosine": 0.4436394901742749}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.3725, "drop": 0.0}, "paraphrase": {"acc": 0.36220472440944884, "drop": 0.010295275590551156, "rotation": {"subspace_principal_angle": 1.1149135781556565, "mean_class_cosine": 0.9771677078579994}}}, "iid_acc": 0.3725, "selectivity": 0.11249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1275037300836335, "mean_class_cosine": 0.5054561634322786}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8771653543307086, "drop": 0.009084645669291347}}, "iid_acc": 0.88625, "selectivity": 0.62625} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9571788413098237, "drop": 0.010321158690176357, "rotation": {"subspace_principal_angle": 1.1895811374243113, "mean_class_cosine": 0.6930412270179037}}}, "iid_acc": 0.9675, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.1204246248516365, "mean_class_cosine": 0.734504756933933}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.28125, "drop": 0.0}, "paraphrase": {"acc": 0.3047858942065491, "drop": -0.023535894206549113, "rotation": {"subspace_principal_angle": 1.566917995749759, "mean_class_cosine": 0.7334608100976151}}}, "iid_acc": 0.28125, "selectivity": 0.22, "iid_split_rotation": {"subspace_principal_angle": 1.5646555108195546, "mean_class_cosine": 0.8084161094056437}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9445843828715366, "drop": 0.010415617128463395}}, "iid_acc": 0.955, "selectivity": 0.8937499999999999} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.60625, "drop": 0.0}, "paraphrase": {"acc": 0.6375488917861799, "drop": -0.03129889178617995, "rotation": {"subspace_principal_angle": 1.2380910520249748, "mean_class_cosine": 0.3266011440007731}}}, "iid_acc": 0.60625, "selectivity": 0.08499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2935899170666836, "mean_class_cosine": 0.2736697785737916}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.4915254237288136, "drop": -0.014025423728813602, "rotation": {"subspace_principal_angle": 1.481666093252446, "mean_class_cosine": -0.9406980138297385}}}, "iid_acc": 0.4775, "selectivity": -0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2591046492876659, "mean_class_cosine": 0.7431471438900024}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5019556714471969, "drop": 0.011794328552803157}}, "iid_acc": 0.51375, "selectivity": -0.007499999999999951} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.5962815405046481, "drop": -0.006281540504648131, "rotation": {"subspace_principal_angle": 1.3616817945475932, "mean_class_cosine": 0.34929473551176465}}}, "iid_acc": 0.59, "selectivity": 0.33499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3738625012283205, "mean_class_cosine": 0.31435666695467485}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.07625, "drop": 0.0}, "paraphrase": {"acc": 0.07702523240371846, "drop": -0.0007752324037184621, "rotation": {"subspace_principal_angle": 1.4014772128715691, "mean_class_cosine": -0.26880588026423574}}}, "iid_acc": 0.07625, "selectivity": -0.17875000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.2922523705095328, "mean_class_cosine": 0.5906582844419805}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.49125, "drop": 0.0}, "paraphrase": {"acc": 0.4913678618857902, "drop": -0.00011786188579016033}}, "iid_acc": 0.49125, "selectivity": 0.23625000000000002} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.765, "drop": 0.0}, "paraphrase": {"acc": 0.7686084142394822, "drop": -0.003608414239482216, "rotation": {"subspace_principal_angle": 1.267703414249137, "mean_class_cosine": 0.2984735606750531}}}, "iid_acc": 0.765, "selectivity": 0.21250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.297520243254995, "mean_class_cosine": 0.26988739363466807}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5355987055016181, "drop": 0.0019012944983818336, "rotation": {"subspace_principal_angle": 1.4289319500576079, "mean_class_cosine": 0.9995585763899515}}}, "iid_acc": 0.5375, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.095765507578661, "mean_class_cosine": 0.9992212510316336}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6828478964401294, "drop": 0.012152103559870508}}, "iid_acc": 0.695, "selectivity": 0.14249999999999996} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6865912762520194, "drop": 0.007158723747980544, "rotation": {"subspace_principal_angle": 1.3189864885611238, "mean_class_cosine": 0.2491571278556936}}}, "iid_acc": 0.69375, "selectivity": 0.19374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3741028601612726, "mean_class_cosine": 0.19542762781810974}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5137318255250404, "drop": 0.015018174474959678, "rotation": {"subspace_principal_angle": 1.410712709591301, "mean_class_cosine": 0.9996250431090286}}}, "iid_acc": 0.52875, "selectivity": 0.028750000000000053, "iid_split_rotation": {"subspace_principal_angle": 0.9117211516634655, "mean_class_cosine": 0.9997105404220764}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.592891760904685, "drop": 0.018358239095314977}}, "iid_acc": 0.61125, "selectivity": 0.11124999999999996} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.6995073891625616, "drop": 0.016742610837438443, "rotation": {"subspace_principal_angle": 1.398637777187281, "mean_class_cosine": 0.17130938688822317}}}, "iid_acc": 0.71625, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4203666415350267, "mean_class_cosine": 0.14986297902276266}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.4975369458128079, "drop": -0.0012869458128078604, "rotation": {"subspace_principal_angle": 1.5334169157311857, "mean_class_cosine": 0.9992809010722258}}}, "iid_acc": 0.49625, "selectivity": -0.08625, "iid_split_rotation": {"subspace_principal_angle": 1.523586797466333, "mean_class_cosine": 0.9461897010476042}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6765188834154351, "drop": 0.01473111658456494}}, "iid_acc": 0.69125, "selectivity": 0.10875000000000001} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.9275, "drop": 0.0}, "paraphrase": {"acc": 0.9181692094313454, "drop": 0.009330790568654601, "rotation": {"subspace_principal_angle": 1.0867093987105656, "mean_class_cosine": 0.4654003933713694}}}, "iid_acc": 0.9275, "selectivity": 0.45875, "iid_split_rotation": {"subspace_principal_angle": 1.0076905925780175, "mean_class_cosine": 0.533814981080676}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5312066574202496, "drop": 4.33425797503606e-05, "rotation": {"subspace_principal_angle": 0.9422006877595401, "mean_class_cosine": 0.9720210108018141}}}, "iid_acc": 0.53125, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 0.9637054547388073, "mean_class_cosine": 0.8521486536674154}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.8987517337031901, "drop": 0.016248266296809977}}, "iid_acc": 0.915, "selectivity": 0.44625000000000004} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9967051070840197, "drop": -0.0017051070840197191, "rotation": {"subspace_principal_angle": 1.004243983718731, "mean_class_cosine": 0.5367262616365928}}}, "iid_acc": 0.995, "selectivity": 0.1725, "iid_split_rotation": {"subspace_principal_angle": 0.8818699720917821, "mean_class_cosine": 0.6357087708516111}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.7166392092257001, "drop": -0.021639209225700173, "rotation": {"subspace_principal_angle": 1.3903038602930748, "mean_class_cosine": 0.999942812375732}}}, "iid_acc": 0.695, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2573644298943418, "mean_class_cosine": 0.9999792008680795}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915}}, "iid_acc": 0.9925, "selectivity": 0.17000000000000004} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6451612903225806, "drop": 0.03733870967741937, "rotation": {"subspace_principal_angle": 1.534693889353481, "mean_class_cosine": 0.03609459538397165}}}, "iid_acc": 0.6825, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.402974256327293, "mean_class_cosine": 0.16703541534575633}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.47375, "drop": 0.0}, "paraphrase": {"acc": 0.49789621318373073, "drop": -0.024146213183730725, "rotation": {"subspace_principal_angle": 1.4387989610109857, "mean_class_cosine": -0.9029123388757152}}}, "iid_acc": 0.47375, "selectivity": -0.12749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5531483539033937, "mean_class_cosine": -0.16026020693951765}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.7124824684431977, "drop": -0.011232468443197696}}, "iid_acc": 0.70125, "selectivity": 0.10000000000000009} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6779661016949152, "drop": -0.009696870925684498, "rotation": {"subspace_principal_angle": 1.3728662084585508, "mean_class_cosine": 0.23801349724373286}}}, "iid_acc": 0.6682692307692307, "selectivity": 0.23076923076923073, "iid_split_rotation": {"subspace_principal_angle": 1.3852570180024775, "mean_class_cosine": 0.240802390739838}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.49038461538461536, "drop": 0.0}, "paraphrase": {"acc": 0.4124293785310734, "drop": 0.07795523685354194, "rotation": {"subspace_principal_angle": 1.093285198742622, "mean_class_cosine": 0.5340692801147456}}}, "iid_acc": 0.49038461538461536, "selectivity": 0.05288461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.1837899565316878, "mean_class_cosine": 0.4155974505600208}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.6298076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6497175141242938, "drop": -0.019909821816601503}}, "iid_acc": 0.6298076923076923, "selectivity": 0.1923076923076923} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.873972602739726, "drop": 0.026027397260274032, "rotation": {"subspace_principal_angle": 1.2965339594699503, "mean_class_cosine": 0.2708369468728177}}}, "iid_acc": 0.9, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.2862423961352993, "mean_class_cosine": 0.28072934769702657}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.575, "drop": 0.0}, "paraphrase": {"acc": 0.5671232876712329, "drop": 0.007876712328767077, "rotation": {"subspace_principal_angle": 1.0126162968840653, "mean_class_cosine": 0.687860663967105}}}, "iid_acc": 0.575, "selectivity": -0.15375000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.5178918149259684, "mean_class_cosine": 0.7969393733558711}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8575342465753425, "drop": 0.003715753424657464}}, "iid_acc": 0.86125, "selectivity": 0.13249999999999995} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7098591549295775, "drop": 0.04639084507042246, "rotation": {"subspace_principal_angle": 1.285187527092102, "mean_class_cosine": 0.2817416408844383}}, "domain": {"acc": 0.71375, "drop": 0.04249999999999998, "rotation": {"subspace_principal_angle": 1.3127911813868958, "mean_class_cosine": 0.25515223416644334}}}, "iid_acc": 0.75625, "selectivity": 0.22124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2283794043097545, "mean_class_cosine": 0.33576468083719035}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5253521126760563, "drop": -0.026602112676056278, "rotation": {"subspace_principal_angle": 1.214807789184398, "mean_class_cosine": 0.9831397387284473}}, "domain": {"acc": 0.4875, "drop": 0.011250000000000038, "rotation": {"subspace_principal_angle": 1.5537344695259054, "mean_class_cosine": 0.34478575371993236}}}, "iid_acc": 0.49875, "selectivity": -0.036250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1073273476869623, "mean_class_cosine": 0.9951775418509261}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7126760563380282, "drop": 0.018573943661971737}, "domain": {"acc": 0.7225, "drop": 0.008749999999999925}}, "iid_acc": 0.73125, "selectivity": 0.19624999999999992} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.82, "drop": 0.0}, "paraphrase": {"acc": 0.8269961977186312, "drop": -0.006996197718631247, "rotation": {"subspace_principal_angle": 1.0415904118081376, "mean_class_cosine": 0.5048480397255118}}, "domain": {"acc": 0.59375, "drop": 0.22624999999999995, "rotation": {"subspace_principal_angle": 1.2834570977197297, "mean_class_cosine": 0.28340154844279125}}, "length": {"acc": 0.8264840182648402, "drop": -0.006484018264840241, "rotation": {"subspace_principal_angle": 0.999883755381926, "mean_class_cosine": 0.5404001186906672}}}, "iid_acc": 0.82, "selectivity": 0.3549999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9949686563230938, "mean_class_cosine": 0.5445291790193694}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6863117870722434, "drop": 0.018688212927756576, "rotation": {"subspace_principal_angle": 1.3283778591084485, "mean_class_cosine": 0.9327609071818541}}, "domain": {"acc": 0.565, "drop": 0.14, "rotation": {"subspace_principal_angle": 1.5278891311014122, "mean_class_cosine": 0.3806568220146318}}, "length": {"acc": 0.7062404870624048, "drop": -0.0012404870624048714, "rotation": {"subspace_principal_angle": 1.3125076791898616, "mean_class_cosine": 0.9017445724833599}}}, "iid_acc": 0.705, "selectivity": 0.23999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4606505742125795, "mean_class_cosine": 0.9081384351848174}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.7975, "drop": 0.0}, "paraphrase": {"acc": 0.7984790874524715, "drop": -0.0009790874524715498}, "domain": {"acc": 0.645, "drop": 0.15249999999999997}, "length": {"acc": 0.7975646879756468, "drop": -6.46879756468488e-05}}, "iid_acc": 0.7975, "selectivity": 0.33249999999999996} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8611544461778471, "drop": 0.0013455538221529562, "rotation": {"subspace_principal_angle": 0.9939269073783296, "mean_class_cosine": 0.6453866739528615}}}, "iid_acc": 0.8625, "selectivity": 0.5362500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4271388898728552, "mean_class_cosine": 0.6524369913147846}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7316692667706708, "drop": 0.00458073322932917, "rotation": {"subspace_principal_angle": 1.3724063354724922, "mean_class_cosine": 0.9469017405917457}}}, "iid_acc": 0.73625, "selectivity": 0.41, "iid_split_rotation": {"subspace_principal_angle": 1.0603683518688838, "mean_class_cosine": 0.950801823774931}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.81625, "drop": 0.0}, "paraphrase": {"acc": 0.8081123244929798, "drop": 0.00813767550702027}}, "iid_acc": 0.81625, "selectivity": 0.49000000000000005} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9160671462829736, "drop": 0.021432853717026412, "rotation": {"subspace_principal_angle": 1.1847135524986785, "mean_class_cosine": 0.706584730242114}}}, "iid_acc": 0.9375, "selectivity": 0.87625, "iid_split_rotation": {"subspace_principal_angle": 1.11150260490418, "mean_class_cosine": 0.7556327352536992}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.4460431654676259, "drop": 0.05520683453237407, "rotation": {"subspace_principal_angle": 1.199717400716062, "mean_class_cosine": 0.8611479821783318}}}, "iid_acc": 0.50125, "selectivity": 0.43999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4104637914876992, "mean_class_cosine": 0.8005163008883979}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8465227817745803, "drop": 0.03222721822541974}}, "iid_acc": 0.87875, "selectivity": 0.8175} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.515032679738562, "drop": 0.009967320261437984, "rotation": {"subspace_principal_angle": 1.566988880756832, "mean_class_cosine": 0.003807436838872146}}}, "iid_acc": 0.525, "selectivity": 0.03750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.515059781737158, "mean_class_cosine": 0.055707691363518934}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5189542483660131, "drop": 0.00729575163398688, "rotation": {"subspace_principal_angle": 1.565436854816222, "mean_class_cosine": 0.11469077846809743}}}, "iid_acc": 0.52625, "selectivity": 0.03875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.519484137603478, "mean_class_cosine": 0.056727235397258546}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5124183006535947, "drop": 0.023831699346405277}}, "iid_acc": 0.53625, "selectivity": 0.048750000000000016} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5485175202156334, "drop": 0.023982479784366628, "rotation": {"subspace_principal_angle": 1.4303692527758758, "mean_class_cosine": 0.35644155459058985}}}, "iid_acc": 0.5725, "selectivity": 0.2575, "iid_split_rotation": {"subspace_principal_angle": 1.4149642770924389, "mean_class_cosine": 0.3618340196725102}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.28125, "drop": 0.0}, "paraphrase": {"acc": 0.28975741239892183, "drop": -0.008507412398921832, "rotation": {"subspace_principal_angle": 1.5249290421568074, "mean_class_cosine": 0.644382717112205}}}, "iid_acc": 0.28125, "selectivity": -0.03375, "iid_split_rotation": {"subspace_principal_angle": 1.5382442039631525, "mean_class_cosine": 0.5744283723940081}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5053908355795148, "drop": 0.009609164420485206}}, "iid_acc": 0.515, "selectivity": 0.2} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.7345575959933222, "drop": -0.01080759599332215, "rotation": {"subspace_principal_angle": 1.2514538362385088, "mean_class_cosine": 0.31394236141299214}}}, "iid_acc": 0.72375, "selectivity": 0.22499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3629405243520136, "mean_class_cosine": 0.20636233074920893}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.6060100166944908, "drop": -0.0297600166944908, "rotation": {"subspace_principal_angle": 1.4254734942092195, "mean_class_cosine": 0.930851220197922}}}, "iid_acc": 0.57625, "selectivity": 0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5027930622930314, "mean_class_cosine": 0.8705083526841852}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7312186978297162, "drop": 0.003781302170283829}}, "iid_acc": 0.735, "selectivity": 0.23624999999999996} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6237623762376238, "drop": 0.058737623762376234, "rotation": {"subspace_principal_angle": 1.3701462267495714, "mean_class_cosine": 0.19930643009504329}}}, "iid_acc": 0.6825, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3138345624755305, "mean_class_cosine": 0.25414324935789645}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.6072607260726073, "drop": -0.007260726072607304, "rotation": {"subspace_principal_angle": 1.420901469694498, "mean_class_cosine": 0.9343142755680263}}}, "iid_acc": 0.6, "selectivity": 0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2228313385316762, "mean_class_cosine": 0.8726714354786228}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.66625, "drop": 0.0}, "paraphrase": {"acc": 0.6254125412541254, "drop": 0.04083745874587463}}, "iid_acc": 0.66625, "selectivity": 0.17625000000000002} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7648902821316614, "drop": -0.03364028213166148, "rotation": {"subspace_principal_angle": 1.1748420004300262, "mean_class_cosine": 0.38568885340073633}}}, "iid_acc": 0.73125, "selectivity": 0.10624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1987833070239917, "mean_class_cosine": 0.3634914913995175}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5094043887147336, "drop": 0.020595611285266435, "rotation": {"subspace_principal_angle": 1.3931515700293393, "mean_class_cosine": 0.8341203196581115}}}, "iid_acc": 0.53, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.5409782777316359, "mean_class_cosine": 0.7395207496379319}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7523510971786834, "drop": -0.026101097178683474}}, "iid_acc": 0.72625, "selectivity": 0.10124999999999995} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.8688981868898187, "drop": 0.046101813110181356, "rotation": {"subspace_principal_angle": 1.0974938475903255, "mean_class_cosine": 0.45582819608799463}}}, "iid_acc": 0.915, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 0.9804764137968498, "mean_class_cosine": 0.5566268231619027}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.8125, "drop": 0.0}, "paraphrase": {"acc": 0.6443514644351465, "drop": 0.16814853556485354, "rotation": {"subspace_principal_angle": 1.3274674971100646, "mean_class_cosine": 0.8179461473320861}}}, "iid_acc": 0.8125, "selectivity": 0.25125, "iid_split_rotation": {"subspace_principal_angle": 1.311349815093557, "mean_class_cosine": 0.9540198561738099}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8521617852161785, "drop": 0.04783821478382155}}, "iid_acc": 0.9, "selectivity": 0.33875} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9852700490998363, "drop": 0.0059799509001636775, "rotation": {"subspace_principal_angle": 0.6333378965948789, "mean_class_cosine": 0.8060565063456264}}}, "iid_acc": 0.99125, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 0.6062508845786333, "mean_class_cosine": 0.8217899986183517}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8772504091653028, "drop": 0.0027495908346971687, "rotation": {"subspace_principal_angle": 1.13183836457679, "mean_class_cosine": 0.9746417417399029}}}, "iid_acc": 0.88, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.0774674925239829, "mean_class_cosine": 0.989696915490704}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9885433715220949, "drop": 0.00020662847790509087}}, "iid_acc": 0.98875, "selectivity": 0.11750000000000005} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.7, "drop": -0.022499999999999964, "rotation": {"subspace_principal_angle": 1.493482753933245, "mean_class_cosine": 0.0772365736671137}}}, "iid_acc": 0.6775, "selectivity": 0.012499999999999956, "iid_split_rotation": {"subspace_principal_angle": 1.5233694038561607, "mean_class_cosine": 0.0474091452723575}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.523943661971831, "drop": 0.003556338028169015, "rotation": {"subspace_principal_angle": 1.364642797562692, "mean_class_cosine": 0.26221087675633176}}}, "iid_acc": 0.5275, "selectivity": -0.13750000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.5645617307273036, "mean_class_cosine": 0.0046736400727034775}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7492957746478873, "drop": -0.019295774647887343}}, "iid_acc": 0.73, "selectivity": 0.06499999999999995} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5722543352601156, "drop": -0.01936971987550018, "rotation": {"subspace_principal_angle": 1.4113114437701857, "mean_class_cosine": 0.18732090600899934}}}, "iid_acc": 0.5528846153846154, "selectivity": 0.1586538461538462, "iid_split_rotation": {"subspace_principal_angle": 1.4113877250233597, "mean_class_cosine": 0.1900533891006572}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.41346153846153844, "drop": 0.0}, "paraphrase": {"acc": 0.44508670520231214, "drop": -0.0316251667407737, "rotation": {"subspace_principal_angle": 1.4614748229703205, "mean_class_cosine": 0.41743847989205585}}}, "iid_acc": 0.41346153846153844, "selectivity": 0.019230769230769218, "iid_split_rotation": {"subspace_principal_angle": 1.1472146976181432, "mean_class_cosine": 0.7289310712969407}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.5336538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": -0.044380835927078754}}, "iid_acc": 0.5336538461538461, "selectivity": 0.13942307692307693} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8848821081830791, "drop": -0.009882108183079108, "rotation": {"subspace_principal_angle": 1.2416497520853838, "mean_class_cosine": 0.32323553402514815}}}, "iid_acc": 0.875, "selectivity": 0.11124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1798700863789469, "mean_class_cosine": 0.3810449400671245}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5423023578363384, "drop": 0.022697642163661547, "rotation": {"subspace_principal_angle": 1.0521118048612201, "mean_class_cosine": 0.46745139889316134}}}, "iid_acc": 0.565, "selectivity": -0.1987500000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8706941393205756, "mean_class_cosine": 0.8476985688696614}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.84875, "drop": 0.0}, "paraphrase": {"acc": 0.855755894590846, "drop": -0.007005894590845996}}, "iid_acc": 0.84875, "selectivity": 0.08499999999999996} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.81375, "drop": 0.0}, "paraphrase": {"acc": 0.7830985915492957, "drop": 0.030651408450704243, "rotation": {"subspace_principal_angle": 1.2474741652442436, "mean_class_cosine": 0.3177183323280257}}, "domain": {"acc": 0.7875, "drop": 0.026249999999999996, "rotation": {"subspace_principal_angle": 1.1724916667604595, "mean_class_cosine": 0.38785627191420236}}}, "iid_acc": 0.81375, "selectivity": 0.27125, "iid_split_rotation": {"subspace_principal_angle": 1.2460595658666687, "mean_class_cosine": 0.31905931609907595}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5450704225352113, "drop": -0.018820422535211323, "rotation": {"subspace_principal_angle": 1.217800297410841, "mean_class_cosine": 0.9938492827921086}}, "domain": {"acc": 0.4875, "drop": 0.03875000000000001, "rotation": {"subspace_principal_angle": 1.4265907543600744, "mean_class_cosine": 0.524717137763018}}}, "iid_acc": 0.52625, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.0792601449175692, "mean_class_cosine": 0.9975233270077284}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.7887323943661971, "drop": 0.03376760563380288}, "domain": {"acc": 0.78875, "drop": 0.03375000000000006}}, "iid_acc": 0.8225, "selectivity": 0.28} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8269961977186312, "drop": 0.02675380228136881, "rotation": {"subspace_principal_angle": 1.0856620278217086, "mean_class_cosine": 0.46632716618126924}}, "domain": {"acc": 0.59125, "drop": 0.26249999999999996, "rotation": {"subspace_principal_angle": 1.2546499825836788, "mean_class_cosine": 0.3109062073482294}}, "length": {"acc": 0.8508371385083714, "drop": 0.0029128614916286155, "rotation": {"subspace_principal_angle": 1.0562705758313071, "mean_class_cosine": 0.492122058178101}}}, "iid_acc": 0.85375, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.04571866168401, "mean_class_cosine": 0.5012802086419837}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.6844106463878327, "drop": -0.03316064638783267, "rotation": {"subspace_principal_angle": 1.2624327988110398, "mean_class_cosine": 0.9443739768139503}}, "domain": {"acc": 0.55625, "drop": 0.09499999999999997, "rotation": {"subspace_principal_angle": 1.2485837708508047, "mean_class_cosine": 0.5140204613064764}}, "length": {"acc": 0.6727549467275494, "drop": -0.021504946727549434, "rotation": {"subspace_principal_angle": 1.3927565265280373, "mean_class_cosine": 0.891203465902359}}}, "iid_acc": 0.65125, "selectivity": 0.17375000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4689242704183265, "mean_class_cosine": 0.9409811928992509}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8136882129277566, "drop": 0.023811787072243407}, "domain": {"acc": 0.6275, "drop": 0.21000000000000008}, "length": {"acc": 0.8340943683409436, "drop": 0.003405631659056385}}, "iid_acc": 0.8375, "selectivity": 0.36000000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8533541341653667, "drop": 0.0066458658346333355, "rotation": {"subspace_principal_angle": 1.3730733013635024, "mean_class_cosine": 0.2667006488076803}}}, "iid_acc": 0.86, "selectivity": 0.59125, "iid_split_rotation": {"subspace_principal_angle": 1.4307610404327311, "mean_class_cosine": 0.277445881216753}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": -0.006445007800311986, "rotation": {"subspace_principal_angle": 1.4820823089446629, "mean_class_cosine": 0.9745838181982655}}}, "iid_acc": 0.86875, "selectivity": 0.6000000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5380315983443016, "mean_class_cosine": 0.9691609768158005}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.40875, "drop": 0.0}, "paraphrase": {"acc": 0.38221528861154447, "drop": 0.026534711388455534}}, "iid_acc": 0.40875, "selectivity": 0.14} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.935251798561151, "drop": 0.012248201438848971, "rotation": {"subspace_principal_angle": 1.1046415070501514, "mean_class_cosine": 0.6973108185521933}}}, "iid_acc": 0.9475, "selectivity": 0.87375, "iid_split_rotation": {"subspace_principal_angle": 1.0670730680867, "mean_class_cosine": 0.747139744826814}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4628297362110312, "drop": 0.03967026378896876, "rotation": {"subspace_principal_angle": 1.5435774572835343, "mean_class_cosine": 0.8716859158620075}}}, "iid_acc": 0.5025, "selectivity": 0.42874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.389055684030504, "mean_class_cosine": 0.7772996017727697}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.92625, "drop": 0.0}, "paraphrase": {"acc": 0.9160671462829736, "drop": 0.01018285371702643}}, "iid_acc": 0.92625, "selectivity": 0.8525} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5176470588235295, "drop": 0.018602941176470544, "rotation": {"subspace_principal_angle": 1.5270884311030197, "mean_class_cosine": 0.04369398057174439}}}, "iid_acc": 0.53625, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.476543368793087, "mean_class_cosine": 0.094113468729472}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.5045751633986928, "drop": 0.011674836601307192, "rotation": {"subspace_principal_angle": 1.5294742379185886, "mean_class_cosine": 0.18570440239474292}}}, "iid_acc": 0.51625, "selectivity": 0.04249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.444417852756359, "mean_class_cosine": 0.20646522396999623}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5333333333333333, "drop": 0.0029166666666666785}}, "iid_acc": 0.53625, "selectivity": 0.0625} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.5431266846361186, "drop": 0.0518733153638814, "rotation": {"subspace_principal_angle": 1.467102949540591, "mean_class_cosine": 0.29891759428990633}}}, "iid_acc": 0.595, "selectivity": 0.31375, "iid_split_rotation": {"subspace_principal_angle": 1.5242732907156675, "mean_class_cosine": 0.27641104971513886}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.1925, "drop": 0.0}, "paraphrase": {"acc": 0.20754716981132076, "drop": -0.01504716981132076, "rotation": {"subspace_principal_angle": 1.3006810428081945, "mean_class_cosine": 0.5938365988587205}}}, "iid_acc": 0.1925, "selectivity": -0.08875, "iid_split_rotation": {"subspace_principal_angle": 1.5008582164851936, "mean_class_cosine": 0.6730802808549522}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5309973045822103, "drop": 0.017752695417789677}}, "iid_acc": 0.54875, "selectivity": 0.26749999999999996} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7429048414023373, "drop": -0.002904841402337266, "rotation": {"subspace_principal_angle": 1.321041649041896, "mean_class_cosine": 0.24716625603971074}}}, "iid_acc": 0.74, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3832981091452556, "mean_class_cosine": 0.18640154565031142}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5976627712854758, "drop": -0.03766277128547579, "rotation": {"subspace_principal_angle": 1.3172594721297197, "mean_class_cosine": 0.9641971900851679}}}, "iid_acc": 0.56, "selectivity": 0.0525000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1295344939397884, "mean_class_cosine": 0.9754037717574164}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7378964941569283, "drop": -0.009146494156928253}}, "iid_acc": 0.72875, "selectivity": 0.22125000000000006} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6551155115511551, "drop": 0.01988448844884494, "rotation": {"subspace_principal_angle": 1.4103931207701927, "mean_class_cosine": 0.1597162496412831}}}, "iid_acc": 0.675, "selectivity": 0.18500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.451975006440924, "mean_class_cosine": 0.1185419210711762}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5396039603960396, "drop": 0.0016460396039603697, "rotation": {"subspace_principal_angle": 1.5446898330353456, "mean_class_cosine": 0.9934586731916171}}}, "iid_acc": 0.54125, "selectivity": 0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3541190216750216, "mean_class_cosine": 0.9952103287013051}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.6975, "drop": 0.0}, "paraphrase": {"acc": 0.6617161716171617, "drop": 0.035783828382838334}}, "iid_acc": 0.6975, "selectivity": 0.20750000000000002} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7523510971786834, "drop": -0.00735109717868343, "rotation": {"subspace_principal_angle": 1.3838768695601789, "mean_class_cosine": 0.18583289760856309}}}, "iid_acc": 0.745, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.411496556459018, "mean_class_cosine": 0.1586268817831123}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.485, "drop": 0.0}, "paraphrase": {"acc": 0.5219435736677116, "drop": -0.036943573667711616, "rotation": {"subspace_principal_angle": 1.4598677944473912, "mean_class_cosine": 0.7189138058825407}}}, "iid_acc": 0.485, "selectivity": -0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.257226644633361, "mean_class_cosine": 0.8721290560249453}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7507836990595611, "drop": 0.005466300940438873}}, "iid_acc": 0.75625, "selectivity": 0.17374999999999996} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9121338912133892, "drop": 0.010366108786610817, "rotation": {"subspace_principal_angle": 1.190236906942168, "mean_class_cosine": 0.3714399247796692}}}, "iid_acc": 0.9225, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1180814498000364, "mean_class_cosine": 0.43740853126331475}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.8019525801952581, "drop": 0.0642974198047419, "rotation": {"subspace_principal_angle": 1.316257361185483, "mean_class_cosine": 0.9428379882047717}}}, "iid_acc": 0.86625, "selectivity": 0.33124999999999993, "iid_split_rotation": {"subspace_principal_angle": 0.8984797392010688, "mean_class_cosine": 0.9658337076254448}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.9107391910739191, "drop": 0.009260808926080943}}, "iid_acc": 0.92, "selectivity": 0.385} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9950900163666121, "drop": 0.0011599836333878732, "rotation": {"subspace_principal_angle": 0.8576615569303944, "mean_class_cosine": 0.6542078543716676}}}, "iid_acc": 0.99625, "selectivity": 0.15000000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9012913552860777, "mean_class_cosine": 0.6205978969084885}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.68125, "drop": 0.0}, "paraphrase": {"acc": 0.5548281505728314, "drop": 0.1264218494271686, "rotation": {"subspace_principal_angle": 0.6861186089833221, "mean_class_cosine": 0.9976598565566228}}}, "iid_acc": 0.68125, "selectivity": -0.16499999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.6195594346223908, "mean_class_cosine": 0.9998339526246938}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": -0.0009533551554827824}}, "iid_acc": 0.9925, "selectivity": 0.1462500000000001} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6915492957746479, "drop": -0.004049295774647854, "rotation": {"subspace_principal_angle": 1.4821835383859001, "mean_class_cosine": 0.08849686598665059}}}, "iid_acc": 0.6875, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5422652050607666, "mean_class_cosine": 0.0285272510510537}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5140845070422535, "drop": 0.03966549295774646, "rotation": {"subspace_principal_angle": 1.411758063021396, "mean_class_cosine": -0.05543427344626703}}}, "iid_acc": 0.55375, "selectivity": -0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1826722547195974, "mean_class_cosine": -0.8190873052106897}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7450704225352113, "drop": -0.01757042253521124}}, "iid_acc": 0.7275, "selectivity": 0.08875} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5673076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.5549132947976878, "drop": 0.012394397510004485, "rotation": {"subspace_principal_angle": 1.435539235192927, "mean_class_cosine": 0.18347039301699317}}}, "iid_acc": 0.5673076923076923, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.3743463171910144, "mean_class_cosine": 0.24089153525080556}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.4326923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.41040462427745666, "drop": 0.022287683414851045, "rotation": {"subspace_principal_angle": 1.3988658583948614, "mean_class_cosine": -0.27577383548106055}}}, "iid_acc": 0.4326923076923077, "selectivity": 0.04326923076923078, "iid_split_rotation": {"subspace_principal_angle": 1.0687827569428319, "mean_class_cosine": 0.8344718355538371}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5953757225433526, "drop": -0.04249110715873716}}, "iid_acc": 0.5528846153846154, "selectivity": 0.1634615384615385} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.9029126213592233, "drop": -0.006662621359223353, "rotation": {"subspace_principal_angle": 1.20453886521433, "mean_class_cosine": 0.3581236367027224}}}, "iid_acc": 0.89625, "selectivity": 0.17374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2309341256728386, "mean_class_cosine": 0.3333571784399262}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5298196948682385, "drop": 0.022680305131761447, "rotation": {"subspace_principal_angle": 1.310916058135534, "mean_class_cosine": -0.07981176620814859}}}, "iid_acc": 0.5525, "selectivity": -0.17000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1896244566676593, "mean_class_cosine": 0.9512750131760109}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.8575, "drop": 0.0}, "paraphrase": {"acc": 0.8640776699029126, "drop": -0.006577669902912531}}, "iid_acc": 0.8575, "selectivity": 0.135} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8056338028169014, "drop": 0.048116197183098564, "rotation": {"subspace_principal_angle": 1.2927332146049364, "mean_class_cosine": 0.2744936747860658}}, "domain": {"acc": 0.82375, "drop": 0.030000000000000027, "rotation": {"subspace_principal_angle": 1.2846999556990648, "mean_class_cosine": 0.28220942740249133}}}, "iid_acc": 0.85375, "selectivity": 0.32125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.220104644031074, "mean_class_cosine": 0.3435474732924775}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.532394366197183, "drop": -0.012394366197183038, "rotation": {"subspace_principal_angle": 0.365026570963859, "mean_class_cosine": 0.9987360435705452}}, "domain": {"acc": 0.4875, "drop": 0.03250000000000003, "rotation": {"subspace_principal_angle": 1.5449752153763945, "mean_class_cosine": 0.4557918147259487}}}, "iid_acc": 0.52, "selectivity": -0.012499999999999956, "iid_split_rotation": {"subspace_principal_angle": 0.34709256477040135, "mean_class_cosine": 0.9995894414380515}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.8225352112676056, "drop": 0.022464788732394325}, "domain": {"acc": 0.8325, "drop": 0.012499999999999956}}, "iid_acc": 0.845, "selectivity": 0.3125} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8688212927756654, "drop": 0.016178707224334565, "rotation": {"subspace_principal_angle": 1.2184200366683278, "mean_class_cosine": 0.34512905932789906}}, "domain": {"acc": 0.7325, "drop": 0.15249999999999997, "rotation": {"subspace_principal_angle": 1.3710819230300715, "mean_class_cosine": 0.1983894193717613}}, "length": {"acc": 0.8919330289193302, "drop": -0.006933028919330231, "rotation": {"subspace_principal_angle": 1.1635770628342312, "mean_class_cosine": 0.3960575192257396}}}, "iid_acc": 0.885, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 1.178058815472859, "mean_class_cosine": 0.38271893642492194}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7585551330798479, "drop": -0.02605513307984786, "rotation": {"subspace_principal_angle": 1.0808603554768208, "mean_class_cosine": 0.9645911983303304}}, "domain": {"acc": 0.5575, "drop": 0.17500000000000004, "rotation": {"subspace_principal_angle": 1.4332455458492723, "mean_class_cosine": 0.4266662959732691}}, "length": {"acc": 0.7595129375951294, "drop": -0.02701293759512935, "rotation": {"subspace_principal_angle": 1.1028359956060956, "mean_class_cosine": 0.9361757660470457}}}, "iid_acc": 0.7325, "selectivity": 0.25375000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.340774604922017, "mean_class_cosine": 0.9544915504856384}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.876425855513308, "drop": 0.01732414448669206}, "domain": {"acc": 0.57875, "drop": 0.31500000000000006}, "length": {"acc": 0.898021308980213, "drop": -0.0042713089802129955}}, "iid_acc": 0.89375, "selectivity": 0.41500000000000004} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": 0.0023049921996879386, "rotation": {"subspace_principal_angle": 1.2939750508328909, "mean_class_cosine": 0.3697587563743038}}}, "iid_acc": 0.8775, "selectivity": 0.60625, "iid_split_rotation": {"subspace_principal_angle": 1.3239667655285696, "mean_class_cosine": 0.3832240775461434}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.57375, "drop": 0.0}, "paraphrase": {"acc": 0.5772230889235569, "drop": -0.0034730889235569107, "rotation": {"subspace_principal_angle": 1.4461389808941905, "mean_class_cosine": 0.8942227001502406}}}, "iid_acc": 0.57375, "selectivity": 0.3025, "iid_split_rotation": {"subspace_principal_angle": 1.5670258105099684, "mean_class_cosine": 0.9555720866412488}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": 0.0010549921996879652}}, "iid_acc": 0.87625, "selectivity": 0.605} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.9625, "drop": 0.0}, "paraphrase": {"acc": 0.9568345323741008, "drop": 0.0056654676258992565, "rotation": {"subspace_principal_angle": 1.0889403294719116, "mean_class_cosine": 0.7097651658028382}}}, "iid_acc": 0.9625, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.0742515810195217, "mean_class_cosine": 0.7520191017637144}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.45083932853717024, "drop": 0.02416067146282974, "rotation": {"subspace_principal_angle": 1.1366627849078705, "mean_class_cosine": 0.861011502467378}}}, "iid_acc": 0.475, "selectivity": 0.41874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5498782331536536, "mean_class_cosine": 0.7592574844685773}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9400479616306955, "drop": 0.014952038369304477}}, "iid_acc": 0.955, "selectivity": 0.8987499999999999} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5830065359477125, "drop": 0.01699346405228752, "rotation": {"subspace_principal_angle": 1.4156934488627833, "mean_class_cosine": 0.1544817430571851}}}, "iid_acc": 0.6, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3875062308714112, "mean_class_cosine": 0.18226553873635268}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5254901960784314, "drop": 0.012009803921568585, "rotation": {"subspace_principal_angle": 0.8046612760829073, "mean_class_cosine": 0.9020520136987834}}}, "iid_acc": 0.5375, "selectivity": 0.03249999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.5910745370085485, "mean_class_cosine": -0.921862888991628}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.5825, "drop": 0.0}, "paraphrase": {"acc": 0.550326797385621, "drop": 0.03217320261437906}}, "iid_acc": 0.5825, "selectivity": 0.07750000000000001} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5876010781671159, "drop": 0.011148921832884073, "rotation": {"subspace_principal_angle": 1.4649394534970999, "mean_class_cosine": 0.27348013078523753}}}, "iid_acc": 0.59875, "selectivity": 0.3075, "iid_split_rotation": {"subspace_principal_angle": 1.5210318775010727, "mean_class_cosine": 0.26014522379767274}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.18375, "drop": 0.0}, "paraphrase": {"acc": 0.1725067385444744, "drop": 0.011243261455525594, "rotation": {"subspace_principal_angle": 1.2577695007550802, "mean_class_cosine": 0.6357274521759256}}}, "iid_acc": 0.18375, "selectivity": -0.10750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4122075752855408, "mean_class_cosine": 0.6982530139834328}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.59375, "drop": 0.0}, "paraphrase": {"acc": 0.5592991913746631, "drop": 0.03445080862533689}}, "iid_acc": 0.59375, "selectivity": 0.3025} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7161936560934892, "drop": 0.0025563439065108273, "rotation": {"subspace_principal_angle": 1.4142647213981934, "mean_class_cosine": 0.1558931614185028}}}, "iid_acc": 0.71875, "selectivity": 0.23375, "iid_split_rotation": {"subspace_principal_angle": 1.4440239039597567, "mean_class_cosine": 0.12643313043512422}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5442404006677797, "drop": -0.014240400667779651, "rotation": {"subspace_principal_angle": 1.438598695547982, "mean_class_cosine": 0.9982834430354565}}}, "iid_acc": 0.53, "selectivity": 0.04500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.312217021230763, "mean_class_cosine": 0.9973052099136428}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7712854757929883, "drop": -0.0075354757929883}}, "iid_acc": 0.76375, "selectivity": 0.27875000000000005} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6567656765676567, "drop": 0.02323432343234333, "rotation": {"subspace_principal_angle": 1.4737611346239565, "mean_class_cosine": 0.09688298605763825}}}, "iid_acc": 0.68, "selectivity": 0.18875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4500535476739749, "mean_class_cosine": 0.12044961172303534}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5379537953795379, "drop": 0.002046204620462122, "rotation": {"subspace_principal_angle": 0.9674520996825667, "mean_class_cosine": 0.9975306978126972}}}, "iid_acc": 0.54, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 0.6210793518048389, "mean_class_cosine": 0.9983833459358811}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.6914191419141914, "drop": 0.041080858085808636}}, "iid_acc": 0.7325, "selectivity": 0.24125000000000002} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7304075235109718, "drop": -0.016657523510971828, "rotation": {"subspace_principal_angle": 1.4688894465039684, "mean_class_cosine": 0.10173058782242872}}}, "iid_acc": 0.71375, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.4124632331913725, "mean_class_cosine": 0.15767237056399389}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.5266457680250783, "drop": -0.030395768025078318, "rotation": {"subspace_principal_angle": 0.7632925636553622, "mean_class_cosine": 0.9245850252469596}}}, "iid_acc": 0.49625, "selectivity": -0.03999999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.8486480460237737, "mean_class_cosine": 0.7937846466717617}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7570532915360502, "drop": -0.0058032915360501924}}, "iid_acc": 0.75125, "selectivity": 0.21499999999999997} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9316596931659693, "drop": 0.0070903068340306685, "rotation": {"subspace_principal_angle": 1.1500093845203234, "mean_class_cosine": 0.40847887501442204}}}, "iid_acc": 0.93875, "selectivity": 0.4149999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.0038823721844266, "mean_class_cosine": 0.5370313385967125}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.62375, "drop": 0.0}, "paraphrase": {"acc": 0.6192468619246861, "drop": 0.004503138075313884, "rotation": {"subspace_principal_angle": 1.486642282369967, "mean_class_cosine": 0.9620795603326425}}}, "iid_acc": 0.62375, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4105065066746125, "mean_class_cosine": 0.9378064089839906}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9386331938633193, "drop": 0.00011680613668063611}}, "iid_acc": 0.93875, "selectivity": 0.4149999999999999} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9819967266775778, "drop": 0.009253273322422206, "rotation": {"subspace_principal_angle": 0.7667082993155465, "mean_class_cosine": 0.7201982449314277}}}, "iid_acc": 0.99125, "selectivity": 0.13374999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.7510140573282249, "mean_class_cosine": 0.7309972720101955}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.939443535188216, "drop": -0.006943535188215999, "rotation": {"subspace_principal_angle": 0.996635177095526, "mean_class_cosine": 0.953156648540542}}}, "iid_acc": 0.9325, "selectivity": 0.07499999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9887430183481719, "mean_class_cosine": 0.9831308468715932}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9819967266775778, "drop": 0.00675327332242226}}, "iid_acc": 0.98875, "selectivity": 0.13124999999999998} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6873239436619718, "drop": 0.02267605633802816, "rotation": {"subspace_principal_angle": 1.4852509416429518, "mean_class_cosine": 0.08544108628330754}}}, "iid_acc": 0.71, "selectivity": 0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4922855379226647, "mean_class_cosine": 0.0784301577092764}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5154929577464789, "drop": 0.03325704225352111, "rotation": {"subspace_principal_angle": 0.5600755724824341, "mean_class_cosine": -0.860265927958924}}}, "iid_acc": 0.54875, "selectivity": -0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.2092432879483297, "mean_class_cosine": 0.6330568261731295}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7464788732394366, "drop": -0.006478873239436633}}, "iid_acc": 0.74, "selectivity": 0.15125} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6184971098265896, "drop": -0.012727879057358837, "rotation": {"subspace_principal_angle": 1.4119061291333692, "mean_class_cosine": 0.1704373905622454}}}, "iid_acc": 0.6057692307692307, "selectivity": 0.24999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4297364412104765, "mean_class_cosine": 0.20248169581710895}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.4567307692307692, "drop": 0.0}, "paraphrase": {"acc": 0.4508670520231214, "drop": 0.005863717207647834, "rotation": {"subspace_principal_angle": 1.3366057576399157, "mean_class_cosine": 0.7129666544008023}}}, "iid_acc": 0.4567307692307692, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.417385145003071, "mean_class_cosine": 0.826148295807795}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.5480769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.5433526011560693, "drop": 0.004724321920853813}}, "iid_acc": 0.5480769230769231, "selectivity": 0.19230769230769235} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8932038834951457, "drop": 0.0017961165048543393, "rotation": {"subspace_principal_angle": 1.311372754664119, "mean_class_cosine": 0.2565234550971107}}}, "iid_acc": 0.895, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.2262739349985579, "mean_class_cosine": 0.337747173335497}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5436893203883495, "drop": 0.022560679611650536, "rotation": {"subspace_principal_angle": 1.2898034193621133, "mean_class_cosine": -0.3258881311565935}}}, "iid_acc": 0.56625, "selectivity": -0.1825, "iid_split_rotation": {"subspace_principal_angle": 1.2264589211138428, "mean_class_cosine": 0.9728987695098255}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.90375, "drop": 0.0}, "paraphrase": {"acc": 0.9042995839112344, "drop": -0.0005495839112343859}}, "iid_acc": 0.90375, "selectivity": 0.15500000000000003} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8323943661971831, "drop": 0.02135563380281691, "rotation": {"subspace_principal_angle": 1.2914523895644616, "mean_class_cosine": 0.27572507644251076}}, "domain": {"acc": 0.87125, "drop": -0.01749999999999996, "rotation": {"subspace_principal_angle": 1.4039511655723063, "mean_class_cosine": 0.16607215124792452}}}, "iid_acc": 0.85375, "selectivity": 0.33125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.250730833076561, "mean_class_cosine": 0.3146287288985242}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5352112676056338, "drop": -0.011461267605633707, "rotation": {"subspace_principal_angle": 0.9170707486812013, "mean_class_cosine": 0.9954078358133513}}, "domain": {"acc": 0.4875, "drop": 0.03625000000000006, "rotation": {"subspace_principal_angle": 1.5523343990205853, "mean_class_cosine": 0.22749359327828106}}}, "iid_acc": 0.52375, "selectivity": 0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.4480914990441032, "mean_class_cosine": 0.9983051052181077}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8436619718309859, "drop": 0.033838028169014045}, "domain": {"acc": 0.82, "drop": 0.057499999999999996}}, "iid_acc": 0.8775, "selectivity": 0.355} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8840304182509505, "drop": 0.014719581749049526, "rotation": {"subspace_principal_angle": 1.3732084918619123, "mean_class_cosine": 0.19630467274275237}}, "domain": {"acc": 0.655, "drop": 0.24375000000000002, "rotation": {"subspace_principal_angle": 1.4452895556036265, "mean_class_cosine": 0.1251775345482053}}, "length": {"acc": 0.8995433789954338, "drop": -0.0007933789954337467, "rotation": {"subspace_principal_angle": 1.335992218473073, "mean_class_cosine": 0.2326524735797551}}}, "iid_acc": 0.89875, "selectivity": 0.39875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3151033264444465, "mean_class_cosine": 0.2529159390133714}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.8365019011406845, "drop": 0.032248098859315566, "rotation": {"subspace_principal_angle": 1.0533503571330138, "mean_class_cosine": 0.9658970994045344}}, "domain": {"acc": 0.5575, "drop": 0.31125, "rotation": {"subspace_principal_angle": 1.3501859541358534, "mean_class_cosine": 0.2666305404277387}}, "length": {"acc": 0.8751902587519026, "drop": -0.006440258751902572, "rotation": {"subspace_principal_angle": 1.4766030164776944, "mean_class_cosine": 0.9618498962901108}}}, "iid_acc": 0.86875, "selectivity": 0.36875, "iid_split_rotation": {"subspace_principal_angle": 0.8666123632485124, "mean_class_cosine": 0.9567597495837794}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.8954372623574145, "drop": 0.023312737642585457}, "domain": {"acc": 0.83125, "drop": 0.08749999999999991}, "length": {"acc": 0.923896499238965, "drop": -0.005146499238965041}}, "iid_acc": 0.91875, "selectivity": 0.41874999999999996} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8829953198127926, "drop": -0.0004953198127926051, "rotation": {"subspace_principal_angle": 1.300584248001103, "mean_class_cosine": 0.33037107823356726}}}, "iid_acc": 0.8825, "selectivity": 0.5887499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3544227368327597, "mean_class_cosine": 0.32442597332270096}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7410296411856474, "drop": -0.013529641185647368, "rotation": {"subspace_principal_angle": 1.0031403826052068, "mean_class_cosine": 0.9483063492551265}}}, "iid_acc": 0.7275, "selectivity": 0.43375, "iid_split_rotation": {"subspace_principal_angle": 0.835980512632536, "mean_class_cosine": 0.9567132219152211}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8939157566302652, "drop": 0.00358424336973473}}, "iid_acc": 0.8975, "selectivity": 0.60375} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.97375, "drop": 0.0}, "paraphrase": {"acc": 0.9592326139088729, "drop": 0.014517386091127071, "rotation": {"subspace_principal_angle": 1.0461394619384272, "mean_class_cosine": 0.7060277954956268}}}, "iid_acc": 0.97375, "selectivity": 0.90375, "iid_split_rotation": {"subspace_principal_angle": 1.0572533041916146, "mean_class_cosine": 0.7322189822775439}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6546762589928058, "drop": 0.005323741007194238, "rotation": {"subspace_principal_angle": 1.3147848288233726, "mean_class_cosine": 0.8625758379093228}}}, "iid_acc": 0.66, "selectivity": 0.5900000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2749765898173022, "mean_class_cosine": 0.8458260421792687}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9616306954436451, "drop": 0.009619304556354846}}, "iid_acc": 0.97125, "selectivity": 0.9012499999999999} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.62875, "drop": 0.0}, "paraphrase": {"acc": 0.6065359477124183, "drop": 0.022214052287581776, "rotation": {"subspace_principal_angle": 1.4471225946917146, "mean_class_cosine": 0.12335870420709219}}}, "iid_acc": 0.62875, "selectivity": 0.14375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.396757975212198, "mean_class_cosine": 0.1731610965171537}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5202614379084968, "drop": 0.019738562091503264, "rotation": {"subspace_principal_angle": 1.017105332000735, "mean_class_cosine": 0.8150921314497699}}}, "iid_acc": 0.54, "selectivity": 0.05500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3938330496938112, "mean_class_cosine": -0.5401181987718008}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.63125, "drop": 0.0}, "paraphrase": {"acc": 0.6143790849673203, "drop": 0.01687091503267968}}, "iid_acc": 0.63125, "selectivity": 0.14625} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6266846361185984, "drop": 0.028315363881401656, "rotation": {"subspace_principal_angle": 1.4639056921866045, "mean_class_cosine": 0.2607364800995969}}}, "iid_acc": 0.655, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5392183501899661, "mean_class_cosine": 0.23109273289109802}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.1925, "drop": 0.0}, "paraphrase": {"acc": 0.18867924528301888, "drop": 0.003820754716981123, "rotation": {"subspace_principal_angle": 1.3025058812372758, "mean_class_cosine": 0.609760025746434}}}, "iid_acc": 0.1925, "selectivity": -0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.380643945283782, "mean_class_cosine": 0.7278668025302554}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6145552560646901, "drop": 0.03419474393530997}}, "iid_acc": 0.64875, "selectivity": 0.38250000000000006} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.775, "drop": 0.0}, "paraphrase": {"acc": 0.7696160267111853, "drop": 0.005383973288814681, "rotation": {"subspace_principal_angle": 1.4245590947156432, "mean_class_cosine": 0.145716567257158}}}, "iid_acc": 0.775, "selectivity": 0.28875, "iid_split_rotation": {"subspace_principal_angle": 1.471068816739365, "mean_class_cosine": 0.09956228431408685}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5425709515859767, "drop": -0.017570951585976657, "rotation": {"subspace_principal_angle": 1.4049876290028978, "mean_class_cosine": 0.8302182537519393}}}, "iid_acc": 0.525, "selectivity": 0.03875000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.628667819566516, "mean_class_cosine": 0.9886859497074707}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.7863105175292153, "drop": 0.001189482470784653}}, "iid_acc": 0.7875, "selectivity": 0.30124999999999996} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": 0.00708333333333333, "rotation": {"subspace_principal_angle": 1.4652126246929733, "mean_class_cosine": 0.10538763834081566}}}, "iid_acc": 0.67375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.453567266755897, "mean_class_cosine": 0.1169607381052836}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5412541254125413, "drop": 0.00874587458745879, "rotation": {"subspace_principal_angle": 1.435416177004221, "mean_class_cosine": -0.8400641400418787}}}, "iid_acc": 0.55, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.353882512217704, "mean_class_cosine": 0.9873913376016163}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.6897689768976898, "drop": 0.026481023102310264}}, "iid_acc": 0.71625, "selectivity": 0.22250000000000003} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7476489028213166, "drop": -0.02014890282131654, "rotation": {"subspace_principal_angle": 1.5031766373236868, "mean_class_cosine": 0.06756817028714371}}}, "iid_acc": 0.7275, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.4985421261579637, "mean_class_cosine": 0.07219134782882808}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5297805642633229, "drop": -0.016030564263322833, "rotation": {"subspace_principal_angle": 1.4269512459642844, "mean_class_cosine": 0.6191468418063922}}}, "iid_acc": 0.51375, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 0.6941036459644455, "mean_class_cosine": 0.6890861724335152}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7586206896551724, "drop": -0.008620689655172376}}, "iid_acc": 0.75, "selectivity": 0.23624999999999996} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9400278940027894, "drop": 0.01122210599721063, "rotation": {"subspace_principal_angle": 1.212406540346009, "mean_class_cosine": 0.3507667834923325}}}, "iid_acc": 0.95125, "selectivity": 0.41875000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.1491505155466, "mean_class_cosine": 0.40926267218064927}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.785, "drop": 0.0}, "paraphrase": {"acc": 0.796373779637378, "drop": -0.011373779637377956, "rotation": {"subspace_principal_angle": 0.5776320485859313, "mean_class_cosine": 0.9642980239049406}}}, "iid_acc": 0.785, "selectivity": 0.25250000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.8609996534230742, "mean_class_cosine": 0.9712015306482873}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9372384937238494, "drop": 0.014011506276150665}}, "iid_acc": 0.95125, "selectivity": 0.41875000000000007} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.9975, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.004046644844517222, "rotation": {"subspace_principal_angle": 0.8761164033837602, "mean_class_cosine": 0.6401395707229799}}}, "iid_acc": 0.9975, "selectivity": 0.1825000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8963642783199154, "mean_class_cosine": 0.6244538122618211}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.4075286415711948, "drop": 0.3112213584288052, "rotation": {"subspace_principal_angle": 1.3896444565022879, "mean_class_cosine": 0.8607515326377927}}}, "iid_acc": 0.71875, "selectivity": -0.09624999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1374176741267867, "mean_class_cosine": 0.9949908284390802}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9918166939443536, "drop": 0.004433306055646402}}, "iid_acc": 0.99625, "selectivity": 0.18125000000000002} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7112676056338029, "drop": 0.02498239436619709, "rotation": {"subspace_principal_angle": 1.512652494056396, "mean_class_cosine": 0.05811107708207547}}}, "iid_acc": 0.73625, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.5022867064091796, "mean_class_cosine": 0.06845604053333303}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5338028169014084, "drop": 0.03119718309859154, "rotation": {"subspace_principal_angle": 1.2680709408982833, "mean_class_cosine": -0.35755644118795693}}}, "iid_acc": 0.565, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.547048981725239, "mean_class_cosine": 0.09058745031393456}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.78125, "drop": 0.0}, "paraphrase": {"acc": 0.7507042253521127, "drop": 0.030545774647887325}}, "iid_acc": 0.78125, "selectivity": 0.19125000000000003} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5961538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5953757225433526, "drop": 0.000778123610493564, "rotation": {"subspace_principal_angle": 1.4398635945392528, "mean_class_cosine": 0.13929059880991962}}}, "iid_acc": 0.5961538461538461, "selectivity": 0.22115384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.4596191174582307, "mean_class_cosine": 0.18093919697525562}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5192307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.5491329479768786, "drop": -0.029902178746109342, "rotation": {"subspace_principal_angle": 1.0835579282901093, "mean_class_cosine": 0.6041012733975983}}}, "iid_acc": 0.5192307692307693, "selectivity": 0.14423076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.0998389268494577, "mean_class_cosine": 0.6208614359967286}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5721153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": -0.0059192974655403185}}, "iid_acc": 0.5721153846153846, "selectivity": 0.19711538461538458} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.9056865464632455, "drop": -0.0006865464632455076, "rotation": {"subspace_principal_angle": 1.3934785192501797, "mean_class_cosine": 0.17639007447692165}}}, "iid_acc": 0.905, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3193210248676257, "mean_class_cosine": 0.24883312785983533}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.57, "drop": 0.0}, "paraphrase": {"acc": 0.5533980582524272, "drop": 0.01660194174757279, "rotation": {"subspace_principal_angle": 1.5354546576355417, "mean_class_cosine": 0.1563657422847014}}}, "iid_acc": 0.57, "selectivity": -0.14250000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.445383158458667, "mean_class_cosine": 0.9556115381599443}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9084604715672677, "drop": 0.006539528432732311}}, "iid_acc": 0.915, "selectivity": 0.2025} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.776056338028169, "drop": 0.033943661971831074, "rotation": {"subspace_principal_angle": 1.3558532015708853, "mean_class_cosine": 0.2132918626421837}}, "domain": {"acc": 0.8, "drop": 0.010000000000000009, "rotation": {"subspace_principal_angle": 1.3213395199914533, "mean_class_cosine": 0.24687761614550957}}}, "iid_acc": 0.81, "selectivity": 0.30000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2949820598326707, "mean_class_cosine": 0.27233051788958473}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5309859154929577, "drop": -0.002235915492957652, "rotation": {"subspace_principal_angle": 0.949671245424435, "mean_class_cosine": 0.9981407854560767}}, "domain": {"acc": 0.4875, "drop": 0.041250000000000064, "rotation": {"subspace_principal_angle": 1.5209675895490495, "mean_class_cosine": 0.5347820929835523}}}, "iid_acc": 0.52875, "selectivity": 0.018750000000000044, "iid_split_rotation": {"subspace_principal_angle": 0.5977677192699846, "mean_class_cosine": 0.9995599104407848}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.825, "drop": 0.0}, "paraphrase": {"acc": 0.7295774647887324, "drop": 0.09542253521126753}, "domain": {"acc": 0.73125, "drop": 0.09375}}, "iid_acc": 0.825, "selectivity": 0.31499999999999995} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.825, "drop": 0.0}, "paraphrase": {"acc": 0.8022813688212928, "drop": 0.022718631178707205, "rotation": {"subspace_principal_angle": 1.3701838535641038, "mean_class_cosine": 0.19926955803809265}}, "domain": {"acc": 0.56, "drop": 0.2649999999999999, "rotation": {"subspace_principal_angle": 1.3671422264532063, "mean_class_cosine": 0.2022492580826582}}, "length": {"acc": 0.8356164383561644, "drop": -0.010616438356164437, "rotation": {"subspace_principal_angle": 1.3466907976398639, "mean_class_cosine": 0.22223434808814393}}}, "iid_acc": 0.825, "selectivity": 0.32499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3129495009920322, "mean_class_cosine": 0.25499915160174697}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.6406844106463878, "drop": -0.04568441064638784, "rotation": {"subspace_principal_angle": 1.5652743111758125, "mean_class_cosine": 0.9561201033378162}}, "domain": {"acc": 0.5575, "drop": 0.03749999999999998, "rotation": {"subspace_principal_angle": 1.307138927137268, "mean_class_cosine": 0.5860658652565125}}, "length": {"acc": 0.6164383561643836, "drop": -0.02143835616438361, "rotation": {"subspace_principal_angle": 0.9884986790030585, "mean_class_cosine": 0.9027214333318725}}}, "iid_acc": 0.595, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2061930031896504, "mean_class_cosine": 0.9278347514353975}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.844106463878327, "drop": 0.009643536121673013}, "domain": {"acc": 0.66875, "drop": 0.18500000000000005}, "length": {"acc": 0.8493150684931506, "drop": 0.004434931506849371}}, "iid_acc": 0.85375, "selectivity": 0.35375} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8642745709828393, "drop": 0.005725429017160666, "rotation": {"subspace_principal_angle": 1.3957840280918499, "mean_class_cosine": 0.27977067989217674}}}, "iid_acc": 0.87, "selectivity": 0.5974999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3533239187236508, "mean_class_cosine": 0.32998309013733307}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5460218408736349, "drop": -0.0035218408736349405, "rotation": {"subspace_principal_angle": 1.5216614620859656, "mean_class_cosine": 0.8734402521973395}}}, "iid_acc": 0.5425, "selectivity": 0.26999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.0674047358789585, "mean_class_cosine": 0.9542985146974496}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.890795631825273, "drop": -0.008295631825273042}}, "iid_acc": 0.8825, "selectivity": 0.6099999999999999} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9376498800959233, "drop": 0.023600119904076733, "rotation": {"subspace_principal_angle": 1.1479338257378555, "mean_class_cosine": 0.6070660513833379}}}, "iid_acc": 0.96125, "selectivity": 0.89375, "iid_split_rotation": {"subspace_principal_angle": 1.1171811389695125, "mean_class_cosine": 0.6559427626785893}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6139088729016786, "drop": 0.008591127098321438, "rotation": {"subspace_principal_angle": 1.2623715388420953, "mean_class_cosine": 0.8819688751478222}}}, "iid_acc": 0.6225, "selectivity": 0.555, "iid_split_rotation": {"subspace_principal_angle": 1.2562343632402415, "mean_class_cosine": 0.8445383158129988}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.95375, "drop": 0.0}, "paraphrase": {"acc": 0.9328537170263789, "drop": 0.020896282973621116}}, "iid_acc": 0.95375, "selectivity": 0.88625} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": 0.011323529411764732, "rotation": {"subspace_principal_angle": 1.4587999072142643, "mean_class_cosine": 0.11176243416412668}}}, "iid_acc": 0.5525, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4400549078317155, "mean_class_cosine": 0.13036926971343207}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.538562091503268, "drop": -0.02231209150326796, "rotation": {"subspace_principal_angle": 0.9958229238018077, "mean_class_cosine": 0.9168202372116476}}}, "iid_acc": 0.51625, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.4327159822922633, "mean_class_cosine": -0.5177790256388579}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.49673202614379086, "drop": 0.054517973856209156}}, "iid_acc": 0.55125, "selectivity": 0.026249999999999996} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5619946091644205, "drop": 0.039255390835579496, "rotation": {"subspace_principal_angle": 1.5685155167026745, "mean_class_cosine": 0.2634805112669454}}}, "iid_acc": 0.60125, "selectivity": 0.3212499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.498476982651726, "mean_class_cosine": 0.251518325093681}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.26125, "drop": 0.0}, "paraphrase": {"acc": 0.261455525606469, "drop": -0.00020552560646902585, "rotation": {"subspace_principal_angle": 1.3916019392156598, "mean_class_cosine": 0.6531664347702235}}}, "iid_acc": 0.26125, "selectivity": -0.018750000000000044, "iid_split_rotation": {"subspace_principal_angle": 1.3668566895557186, "mean_class_cosine": 0.6689391992705035}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5700808625336927, "drop": 0.016169137466307326}}, "iid_acc": 0.58625, "selectivity": 0.30625} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7195325542570952, "drop": -0.009532554257095205, "rotation": {"subspace_principal_angle": 1.4001021400462081, "mean_class_cosine": 0.16986648813286045}}}, "iid_acc": 0.71, "selectivity": 0.21249999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.397132197602315, "mean_class_cosine": 0.17279251519584937}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.5859766277128547, "drop": -0.009726627712854707, "rotation": {"subspace_principal_angle": 1.328491446497459, "mean_class_cosine": 0.9377244571178811}}}, "iid_acc": 0.57625, "selectivity": 0.07875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4036636361795571, "mean_class_cosine": 0.938290713035285}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7479131886477463, "drop": -0.00416318864774623}}, "iid_acc": 0.74375, "selectivity": 0.24625000000000002} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.6375, "drop": 0.0}, "paraphrase": {"acc": 0.6105610561056105, "drop": 0.026938943894389444, "rotation": {"subspace_principal_angle": 1.490662860294879, "mean_class_cosine": 0.08004773289195696}}}, "iid_acc": 0.6375, "selectivity": 0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.516170603318211, "mean_class_cosine": 0.054598560612693414}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5412541254125413, "drop": 0.0049958745874587596, "rotation": {"subspace_principal_angle": 1.018522615405077, "mean_class_cosine": 0.998295366669647}}}, "iid_acc": 0.54625, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.799728898586944, "mean_class_cosine": 0.9985173926775679}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.68875, "drop": 0.0}, "paraphrase": {"acc": 0.6831683168316832, "drop": 0.005581683168316753}}, "iid_acc": 0.68875, "selectivity": 0.19874999999999998} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.68125, "drop": 0.0}, "paraphrase": {"acc": 0.677115987460815, "drop": 0.004134012539184995, "rotation": {"subspace_principal_angle": 1.513447961089972, "mean_class_cosine": 0.05731693598787185}}}, "iid_acc": 0.68125, "selectivity": 0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5579856772962786, "mean_class_cosine": 0.012810299103026977}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5250783699059561, "drop": -0.02382836990595616, "rotation": {"subspace_principal_angle": 1.5103655352700989, "mean_class_cosine": 0.7751093816111925}}}, "iid_acc": 0.50125, "selectivity": -0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9404374749778482, "mean_class_cosine": 0.48400831659040194}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7570532915360502, "drop": -0.027053291536050184}}, "iid_acc": 0.73, "selectivity": 0.15749999999999997} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.91625, "drop": 0.0}, "paraphrase": {"acc": 0.9135285913528591, "drop": 0.002721408647140877, "rotation": {"subspace_principal_angle": 1.2733184253579535, "mean_class_cosine": 0.29310981653055773}}}, "iid_acc": 0.91625, "selectivity": 0.37250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1920176049103017, "mean_class_cosine": 0.36978603528128035}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.83, "drop": 0.0}, "paraphrase": {"acc": 0.8145048814504882, "drop": 0.0154951185495118, "rotation": {"subspace_principal_angle": 1.4437573872568787, "mean_class_cosine": 0.9676721148007581}}}, "iid_acc": 0.83, "selectivity": 0.28625, "iid_split_rotation": {"subspace_principal_angle": 1.4314363898953524, "mean_class_cosine": 0.9613334451017563}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.91625, "drop": 0.0}, "paraphrase": {"acc": 0.9135285913528591, "drop": 0.002721408647140877}}, "iid_acc": 0.91625, "selectivity": 0.37250000000000005} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.0015466448445171643, "rotation": {"subspace_principal_angle": 1.0288570623244413, "mean_class_cosine": 0.5157983478150161}}}, "iid_acc": 0.995, "selectivity": 0.22999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.069772811988795, "mean_class_cosine": 0.4803235060741538}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.734860883797054, "drop": -0.03611088379705407, "rotation": {"subspace_principal_angle": 1.4159558052810821, "mean_class_cosine": 0.998661624282861}}}, "iid_acc": 0.69875, "selectivity": -0.06625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.1246427235862093, "mean_class_cosine": 0.9995471474354687}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9918166939443536, "drop": -0.0018166939443535757}}, "iid_acc": 0.99, "selectivity": 0.22499999999999998} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6845070422535211, "drop": 0.009242957746478875, "rotation": {"subspace_principal_angle": 1.5353921160260495, "mean_class_cosine": 0.035396814949682484}}}, "iid_acc": 0.69375, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4561316678570049, "mean_class_cosine": 0.11441355588600202}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5295774647887324, "drop": -0.012077464788732395, "rotation": {"subspace_principal_angle": 1.1952282463550592, "mean_class_cosine": 0.34100773822159186}}}, "iid_acc": 0.5175, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.5260408088060127, "mean_class_cosine": 0.9004767260950717}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7535211267605634, "drop": -0.017271126760563416}}, "iid_acc": 0.73625, "selectivity": 0.09124999999999994} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.6127167630057804, "drop": -0.050216763005780374, "rotation": {"subspace_principal_angle": 1.4402506816259675, "mean_class_cosine": 0.1541271484023787}}}, "iid_acc": 0.5625, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.373294097014231, "mean_class_cosine": 0.19890553489034377}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.49710982658959535, "drop": -0.035571365051133785, "rotation": {"subspace_principal_angle": 1.1768674031248585, "mean_class_cosine": 0.7104337840281887}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.07692307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.1855767421726293, "mean_class_cosine": 0.8092177583847361}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5336538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5549132947976878, "drop": -0.02125944864384166}}, "iid_acc": 0.5336538461538461, "selectivity": 0.1490384615384615} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.9001386962552012, "drop": -0.01013869625520114, "rotation": {"subspace_principal_angle": 1.3883765772077856, "mean_class_cosine": 0.18140970234637815}}}, "iid_acc": 0.89, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3442390049803465, "mean_class_cosine": 0.22462415910132738}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5436893203883495, "drop": 0.018810679611650505, "rotation": {"subspace_principal_angle": 0.9241498843813925, "mean_class_cosine": -0.17909041195198278}}}, "iid_acc": 0.5625, "selectivity": -0.135, "iid_split_rotation": {"subspace_principal_angle": 0.7140290968434547, "mean_class_cosine": 0.9686729997957932}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8890429958391124, "drop": -0.010292995839112362}}, "iid_acc": 0.87875, "selectivity": 0.18125000000000002} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8325, "drop": 0.0}, "paraphrase": {"acc": 0.7746478873239436, "drop": 0.05785211267605639, "rotation": {"subspace_principal_angle": 1.2990940486645208, "mean_class_cosine": 0.26837165555553494}}, "domain": {"acc": 0.7975, "drop": 0.03500000000000003, "rotation": {"subspace_principal_angle": 1.3921776610825947, "mean_class_cosine": 0.17767038602968485}}}, "iid_acc": 0.8325, "selectivity": 0.32625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2834752018104276, "mean_class_cosine": 0.28338418654918335}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5309859154929577, "drop": -0.002235915492957652, "rotation": {"subspace_principal_angle": 1.0586485465355164, "mean_class_cosine": 0.9956582503304628}}, "domain": {"acc": 0.4875, "drop": 0.041250000000000064, "rotation": {"subspace_principal_angle": 1.361095756118237, "mean_class_cosine": 0.3175504588818687}}}, "iid_acc": 0.52875, "selectivity": 0.022500000000000075, "iid_split_rotation": {"subspace_principal_angle": 0.8302811510741681, "mean_class_cosine": 0.9988764885217147}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.819718309859155, "drop": 0.04653169014084502}, "domain": {"acc": 0.8425, "drop": 0.023749999999999938}}, "iid_acc": 0.86625, "selectivity": 0.36} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8829953198127926, "drop": -0.0029953198127925518, "rotation": {"subspace_principal_angle": 1.3716186408201203, "mean_class_cosine": 0.26184135314471124}}}, "iid_acc": 0.88, "selectivity": 0.59875, "iid_split_rotation": {"subspace_principal_angle": 1.3986787504321423, "mean_class_cosine": 0.27403222616028045}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.80625, "drop": 0.0}, "paraphrase": {"acc": 0.7862714508580343, "drop": 0.01997854914196573, "rotation": {"subspace_principal_angle": 1.1213682836223715, "mean_class_cosine": 0.9645304775406084}}}, "iid_acc": 0.80625, "selectivity": 0.525, "iid_split_rotation": {"subspace_principal_angle": 1.190218420873915, "mean_class_cosine": 0.9648000749883475}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8923556942277691, "drop": 0.00014430577223090157}}, "iid_acc": 0.8925, "selectivity": 0.61125} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5437908496732026, "drop": -0.002540849673202561, "rotation": {"subspace_principal_angle": 1.4224273033339936, "mean_class_cosine": 0.14782527199893394}}}, "iid_acc": 0.54125, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.4416538946189268, "mean_class_cosine": 0.1287837634804017}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.538562091503268, "drop": -0.021062091503267988, "rotation": {"subspace_principal_angle": 1.2749369531499732, "mean_class_cosine": 0.8687066652493326}}}, "iid_acc": 0.5175, "selectivity": 0.024999999999999967, "iid_split_rotation": {"subspace_principal_angle": 1.539011739741241, "mean_class_cosine": -0.5381342265082592}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": 0.016838235294117654}}, "iid_acc": 0.54625, "selectivity": 0.05375000000000002} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.5916442048517521, "drop": 0.023355795148247926, "rotation": {"subspace_principal_angle": 1.3790979385283946, "mean_class_cosine": 0.3195081442711161}}}, "iid_acc": 0.615, "selectivity": 0.36375, "iid_split_rotation": {"subspace_principal_angle": 1.4199621331697765, "mean_class_cosine": 0.29428965873721996}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.2875, "drop": 0.0}, "paraphrase": {"acc": 0.2803234501347709, "drop": 0.007176549865229087, "rotation": {"subspace_principal_angle": 1.1732397784339854, "mean_class_cosine": 0.6776335920045801}}}, "iid_acc": 0.2875, "selectivity": 0.036250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3510532914197495, "mean_class_cosine": 0.6935310313683204}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.63, "drop": 0.0}, "paraphrase": {"acc": 0.6226415094339622, "drop": 0.0073584905660377675}}, "iid_acc": 0.63, "selectivity": 0.37875000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.70375, "drop": 0.0}, "paraphrase": {"acc": 0.6828046744574291, "drop": 0.020945325542570892, "rotation": {"subspace_principal_angle": 1.4506727333643843, "mean_class_cosine": 0.1198349109994324}}}, "iid_acc": 0.70375, "selectivity": 0.20375, "iid_split_rotation": {"subspace_principal_angle": 1.4560638773991705, "mean_class_cosine": 0.1144809009150342}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.5442404006677797, "drop": -0.021740400667779713, "rotation": {"subspace_principal_angle": 1.3673823057780032, "mean_class_cosine": 0.9981711848367865}}}, "iid_acc": 0.5225, "selectivity": 0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 1.1930110372511171, "mean_class_cosine": 0.9981828476111254}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7362270450751253, "drop": 0.013772954924874736}}, "iid_acc": 0.75, "selectivity": 0.25} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.7084639498432602, "drop": -0.02096394984326022, "rotation": {"subspace_principal_angle": 1.5051962181158443, "mean_class_cosine": 0.06555306849859195}}}, "iid_acc": 0.6875, "selectivity": 0.14125, "iid_split_rotation": {"subspace_principal_angle": 1.5172834822024313, "mean_class_cosine": 0.053487308133283416}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5658307210031348, "drop": -0.010830721003134758, "rotation": {"subspace_principal_angle": 1.1030041242538011, "mean_class_cosine": 0.6806746541237996}}}, "iid_acc": 0.555, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.0411245457901148, "mean_class_cosine": 0.7191236806689514}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.7492163009404389, "drop": -0.0029663009404389262}}, "iid_acc": 0.74625, "selectivity": 0.19999999999999996} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9288702928870293, "drop": 0.012379707112970761, "rotation": {"subspace_principal_angle": 1.2301430406581242, "mean_class_cosine": 0.3341029094871022}}}, "iid_acc": 0.94125, "selectivity": 0.41500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.153997330636452, "mean_class_cosine": 0.4048355678035365}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7642956764295676, "drop": -0.005545676429567603, "rotation": {"subspace_principal_angle": 1.5114298942376507, "mean_class_cosine": 0.967424461326506}}}, "iid_acc": 0.75875, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9983998428171448, "mean_class_cosine": 0.9658650662106624}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.9428172942817294, "drop": 0.005932705718270537}}, "iid_acc": 0.94875, "selectivity": 0.4225} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.70375, "drop": 0.0}, "paraphrase": {"acc": 0.6732394366197183, "drop": 0.030510563380281686, "rotation": {"subspace_principal_angle": 1.5542729001951894, "mean_class_cosine": 0.01652267472899558}}}, "iid_acc": 0.70375, "selectivity": 0.13124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4874758177302438, "mean_class_cosine": 0.08322413642773308}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5464788732394367, "drop": 0.02602112676056334, "rotation": {"subspace_principal_angle": 1.5278946651673369, "mean_class_cosine": -0.5372610043917487}}}, "iid_acc": 0.5725, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.4388173212776232, "mean_class_cosine": 0.7758844286751576}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7183098591549296, "drop": 0.03419014084507033}}, "iid_acc": 0.7525, "selectivity": 0.17999999999999994} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5606936416184971, "drop": 0.0018063583815028927, "rotation": {"subspace_principal_angle": 1.4713914583801808, "mean_class_cosine": 0.1086073697472787}}}, "iid_acc": 0.5625, "selectivity": 0.20192307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.4025179262636047, "mean_class_cosine": 0.18370954160961964}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.47398843930635837, "drop": -0.012449977767896803, "rotation": {"subspace_principal_angle": 1.1101897330656618, "mean_class_cosine": 0.7275819529837535}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.10096153846153849, "iid_split_rotation": {"subspace_principal_angle": 1.5497284875073214, "mean_class_cosine": 0.7930942268672401}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.6009615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": 0.022926856380613536}}, "iid_acc": 0.6009615384615384, "selectivity": 0.24038461538461536} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8987517337031901, "drop": -1.733703190009983e-06, "rotation": {"subspace_principal_angle": 1.403522863323226, "mean_class_cosine": 0.16649449067720934}}}, "iid_acc": 0.89875, "selectivity": 0.21750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.420138412202417, "mean_class_cosine": 0.15008862700519784}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.57125, "drop": 0.0}, "paraphrase": {"acc": 0.5617198335644937, "drop": 0.009530166435506304, "rotation": {"subspace_principal_angle": 0.94423285674327, "mean_class_cosine": 0.24684962591703655}}}, "iid_acc": 0.57125, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5221499477378648, "mean_class_cosine": 0.9517198734637033}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9098474341192788, "drop": 0.012652565880721167}}, "iid_acc": 0.9225, "selectivity": 0.24124999999999996} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.85625, "drop": 0.0}, "paraphrase": {"acc": 0.8323943661971831, "drop": 0.023855633802816856, "rotation": {"subspace_principal_angle": 1.2594449210343017, "mean_class_cosine": 0.3063453465601831}}, "domain": {"acc": 0.8675, "drop": -0.011250000000000093, "rotation": {"subspace_principal_angle": 1.2223752440939522, "mean_class_cosine": 0.3414141888265658}}}, "iid_acc": 0.85625, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 1.1803322078545402, "mean_class_cosine": 0.38061764197303105}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5366197183098591, "drop": -0.004119718309859133, "rotation": {"subspace_principal_angle": 1.0843410916453302, "mean_class_cosine": 0.9995443885101234}}, "domain": {"acc": 0.4875, "drop": 0.044999999999999984, "rotation": {"subspace_principal_angle": 1.5626024602699073, "mean_class_cosine": 0.4830036010251207}}}, "iid_acc": 0.5325, "selectivity": 0.030000000000000027, "iid_split_rotation": {"subspace_principal_angle": 1.1107445344198659, "mean_class_cosine": 0.9998835461376956}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8028169014084507, "drop": 0.03468309859154928}, "domain": {"acc": 0.8225, "drop": 0.015000000000000013}}, "iid_acc": 0.8375, "selectivity": 0.3350000000000001} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.9125, "drop": 0.0}, "paraphrase": {"acc": 0.8954372623574145, "drop": 0.01706273764258548, "rotation": {"subspace_principal_angle": 1.0306140943276123, "mean_class_cosine": 0.5142922854845775}}, "domain": {"acc": 0.77875, "drop": 0.13374999999999992, "rotation": {"subspace_principal_angle": 1.2652458098581651, "mean_class_cosine": 0.30081823858041606}}, "length": {"acc": 0.9193302891933028, "drop": -0.006830289193302863, "rotation": {"subspace_principal_angle": 0.994782867288978, "mean_class_cosine": 0.5446849986678566}}}, "iid_acc": 0.9125, "selectivity": 0.4125, "iid_split_rotation": {"subspace_principal_angle": 1.0394992159651213, "mean_class_cosine": 0.5066520720072085}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.58125, "drop": 0.0}, "paraphrase": {"acc": 0.6254752851711026, "drop": -0.04422528517110258, "rotation": {"subspace_principal_angle": 0.9837366616608404, "mean_class_cosine": 0.9776968730440565}}, "domain": {"acc": 0.5575, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.3537717360413388, "mean_class_cosine": 0.5378898485059446}}, "length": {"acc": 0.6012176560121766, "drop": -0.019967656012176538, "rotation": {"subspace_principal_angle": 1.0664869055781216, "mean_class_cosine": 0.8881910220135161}}}, "iid_acc": 0.58125, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5132821504454912, "mean_class_cosine": 0.9660599604667153}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.91, "drop": 0.0}, "paraphrase": {"acc": 0.9049429657794676, "drop": 0.005057034220532386}, "domain": {"acc": 0.755, "drop": 0.15500000000000003}, "length": {"acc": 0.9178082191780822, "drop": -0.007808219178082165}}, "iid_acc": 0.91, "selectivity": 0.41000000000000003} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8783151326053042, "drop": 0.005434867394695786, "rotation": {"subspace_principal_angle": 1.4018526090844314, "mean_class_cosine": 0.21584227023789393}}}, "iid_acc": 0.88375, "selectivity": 0.6412500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3568370650579906, "mean_class_cosine": 0.277933277471287}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8798751950078003, "drop": 0.001374804992199663, "rotation": {"subspace_principal_angle": 1.5437428796709174, "mean_class_cosine": 0.9838278021789155}}}, "iid_acc": 0.88125, "selectivity": 0.6387499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5098184352195878, "mean_class_cosine": 0.9768614686576095}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8923556942277691, "drop": 0.007644305772230964}}, "iid_acc": 0.9, "selectivity": 0.6575} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.965, "drop": 0.0}, "paraphrase": {"acc": 0.9496402877697842, "drop": 0.015359712230215816, "rotation": {"subspace_principal_angle": 0.9402737529008706, "mean_class_cosine": 0.7722383597549551}}}, "iid_acc": 0.965, "selectivity": 0.9075, "iid_split_rotation": {"subspace_principal_angle": 0.8971810536010522, "mean_class_cosine": 0.8007298180881067}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.13875, "drop": 0.0}, "paraphrase": {"acc": 0.14628297362110312, "drop": -0.007532973621103112, "rotation": {"subspace_principal_angle": 1.1315401402990097, "mean_class_cosine": 0.6941035532694599}}}, "iid_acc": 0.13875, "selectivity": 0.08125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.385883009922417, "mean_class_cosine": 0.6014609571780535}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9136690647482014, "drop": 0.025080935251798553}}, "iid_acc": 0.93875, "selectivity": 0.88125} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.6300653594771242, "drop": 0.008684640522875875, "rotation": {"subspace_principal_angle": 1.2787080353422497, "mean_class_cosine": 0.28795269263212087}}}, "iid_acc": 0.63875, "selectivity": 0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.306568966622189, "mean_class_cosine": 0.2611635211442275}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": -0.028676470588235303, "rotation": {"subspace_principal_angle": 1.4646247494771925, "mean_class_cosine": 0.7483798246130362}}}, "iid_acc": 0.5125, "selectivity": -0.02750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5194053435641155, "mean_class_cosine": -0.9436278317396236}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5333333333333333, "drop": 0.0004166666666666208}}, "iid_acc": 0.53375, "selectivity": -0.006250000000000089} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.62, "drop": 0.0}, "paraphrase": {"acc": 0.5876010781671159, "drop": 0.032398921832884064, "rotation": {"subspace_principal_angle": 1.3946988954670643, "mean_class_cosine": 0.4588645470889185}}}, "iid_acc": 0.62, "selectivity": 0.31625, "iid_split_rotation": {"subspace_principal_angle": 1.4814328069291833, "mean_class_cosine": 0.3908719413079447}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.1825, "drop": 0.0}, "paraphrase": {"acc": 0.1765498652291105, "drop": 0.005950134770889487, "rotation": {"subspace_principal_angle": 1.428737416987744, "mean_class_cosine": 0.6603620114391797}}}, "iid_acc": 0.1825, "selectivity": -0.12125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3464001687915095, "mean_class_cosine": 0.917084029092293}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5175202156334232, "drop": 0.012479784366576818}}, "iid_acc": 0.53, "selectivity": 0.22625} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7612687813021702, "drop": -0.0037687813021702876, "rotation": {"subspace_principal_angle": 1.0841355564151438, "mean_class_cosine": 0.46767695778639146}}}, "iid_acc": 0.7575, "selectivity": 0.24624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1201130263537538, "mean_class_cosine": 0.4355807084230253}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6828046744574291, "drop": -0.012804674457429055, "rotation": {"subspace_principal_angle": 1.559246280818732, "mean_class_cosine": 0.8724262003165955}}}, "iid_acc": 0.67, "selectivity": 0.15875000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.208049879221657, "mean_class_cosine": 0.8740158426804394}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7395659432387313, "drop": 0.01418405676126877}}, "iid_acc": 0.75375, "selectivity": 0.24250000000000005} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6287128712871287, "drop": 0.05503712871287125, "rotation": {"subspace_principal_angle": 1.4304093704815346, "mean_class_cosine": 0.1399262745187435}}}, "iid_acc": 0.68375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4264870806603527, "mean_class_cosine": 0.14380889026388544}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5610561056105611, "drop": -0.008556105610561082, "rotation": {"subspace_principal_angle": 1.4505633244788272, "mean_class_cosine": 0.9993707117635033}}}, "iid_acc": 0.5525, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.562613516437802, "mean_class_cosine": 0.9992208351646927}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6171617161716172, "drop": 0.04283828382838284}}, "iid_acc": 0.66, "selectivity": 0.15625} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7429467084639498, "drop": -0.011696708463949879, "rotation": {"subspace_principal_angle": 1.4343600588327907, "mean_class_cosine": 0.1360133715328124}}}, "iid_acc": 0.73125, "selectivity": 0.12874999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3984660781436913, "mean_class_cosine": 0.17147854522739675}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5266457680250783, "drop": -0.017895768025078307, "rotation": {"subspace_principal_angle": 1.5522402276263103, "mean_class_cosine": 0.9681835964660888}}}, "iid_acc": 0.50875, "selectivity": -0.09375, "iid_split_rotation": {"subspace_principal_angle": 1.278589503289651, "mean_class_cosine": 0.36941996233066643}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7429467084639498, "drop": -0.009196708463949821}}, "iid_acc": 0.73375, "selectivity": 0.13124999999999998} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9163179916317992, "drop": 0.021182008368200833, "rotation": {"subspace_principal_angle": 1.190401875207584, "mean_class_cosine": 0.371286753786929}}}, "iid_acc": 0.9375, "selectivity": 0.39625, "iid_split_rotation": {"subspace_principal_angle": 1.0454137579769693, "mean_class_cosine": 0.5015440139447017}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5425383542538355, "drop": 0.008711645746164565, "rotation": {"subspace_principal_angle": 0.70122334453455, "mean_class_cosine": 0.9845649431106505}}}, "iid_acc": 0.55125, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 0.5014314350811042, "mean_class_cosine": 0.9308041543855714}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.945, "drop": 0.0}, "paraphrase": {"acc": 0.9456066945606695, "drop": -0.0006066945606695295}}, "iid_acc": 0.945, "selectivity": 0.40374999999999994} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9950900163666121, "drop": -9.001636661210011e-05, "rotation": {"subspace_principal_angle": 0.9545129430981294, "mean_class_cosine": 0.578006280576806}}}, "iid_acc": 0.995, "selectivity": 0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9634855001163742, "mean_class_cosine": 0.5706612157863881}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.723404255319149, "drop": -0.024654255319149, "rotation": {"subspace_principal_angle": 0.9334539669903851, "mean_class_cosine": 0.9999542616831976}}}, "iid_acc": 0.69875, "selectivity": -0.12875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4894565210879196, "mean_class_cosine": 0.9999767868513006}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.000296644844517191}}, "iid_acc": 0.99375, "selectivity": 0.16625} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7183098591549296, "drop": 0.025440140845070403, "rotation": {"subspace_principal_angle": 1.4257021861565011, "mean_class_cosine": 0.14458558178778447}}}, "iid_acc": 0.74375, "selectivity": 0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4515121208658293, "mean_class_cosine": 0.11900153014654812}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5183098591549296, "drop": 0.04419014084507045, "rotation": {"subspace_principal_angle": 1.5128863516688837, "mean_class_cosine": -0.06917035187866413}}}, "iid_acc": 0.5625, "selectivity": -0.03374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3373195084183762, "mean_class_cosine": 0.802090774256028}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.752112676056338, "drop": -0.01711267605633804}}, "iid_acc": 0.735, "selectivity": 0.13875000000000004} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.630057803468208, "drop": -0.005057803468208055, "rotation": {"subspace_principal_angle": 1.3578016617138275, "mean_class_cosine": 0.22714259169758785}}}, "iid_acc": 0.625, "selectivity": 0.24038461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.34596465147534, "mean_class_cosine": 0.24687761314743237}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.4519230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.45664739884393063, "drop": -0.004724321920853702, "rotation": {"subspace_principal_angle": 1.4589762703207936, "mean_class_cosine": 0.9288307050828664}}}, "iid_acc": 0.4519230769230769, "selectivity": 0.06730769230769229, "iid_split_rotation": {"subspace_principal_angle": 1.2273866113889293, "mean_class_cosine": 0.9582100809299986}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.5865384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.5838150289017341, "drop": 0.002723432636727474}}, "iid_acc": 0.5865384615384616, "selectivity": 0.20192307692307693} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8904299583911235, "drop": 0.0045700416088765294, "rotation": {"subspace_principal_angle": 1.3187059663512182, "mean_class_cosine": 0.24942879345042712}}}, "iid_acc": 0.895, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2441873711885174, "mean_class_cosine": 0.3208330998449743}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5325936199722607, "drop": 0.01740638002773931, "rotation": {"subspace_principal_angle": 1.2286707925235416, "mean_class_cosine": -0.644580232070691}}}, "iid_acc": 0.55, "selectivity": -0.20875, "iid_split_rotation": {"subspace_principal_angle": 0.7593618744036684, "mean_class_cosine": 0.9814176976119293}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8945908460471568, "drop": 0.004159153952843275}}, "iid_acc": 0.89875, "selectivity": 0.14} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7098150782361309, "drop": 0.04768492176386907, "rotation": {"subspace_principal_angle": 1.2799135817505813, "mean_class_cosine": 0.28679799861480304}}, "domain": {"acc": 0.73375, "drop": 0.023749999999999938, "rotation": {"subspace_principal_angle": 1.206450803740472, "mean_class_cosine": 0.3563378556067195}}}, "iid_acc": 0.7575, "selectivity": 0.2587499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2152898580705969, "mean_class_cosine": 0.34806520925621826}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.5889046941678521, "drop": -0.03265469416785205, "rotation": {"subspace_principal_angle": 1.51956074817224, "mean_class_cosine": 0.8695877050620235}}, "domain": {"acc": 0.54125, "drop": 0.015000000000000013, "rotation": {"subspace_principal_angle": 1.5651562278626985, "mean_class_cosine": 0.5820640074758123}}}, "iid_acc": 0.55625, "selectivity": 0.057499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.548228069141704, "mean_class_cosine": 0.8592588697201281}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.7475, "drop": 0.0}, "paraphrase": {"acc": 0.7226173541963016, "drop": 0.024882645803698478}, "domain": {"acc": 0.7575, "drop": -0.009999999999999898}}, "iid_acc": 0.7475, "selectivity": 0.24875000000000003} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.807843137254902, "drop": -0.005343137254901986, "rotation": {"subspace_principal_angle": 0.9349235883028191, "mean_class_cosine": 0.5938799054215346}}, "domain": {"acc": 0.66, "drop": 0.14249999999999996, "rotation": {"subspace_principal_angle": 1.2644579436182715, "mean_class_cosine": 0.30156951858956016}}, "length": {"acc": 0.802962962962963, "drop": -0.0004629629629629983, "rotation": {"subspace_principal_angle": 0.8439392303466156, "mean_class_cosine": 0.6645243339696073}}}, "iid_acc": 0.8025, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 0.8932244951414783, "mean_class_cosine": 0.626903094729581}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.7137254901960784, "drop": -0.017475490196078414, "rotation": {"subspace_principal_angle": 1.1712305596628225, "mean_class_cosine": 0.9467024508176543}}, "domain": {"acc": 0.57875, "drop": 0.11750000000000005, "rotation": {"subspace_principal_angle": 1.3012077657528451, "mean_class_cosine": 0.356404654810238}}, "length": {"acc": 0.7051851851851851, "drop": -0.008935185185185102, "rotation": {"subspace_principal_angle": 1.1652413786828095, "mean_class_cosine": 0.9420615872580638}}}, "iid_acc": 0.69625, "selectivity": 0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4630243777096839, "mean_class_cosine": 0.8983235883657518}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.78375, "drop": 0.0}, "paraphrase": {"acc": 0.7862745098039216, "drop": -0.002524509803921604}, "domain": {"acc": 0.61125, "drop": 0.1725}, "length": {"acc": 0.7881481481481482, "drop": -0.0043981481481482065}}, "iid_acc": 0.78375, "selectivity": 0.30999999999999994} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.8475, "drop": 0.0}, "paraphrase": {"acc": 0.8348484848484848, "drop": 0.012651515151515191, "rotation": {"subspace_principal_angle": 1.4031997490020873, "mean_class_cosine": 0.28396714753242047}}}, "iid_acc": 0.8475, "selectivity": 0.5975, "iid_split_rotation": {"subspace_principal_angle": 1.3806377410186017, "mean_class_cosine": 0.32568678636943665}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8181818181818182, "drop": 0.008068181818181808, "rotation": {"subspace_principal_angle": 1.4758819738130697, "mean_class_cosine": 0.9741084999680931}}}, "iid_acc": 0.82625, "selectivity": 0.57625, "iid_split_rotation": {"subspace_principal_angle": 1.1310624243298266, "mean_class_cosine": 0.9619284246782831}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.865, "drop": 0.0}, "paraphrase": {"acc": 0.8545454545454545, "drop": 0.010454545454545494}}, "iid_acc": 0.865, "selectivity": 0.615} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9343065693430657, "drop": 0.011943430656934373, "rotation": {"subspace_principal_angle": 1.1865093507054412, "mean_class_cosine": 0.7073522515216414}}}, "iid_acc": 0.94625, "selectivity": 0.865, "iid_split_rotation": {"subspace_principal_angle": 1.127510316714073, "mean_class_cosine": 0.7428054255764255}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5571776155717761, "drop": 0.00907238442822389, "rotation": {"subspace_principal_angle": 1.3920610393607629, "mean_class_cosine": 0.7826885130462339}}}, "iid_acc": 0.56625, "selectivity": 0.48500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3330395256421586, "mean_class_cosine": 0.8909496426331701}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8491484184914841, "drop": 0.022101581508515822}}, "iid_acc": 0.87125, "selectivity": 0.7899999999999999} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.4967490247074122, "drop": 0.027000975292587825, "rotation": {"subspace_principal_angle": 1.56369432512863, "mean_class_cosine": 0.0071019419641176074}}}, "iid_acc": 0.52375, "selectivity": -0.011249999999999982, "iid_split_rotation": {"subspace_principal_angle": 1.5297750902165248, "mean_class_cosine": 0.04100973285438006}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.48764629388816644, "drop": 0.014853706111833509, "rotation": {"subspace_principal_angle": 1.4431237989109693, "mean_class_cosine": -0.12706524234551198}}}, "iid_acc": 0.5025, "selectivity": -0.032500000000000084, "iid_split_rotation": {"subspace_principal_angle": 1.5637689056526396, "mean_class_cosine": -0.006306566848071626}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5006501950585176, "drop": 0.01684980494148236}}, "iid_acc": 0.5175, "selectivity": -0.01750000000000007} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5309973045822103, "drop": 0.0690026954177897, "rotation": {"subspace_principal_angle": 1.38978118980255, "mean_class_cosine": 0.3392687986531371}}}, "iid_acc": 0.6, "selectivity": 0.33875, "iid_split_rotation": {"subspace_principal_angle": 1.4386478796411275, "mean_class_cosine": 0.3036790886331675}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.1375, "drop": 0.0}, "paraphrase": {"acc": 0.1792452830188679, "drop": -0.0417452830188679, "rotation": {"subspace_principal_angle": 1.4906453725067592, "mean_class_cosine": 0.7942802395915041}}}, "iid_acc": 0.1375, "selectivity": -0.12374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4328823629930647, "mean_class_cosine": 0.6275373859348595}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.49326145552560646, "drop": 0.014238544474393489}}, "iid_acc": 0.5075, "selectivity": 0.24624999999999997} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7259136212624585, "drop": 0.0015863787375415273, "rotation": {"subspace_principal_angle": 1.3191750526863224, "mean_class_cosine": 0.24897450603487703}}}, "iid_acc": 0.7275, "selectivity": 0.21750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3632129148081515, "mean_class_cosine": 0.2060957956694738}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5647840531561462, "drop": -0.016034053156146255, "rotation": {"subspace_principal_angle": 1.2764535135475243, "mean_class_cosine": 0.9760866896549567}}}, "iid_acc": 0.54875, "selectivity": 0.03874999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4318977466650444, "mean_class_cosine": 0.980247865923659}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7325581395348837, "drop": 0.012441860465116306}}, "iid_acc": 0.745, "selectivity": 0.235} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.603648424543947, "drop": 0.047601575456053036, "rotation": {"subspace_principal_angle": 1.439031942391082, "mean_class_cosine": 0.13138343627378768}}}, "iid_acc": 0.65125, "selectivity": 0.12875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.281323973512102, "mean_class_cosine": 0.2854465710869707}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.615257048092869, "drop": -0.00025704809286897223, "rotation": {"subspace_principal_angle": 1.3313954355137612, "mean_class_cosine": 0.9237404825006307}}}, "iid_acc": 0.615, "selectivity": 0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.5707267568617127, "mean_class_cosine": 0.8603555241845086}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.6119402985074627, "drop": 0.05680970149253728}}, "iid_acc": 0.66875, "selectivity": 0.14625} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.7037643207855974, "drop": 0.013735679214402663, "rotation": {"subspace_principal_angle": 1.2968057041198295, "mean_class_cosine": 0.27057534861786225}}}, "iid_acc": 0.7175, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2964308527675759, "mean_class_cosine": 0.2709361985468884}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.485, "drop": 0.0}, "paraphrase": {"acc": 0.5073649754500819, "drop": -0.02236497545008187, "rotation": {"subspace_principal_angle": 1.0405608352231381, "mean_class_cosine": 0.7548288630584341}}}, "iid_acc": 0.485, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4829069517857016, "mean_class_cosine": 0.8524232187677423}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.6988543371522095, "drop": 0.03364566284779058}}, "iid_acc": 0.7325, "selectivity": 0.12} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8884353741496599, "drop": 0.017814625850340104, "rotation": {"subspace_principal_angle": 1.0443295050595705, "mean_class_cosine": 0.5024817409879795}}}, "iid_acc": 0.90625, "selectivity": 0.415, "iid_split_rotation": {"subspace_principal_angle": 1.0612122822978838, "mean_class_cosine": 0.4878141817807545}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.7877551020408163, "drop": 0.05724489795918364, "rotation": {"subspace_principal_angle": 1.3338632968935769, "mean_class_cosine": 0.8769564447689145}}}, "iid_acc": 0.845, "selectivity": 0.35374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3069371602368167, "mean_class_cosine": 0.9412077325237846}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.8789115646258503, "drop": 0.026088435374149732}}, "iid_acc": 0.905, "selectivity": 0.41375} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.0027821252059307966, "rotation": {"subspace_principal_angle": 0.6856933607020452, "mean_class_cosine": 0.7739801903730489}}}, "iid_acc": 0.99125, "selectivity": 0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6339194714608114, "mean_class_cosine": 0.8057121716043742}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8517298187808896, "drop": 0.007020181219110411, "rotation": {"subspace_principal_angle": 1.4167093648556626, "mean_class_cosine": 0.9633020957225444}}}, "iid_acc": 0.85875, "selectivity": -0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 1.5217003295605112, "mean_class_cosine": 0.9881251646589356}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.985, "drop": 0.0}, "paraphrase": {"acc": 0.9835255354200988, "drop": 0.0014744645799011913}}, "iid_acc": 0.985, "selectivity": 0.10375000000000001} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.65875, "drop": 0.0}, "paraphrase": {"acc": 0.6511299435028248, "drop": 0.007620056497175121, "rotation": {"subspace_principal_angle": 1.5394490513572545, "mean_class_cosine": 0.031342141781080114}}}, "iid_acc": 0.65875, "selectivity": 0.00374999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.529266372043682, "mean_class_cosine": 0.04151801773809659}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5268361581920904, "drop": -0.011836158192090385, "rotation": {"subspace_principal_angle": 1.5178073225930298, "mean_class_cosine": 0.21529478327501697}}}, "iid_acc": 0.515, "selectivity": -0.14, "iid_split_rotation": {"subspace_principal_angle": 1.0749998948016042, "mean_class_cosine": -0.653032824583841}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.7231638418079096, "drop": -0.0031638418079096287}}, "iid_acc": 0.72, "selectivity": 0.06499999999999995} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.6105769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": -0.004807692307692291, "rotation": {"subspace_principal_angle": 1.399178613094325, "mean_class_cosine": 0.18957379279214578}}}, "iid_acc": 0.6105769230769231, "selectivity": 0.13942307692307698, "iid_split_rotation": {"subspace_principal_angle": 1.4319729985001466, "mean_class_cosine": 0.17742658711691725}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.4951923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.5443786982248521, "drop": -0.049186390532544366, "rotation": {"subspace_principal_angle": 1.3438575364804408, "mean_class_cosine": 0.641655124351597}}}, "iid_acc": 0.4951923076923077, "selectivity": 0.024038461538461564, "iid_split_rotation": {"subspace_principal_angle": 1.2506869110680288, "mean_class_cosine": 0.4854077602261109}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": -0.033653846153846145}}, "iid_acc": 0.5817307692307693, "selectivity": 0.11057692307692313} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8706896551724138, "drop": 0.009310344827586192, "rotation": {"subspace_principal_angle": 1.2345926633445847, "mean_class_cosine": 0.3299056836099217}}}, "iid_acc": 0.88, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.219582897698587, "mean_class_cosine": 0.3440374169146968}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.735632183908046, "drop": 0.010617816091954002, "rotation": {"subspace_principal_angle": 1.5495300864536512, "mean_class_cosine": 0.7777397530079471}}}, "iid_acc": 0.74625, "selectivity": -0.028750000000000053, "iid_split_rotation": {"subspace_principal_angle": 1.4860573861271493, "mean_class_cosine": 0.7695588747682932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.85, "drop": 0.0}, "paraphrase": {"acc": 0.8520114942528736, "drop": -0.0020114942528736135}}, "iid_acc": 0.85, "selectivity": 0.07499999999999996} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.78, "drop": 0.0}, "paraphrase": {"acc": 0.732574679943101, "drop": 0.047425320056899034, "rotation": {"subspace_principal_angle": 1.2608414774243961, "mean_class_cosine": 0.30501563767234113}}, "domain": {"acc": 0.73, "drop": 0.050000000000000044, "rotation": {"subspace_principal_angle": 1.2628486007462063, "mean_class_cosine": 0.30310354600989187}}}, "iid_acc": 0.78, "selectivity": 0.275, "iid_split_rotation": {"subspace_principal_angle": 1.2752896290640992, "mean_class_cosine": 0.29122462225458356}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5419630156472262, "drop": -0.028213015647226114, "rotation": {"subspace_principal_angle": 1.2025920048727983, "mean_class_cosine": 0.9965846798687883}}, "domain": {"acc": 0.54125, "drop": -0.02749999999999997, "rotation": {"subspace_principal_angle": 1.4516540873608994, "mean_class_cosine": 0.4073452915520843}}}, "iid_acc": 0.51375, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 0.9220355133014713, "mean_class_cosine": 0.9960516328052383}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.7695590327169275, "drop": 0.04044096728307256}, "domain": {"acc": 0.81, "drop": 0.0}}, "iid_acc": 0.81, "selectivity": 0.30500000000000005} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8475, "drop": 0.0}, "paraphrase": {"acc": 0.8176470588235294, "drop": 0.029852941176470638, "rotation": {"subspace_principal_angle": 1.2030769237259111, "mean_class_cosine": 0.35948823052238327}}, "domain": {"acc": 0.52875, "drop": 0.31875, "rotation": {"subspace_principal_angle": 1.3485990887438482, "mean_class_cosine": 0.2203733735747998}}, "length": {"acc": 0.845925925925926, "drop": 0.001574074074074061, "rotation": {"subspace_principal_angle": 1.1802807331888672, "mean_class_cosine": 0.3806652417654258}}}, "iid_acc": 0.8475, "selectivity": 0.32125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1501816602046397, "mean_class_cosine": 0.40832162126076316}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.6941176470588235, "drop": 0.017132352941176543, "rotation": {"subspace_principal_angle": 1.2416590458376622, "mean_class_cosine": 0.9484259293470975}}, "domain": {"acc": 0.555, "drop": 0.15625, "rotation": {"subspace_principal_angle": 1.321419755161037, "mean_class_cosine": 0.2893218657278887}}, "length": {"acc": 0.72, "drop": -0.008749999999999925, "rotation": {"subspace_principal_angle": 1.35683655045995, "mean_class_cosine": 0.9567589393114461}}}, "iid_acc": 0.71125, "selectivity": 0.18500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5633547700573491, "mean_class_cosine": 0.9200623797757711}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8117647058823529, "drop": 0.014485294117647096}, "domain": {"acc": 0.58125, "drop": 0.245}, "length": {"acc": 0.8266666666666667, "drop": -0.0004166666666666208}}, "iid_acc": 0.82625, "selectivity": 0.30000000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.8575, "drop": 0.0}, "paraphrase": {"acc": 0.8515151515151516, "drop": 0.005984848484848482, "rotation": {"subspace_principal_angle": 1.4452099258783577, "mean_class_cosine": 0.24580852369730838}}}, "iid_acc": 0.8575, "selectivity": 0.6312500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3943169835208806, "mean_class_cosine": 0.29648597603430116}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.83625, "drop": 0.0}, "paraphrase": {"acc": 0.8181818181818182, "drop": 0.018068181818181817, "rotation": {"subspace_principal_angle": 1.4625849696271982, "mean_class_cosine": 0.976912859464512}}}, "iid_acc": 0.83625, "selectivity": 0.6100000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3260532260143207, "mean_class_cosine": 0.9681319283557925}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.8727272727272727, "drop": 0.01727272727272733}}, "iid_acc": 0.89, "selectivity": 0.6637500000000001} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9343065693430657, "drop": 0.020693430656934297, "rotation": {"subspace_principal_angle": 1.197650762223889, "mean_class_cosine": 0.684597725874047}}}, "iid_acc": 0.955, "selectivity": 0.8875, "iid_split_rotation": {"subspace_principal_angle": 1.55650137949423, "mean_class_cosine": 0.7066198619660665}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5328467153284672, "drop": -0.00909671532846712, "rotation": {"subspace_principal_angle": 1.3016766284172454, "mean_class_cosine": 0.7910526489249412}}}, "iid_acc": 0.52375, "selectivity": 0.45625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2176842386779647, "mean_class_cosine": 0.8907647870321528}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.8953771289537713, "drop": 0.024622871046228734}}, "iid_acc": 0.92, "selectivity": 0.8525} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5110533159947984, "drop": 0.025196684005201586, "rotation": {"subspace_principal_angle": 1.5229524083126806, "mean_class_cosine": 0.04782566779303716}}}, "iid_acc": 0.53625, "selectivity": 0.07124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.447992810822192, "mean_class_cosine": 0.12249508806050173}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.48875, "drop": 0.0}, "paraphrase": {"acc": 0.4967490247074122, "drop": -0.007999024707412206, "rotation": {"subspace_principal_angle": 1.0259677749536884, "mean_class_cosine": -0.6292438791105933}}}, "iid_acc": 0.48875, "selectivity": 0.023749999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.512376005309043, "mean_class_cosine": 0.4076926470016709}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.4759427828348505, "drop": 0.021557217165149523}}, "iid_acc": 0.4975, "selectivity": 0.03249999999999997} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.64, "drop": 0.0}, "paraphrase": {"acc": 0.6078167115902965, "drop": 0.032183288409703525, "rotation": {"subspace_principal_angle": 1.5433122888881867, "mean_class_cosine": 0.42874421872110796}}}, "iid_acc": 0.64, "selectivity": 0.34750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.5081587697047814, "mean_class_cosine": 0.4198126394888509}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.3575, "drop": 0.0}, "paraphrase": {"acc": 0.31266846361185985, "drop": 0.044831536388140136, "rotation": {"subspace_principal_angle": 1.530582416292941, "mean_class_cosine": 0.6444087690823997}}}, "iid_acc": 0.3575, "selectivity": 0.065, "iid_split_rotation": {"subspace_principal_angle": 1.459558734848993, "mean_class_cosine": 0.588656959531061}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.5053908355795148, "drop": -0.0003908355795148033}}, "iid_acc": 0.505, "selectivity": 0.21250000000000002} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.7408637873754153, "drop": -0.023363787375415224, "rotation": {"subspace_principal_angle": 1.3639308415465492, "mean_class_cosine": 0.20539322845895566}}}, "iid_acc": 0.7175, "selectivity": 0.20500000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.3710931416929857, "mean_class_cosine": 0.19837842368661718}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.6212624584717608, "drop": -0.006262458471760768, "rotation": {"subspace_principal_angle": 1.5406708151005613, "mean_class_cosine": 0.8985506716369838}}}, "iid_acc": 0.615, "selectivity": 0.10250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5633757536581958, "mean_class_cosine": 0.9281917050542527}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7558139534883721, "drop": -0.0020639534883720723}}, "iid_acc": 0.75375, "selectivity": 0.24125000000000008} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6086235489220564, "drop": 0.02637645107794362, "rotation": {"subspace_principal_angle": 1.490593398738083, "mean_class_cosine": 0.08011697135563739}}}, "iid_acc": 0.635, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 1.4638090458782234, "mean_class_cosine": 0.10678329666310886}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.593698175787728, "drop": -0.0036981757877280197, "rotation": {"subspace_principal_angle": 1.5381806288996867, "mean_class_cosine": 0.9646917936431184}}}, "iid_acc": 0.59, "selectivity": 0.07499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.47230491449444, "mean_class_cosine": 0.9335513939805022}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6650082918739635, "drop": 0.014991708126036563}}, "iid_acc": 0.68, "selectivity": 0.16500000000000004} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.6939443535188216, "drop": 0.02605564648117842, "rotation": {"subspace_principal_angle": 1.3786060400308688, "mean_class_cosine": 0.19100931112364286}}}, "iid_acc": 0.72, "selectivity": 0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.264300653737959, "mean_class_cosine": 0.3017194819812067}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.4828150572831424, "drop": 0.01718494271685761, "rotation": {"subspace_principal_angle": 1.5623310465881193, "mean_class_cosine": 0.5999071653626515}}}, "iid_acc": 0.5, "selectivity": -0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4335616759166658, "mean_class_cosine": 0.882837690979637}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7217675941080196, "drop": 0.014482405891980354}}, "iid_acc": 0.73625, "selectivity": 0.09375} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.92875, "drop": 0.0}, "paraphrase": {"acc": 0.9224489795918367, "drop": 0.0063010204081632715, "rotation": {"subspace_principal_angle": 1.0136784498875773, "mean_class_cosine": 0.5287421015976792}}}, "iid_acc": 0.92875, "selectivity": 0.46124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.0680431153367775, "mean_class_cosine": 0.4818398888533543}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.5931972789115646, "drop": 0.06680272108843543, "rotation": {"subspace_principal_angle": 1.4285742922049103, "mean_class_cosine": 0.8425489186964704}}}, "iid_acc": 0.66, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.0889790851620134, "mean_class_cosine": 0.7904713230276075}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.9414965986394558, "drop": -0.02024659863945577}}, "iid_acc": 0.92125, "selectivity": 0.45375} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9835255354200988, "drop": 0.008974464579901253, "rotation": {"subspace_principal_angle": 1.0095289992802334, "mean_class_cosine": 0.5322595207735084}}}, "iid_acc": 0.9925, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9575577798603117, "mean_class_cosine": 0.5755189198683394}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.68875, "drop": 0.0}, "paraphrase": {"acc": 0.5535420098846787, "drop": 0.13520799011532125, "rotation": {"subspace_principal_angle": 0.9140922175459604, "mean_class_cosine": 0.9973997191189731}}}, "iid_acc": 0.68875, "selectivity": -0.17000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2996635368722853, "mean_class_cosine": 0.9996246655982532}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.001929571663920937}}, "iid_acc": 0.98875, "selectivity": 0.13} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6511299435028248, "drop": 0.022620056497175134, "rotation": {"subspace_principal_angle": 1.4877740078390536, "mean_class_cosine": 0.08292697708829477}}}, "iid_acc": 0.67375, "selectivity": 0.039999999999999925, "iid_split_rotation": {"subspace_principal_angle": 1.4794201635002593, "mean_class_cosine": 0.09124905725263827}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.4971751412429379, "drop": 0.03032485875706209, "rotation": {"subspace_principal_angle": 1.4466641276425085, "mean_class_cosine": 0.1695136595431389}}}, "iid_acc": 0.5275, "selectivity": -0.10625000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.9893876475048883, "mean_class_cosine": -0.7941063384626064}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.730225988700565, "drop": -0.006475988700564983}}, "iid_acc": 0.72375, "selectivity": 0.08999999999999997} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.6346153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.6449704142011834, "drop": -0.010355029585798814, "rotation": {"subspace_principal_angle": 1.401130641487477, "mean_class_cosine": 0.20190956693568948}}}, "iid_acc": 0.6346153846153846, "selectivity": 0.14423076923076922, "iid_split_rotation": {"subspace_principal_angle": 1.5640320536219627, "mean_class_cosine": 0.22919831492919082}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.5480769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.5739644970414202, "drop": -0.025887573964497035, "rotation": {"subspace_principal_angle": 1.3084230282064335, "mean_class_cosine": 0.6381937638179063}}}, "iid_acc": 0.5480769230769231, "selectivity": 0.057692307692307765, "iid_split_rotation": {"subspace_principal_angle": 1.1829697589881627, "mean_class_cosine": 0.46077624553615687}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.5913461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.621301775147929, "drop": -0.029955621301775093}}, "iid_acc": 0.5913461538461539, "selectivity": 0.10096153846153849} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.896551724137931, "drop": 0.005948275862068919, "rotation": {"subspace_principal_angle": 1.1447057554079878, "mean_class_cosine": 0.41331408912500195}}}, "iid_acc": 0.9025, "selectivity": 0.11624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.074604861072015, "mean_class_cosine": 0.4760797664026806}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7775, "drop": 0.0}, "paraphrase": {"acc": 0.7873563218390804, "drop": -0.009856321839080473, "rotation": {"subspace_principal_angle": 1.4653362403173142, "mean_class_cosine": 0.5927810736166856}}}, "iid_acc": 0.7775, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.5560355491414277, "mean_class_cosine": 0.4882192393355668}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.81875, "drop": 0.0}, "paraphrase": {"acc": 0.8117816091954023, "drop": 0.006968390804597657}}, "iid_acc": 0.81875, "selectivity": 0.03249999999999997} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7937411095305832, "drop": 0.010008890469416776, "rotation": {"subspace_principal_angle": 1.3185889844917067, "mean_class_cosine": 0.2495420761739433}}, "domain": {"acc": 0.8825, "drop": -0.07874999999999999, "rotation": {"subspace_principal_angle": 1.3325721583114072, "mean_class_cosine": 0.23597731993808926}}}, "iid_acc": 0.80375, "selectivity": 0.29874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.269745108593276, "mean_class_cosine": 0.2965243102986965}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5391180654338549, "drop": -0.02661806543385492, "rotation": {"subspace_principal_angle": 0.6582429807151361, "mean_class_cosine": 0.99926299379126}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.5312218618656819, "mean_class_cosine": 0.3356468163658598}}}, "iid_acc": 0.5125, "selectivity": 0.007499999999999951, "iid_split_rotation": {"subspace_principal_angle": 0.8629499226938586, "mean_class_cosine": 0.9991531829071234}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.802275960170697, "drop": 0.008974039829303004}, "domain": {"acc": 0.84625, "drop": -0.03499999999999992}}, "iid_acc": 0.81125, "selectivity": 0.30625} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8666666666666667, "drop": 0.0008333333333333526, "rotation": {"subspace_principal_angle": 1.1792412886091128, "mean_class_cosine": 0.38162622377493116}}, "domain": {"acc": 0.54125, "drop": 0.32625000000000004, "rotation": {"subspace_principal_angle": 1.3149401585776963, "mean_class_cosine": 0.25307379864091606}}, "length": {"acc": 0.8725925925925926, "drop": -0.0050925925925925375, "rotation": {"subspace_principal_angle": 1.129874880545573, "mean_class_cosine": 0.42677296414992083}}}, "iid_acc": 0.8675, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.1059112988008373, "mean_class_cosine": 0.44832003402375775}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7941176470588235, "drop": 0.00963235294117648, "rotation": {"subspace_principal_angle": 1.2208647478151893, "mean_class_cosine": 0.9486786112928252}}, "domain": {"acc": 0.555, "drop": 0.24874999999999992, "rotation": {"subspace_principal_angle": 1.4628719512711352, "mean_class_cosine": 0.16922760359562494}}, "length": {"acc": 0.8103703703703704, "drop": -0.006620370370370443, "rotation": {"subspace_principal_angle": 1.4518778845432394, "mean_class_cosine": 0.9386719672471594}}}, "iid_acc": 0.80375, "selectivity": 0.31249999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.1012982204245179, "mean_class_cosine": 0.8734916313973369}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8666666666666667, "drop": 0.01083333333333325}, "domain": {"acc": 0.525, "drop": 0.3524999999999999}, "length": {"acc": 0.8740740740740741, "drop": 0.0034259259259258323}}, "iid_acc": 0.8775, "selectivity": 0.3862499999999999} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.8681818181818182, "drop": -0.0019318181818182012, "rotation": {"subspace_principal_angle": 1.348541899340855, "mean_class_cosine": 0.33742946528630835}}}, "iid_acc": 0.86625, "selectivity": 0.6087499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3243203014914424, "mean_class_cosine": 0.3890416789316658}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5772727272727273, "drop": 0.011477272727272725, "rotation": {"subspace_principal_angle": 1.1652446863372317, "mean_class_cosine": 0.943376790057927}}}, "iid_acc": 0.58875, "selectivity": 0.33125, "iid_split_rotation": {"subspace_principal_angle": 1.0212787219956532, "mean_class_cosine": 0.9231286978282006}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8727272727272727, "drop": 0.002272727272727315}}, "iid_acc": 0.875, "selectivity": 0.6174999999999999} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.97, "drop": 0.0}, "paraphrase": {"acc": 0.9586374695863747, "drop": 0.011362530413625227, "rotation": {"subspace_principal_angle": 1.0788515585794918, "mean_class_cosine": 0.680525963213341}}}, "iid_acc": 0.97, "selectivity": 0.91625, "iid_split_rotation": {"subspace_principal_angle": 1.0595584522895658, "mean_class_cosine": 0.7030153073160818}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.732360097323601, "drop": 0.0026399026763990197, "rotation": {"subspace_principal_angle": 1.5116341861027407, "mean_class_cosine": 0.873531191022212}}}, "iid_acc": 0.735, "selectivity": 0.68125, "iid_split_rotation": {"subspace_principal_angle": 1.4944238666561842, "mean_class_cosine": 0.9319808959810626}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9513381995133819, "drop": 0.004911800486618101}}, "iid_acc": 0.95625, "selectivity": 0.9025000000000001} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.56125, "drop": 0.0}, "paraphrase": {"acc": 0.5565669700910273, "drop": 0.004683029908972736, "rotation": {"subspace_principal_angle": 1.4553250675447744, "mean_class_cosine": 0.11521482211569181}}}, "iid_acc": 0.56125, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.4999753911814653, "mean_class_cosine": 0.0707617488193961}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5097529258777633, "drop": -0.01975292587776334, "rotation": {"subspace_principal_angle": 1.5663024506899672, "mean_class_cosine": 0.00763914984376122}}}, "iid_acc": 0.49, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.2427849178014092, "mean_class_cosine": 0.3406471787964678}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.48, "drop": 0.0}, "paraphrase": {"acc": 0.49154746423927176, "drop": -0.011547464239271776}}, "iid_acc": 0.48, "selectivity": -0.018750000000000044} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.610512129380054, "drop": 0.028237870619946093, "rotation": {"subspace_principal_angle": 1.430640605755257, "mean_class_cosine": 0.2631769683858712}}}, "iid_acc": 0.63875, "selectivity": 0.39625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.378148191691318, "mean_class_cosine": 0.297004025063675}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.12, "drop": 0.0}, "paraphrase": {"acc": 0.1280323450134771, "drop": -0.008032345013477105, "rotation": {"subspace_principal_angle": 1.543620776429251, "mean_class_cosine": 0.9288205176328912}}}, "iid_acc": 0.12, "selectivity": -0.1225, "iid_split_rotation": {"subspace_principal_angle": 1.43244607202274, "mean_class_cosine": 0.6582331054084928}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.6185983827493261, "drop": -0.0198483827493261}}, "iid_acc": 0.59875, "selectivity": 0.35625} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7657807308970099, "drop": -0.024530730897009967, "rotation": {"subspace_principal_angle": 1.3660163460568273, "mean_class_cosine": 0.20335174270412884}}}, "iid_acc": 0.74125, "selectivity": 0.18999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.413083036491105, "mean_class_cosine": 0.1570602898398632}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5475, "drop": 0.0}, "paraphrase": {"acc": 0.5647840531561462, "drop": -0.01728405315614623, "rotation": {"subspace_principal_angle": 0.8355530661327796, "mean_class_cosine": 0.9953186629963564}}}, "iid_acc": 0.5475, "selectivity": -0.003750000000000031, "iid_split_rotation": {"subspace_principal_angle": 1.4453486412673606, "mean_class_cosine": 0.996980243254297}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7774086378737541, "drop": -0.004908637873754174}}, "iid_acc": 0.7725, "selectivity": 0.22124999999999995} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6252072968490879, "drop": 0.044792703150912105, "rotation": {"subspace_principal_angle": 1.4695161231652571, "mean_class_cosine": 0.10110714243323128}}}, "iid_acc": 0.67, "selectivity": 0.17250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4419616097046257, "mean_class_cosine": 0.12847860473843392}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5174129353233831, "drop": 0.010087064676616908, "rotation": {"subspace_principal_angle": 0.9292634922449805, "mean_class_cosine": 0.9948800762828822}}}, "iid_acc": 0.5275, "selectivity": 0.02999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.044667492526286, "mean_class_cosine": 0.9982719754571383}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6882255389718076, "drop": 0.005524461028192373}}, "iid_acc": 0.69375, "selectivity": 0.19624999999999998} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.6890343698854338, "drop": 0.022215630114566287, "rotation": {"subspace_principal_angle": 1.44866296261917, "mean_class_cosine": 0.12182995563984489}}}, "iid_acc": 0.71125, "selectivity": 0.17875000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4353636321031356, "mean_class_cosine": 0.13501905614072562}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.45625, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.018381751227495913, "rotation": {"subspace_principal_angle": 1.4711476398340722, "mean_class_cosine": -0.12720149911565562}}}, "iid_acc": 0.45625, "selectivity": -0.07624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3619831712096353, "mean_class_cosine": 0.9877340756601919}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7495908346972177, "drop": 0.007909165302782295}}, "iid_acc": 0.7575, "selectivity": 0.22499999999999998} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9333333333333333, "drop": 0.009166666666666656, "rotation": {"subspace_principal_angle": 1.1452162894508524, "mean_class_cosine": 0.41284914883949264}}}, "iid_acc": 0.9425, "selectivity": 0.45375, "iid_split_rotation": {"subspace_principal_angle": 1.1544129671241614, "mean_class_cosine": 0.4044554792046704}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.82125, "drop": 0.0}, "paraphrase": {"acc": 0.8108843537414966, "drop": 0.010365646258503447, "rotation": {"subspace_principal_angle": 1.4778841101146312, "mean_class_cosine": 0.9112587174612317}}}, "iid_acc": 0.82125, "selectivity": 0.3325, "iid_split_rotation": {"subspace_principal_angle": 1.2685804950112414, "mean_class_cosine": 0.7130834799426322}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9374149659863945, "drop": 0.005085034013605494}}, "iid_acc": 0.9425, "selectivity": 0.45375} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9818780889621087, "drop": 0.011871911037891314, "rotation": {"subspace_principal_angle": 1.1022275614770123, "mean_class_cosine": 0.4516097785057144}}}, "iid_acc": 0.99375, "selectivity": 0.23875000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.0082315927465428, "mean_class_cosine": 0.5333574325360171}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9654036243822076, "drop": -0.00915362438220757, "rotation": {"subspace_principal_angle": 1.43644764939401, "mean_class_cosine": 0.9461248783860318}}}, "iid_acc": 0.95625, "selectivity": 0.20125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5120745206596544, "mean_class_cosine": 0.9885336177553897}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.008577018121911029}}, "iid_acc": 0.99375, "selectivity": 0.23875000000000002} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.693502824858757, "drop": 0.01774717514124302, "rotation": {"subspace_principal_angle": 1.4474830685331326, "mean_class_cosine": 0.12300097560505913}}}, "iid_acc": 0.71125, "selectivity": 0.09625000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3987832690834892, "mean_class_cosine": 0.1711660439449748}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5155367231638418, "drop": -0.015536723163841804, "rotation": {"subspace_principal_angle": 1.3401905016501925, "mean_class_cosine": 0.29339747507161273}}}, "iid_acc": 0.5, "selectivity": -0.11499999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6179518034185482, "mean_class_cosine": -0.9377606995696546}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.7259887005649718, "drop": -0.009738700564971725}}, "iid_acc": 0.71625, "selectivity": 0.10125000000000006} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.6586538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.6568047337278107, "drop": 0.0018491124260354708, "rotation": {"subspace_principal_angle": 1.4172417633253025, "mean_class_cosine": 0.17412301577924724}}}, "iid_acc": 0.6586538461538461, "selectivity": 0.21634615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.3955447682800144, "mean_class_cosine": 0.20505793596312397}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.49038461538461536, "drop": 0.0}, "paraphrase": {"acc": 0.4319526627218935, "drop": 0.05843195266272189, "rotation": {"subspace_principal_angle": 1.0501848723626588, "mean_class_cosine": 0.8199704533724792}}}, "iid_acc": 0.49038461538461536, "selectivity": 0.04807692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.1738154155015759, "mean_class_cosine": 0.567012811542814}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.6778846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.0625}}, "iid_acc": 0.6778846153846154, "selectivity": 0.23557692307692313} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.8951149425287356, "drop": 0.016135057471264425, "rotation": {"subspace_principal_angle": 1.3118705655633425, "mean_class_cosine": 0.2560422701778298}}}, "iid_acc": 0.91125, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3182833995213499, "mean_class_cosine": 0.24983798198888385}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5804597701149425, "drop": 0.02079022988505741, "rotation": {"subspace_principal_angle": 1.0152273026792804, "mean_class_cosine": 0.8731250404929112}}}, "iid_acc": 0.60125, "selectivity": -0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9386885075423576, "mean_class_cosine": 0.5900924520237005}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.9037356321839081, "drop": 0.00501436781609188}}, "iid_acc": 0.90875, "selectivity": 0.15999999999999992} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.8421052631578947, "drop": 0.026644736842105332, "rotation": {"subspace_principal_angle": 1.3051008608954664, "mean_class_cosine": 0.262580393680383}}, "domain": {"acc": 0.865, "drop": 0.003750000000000031, "rotation": {"subspace_principal_angle": 1.4287631781026644, "mean_class_cosine": 0.1415560812007921}}}, "iid_acc": 0.86875, "selectivity": 0.35750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2792444334922417, "mean_class_cosine": 0.28743897242092714}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.51125, "drop": 0.0}, "paraphrase": {"acc": 0.5391180654338549, "drop": -0.027868065433854894, "rotation": {"subspace_principal_angle": 1.5601359902088487, "mean_class_cosine": 0.9979053013441271}}, "domain": {"acc": 0.54125, "drop": -0.030000000000000027, "rotation": {"subspace_principal_angle": 1.4395920974393557, "mean_class_cosine": 0.23107881587582263}}}, "iid_acc": 0.51125, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.1884475439272368, "mean_class_cosine": 0.9977411853469901}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8492176386913229, "drop": 0.020782361308677055}, "domain": {"acc": 0.83375, "drop": 0.036250000000000004}}, "iid_acc": 0.87, "selectivity": 0.35875} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8647058823529412, "drop": 0.02904411764705883, "rotation": {"subspace_principal_angle": 1.3891107177208437, "mean_class_cosine": 0.1806876941601903}}, "domain": {"acc": 0.49, "drop": 0.40375000000000005, "rotation": {"subspace_principal_angle": 1.470272439761965, "mean_class_cosine": 0.10035467270286502}}, "length": {"acc": 0.8933333333333333, "drop": 0.0004166666666667318, "rotation": {"subspace_principal_angle": 1.336894896267935, "mean_class_cosine": 0.23177447065599882}}}, "iid_acc": 0.89375, "selectivity": 0.40125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.274478152252649, "mean_class_cosine": 0.2920008293919314}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8117647058823529, "drop": 0.0194852941176471, "rotation": {"subspace_principal_angle": 1.3046911642334313, "mean_class_cosine": 0.9572532608121695}}, "domain": {"acc": 0.48125, "drop": 0.35000000000000003, "rotation": {"subspace_principal_angle": 1.2118756513387043, "mean_class_cosine": -0.011293529250558751}}, "length": {"acc": 0.8355555555555556, "drop": -0.0043055555555555625, "rotation": {"subspace_principal_angle": 1.2439536531709152, "mean_class_cosine": 0.9646290213720963}}}, "iid_acc": 0.83125, "selectivity": 0.33875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5181831450402061, "mean_class_cosine": 0.9548748861745605}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8862745098039215, "drop": 0.007475490196078516}, "domain": {"acc": 0.7925, "drop": 0.10125000000000006}, "length": {"acc": 0.8962962962962963, "drop": -0.0025462962962962132}}, "iid_acc": 0.89375, "selectivity": 0.40125000000000005} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8818181818181818, "drop": 0.009431818181818152, "rotation": {"subspace_principal_angle": 1.4369187993598345, "mean_class_cosine": 0.22041085502998395}}}, "iid_acc": 0.89125, "selectivity": 0.675, "iid_split_rotation": {"subspace_principal_angle": 1.4078654977043972, "mean_class_cosine": 0.2586407992914274}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.7727272727272727, "drop": 0.037272727272727346, "rotation": {"subspace_principal_angle": 1.0175502832205285, "mean_class_cosine": 0.972353587476467}}}, "iid_acc": 0.81, "selectivity": 0.59375, "iid_split_rotation": {"subspace_principal_angle": 0.7420502510291406, "mean_class_cosine": 0.9634510604904396}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8833333333333333, "drop": 0.004166666666666652}}, "iid_acc": 0.8875, "selectivity": 0.6712499999999999} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9683698296836983, "drop": 0.007880170316301616, "rotation": {"subspace_principal_angle": 1.0992412190119936, "mean_class_cosine": 0.6867400981100495}}}, "iid_acc": 0.97625, "selectivity": 0.91875, "iid_split_rotation": {"subspace_principal_angle": 1.0034069622469282, "mean_class_cosine": 0.7160948548382569}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.80125, "drop": 0.0}, "paraphrase": {"acc": 0.8175182481751825, "drop": -0.016268248175182465, "rotation": {"subspace_principal_angle": 1.379334954656771, "mean_class_cosine": 0.8678086437576017}}}, "iid_acc": 0.80125, "selectivity": 0.74375, "iid_split_rotation": {"subspace_principal_angle": 0.8811167846778503, "mean_class_cosine": 0.9382525612852962}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9635036496350365, "drop": 0.01274635036496341}}, "iid_acc": 0.97625, "selectivity": 0.91875} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.5925, "drop": 0.0}, "paraphrase": {"acc": 0.599479843953186, "drop": -0.00697984395318596, "rotation": {"subspace_principal_angle": 1.4685167248622502, "mean_class_cosine": 0.10210136870418976}}}, "iid_acc": 0.5925, "selectivity": 0.09625, "iid_split_rotation": {"subspace_principal_angle": 1.492714330722181, "mean_class_cosine": 0.07800267856136348}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5097529258777633, "drop": -0.01975292587776334, "rotation": {"subspace_principal_angle": 1.3905538624434135, "mean_class_cosine": 0.43937650291700403}}}, "iid_acc": 0.49, "selectivity": -0.006250000000000033, "iid_split_rotation": {"subspace_principal_angle": 1.4581021828043095, "mean_class_cosine": -0.1678392152718658}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.577373211963589, "drop": -0.017373211963588986}}, "iid_acc": 0.56, "selectivity": 0.06375000000000003} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.65625, "drop": 0.0}, "paraphrase": {"acc": 0.6320754716981132, "drop": 0.024174528301886822, "rotation": {"subspace_principal_angle": 1.4488303719308924, "mean_class_cosine": 0.2608030814489897}}}, "iid_acc": 0.65625, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.4365260084834874, "mean_class_cosine": 0.27287586977099626}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.12, "drop": 0.0}, "paraphrase": {"acc": 0.12533692722371967, "drop": -0.005336927223719673, "rotation": {"subspace_principal_angle": 1.4252705910030379, "mean_class_cosine": 0.9453360722802503}}}, "iid_acc": 0.12, "selectivity": -0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.444135686833852, "mean_class_cosine": 0.6596132535425966}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6792452830188679, "drop": 0.0007547169811321641}}, "iid_acc": 0.68, "selectivity": 0.41375000000000006} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.760797342192691, "drop": -0.007047342192690986, "rotation": {"subspace_principal_angle": 1.4082028248717875, "mean_class_cosine": 0.16187804379501777}}}, "iid_acc": 0.75375, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.4629230727487192, "mean_class_cosine": 0.10766416205491924}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5813953488372093, "drop": -0.021395348837209283, "rotation": {"subspace_principal_angle": 0.7137539003362412, "mean_class_cosine": 0.9509360173967503}}}, "iid_acc": 0.56, "selectivity": 0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.0125517193691644, "mean_class_cosine": 0.9691372563411869}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.782392026578073, "drop": -0.008642026578073003}}, "iid_acc": 0.77375, "selectivity": 0.23875000000000002} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6451077943615257, "drop": 0.032392205638474336, "rotation": {"subspace_principal_angle": 1.4944467452146757, "mean_class_cosine": 0.0762754262877726}}}, "iid_acc": 0.6775, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.510090306300586, "mean_class_cosine": 0.06066874151454521}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5190713101160862, "drop": 0.005928689883913818, "rotation": {"subspace_principal_angle": 0.9851631097477104, "mean_class_cosine": 0.9870998388876457}}}, "iid_acc": 0.525, "selectivity": 0.01375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.8525820765093011, "mean_class_cosine": 0.9943272408460271}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.681592039800995, "drop": 0.028407960199004934}}, "iid_acc": 0.71, "selectivity": 0.19874999999999998} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7184942716857611, "drop": 0.01650572831423891, "rotation": {"subspace_principal_angle": 1.4867843877013787, "mean_class_cosine": 0.08391314783631819}}}, "iid_acc": 0.735, "selectivity": 0.18499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4727261303836074, "mean_class_cosine": 0.09791306933287178}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.4575, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.017131751227495884, "rotation": {"subspace_principal_angle": 1.4184867997055783, "mean_class_cosine": -0.12546411952087166}}}, "iid_acc": 0.4575, "selectivity": -0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0826072069356356, "mean_class_cosine": 0.9756432630642227}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7446808510638298, "drop": -0.004680851063829761}}, "iid_acc": 0.74, "selectivity": 0.18999999999999995} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9401360544217687, "drop": 0.01861394557823126, "rotation": {"subspace_principal_angle": 1.1843775125331726, "mean_class_cosine": 0.3768737131337715}}}, "iid_acc": 0.95875, "selectivity": 0.5037499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1630113814353649, "mean_class_cosine": 0.3965768790298291}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.7755102040816326, "drop": 0.026989795918367365, "rotation": {"subspace_principal_angle": 0.9611960505376331, "mean_class_cosine": 0.9267288005728522}}}, "iid_acc": 0.8025, "selectivity": 0.3475, "iid_split_rotation": {"subspace_principal_angle": 0.6069137112776986, "mean_class_cosine": 0.8431794446206526}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9523809523809523, "drop": 0.0038690476190477163}}, "iid_acc": 0.95625, "selectivity": 0.50125} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.005282125205930854, "rotation": {"subspace_principal_angle": 1.0067410745718517, "mean_class_cosine": 0.5346176539849523}}}, "iid_acc": 0.99375, "selectivity": 0.16125, "iid_split_rotation": {"subspace_principal_angle": 0.9448974604279329, "mean_class_cosine": 0.5858259839474748}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.66625, "drop": 0.0}, "paraphrase": {"acc": 0.40362438220757824, "drop": 0.2626256177924218, "rotation": {"subspace_principal_angle": 0.5567698167152252, "mean_class_cosine": 0.9809168363033169}}}, "iid_acc": 0.66625, "selectivity": -0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.4341855078247439, "mean_class_cosine": 0.9999348952126837}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.006077018121910971}}, "iid_acc": 0.99125, "selectivity": 0.15874999999999995} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.6878531073446328, "drop": 0.030896892655367214, "rotation": {"subspace_principal_angle": 1.4853872573002862, "mean_class_cosine": 0.08530526830833424}}}, "iid_acc": 0.71875, "selectivity": 0.11499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3987163989876108, "mean_class_cosine": 0.1712319268023629}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.5240112994350282, "drop": -0.019011299435028217, "rotation": {"subspace_principal_angle": 1.4845790668606034, "mean_class_cosine": 0.3310855041248773}}}, "iid_acc": 0.505, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.4077945456675773, "mean_class_cosine": -0.8399837311477907}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7372881355932204, "drop": 0.026461864406779667}}, "iid_acc": 0.76375, "selectivity": 0.16000000000000003} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6686390532544378, "drop": 0.018860946745562157, "rotation": {"subspace_principal_angle": 1.4431099310758673, "mean_class_cosine": 0.13799385145730425}}}, "iid_acc": 0.6875, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.4632019943344636, "mean_class_cosine": 0.12184824535087391}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.5913461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.5502958579881657, "drop": 0.04105029585798814, "rotation": {"subspace_principal_angle": 1.0140814812234873, "mean_class_cosine": 0.7438713263454885}}}, "iid_acc": 0.5913461538461539, "selectivity": 0.1634615384615385, "iid_split_rotation": {"subspace_principal_angle": 0.9988210743938787, "mean_class_cosine": 0.6629369773883759}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6745562130177515, "drop": 0.012943786982248517}}, "iid_acc": 0.6875, "selectivity": 0.25961538461538464} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8936781609195402, "drop": 0.013821839080459752, "rotation": {"subspace_principal_angle": 1.3954839068042866, "mean_class_cosine": 0.17441577734760833}}}, "iid_acc": 0.9075, "selectivity": 0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3326472336535684, "mean_class_cosine": 0.23590436416653315}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6135057471264368, "drop": 0.021494252873563213, "rotation": {"subspace_principal_angle": 0.9373099242351036, "mean_class_cosine": 0.8137785112536557}}}, "iid_acc": 0.635, "selectivity": -0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3179859570982269, "mean_class_cosine": 0.6358483449489087}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9051724137931034, "drop": 0.00982758620689661}}, "iid_acc": 0.915, "selectivity": 0.17500000000000004} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.795, "drop": 0.0}, "paraphrase": {"acc": 0.786628733997155, "drop": 0.00837126600284499, "rotation": {"subspace_principal_angle": 1.310108733461078, "mean_class_cosine": 0.2577449743747703}}, "domain": {"acc": 0.79, "drop": 0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.3864147968628893, "mean_class_cosine": 0.18333858173678275}}}, "iid_acc": 0.795, "selectivity": 0.28500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3577658636568546, "mean_class_cosine": 0.2114228247278906}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5362731152204836, "drop": -0.023773115220483643, "rotation": {"subspace_principal_angle": 1.4911474327942673, "mean_class_cosine": 0.9991258530703471}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.4889114257860925, "mean_class_cosine": 0.43982299953686566}}}, "iid_acc": 0.5125, "selectivity": 0.0024999999999999467, "iid_split_rotation": {"subspace_principal_angle": 1.5599854028146705, "mean_class_cosine": 0.9990812404452749}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.82, "drop": 0.0}, "paraphrase": {"acc": 0.7923186344238976, "drop": 0.027681365576102346}, "domain": {"acc": 0.83125, "drop": -0.011250000000000093}}, "iid_acc": 0.82, "selectivity": 0.30999999999999994} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8058823529411765, "drop": 0.03161764705882353, "rotation": {"subspace_principal_angle": 1.3708466120257015, "mean_class_cosine": 0.19862004766983304}}, "domain": {"acc": 0.68125, "drop": 0.15625, "rotation": {"subspace_principal_angle": 1.384027843480055, "mean_class_cosine": 0.1856845513313584}}, "length": {"acc": 0.8488888888888889, "drop": -0.011388888888888893, "rotation": {"subspace_principal_angle": 1.3282357766524422, "mean_class_cosine": 0.24018900406085658}}}, "iid_acc": 0.8375, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 1.3949678570258965, "mean_class_cosine": 0.17492389391787147}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.77625, "drop": 0.0}, "paraphrase": {"acc": 0.7823529411764706, "drop": -0.006102941176470589, "rotation": {"subspace_principal_angle": 0.8158694259835028, "mean_class_cosine": 0.9437326197394826}}, "domain": {"acc": 0.555, "drop": 0.22124999999999995, "rotation": {"subspace_principal_angle": 1.4680655032048706, "mean_class_cosine": 0.22680383044754415}}, "length": {"acc": 0.7866666666666666, "drop": -0.01041666666666663, "rotation": {"subspace_principal_angle": 1.075088915720483, "mean_class_cosine": 0.9383123548837031}}}, "iid_acc": 0.77625, "selectivity": 0.26749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3496789801104079, "mean_class_cosine": 0.92006570726007}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.8431372549019608, "drop": 0.008112745098039165}, "domain": {"acc": 0.55, "drop": 0.3012499999999999}, "length": {"acc": 0.8548148148148148, "drop": -0.003564814814814854}}, "iid_acc": 0.85125, "selectivity": 0.3424999999999999} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8363636363636363, "drop": 0.0023863636363636864, "rotation": {"subspace_principal_angle": 1.3907909651765882, "mean_class_cosine": 0.27160223854429794}}}, "iid_acc": 0.83875, "selectivity": 0.58375, "iid_split_rotation": {"subspace_principal_angle": 1.374984073519207, "mean_class_cosine": 0.3084279719861868}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5393939393939394, "drop": -0.008143939393939426, "rotation": {"subspace_principal_angle": 1.3095900045768096, "mean_class_cosine": 0.9315569392551215}}}, "iid_acc": 0.53125, "selectivity": 0.27625, "iid_split_rotation": {"subspace_principal_angle": 1.3512999965296404, "mean_class_cosine": 0.9302928191576891}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8712121212121212, "drop": 0.002537878787878811}}, "iid_acc": 0.87375, "selectivity": 0.61875} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9416058394160584, "drop": 0.017144160583941637, "rotation": {"subspace_principal_angle": 1.169967417171358, "mean_class_cosine": 0.6021107227776437}}}, "iid_acc": 0.95875, "selectivity": 0.88, "iid_split_rotation": {"subspace_principal_angle": 1.0879671022275426, "mean_class_cosine": 0.6463882048277961}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5717761557177615, "drop": -0.005526155717761494, "rotation": {"subspace_principal_angle": 1.5297708862576174, "mean_class_cosine": 0.8076540626245426}}}, "iid_acc": 0.56625, "selectivity": 0.48750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.31250485165038, "mean_class_cosine": 0.8967378658864777}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9245742092457421, "drop": 0.03417579075425792}}, "iid_acc": 0.95875, "selectivity": 0.88} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5370611183355006, "drop": 0.030438881664499373, "rotation": {"subspace_principal_angle": 1.4333280858176263, "mean_class_cosine": 0.13703568181783926}}}, "iid_acc": 0.5675, "selectivity": 0.08000000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5372359367014987, "mean_class_cosine": 0.03355409060481619}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5019505851755527, "drop": -0.0019505851755526882, "rotation": {"subspace_principal_angle": 1.5166639134271087, "mean_class_cosine": 0.7284040860032104}}}, "iid_acc": 0.5, "selectivity": 0.012500000000000011, "iid_split_rotation": {"subspace_principal_angle": 1.5345005286615787, "mean_class_cosine": -0.9244877025726085}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5136540962288687, "drop": -0.0011540962288687506}}, "iid_acc": 0.5125, "selectivity": 0.024999999999999967} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.6475, "drop": 0.0}, "paraphrase": {"acc": 0.578167115902965, "drop": 0.06933288409703497, "rotation": {"subspace_principal_angle": 1.4396284306360443, "mean_class_cosine": 0.2858703332393371}}}, "iid_acc": 0.6475, "selectivity": 0.39249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4241221420081531, "mean_class_cosine": 0.3414796507506154}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.2, "drop": 0.0}, "paraphrase": {"acc": 0.2169811320754717, "drop": -0.016981132075471694, "rotation": {"subspace_principal_angle": 1.478985421617415, "mean_class_cosine": 0.7260803181295756}}}, "iid_acc": 0.2, "selectivity": -0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3657973968851722, "mean_class_cosine": 0.6865447650543287}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.62375, "drop": 0.0}, "paraphrase": {"acc": 0.5956873315363881, "drop": 0.02806266846361194}}, "iid_acc": 0.62375, "selectivity": 0.36875} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7541528239202658, "drop": -0.004152823920265836, "rotation": {"subspace_principal_angle": 1.3055947110598365, "mean_class_cosine": 0.26210384065900993}}}, "iid_acc": 0.75, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4135255349018365, "mean_class_cosine": 0.15662326790731992}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6295681063122923, "drop": -0.007068106312292266, "rotation": {"subspace_principal_angle": 1.412398245668166, "mean_class_cosine": 0.9872471083520573}}}, "iid_acc": 0.6225, "selectivity": 0.06875000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.0308446538703597, "mean_class_cosine": 0.9830930319085469}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.7458471760797342, "drop": 0.0004028239202658046}}, "iid_acc": 0.74625, "selectivity": 0.1925} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.64625, "drop": 0.0}, "paraphrase": {"acc": 0.593698175787728, "drop": 0.052551824212272, "rotation": {"subspace_principal_angle": 1.4774953080617963, "mean_class_cosine": 0.09316571216572649}}}, "iid_acc": 0.64625, "selectivity": 0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4132962154086606, "mean_class_cosine": 0.15684975311512517}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5174129353233831, "drop": 0.010087064676616908, "rotation": {"subspace_principal_angle": 1.4727791381339745, "mean_class_cosine": 0.9962811296518983}}}, "iid_acc": 0.5275, "selectivity": 0.023749999999999938, "iid_split_rotation": {"subspace_principal_angle": 1.324949369047214, "mean_class_cosine": 0.9986827286127624}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.6633499170812603, "drop": 0.0379000829187397}}, "iid_acc": 0.70125, "selectivity": 0.1975} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6972176759410802, "drop": 0.007782324058919765, "rotation": {"subspace_principal_angle": 1.5034674219806008, "mean_class_cosine": 0.06727804731954892}}}, "iid_acc": 0.705, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4178710295321864, "mean_class_cosine": 0.1523299382782468}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.46125, "drop": 0.0}, "paraphrase": {"acc": 0.469721767594108, "drop": -0.008471767594108004, "rotation": {"subspace_principal_angle": 1.2701697435580983, "mean_class_cosine": -0.4177592757825601}}}, "iid_acc": 0.46125, "selectivity": -0.10749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4728680034154755, "mean_class_cosine": 0.9927211138662637}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.723404255319149, "drop": 0.0015957446808509967}}, "iid_acc": 0.725, "selectivity": 0.15625} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8993197278911564, "drop": 0.00693027210884356, "rotation": {"subspace_principal_angle": 1.2974026238207301, "mean_class_cosine": 0.2700006465670901}}}, "iid_acc": 0.90625, "selectivity": 0.44125, "iid_split_rotation": {"subspace_principal_angle": 1.2752165891310814, "mean_class_cosine": 0.29129449547307984}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8761904761904762, "drop": -0.0024404761904761596, "rotation": {"subspace_principal_angle": 1.0942486888882677, "mean_class_cosine": 0.942913260405956}}}, "iid_acc": 0.87375, "selectivity": 0.40875, "iid_split_rotation": {"subspace_principal_angle": 0.9061762984312886, "mean_class_cosine": 0.8552558967615458}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9306122448979591, "drop": -0.008112244897959142}}, "iid_acc": 0.9225, "selectivity": 0.45749999999999996} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.007327018121911055, "rotation": {"subspace_principal_angle": 1.072930156314306, "mean_class_cosine": 0.4775518369824139}}}, "iid_acc": 0.9925, "selectivity": 0.22375, "iid_split_rotation": {"subspace_principal_angle": 1.0267208542859751, "mean_class_cosine": 0.5176272803692485}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.7100494233937397, "drop": -0.022549423393739665, "rotation": {"subspace_principal_angle": 1.2433728178446086, "mean_class_cosine": 0.997937410347562}}}, "iid_acc": 0.6875, "selectivity": -0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4234682182286518, "mean_class_cosine": 0.9995278492579525}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.005679571663920968}}, "iid_acc": 0.9925, "selectivity": 0.22375} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.6675, "drop": 0.0}, "paraphrase": {"acc": 0.6652542372881356, "drop": 0.0022457627118643853, "rotation": {"subspace_principal_angle": 1.5101964181742233, "mean_class_cosine": 0.0605628247623603}}}, "iid_acc": 0.6675, "selectivity": 0.06499999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4116411659990913, "mean_class_cosine": 0.1584841015440742}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.495, "drop": 0.0}, "paraphrase": {"acc": 0.5112994350282486, "drop": -0.0162994350282486, "rotation": {"subspace_principal_angle": 1.2440118776947484, "mean_class_cosine": 0.4397585570740069}}}, "iid_acc": 0.495, "selectivity": -0.10750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.0189797005862058, "mean_class_cosine": -0.9017093672978498}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7090395480225988, "drop": 0.01721045197740112}}, "iid_acc": 0.72625, "selectivity": 0.12374999999999992} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.5976331360946746, "drop": 0.04178994082840226, "rotation": {"subspace_principal_angle": 1.3670769632091069, "mean_class_cosine": 0.221961109804074}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.13942307692307687, "iid_split_rotation": {"subspace_principal_angle": 1.3869516790493233, "mean_class_cosine": 0.2430729393469231}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.3269230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.35502958579881655, "drop": -0.028106508875739622, "rotation": {"subspace_principal_angle": 1.4355728624955333, "mean_class_cosine": 0.7267142033085312}}}, "iid_acc": 0.3269230769230769, "selectivity": -0.17307692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.1849716820736076, "mean_class_cosine": 0.7548318595334315}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.033653846153846145}}, "iid_acc": 0.6490384615384616, "selectivity": 0.14903846153846156} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8936781609195402, "drop": 0.007571839080459775, "rotation": {"subspace_principal_angle": 1.404231999409768, "mean_class_cosine": 0.1657952106350371}}}, "iid_acc": 0.90125, "selectivity": 0.19125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.415744792030871, "mean_class_cosine": 0.1544310160272101}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5919540229885057, "drop": 0.011795977011494263, "rotation": {"subspace_principal_angle": 0.829164366422975, "mean_class_cosine": 0.8319128806446898}}}, "iid_acc": 0.60375, "selectivity": -0.10624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1149004124569097, "mean_class_cosine": 0.34312396741171547}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9123563218390804, "drop": 0.00514367816091954}}, "iid_acc": 0.9175, "selectivity": 0.20750000000000002} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.7837837837837838, "drop": 0.04246621621621627, "rotation": {"subspace_principal_angle": 1.3737563564873747, "mean_class_cosine": 0.1957674385129709}}, "domain": {"acc": 0.80125, "drop": 0.025000000000000022, "rotation": {"subspace_principal_angle": 1.376738696045229, "mean_class_cosine": 0.19284193990423557}}}, "iid_acc": 0.82625, "selectivity": 0.32000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3152150432578507, "mean_class_cosine": 0.2528078527302202}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5362731152204836, "drop": -0.023773115220483643, "rotation": {"subspace_principal_angle": 0.8284726401777219, "mean_class_cosine": 0.9987588658556924}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.559015217957762, "mean_class_cosine": 0.37022421141777156}}}, "iid_acc": 0.5125, "selectivity": 0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.4590601880015512, "mean_class_cosine": 0.9985735288523132}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.84125, "drop": 0.0}, "paraphrase": {"acc": 0.7809388335704125, "drop": 0.06031116642958756}, "domain": {"acc": 0.84125, "drop": 0.0}}, "iid_acc": 0.84125, "selectivity": 0.3350000000000001} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8606060606060606, "drop": 0.019393939393939408, "rotation": {"subspace_principal_angle": 1.3448125803660131, "mean_class_cosine": 0.38575539047723395}}}, "iid_acc": 0.88, "selectivity": 0.67125, "iid_split_rotation": {"subspace_principal_angle": 1.2768731344200692, "mean_class_cosine": 0.4190057988176132}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.355, "drop": 0.0}, "paraphrase": {"acc": 0.3106060606060606, "drop": 0.044393939393939374, "rotation": {"subspace_principal_angle": 1.1355626344445542, "mean_class_cosine": 0.905045264153325}}}, "iid_acc": 0.355, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.5495618026124416, "mean_class_cosine": 0.9314215323578507}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8833333333333333, "drop": 0.0016666666666667052}}, "iid_acc": 0.885, "selectivity": 0.67625} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.56375, "drop": 0.0}, "paraphrase": {"acc": 0.540962288686606, "drop": 0.022787711313393966, "rotation": {"subspace_principal_angle": 1.4370366006567117, "mean_class_cosine": 0.13336121878297752}}}, "iid_acc": 0.56375, "selectivity": 0.039999999999999925, "iid_split_rotation": {"subspace_principal_angle": 1.49957415048444, "mean_class_cosine": 0.07116197799766466}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5019505851755527, "drop": -0.0019505851755526882, "rotation": {"subspace_principal_angle": 1.5354758515901374, "mean_class_cosine": 0.6588408805283696}}}, "iid_acc": 0.5, "selectivity": -0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1055065787936926, "mean_class_cosine": -0.9020263448542752}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.4837451235370611, "drop": 0.04625487646293891}}, "iid_acc": 0.53, "selectivity": 0.006249999999999978} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6226415094339622, "drop": 0.054858490566037754, "rotation": {"subspace_principal_angle": 1.4132260830128582, "mean_class_cosine": 0.315857679610795}}}, "iid_acc": 0.6775, "selectivity": 0.4175, "iid_split_rotation": {"subspace_principal_angle": 1.41912790604836, "mean_class_cosine": 0.32533634812236173}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.245, "drop": 0.0}, "paraphrase": {"acc": 0.2628032345013477, "drop": -0.017803234501347687, "rotation": {"subspace_principal_angle": 1.2435931646346163, "mean_class_cosine": 0.704084265739651}}}, "iid_acc": 0.245, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.2112395431952154, "mean_class_cosine": 0.6851956806446887}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.6185983827493261, "drop": 0.020151617250673937}}, "iid_acc": 0.63875, "selectivity": 0.37875000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7475083056478405, "drop": -0.01750830564784056, "rotation": {"subspace_principal_angle": 1.4592105747301425, "mean_class_cosine": 0.11135433008645393}}}, "iid_acc": 0.73, "selectivity": 0.23249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.5311242177574582, "mean_class_cosine": 0.03966170335806373}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5830564784053156, "drop": -0.018056478405315657, "rotation": {"subspace_principal_angle": 1.3688794358413638, "mean_class_cosine": 0.9651761136342984}}}, "iid_acc": 0.565, "selectivity": 0.06749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3920800768445374, "mean_class_cosine": 0.9821284088039223}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.7625, "drop": 0.0}, "paraphrase": {"acc": 0.7790697674418605, "drop": -0.01656976744186056}}, "iid_acc": 0.7625, "selectivity": 0.26499999999999996} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6677577741407529, "drop": 0.009742225859247111, "rotation": {"subspace_principal_angle": 1.5176621403016026, "mean_class_cosine": 0.05310918824653445}}}, "iid_acc": 0.6775, "selectivity": 0.11250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.457199891551373, "mean_class_cosine": 0.11335228191779778}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.46125, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.013381751227495908, "rotation": {"subspace_principal_angle": 1.5461043074662548, "mean_class_cosine": -0.13586943898113613}}}, "iid_acc": 0.46125, "selectivity": -0.10374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5005876536533236, "mean_class_cosine": 0.9790471285771307}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7364975450081833, "drop": 0.008502454991816677}}, "iid_acc": 0.745, "selectivity": 0.18000000000000005} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.9333333333333333, "drop": 0.0016666666666667052, "rotation": {"subspace_principal_angle": 1.2444931970939492, "mean_class_cosine": 0.32054342619861337}}}, "iid_acc": 0.935, "selectivity": 0.43750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2405200440265627, "mean_class_cosine": 0.32430439122172083}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.85, "drop": 0.0}, "paraphrase": {"acc": 0.8448979591836735, "drop": 0.005102040816326481, "rotation": {"subspace_principal_angle": 0.8606756929177377, "mean_class_cosine": 0.946865026478741}}}, "iid_acc": 0.85, "selectivity": 0.3525, "iid_split_rotation": {"subspace_principal_angle": 1.1743193451678584, "mean_class_cosine": 0.8825257604902363}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9442176870748299, "drop": 0.0020323129251701433}}, "iid_acc": 0.94625, "selectivity": 0.44875000000000004} +{"model": "gpt2-medium", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": 0.0011346787479407094, "rotation": {"subspace_principal_angle": 1.075164913899876, "mean_class_cosine": 0.4755871801312971}}}, "iid_acc": 0.99125, "selectivity": 0.20875, "iid_split_rotation": {"subspace_principal_angle": 0.9579353285296144, "mean_class_cosine": 0.575210123910989}} +{"model": "gpt2-medium", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6725, "drop": 0.0}, "paraphrase": {"acc": 0.6935749588138386, "drop": -0.021074958813838585, "rotation": {"subspace_principal_angle": 1.2610093649983862, "mean_class_cosine": 0.9991933388093968}}}, "iid_acc": 0.6725, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2635441752944665, "mean_class_cosine": 0.9997669069494608}} +{"model": "gpt2-medium", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.004429571663920884}}, "iid_acc": 0.99125, "selectivity": 0.20875} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.655367231638418, "drop": 0.03088276836158199, "rotation": {"subspace_principal_angle": 1.5026236950761325, "mean_class_cosine": 0.06811983851611561}}}, "iid_acc": 0.68625, "selectivity": 0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4122166058344536, "mean_class_cosine": 0.15791590818006312}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5098870056497176, "drop": -0.011137005649717535, "rotation": {"subspace_principal_angle": 1.2304337374701209, "mean_class_cosine": 0.42959183939934587}}}, "iid_acc": 0.49875, "selectivity": -0.08999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.504970426427495, "mean_class_cosine": -0.8759318900080222}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.7189265536723164, "drop": 0.0010734463276835804}}, "iid_acc": 0.72, "selectivity": 0.13124999999999998} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6449704142011834, "drop": -0.0007396449704141217, "rotation": {"subspace_principal_angle": 1.4713950529119475, "mean_class_cosine": 0.14817274811366463}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2211538461538462, "iid_split_rotation": {"subspace_principal_angle": 1.5076060847864945, "mean_class_cosine": 0.10760657740009509}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5048076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.4319526627218935, "drop": 0.07285502958579881, "rotation": {"subspace_principal_angle": 1.3335837272128281, "mean_class_cosine": 0.760631306748815}}}, "iid_acc": 0.5048076923076923, "selectivity": 0.08173076923076922, "iid_split_rotation": {"subspace_principal_angle": 1.2437180651878326, "mean_class_cosine": 0.6436823456850219}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6568047337278107, "drop": 0.030695266272189325}}, "iid_acc": 0.6875, "selectivity": 0.2644230769230769} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8836206896551724, "drop": 0.010129310344827669, "rotation": {"subspace_principal_angle": 1.3913906147486017, "mean_class_cosine": 0.17844485539062008}}}, "iid_acc": 0.89375, "selectivity": 0.1725000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4139643022148158, "mean_class_cosine": 0.15618990061442609}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.6120689655172413, "drop": 0.014181034482758648, "rotation": {"subspace_principal_angle": 0.7525157984620534, "mean_class_cosine": 0.7829464834163522}}}, "iid_acc": 0.62625, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6467108614757874, "mean_class_cosine": 0.4684027921683622}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9051724137931034, "drop": 0.01732758620689656}}, "iid_acc": 0.9225, "selectivity": 0.20125000000000004} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8264580369843528, "drop": 0.026041963015647207, "rotation": {"subspace_principal_angle": 1.205280976722454, "mean_class_cosine": 0.3574306478032766}}, "domain": {"acc": 0.85, "drop": 0.0025000000000000577, "rotation": {"subspace_principal_angle": 1.1833422855356077, "mean_class_cosine": 0.37783240518966293}}}, "iid_acc": 0.8525, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 1.1308070932804455, "mean_class_cosine": 0.4259297240456055}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5405405405405406, "drop": -0.025540540540540557, "rotation": {"subspace_principal_angle": 1.3060835605730696, "mean_class_cosine": 0.9997239873200385}}, "domain": {"acc": 0.54125, "drop": -0.026249999999999996, "rotation": {"subspace_principal_angle": 1.4432212340779689, "mean_class_cosine": 0.4463453977855254}}}, "iid_acc": 0.515, "selectivity": 0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.4900771909518655, "mean_class_cosine": 0.9997738232242801}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8122332859174964, "drop": 0.026516714082503556}, "domain": {"acc": 0.835, "drop": 0.003750000000000031}}, "iid_acc": 0.83875, "selectivity": 0.33999999999999997} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8333333333333334, "drop": 0.03416666666666668, "rotation": {"subspace_principal_angle": 1.3661248872725396, "mean_class_cosine": 0.20324546817811073}}, "domain": {"acc": 0.56875, "drop": 0.29875000000000007, "rotation": {"subspace_principal_angle": 1.424143642480157, "mean_class_cosine": 0.14612757252528197}}, "length": {"acc": 0.8725925925925926, "drop": -0.0050925925925925375, "rotation": {"subspace_principal_angle": 1.304086875035765, "mean_class_cosine": 0.26355866374052267}}}, "iid_acc": 0.8675, "selectivity": 0.34750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2174020916768917, "mean_class_cosine": 0.34608427795190094}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.7941176470588235, "drop": 0.017132352941176543, "rotation": {"subspace_principal_angle": 1.0543437038158343, "mean_class_cosine": 0.9698412280291595}}, "domain": {"acc": 0.5875, "drop": 0.22375, "rotation": {"subspace_principal_angle": 1.5050071211756375, "mean_class_cosine": 0.07636329311882721}}, "length": {"acc": 0.8177777777777778, "drop": -0.006527777777777799, "rotation": {"subspace_principal_angle": 1.0067642663923582, "mean_class_cosine": 0.9690199508303958}}}, "iid_acc": 0.81125, "selectivity": 0.29125, "iid_split_rotation": {"subspace_principal_angle": 1.3667918849667973, "mean_class_cosine": 0.9283366058714894}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8725490196078431, "drop": 0.004950980392156801}, "domain": {"acc": 0.53125, "drop": 0.34624999999999995}, "length": {"acc": 0.8785185185185185, "drop": -0.0010185185185185297}}, "iid_acc": 0.8775, "selectivity": 0.35749999999999993} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8803030303030303, "drop": 0.01344696969696979, "rotation": {"subspace_principal_angle": 1.3932379682070741, "mean_class_cosine": 0.2875718296573594}}}, "iid_acc": 0.89375, "selectivity": 0.6375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3409039719806688, "mean_class_cosine": 0.32901256562331466}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.5975, "drop": 0.0}, "paraphrase": {"acc": 0.5787878787878787, "drop": 0.01871212121212129, "rotation": {"subspace_principal_angle": 0.6350926877279145, "mean_class_cosine": 0.9549271920995692}}}, "iid_acc": 0.5975, "selectivity": 0.34125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.562730533540774, "mean_class_cosine": 0.945283431787009}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8712121212121212, "drop": 0.017537878787878824}}, "iid_acc": 0.88875, "selectivity": 0.6325000000000001} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.96625, "drop": 0.0}, "paraphrase": {"acc": 0.9586374695863747, "drop": 0.007612530413625307, "rotation": {"subspace_principal_angle": 1.1642922551102963, "mean_class_cosine": 0.6176664792575828}}}, "iid_acc": 0.96625, "selectivity": 0.89625, "iid_split_rotation": {"subspace_principal_angle": 1.0828409682713176, "mean_class_cosine": 0.6471064246781972}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8321167883211679, "drop": 0.006633211678832129, "rotation": {"subspace_principal_angle": 1.5226979113961525, "mean_class_cosine": 0.9230064437182347}}}, "iid_acc": 0.83875, "selectivity": 0.76875, "iid_split_rotation": {"subspace_principal_angle": 1.319128362171639, "mean_class_cosine": 0.9669627199357128}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.9625, "drop": 0.0}, "paraphrase": {"acc": 0.9464720194647201, "drop": 0.016027980535279873}}, "iid_acc": 0.9625, "selectivity": 0.8925000000000001} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.635890767230169, "drop": -0.009640767230169045, "rotation": {"subspace_principal_angle": 1.2265731114344571, "mean_class_cosine": 0.3374655623320725}}}, "iid_acc": 0.62625, "selectivity": 0.14125, "iid_split_rotation": {"subspace_principal_angle": 1.3292992619939612, "mean_class_cosine": 0.239156515406049}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.48875, "drop": 0.0}, "paraphrase": {"acc": 0.5084525357607282, "drop": -0.019702535760728224, "rotation": {"subspace_principal_angle": 1.5180213673447036, "mean_class_cosine": 0.7337434200066699}}}, "iid_acc": 0.48875, "selectivity": 0.003750000000000031, "iid_split_rotation": {"subspace_principal_angle": 0.9576309376243967, "mean_class_cosine": -0.9826509562242909}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5149544863459038, "drop": -0.002454486345903839}}, "iid_acc": 0.5125, "selectivity": 0.02749999999999997} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.62125, "drop": 0.0}, "paraphrase": {"acc": 0.5700808625336927, "drop": 0.051169137466307246, "rotation": {"subspace_principal_angle": 1.3744117706906038, "mean_class_cosine": 0.4523473567422777}}}, "iid_acc": 0.62125, "selectivity": 0.33499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3212695884395511, "mean_class_cosine": 0.4413545946757409}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.11875, "drop": 0.0}, "paraphrase": {"acc": 0.11455525606469003, "drop": 0.004194743935309969, "rotation": {"subspace_principal_angle": 1.3057209282680173, "mean_class_cosine": 0.9958332765626513}}}, "iid_acc": 0.11875, "selectivity": -0.1675, "iid_split_rotation": {"subspace_principal_angle": 1.5288056782240633, "mean_class_cosine": 0.6645736917680091}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5619946091644205, "drop": -0.006994609164420407}}, "iid_acc": 0.555, "selectivity": 0.26875000000000004} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7275747508305648, "drop": 0.007425249169435211, "rotation": {"subspace_principal_angle": 1.4274592750130581, "mean_class_cosine": 0.14284673359650063}}}, "iid_acc": 0.735, "selectivity": 0.21250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4743386316433527, "mean_class_cosine": 0.09630818958862122}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5631229235880398, "drop": -0.003122923588039783, "rotation": {"subspace_principal_angle": 1.2083317440135473, "mean_class_cosine": 0.995078207599845}}}, "iid_acc": 0.56, "selectivity": 0.03750000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.2327106774088032, "mean_class_cosine": 0.9974410369073173}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7574750830564784, "drop": -0.0012250830564783932}}, "iid_acc": 0.75625, "selectivity": 0.23375} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.648424543946932, "drop": 0.04407545605306795, "rotation": {"subspace_principal_angle": 1.4115718181974544, "mean_class_cosine": 0.15855257251441335}}}, "iid_acc": 0.6925, "selectivity": 0.20625, "iid_split_rotation": {"subspace_principal_angle": 1.3640799568494573, "mean_class_cosine": 0.20524729008148926}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5223880597014925, "drop": 0.001361940298507558, "rotation": {"subspace_principal_angle": 1.3460720235249697, "mean_class_cosine": 0.9994305298973165}}}, "iid_acc": 0.52375, "selectivity": 0.03750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.238382906553322, "mean_class_cosine": 0.999840093834897}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6401326699834162, "drop": -0.005132669983416216}}, "iid_acc": 0.635, "selectivity": 0.14875} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.6988543371522095, "drop": 0.018645662847790567, "rotation": {"subspace_principal_angle": 1.4345047521354637, "mean_class_cosine": 0.13587002144149035}}}, "iid_acc": 0.7175, "selectivity": 0.14875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.448693729965042, "mean_class_cosine": 0.12179941742267139}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.4675, "drop": 0.0}, "paraphrase": {"acc": 0.4779050736497545, "drop": -0.01040507364975446, "rotation": {"subspace_principal_angle": 1.4886165492494845, "mean_class_cosine": -0.5750441624077374}}}, "iid_acc": 0.4675, "selectivity": -0.10124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4864385355642995, "mean_class_cosine": 0.995645299950886}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7217675941080196, "drop": 0.003232405891980372}}, "iid_acc": 0.725, "selectivity": 0.15625} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9414965986394558, "drop": 0.0010034013605442205, "rotation": {"subspace_principal_angle": 1.0554320916574222, "mean_class_cosine": 0.49285180754683094}}}, "iid_acc": 0.9425, "selectivity": 0.50875, "iid_split_rotation": {"subspace_principal_angle": 1.0718759182874942, "mean_class_cosine": 0.4784778287309053}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5360544217687074, "drop": -0.0023044217687074875, "rotation": {"subspace_principal_angle": 0.6230061876035811, "mean_class_cosine": 0.8605846805322723}}}, "iid_acc": 0.53375, "selectivity": 0.09999999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.7433840996009756, "mean_class_cosine": 0.3607772763702933}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.9401360544217687, "drop": -0.007636054421768734}}, "iid_acc": 0.9325, "selectivity": 0.49874999999999997} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": 0.003634678747940767, "rotation": {"subspace_principal_angle": 1.0513045046339702, "mean_class_cosine": 0.49643906722894926}}}, "iid_acc": 0.99375, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9685923811624354, "mean_class_cosine": 0.5664600954062968}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.685337726523888, "drop": -0.01033772652388798, "rotation": {"subspace_principal_angle": 1.3240904740257078, "mean_class_cosine": 0.999937048647147}}}, "iid_acc": 0.675, "selectivity": -0.17374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5089394180470712, "mean_class_cosine": 0.9999798272733277}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.0015321252059308232}}, "iid_acc": 0.99, "selectivity": 0.14125} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7259887005649718, "drop": 0.005261299435028177, "rotation": {"subspace_principal_angle": 1.48146441422371, "mean_class_cosine": 0.08921314568865987}}}, "iid_acc": 0.73125, "selectivity": 0.1087499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4204367412204741, "mean_class_cosine": 0.1497936706206605}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5141242937853108, "drop": 0.018375706214689203, "rotation": {"subspace_principal_angle": 1.4237156435330802, "mean_class_cosine": -0.529348589575781}}}, "iid_acc": 0.5325, "selectivity": -0.09000000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.0016401557930925, "mean_class_cosine": -0.9687547350705455}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7245762711864406, "drop": -0.005826271186440635}}, "iid_acc": 0.71875, "selectivity": 0.09624999999999995} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6627218934911243, "drop": 0.010355029585798814, "rotation": {"subspace_principal_angle": 1.3926151506687268, "mean_class_cosine": 0.16905691896718814}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.21153846153846156, "iid_split_rotation": {"subspace_principal_angle": 1.455424707460169, "mean_class_cosine": 0.13141358197406708}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.5096153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.4378698224852071, "drop": 0.07174556213017746, "rotation": {"subspace_principal_angle": 1.2579697526624707, "mean_class_cosine": 0.8469173188143175}}}, "iid_acc": 0.5096153846153846, "selectivity": 0.04807692307692302, "iid_split_rotation": {"subspace_principal_angle": 1.545165675478419, "mean_class_cosine": 0.5995767861930784}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.024038461538461453}}, "iid_acc": 0.6394230769230769, "selectivity": 0.1778846153846153} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8922413793103449, "drop": 0.009008620689655134, "rotation": {"subspace_principal_angle": 1.2161263841678638, "mean_class_cosine": 0.34728086917165485}}}, "iid_acc": 0.90125, "selectivity": 0.11250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.242802856120801, "mean_class_cosine": 0.3221441156385047}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5975, "drop": 0.0}, "paraphrase": {"acc": 0.5747126436781609, "drop": 0.022787356321839147, "rotation": {"subspace_principal_angle": 1.0414849280010454, "mean_class_cosine": 0.8882768003569734}}}, "iid_acc": 0.5975, "selectivity": -0.19124999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.075377185201994, "mean_class_cosine": 0.00826524737963924}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8836206896551724, "drop": 0.01762931034482762}}, "iid_acc": 0.90125, "selectivity": 0.11250000000000004} diff --git a/repro_bundle/results_A/results/predictors.jsonl b/repro_bundle/results_A/results/predictors.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..bc5227abf3302ef9d7c185585f63b50ee47870b3 --- /dev/null +++ b/repro_bundle/results_A/results/predictors.jsonl @@ -0,0 +1,472 @@ +{"model": "pythia-70m", "dataset": "sst2", "seed": 0, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921994833, "raptor_stability": 0.8375464081764221, "fragility": 0.25, "augmentation_robustness": 0.6545744286726296, "xie_feature_dispersion": 0.8227753483650888, "pac": 0.7460604184245259, "whitened_cosine_id": 0.8294597268104553}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 0, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998517, "raptor_stability": 0.9239972829818726, "fragility": 1.0, "augmentation_robustness": 0.9364739773993105, "xie_feature_dispersion": 1.0993059236404399, "pac": 0.9302356301905915, "whitened_cosine_id": 0.8396446704864502}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 0, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6360429269373853, "raptor_stability": 0.8923418521881104, "fragility": 1.5, "augmentation_robustness": 0.9079627357769793, "xie_feature_dispersion": 3.4534575785421744, "pac": 0.9001522939825448, "whitened_cosine_id": 0.9029156565666199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 0, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14189668585488155, "raptor_stability": 0.9457845091819763, "fragility": 1.5, "augmentation_robustness": 0.8839691673684479, "xie_feature_dispersion": 5.2630010355712145, "pac": 0.9148768382752122, "whitened_cosine_id": 0.9317190051078796}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 0, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974287, "raptor_stability": 0.7641952633857727, "fragility": 1.0, "augmentation_robustness": 0.7320095613414648, "xie_feature_dispersion": 0.4779918901319739, "pac": 0.7481024123636187, "whitened_cosine_id": 0.6775354743003845}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 0, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5953587445214507, "raptor_stability": 0.8714625239372253, "fragility": 0.5, "augmentation_robustness": 0.8065598522498184, "xie_feature_dispersion": 1.5333196894740788, "pac": 0.8390111880935218, "whitened_cosine_id": 0.8480013012886047}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 0, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994247, "raptor_stability": 0.8230767250061035, "fragility": 0.5, "augmentation_robustness": 0.7221715232065848, "xie_feature_dispersion": 1.779944270255231, "pac": 0.7726241241063442, "whitened_cosine_id": 0.8223811388015747}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 0, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.711632992199055, "raptor_stability": 0.8190855383872986, "fragility": 0.25, "augmentation_robustness": 0.676819096553469, "xie_feature_dispersion": 2.512801472259043, "pac": 0.7479523174703837, "whitened_cosine_id": 0.7729045748710632}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989692, "raptor_stability": 0.8473518490791321, "fragility": 1.0, "augmentation_robustness": 0.8442697431706384, "xie_feature_dispersion": 1.163949469847951, "pac": 0.8458107961248853, "whitened_cosine_id": 0.7645858526229858}} +{"model": "pythia-70m", "dataset": "subj", "seed": 0, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002171, "raptor_stability": 0.9112874269485474, "fragility": 1.5, "augmentation_robustness": 0.8646995541446441, "xie_feature_dispersion": 2.3756762827181857, "pac": 0.8879934905465957, "whitened_cosine_id": 0.9034857153892517}} +{"model": "pythia-70m", "dataset": "spam", "seed": 0, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004674, "raptor_stability": 0.9211491942405701, "fragility": 1.5, "augmentation_robustness": 0.884000762861815, "xie_feature_dispersion": 10.408354824969607, "pac": 0.9025749785511925, "whitened_cosine_id": 0.9505216479301453}} +{"model": "pythia-70m", "dataset": "cola", "seed": 0, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921980065, "raptor_stability": 0.7842410206794739, "fragility": 1.0, "augmentation_robustness": 0.751228525306698, "xie_feature_dispersion": 0.5820240387291664, "pac": 0.7677347729930859, "whitened_cosine_id": 0.7026472687721252}} +{"model": "pythia-70m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.7516991834926783, "raptor_stability": 0.8351565003395081, "fragility": 1.0, "augmentation_robustness": 0.8460384036839802, "xie_feature_dispersion": 1.9650544929284912, "pac": 0.8405974520117441, "whitened_cosine_id": 0.8273500800132751}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 0, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998, "raptor_stability": 0.8865668177604675, "fragility": 1.5, "augmentation_robustness": 0.8695133740159424, "xie_feature_dispersion": 1.822712974246315, "pac": 0.878040095888205, "whitened_cosine_id": 0.8655645847320557}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350087, "raptor_stability": 0.8526354432106018, "fragility": 0.25, "augmentation_robustness": 0.6188354080698831, "xie_feature_dispersion": 1.4101369914642357, "pac": 0.7357354256402424, "whitened_cosine_id": 0.8668140769004822}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 0, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350252, "raptor_stability": 0.8803290128707886, "fragility": 1.0, "augmentation_robustness": 0.8724963955930902, "xie_feature_dispersion": 1.3391749548314755, "pac": 0.8764127042319394, "whitened_cosine_id": 0.8713250756263733}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3953680487548481, "raptor_stability": 0.8455236554145813, "fragility": 0.25, "augmentation_robustness": 0.8483093395593638, "xie_feature_dispersion": 0.9318287616318928, "pac": 0.8469164974869725, "whitened_cosine_id": 0.8953917026519775}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 0, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10457984207175601, "raptor_stability": 0.9457212686538696, "fragility": 1.5, "augmentation_robustness": 0.8857764499429269, "xie_feature_dispersion": 5.199477497190765, "pac": 0.9157488592983982, "whitened_cosine_id": 0.9181342720985413}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 0, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342227, "raptor_stability": 0.7698441743850708, "fragility": 0.25, "augmentation_robustness": 0.6816650768598687, "xie_feature_dispersion": 0.25779876141354974, "pac": 0.7257546256224697, "whitened_cosine_id": 0.7203242182731628}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.574379658862361, "raptor_stability": 0.8597204685211182, "fragility": 0.25, "augmentation_robustness": 0.7682780634518579, "xie_feature_dispersion": 1.6192479136149953, "pac": 0.8139992659864881, "whitened_cosine_id": 0.8496574759483337}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348304, "raptor_stability": 0.8552483320236206, "fragility": 1.0, "augmentation_robustness": 0.7851229050246351, "xie_feature_dispersion": 2.2631913339622987, "pac": 0.8201856185241279, "whitened_cosine_id": 0.8250994682312012}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 0, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934627, "raptor_stability": 0.8244509100914001, "fragility": 0.5, "augmentation_robustness": 0.7136331067294028, "xie_feature_dispersion": 1.9587240977086187, "pac": 0.7690420084104015, "whitened_cosine_id": 0.7910383939743042}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934696, "raptor_stability": 0.8365420699119568, "fragility": 1.0, "augmentation_robustness": 0.7890852919390294, "xie_feature_dispersion": 1.5342572755652866, "pac": 0.8128136809254931, "whitened_cosine_id": 0.8055051565170288}} +{"model": "pythia-160m", "dataset": "subj", "seed": 0, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351844, "raptor_stability": 0.8689209222793579, "fragility": 1.0, "augmentation_robustness": 0.7746642653183333, "xie_feature_dispersion": 1.77134285145721, "pac": 0.8217925937988456, "whitened_cosine_id": 0.9039409160614014}} +{"model": "pythia-160m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352976, "raptor_stability": 0.943769633769989, "fragility": 3.0, "augmentation_robustness": 0.9172945300081299, "xie_feature_dispersion": 6.963799146343127, "pac": 0.9305320818890594, "whitened_cosine_id": 0.9455703496932983}} +{"model": "pythia-160m", "dataset": "cola", "seed": 0, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344465, "raptor_stability": 0.7988646626472473, "fragility": 0.25, "augmentation_robustness": 0.5972243911972416, "xie_feature_dispersion": 0.7758972276639047, "pac": 0.6980445269222444, "whitened_cosine_id": 0.7851360440254211}} +{"model": "pythia-160m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.4494122497402506, "raptor_stability": 0.8332012295722961, "fragility": 1.5, "augmentation_robustness": 0.8344763697080788, "xie_feature_dispersion": 2.5953837186456754, "pac": 0.8338387996401875, "whitened_cosine_id": 0.8253653049468994}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 0, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935011, "raptor_stability": 0.866767942905426, "fragility": 0.5, "augmentation_robustness": 0.7943799335400991, "xie_feature_dispersion": 1.5213101554825748, "pac": 0.8305739382227626, "whitened_cosine_id": 0.8781046867370605}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 0, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881098, "raptor_stability": 0.8615295886993408, "fragility": 0.25, "augmentation_robustness": 0.64759763388926, "xie_feature_dispersion": 1.6884023301086555, "pac": 0.7545636112943004, "whitened_cosine_id": 0.8637628555297852}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881813, "raptor_stability": 0.8620725274085999, "fragility": 1.0, "augmentation_robustness": 0.8437283807712266, "xie_feature_dispersion": 1.923545301649778, "pac": 0.8529004540899132, "whitened_cosine_id": 0.8767724633216858}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.32280286192962354, "raptor_stability": 0.8491001725196838, "fragility": 1.0, "augmentation_robustness": 0.8717545371887874, "xie_feature_dispersion": 6.130394237589204, "pac": 0.8604273548542356, "whitened_cosine_id": 0.8844512701034546}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.1116614706117857, "raptor_stability": 0.945551872253418, "fragility": 2.0, "augmentation_robustness": 0.8856969264519378, "xie_feature_dispersion": 7.687914979274747, "pac": 0.9156243993526778, "whitened_cosine_id": 0.8968205451965332}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 0, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879029, "raptor_stability": 0.787240743637085, "fragility": 0.25, "augmentation_robustness": 0.7337481325439938, "xie_feature_dispersion": 0.12206039095754939, "pac": 0.7604944380905394, "whitened_cosine_id": 0.7804538607597351}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 0, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.4882081459032719, "raptor_stability": 0.8632135391235352, "fragility": 0.25, "augmentation_robustness": 0.7884647975902315, "xie_feature_dispersion": 2.173352441480384, "pac": 0.8258391683568833, "whitened_cosine_id": 0.8471425175666809}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 0, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880458, "raptor_stability": 0.8126577138900757, "fragility": 0.25, "augmentation_robustness": 0.7034424602081344, "xie_feature_dispersion": 4.485569514407045, "pac": 0.758050087049105, "whitened_cosine_id": 0.827558696269989}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 0, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879928, "raptor_stability": 0.7987161874771118, "fragility": 0.25, "augmentation_robustness": 0.6753132678040145, "xie_feature_dispersion": 2.5329861784126293, "pac": 0.7370147276405632, "whitened_cosine_id": 0.810127854347229}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 0, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880347, "raptor_stability": 0.814401388168335, "fragility": 0.25, "augmentation_robustness": 0.6923253341445008, "xie_feature_dispersion": 0.7524282237703519, "pac": 0.7533633611564179, "whitened_cosine_id": 0.8191361427307129}} +{"model": "pythia-410m", "dataset": "subj", "seed": 0, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882566, "raptor_stability": 0.8862898945808411, "fragility": 0.5, "augmentation_robustness": 0.7988711466010014, "xie_feature_dispersion": 1.428972872571398, "pac": 0.8425805205909213, "whitened_cosine_id": 0.8959206342697144}} +{"model": "pythia-410m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882899, "raptor_stability": 0.9429970979690552, "fragility": 3.0, "augmentation_robustness": 0.9052157104985206, "xie_feature_dispersion": 6.72893787243834, "pac": 0.9241064042337879, "whitened_cosine_id": 0.933451235294342}} +{"model": "pythia-410m", "dataset": "cola", "seed": 0, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879473, "raptor_stability": 0.7842450737953186, "fragility": 0.25, "augmentation_robustness": 0.5039294079390225, "xie_feature_dispersion": 1.1149557488538653, "pac": 0.6440872408671705, "whitened_cosine_id": 0.8007057905197144}} +{"model": "pythia-410m", "dataset": "stance", "seed": 0, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.5795818589574081, "raptor_stability": 0.8285464644432068, "fragility": 1.0, "augmentation_robustness": 0.8175707678986336, "xie_feature_dispersion": 1.8361604596405292, "pac": 0.8230586161709201, "whitened_cosine_id": 0.8221842646598816}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 0, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881513, "raptor_stability": 0.8521958589553833, "fragility": 0.5, "augmentation_robustness": 0.7625019501536046, "xie_feature_dispersion": 1.5630966219328584, "pac": 0.8073489045544939, "whitened_cosine_id": 0.8685176968574524}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.8599012494087219, "fragility": 0.25, "augmentation_robustness": 0.5848209179858852, "xie_feature_dispersion": 2.122662038410848, "pac": 0.7223610836973036, "whitened_cosine_id": 0.8232980370521545}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.825421154499054, "fragility": 1.0, "augmentation_robustness": 0.7804717428621477, "xie_feature_dispersion": 2.576726583434504, "pac": 0.8029464486806008, "whitened_cosine_id": 0.8465724587440491}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 0, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22041139949027433, "raptor_stability": 0.8484893441200256, "fragility": 1.0, "augmentation_robustness": 0.8650683659984831, "xie_feature_dispersion": 5.917035974988243, "pac": 0.8567788550592543, "whitened_cosine_id": 0.8215498328208923}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06228035557873579, "raptor_stability": 0.9404226541519165, "fragility": 3.0, "augmentation_robustness": 0.8779983715906402, "xie_feature_dispersion": 9.493553367505593, "pac": 0.9092105128712784, "whitened_cosine_id": 0.8300829529762268}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 0, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.7848396301269531, "fragility": 0.25, "augmentation_robustness": 0.7408315387590054, "xie_feature_dispersion": 0.2653617366201949, "pac": 0.7628355844429793, "whitened_cosine_id": 0.817557156085968}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.019767617442798265, "raptor_stability": 0.8671453595161438, "fragility": 0.25, "augmentation_robustness": 0.755285317721517, "xie_feature_dispersion": 2.8877683311134636, "pac": 0.8112153386188303, "whitened_cosine_id": 0.830573558807373}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 0, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010617, "raptor_stability": 0.8155287504196167, "fragility": 0.25, "augmentation_robustness": 0.6718229230312025, "xie_feature_dispersion": 5.342310320027777, "pac": 0.7436758367254096, "whitened_cosine_id": 0.8247697353363037}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 0, "layer": 22, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.7958077192306519, "fragility": 0.25, "augmentation_robustness": 0.6470101589682584, "xie_feature_dispersion": 2.217429056075072, "pac": 0.7214089390994551, "whitened_cosine_id": 0.8194364309310913}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 0, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8089179396629333, "fragility": 0.5, "augmentation_robustness": 0.7119613985713426, "xie_feature_dispersion": 1.5482508686542642, "pac": 0.760439669117138, "whitened_cosine_id": 0.8221121430397034}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 0, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010872, "raptor_stability": 0.8672937750816345, "fragility": 1.0, "augmentation_robustness": 0.7720944652439713, "xie_feature_dispersion": 2.5476390335776844, "pac": 0.819694120162803, "whitened_cosine_id": 0.8334078788757324}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 0, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010434, "raptor_stability": 0.9216615557670593, "fragility": 1.0, "augmentation_robustness": 0.895838461689926, "xie_feature_dispersion": 23.432722674772293, "pac": 0.9087500087284927, "whitened_cosine_id": 0.8588989973068237}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 0, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010818, "raptor_stability": 0.8071193099021912, "fragility": 0.25, "augmentation_robustness": 0.5249275775049415, "xie_feature_dispersion": 1.549092757263006, "pac": 0.6660234437035664, "whitened_cosine_id": 0.8179035186767578}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 0, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.36672114356076757, "raptor_stability": 0.8304018974304199, "fragility": 1.0, "augmentation_robustness": 0.8219455087263939, "xie_feature_dispersion": 3.593233328751099, "pac": 0.8261737030784069, "whitened_cosine_id": 0.8171507716178894}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8476687073707581, "fragility": 0.5, "augmentation_robustness": 0.7513937676225657, "xie_feature_dispersion": 2.045523026748139, "pac": 0.7995312374966619, "whitened_cosine_id": 0.8353421092033386}} +{"model": "gpt2", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349679, "raptor_stability": 0.8233915567398071, "fragility": 0.25, "augmentation_robustness": 0.5123065291062753, "xie_feature_dispersion": 1.6815976407755642, "pac": 0.6678490429230413, "whitened_cosine_id": 0.857738196849823}} +{"model": "gpt2", "dataset": "imdb", "seed": 0, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350731, "raptor_stability": 0.7829911708831787, "fragility": 0.25, "augmentation_robustness": 0.6517344633023527, "xie_feature_dispersion": 1.6579993440971876, "pac": 0.7173628170927657, "whitened_cosine_id": 0.8754125237464905}} +{"model": "gpt2", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.34635368740834577, "raptor_stability": 0.8845524787902832, "fragility": 0.25, "augmentation_robustness": 0.9187097662639041, "xie_feature_dispersion": 0.8678892394256513, "pac": 0.9016311225270937, "whitened_cosine_id": 0.910406231880188}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 0, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.12115672881499416, "raptor_stability": 0.9302389025688171, "fragility": 1.5, "augmentation_robustness": 0.8321901130650252, "xie_feature_dispersion": 6.121815304004312, "pac": 0.8812145078169211, "whitened_cosine_id": 0.9043624997138977}} +{"model": "gpt2", "dataset": "counterfact", "seed": 0, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859340033, "raptor_stability": 0.7690413594245911, "fragility": 0.25, "augmentation_robustness": 0.809669912657523, "xie_feature_dispersion": 0.1095429604074235, "pac": 0.789355636041057, "whitened_cosine_id": 0.7072374820709229}} +{"model": "gpt2", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.7366388390202255, "raptor_stability": 0.8725014328956604, "fragility": 0.25, "augmentation_robustness": 0.8145765018284364, "xie_feature_dispersion": 1.77009486868241, "pac": 0.8435389673620484, "whitened_cosine_id": 0.8444538116455078}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347128, "raptor_stability": 0.7693442106246948, "fragility": 0.25, "augmentation_robustness": 0.551602767544549, "xie_feature_dispersion": 3.694561283935865, "pac": 0.6604734890846219, "whitened_cosine_id": 0.8187122344970703}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 0, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346073, "raptor_stability": 0.7867448925971985, "fragility": 0.25, "augmentation_robustness": 0.6861444856119838, "xie_feature_dispersion": 2.309585917543921, "pac": 0.7364446891045912, "whitened_cosine_id": 0.8086420297622681}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 0, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346306, "raptor_stability": 0.7805789113044739, "fragility": 0.25, "augmentation_robustness": 0.657242871403201, "xie_feature_dispersion": 0.25866709333672444, "pac": 0.7189108913538375, "whitened_cosine_id": 0.803403377532959}} +{"model": "gpt2", "dataset": "subj", "seed": 0, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352106, "raptor_stability": 0.8574187755584717, "fragility": 0.25, "augmentation_robustness": 0.7295931094481585, "xie_feature_dispersion": 0.951143558340801, "pac": 0.7935059425033151, "whitened_cosine_id": 0.9090107679367065}} +{"model": "gpt2", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593527, "raptor_stability": 0.9161890149116516, "fragility": 2.0, "augmentation_robustness": 0.9103539019844458, "xie_feature_dispersion": 6.497801646724056, "pac": 0.9132714584480487, "whitened_cosine_id": 0.9386647939682007}} +{"model": "gpt2", "dataset": "cola", "seed": 0, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859346528, "raptor_stability": 0.7938734889030457, "fragility": 0.25, "augmentation_robustness": 0.49778314433790527, "xie_feature_dispersion": 0.9649847502240455, "pac": 0.6458283166204755, "whitened_cosine_id": 0.807263970375061}} +{"model": "gpt2", "dataset": "stance", "seed": 0, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.490010700608385, "raptor_stability": 0.8106656670570374, "fragility": 0.25, "augmentation_robustness": 0.7954562938419368, "xie_feature_dispersion": 2.664520632491477, "pac": 0.803060980449487, "whitened_cosine_id": 0.8153484463691711}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935004, "raptor_stability": 0.8306007981300354, "fragility": 0.25, "augmentation_robustness": 0.6972854536891078, "xie_feature_dispersion": 1.1399162300669312, "pac": 0.7639431259095716, "whitened_cosine_id": 0.8771829605102539}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 0, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688191, "raptor_stability": 0.8413395881652832, "fragility": 0.25, "augmentation_robustness": 0.5985845188957462, "xie_feature_dispersion": 1.7719582872902115, "pac": 0.7199620535305147, "whitened_cosine_id": 0.8595471382141113}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 0, "layer": 3, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878758, "raptor_stability": 0.7550429105758667, "fragility": 0.25, "augmentation_robustness": 0.8450710203949588, "xie_feature_dispersion": 0.2304978644612739, "pac": 0.8000569654854128, "whitened_cosine_id": 0.7729927897453308}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 0, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6820984754002586, "raptor_stability": 0.8745980262756348, "fragility": 0.25, "augmentation_robustness": 0.7780721034137521, "xie_feature_dispersion": 0.8621445356293869, "pac": 0.8263350648446934, "whitened_cosine_id": 0.8492467999458313}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880176, "raptor_stability": 0.8115680813789368, "fragility": 0.25, "augmentation_robustness": 0.6729005392577236, "xie_feature_dispersion": 1.137989115163849, "pac": 0.7422343103183302, "whitened_cosine_id": 0.8243570923805237}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 0, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878862, "raptor_stability": 0.777171790599823, "fragility": 0.25, "augmentation_robustness": 0.6061778394389484, "xie_feature_dispersion": 2.1506698277748155, "pac": 0.6916748150193857, "whitened_cosine_id": 0.8085216283798218}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 0, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880017, "raptor_stability": 0.7901369333267212, "fragility": 0.25, "augmentation_robustness": 0.6555666038259415, "xie_feature_dispersion": 0.462184243542827, "pac": 0.7228517685763314, "whitened_cosine_id": 0.8134031891822815}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 0, "layer": 24, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688252, "raptor_stability": 0.8832855224609375, "fragility": 0.25, "augmentation_robustness": 0.7951101769713113, "xie_feature_dispersion": 0.4906864607581503, "pac": 0.8391978497161243, "whitened_cosine_id": 0.8952683210372925}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 0, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880159, "raptor_stability": 0.7916848063468933, "fragility": 0.25, "augmentation_robustness": 0.5224171661394365, "xie_feature_dispersion": 0.79552886696272, "pac": 0.6570509862431648, "whitened_cosine_id": 0.8178503513336182}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 0, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.42249695846338925, "raptor_stability": 0.8128862977027893, "fragility": 0.5, "augmentation_robustness": 0.8037561319272016, "xie_feature_dispersion": 2.2610696467251814, "pac": 0.8083212148149954, "whitened_cosine_id": 0.7861542701721191}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 0, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881549, "raptor_stability": 0.8164692521095276, "fragility": 0.25, "augmentation_robustness": 0.6904412835869388, "xie_feature_dispersion": 1.1005465582305016, "pac": 0.7534552678482331, "whitened_cosine_id": 0.8596013784408569}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648414, "raptor_stability": 0.8818655014038086, "fragility": 0.25, "augmentation_robustness": 0.67470015678154, "xie_feature_dispersion": 1.8204152059660144, "pac": 0.7782828290926743, "whitened_cosine_id": 0.879472017288208}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 0, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.29387292376502, "raptor_stability": 0.8775492906570435, "fragility": 1.0, "augmentation_robustness": 0.8720676344266011, "xie_feature_dispersion": 2.218615090585379, "pac": 0.8748084625418222, "whitened_cosine_id": 0.8882995247840881}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.33450754332538074, "raptor_stability": 0.8566421270370483, "fragility": 1.0, "augmentation_robustness": 0.8744251513008834, "xie_feature_dispersion": 3.7561628051141023, "pac": 0.8655336391689659, "whitened_cosine_id": 0.9043422937393188}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07399960414320991, "raptor_stability": 0.9483939409255981, "fragility": 1.5, "augmentation_robustness": 0.891918719098521, "xie_feature_dispersion": 7.327642739764713, "pac": 0.9201563300120597, "whitened_cosine_id": 0.9209960699081421}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 0, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237645932, "raptor_stability": 0.8066194653511047, "fragility": 0.25, "augmentation_robustness": 0.8373520059306747, "xie_feature_dispersion": 0.06081582657176685, "pac": 0.8219857356408897, "whitened_cosine_id": 0.7630186080932617}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6131208141710327, "raptor_stability": 0.8868826031684875, "fragility": 0.25, "augmentation_robustness": 0.8904501661922035, "xie_feature_dispersion": 1.992273238599221, "pac": 0.8886663846803455, "whitened_cosine_id": 0.8511517643928528}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645996, "raptor_stability": 0.8185003995895386, "fragility": 0.25, "augmentation_robustness": 0.7607385932812474, "xie_feature_dispersion": 4.466765724760196, "pac": 0.789619496435393, "whitened_cosine_id": 0.8152397871017456}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 0, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647149, "raptor_stability": 0.811665415763855, "fragility": 0.25, "augmentation_robustness": 0.7276199921424138, "xie_feature_dispersion": 2.8843177774025395, "pac": 0.7696427039531344, "whitened_cosine_id": 0.8048377633094788}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 0, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647015, "raptor_stability": 0.8171485662460327, "fragility": 0.25, "augmentation_robustness": 0.7939897357140582, "xie_feature_dispersion": 0.15381271196675575, "pac": 0.8055691509800454, "whitened_cosine_id": 0.8043486475944519}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 0, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651123, "raptor_stability": 0.8921712040901184, "fragility": 0.5, "augmentation_robustness": 0.8243175866171568, "xie_feature_dispersion": 0.9955751494789675, "pac": 0.8582443953536376, "whitened_cosine_id": 0.9154570698738098}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 0, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923765166, "raptor_stability": 0.9529117941856384, "fragility": 3.0, "augmentation_robustness": 0.94099314226979, "xie_feature_dispersion": 9.854132791030557, "pac": 0.9469524682277142, "whitened_cosine_id": 0.9523190855979919}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 0, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764728, "raptor_stability": 0.8035885691642761, "fragility": 0.25, "augmentation_robustness": 0.5589292088936479, "xie_feature_dispersion": 1.0666210514085848, "pac": 0.681258889028962, "whitened_cosine_id": 0.818631649017334}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 0, "layer": 20, "concept": "stance", "predictors": {"sip_eigengap": 0.4157006657150525, "raptor_stability": 0.8240557312965393, "fragility": 0.5, "augmentation_robustness": 0.8015463605814978, "xie_feature_dispersion": 3.234317390572246, "pac": 0.8128010459390185, "whitened_cosine_id": 0.8179960250854492}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 0, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649436, "raptor_stability": 0.8530629873275757, "fragility": 0.25, "augmentation_robustness": 0.7821972078583704, "xie_feature_dispersion": 1.1742203784808831, "pac": 0.8176300975929731, "whitened_cosine_id": 0.8717560768127441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921995341, "raptor_stability": 0.8336094617843628, "fragility": 0.25, "augmentation_robustness": 0.6377618570084167, "xie_feature_dispersion": 1.7265608037193576, "pac": 0.7356856593963897, "whitened_cosine_id": 0.8218536972999573}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998223, "raptor_stability": 0.8874242901802063, "fragility": 1.0, "augmentation_robustness": 0.8992101261321568, "xie_feature_dispersion": 1.271550857614664, "pac": 0.8933172081561815, "whitened_cosine_id": 0.8205482363700867}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6806422452118264, "raptor_stability": 0.8989484906196594, "fragility": 1.5, "augmentation_robustness": 0.9168698848955824, "xie_feature_dispersion": 3.4157347611498774, "pac": 0.9079091877576209, "whitened_cosine_id": 0.9129272699356079}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.18807531413389164, "raptor_stability": 0.9559001326560974, "fragility": 2.0, "augmentation_robustness": 0.9200740132537277, "xie_feature_dispersion": 5.534601025030786, "pac": 0.9379870729549126, "whitened_cosine_id": 0.933467447757721}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 1, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.777370810508728, "fragility": 0.5, "augmentation_robustness": 0.7673906542908651, "xie_feature_dispersion": 0.6104756443427384, "pac": 0.7723807323997965, "whitened_cosine_id": 0.6849023103713989}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 1, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.5332073250620768, "raptor_stability": 0.8536012172698975, "fragility": 0.25, "augmentation_robustness": 0.7542190503987464, "xie_feature_dispersion": 1.7320048864765314, "pac": 0.803910133834322, "whitened_cosine_id": 0.8371078372001648}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 1, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994555, "raptor_stability": 0.8290720582008362, "fragility": 0.5, "augmentation_robustness": 0.7840608903875346, "xie_feature_dispersion": 2.6910245205930696, "pac": 0.8065664742941854, "whitened_cosine_id": 0.8217921853065491}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 1, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987687, "raptor_stability": 0.819472074508667, "fragility": 1.0, "augmentation_robustness": 0.698958315186553, "xie_feature_dispersion": 2.073055712645641, "pac": 0.75921519484761, "whitened_cosine_id": 0.7341217398643494}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 1, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989486, "raptor_stability": 0.7951779961585999, "fragility": 0.5, "augmentation_robustness": 0.7589872979724266, "xie_feature_dispersion": 1.2637336763777984, "pac": 0.7770826470655132, "whitened_cosine_id": 0.7604337334632874}} +{"model": "pythia-70m", "dataset": "subj", "seed": 1, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922001947, "raptor_stability": 0.8965046405792236, "fragility": 1.5, "augmentation_robustness": 0.8297915129525963, "xie_feature_dispersion": 2.3654705725557443, "pac": 0.86314807676591, "whitened_cosine_id": 0.9005034565925598}} +{"model": "pythia-70m", "dataset": "spam", "seed": 1, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004856, "raptor_stability": 0.9503062963485718, "fragility": 3.0, "augmentation_robustness": 0.9222602495310279, "xie_feature_dispersion": 6.604643787479847, "pac": 0.9362832729397998, "whitened_cosine_id": 0.9495452046394348}} +{"model": "pythia-70m", "dataset": "cola", "seed": 1, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.7986350655555725, "fragility": 1.0, "augmentation_robustness": 0.720600733310639, "xie_feature_dispersion": 0.7104026437376014, "pac": 0.7596178994331058, "whitened_cosine_id": 0.7400056719779968}} +{"model": "pythia-70m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6754527167162487, "raptor_stability": 0.8382856249809265, "fragility": 1.0, "augmentation_robustness": 0.8615565989812328, "xie_feature_dispersion": 1.8257872743628552, "pac": 0.8499211119810797, "whitened_cosine_id": 0.8341732621192932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 1, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997502, "raptor_stability": 0.8919016122817993, "fragility": 1.5, "augmentation_robustness": 0.8621425329457455, "xie_feature_dispersion": 1.7499668565574926, "pac": 0.8770220726137724, "whitened_cosine_id": 0.8624804019927979}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 1, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349208, "raptor_stability": 0.8506006002426147, "fragility": 0.25, "augmentation_robustness": 0.637601559557224, "xie_feature_dispersion": 1.9524296354090338, "pac": 0.7441010798999194, "whitened_cosine_id": 0.8579458594322205}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 1, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350347, "raptor_stability": 0.8709588050842285, "fragility": 1.0, "augmentation_robustness": 0.8681998579081558, "xie_feature_dispersion": 1.4529956718728516, "pac": 0.8695793314961922, "whitened_cosine_id": 0.8694334030151367}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 1, "layer": 7, "concept": "topic", "predictors": {"sip_eigengap": 0.45824187669700944, "raptor_stability": 0.8648454546928406, "fragility": 1.0, "augmentation_robustness": 0.883777230516235, "xie_feature_dispersion": 2.411760223810931, "pac": 0.8743113426045378, "whitened_cosine_id": 0.9022970199584961}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.12810819210433996, "raptor_stability": 0.9377708435058594, "fragility": 3.0, "augmentation_robustness": 0.8310133850459603, "xie_feature_dispersion": 10.50628418134277, "pac": 0.8843921142759099, "whitened_cosine_id": 0.9117757678031921}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 1, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341767, "raptor_stability": 0.7650079131126404, "fragility": 0.5, "augmentation_robustness": 0.7779032363265277, "xie_feature_dispersion": 0.6428540179632124, "pac": 0.771455574719584, "whitened_cosine_id": 0.7203051447868347}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 1, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5407329041738655, "raptor_stability": 0.8731709122657776, "fragility": 0.5, "augmentation_robustness": 0.8006721595864167, "xie_feature_dispersion": 1.6646203889081308, "pac": 0.8369215359260971, "whitened_cosine_id": 0.854195773601532}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 1, "layer": 3, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348242, "raptor_stability": 0.8365353345870972, "fragility": 1.0, "augmentation_robustness": 0.7581229958494136, "xie_feature_dispersion": 1.9822842109390966, "pac": 0.7973291652182554, "whitened_cosine_id": 0.8330134749412537}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 1, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346446, "raptor_stability": 0.8064267039299011, "fragility": 0.25, "augmentation_robustness": 0.6783393674021946, "xie_feature_dispersion": 3.1602498907933048, "pac": 0.7423830356660479, "whitened_cosine_id": 0.7950520515441895}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 1, "layer": 8, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934659, "raptor_stability": 0.7950870990753174, "fragility": 0.25, "augmentation_robustness": 0.7522003888311254, "xie_feature_dispersion": 1.0110021271866068, "pac": 0.7736437439532213, "whitened_cosine_id": 0.7926974892616272}} +{"model": "pythia-160m", "dataset": "subj", "seed": 1, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351837, "raptor_stability": 0.8696022033691406, "fragility": 1.0, "augmentation_robustness": 0.7639011148909334, "xie_feature_dispersion": 2.2528502880398076, "pac": 0.816751659130037, "whitened_cosine_id": 0.9035648703575134}} +{"model": "pythia-160m", "dataset": "spam", "seed": 1, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352847, "raptor_stability": 0.9360153079032898, "fragility": 3.0, "augmentation_robustness": 0.9010665268870804, "xie_feature_dispersion": 7.136742674643747, "pac": 0.9185409173951851, "whitened_cosine_id": 0.9397127628326416}} +{"model": "pythia-160m", "dataset": "cola", "seed": 1, "layer": 8, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343877, "raptor_stability": 0.799544095993042, "fragility": 0.25, "augmentation_robustness": 0.6070554401031638, "xie_feature_dispersion": 0.34675513205319614, "pac": 0.7032997680481029, "whitened_cosine_id": 0.7765609622001648}} +{"model": "pythia-160m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6971690127369661, "raptor_stability": 0.8396480679512024, "fragility": 1.0, "augmentation_robustness": 0.8287048173260642, "xie_feature_dispersion": 2.4375720598203166, "pac": 0.8341764426386333, "whitened_cosine_id": 0.8327767848968506}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 1, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349912, "raptor_stability": 0.8510211110115051, "fragility": 0.5, "augmentation_robustness": 0.7777689965959116, "xie_feature_dispersion": 0.793449068148806, "pac": 0.8143950538037084, "whitened_cosine_id": 0.873329222202301}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 1, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688131, "raptor_stability": 0.8608813881874084, "fragility": 0.25, "augmentation_robustness": 0.6344172991669839, "xie_feature_dispersion": 2.4202065461787297, "pac": 0.7476493436771962, "whitened_cosine_id": 0.8642251491546631}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 1, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881893, "raptor_stability": 0.8481273055076599, "fragility": 1.0, "augmentation_robustness": 0.8003220169247562, "xie_feature_dispersion": 2.320769849966657, "pac": 0.8242246612162081, "whitened_cosine_id": 0.8769779801368713}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.31678079404561216, "raptor_stability": 0.8443418741226196, "fragility": 1.5, "augmentation_robustness": 0.8372749145627374, "xie_feature_dispersion": 6.763574998208828, "pac": 0.8408083943426785, "whitened_cosine_id": 0.8846130967140198}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.08895215281835345, "raptor_stability": 0.9383633732795715, "fragility": 3.0, "augmentation_robustness": 0.8350381950983168, "xie_feature_dispersion": 10.874814823071377, "pac": 0.8867007841889442, "whitened_cosine_id": 0.8853277564048767}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 1, "layer": 16, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878747, "raptor_stability": 0.7895128726959229, "fragility": 0.25, "augmentation_robustness": 0.7721623022098749, "xie_feature_dispersion": 0.6522465207916613, "pac": 0.7808375874528989, "whitened_cosine_id": 0.7736839652061462}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 1, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5847740085808558, "raptor_stability": 0.8644301295280457, "fragility": 1.0, "augmentation_robustness": 0.7506737064210864, "xie_feature_dispersion": 2.0251662058673543, "pac": 0.807551917974566, "whitened_cosine_id": 0.8443336486816406}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879302, "raptor_stability": 0.8073345422744751, "fragility": 0.25, "augmentation_robustness": 0.7273212412855212, "xie_feature_dispersion": 5.356927418014672, "pac": 0.7673278917799982, "whitened_cosine_id": 0.8217640519142151}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 1, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956880057, "raptor_stability": 0.8053412437438965, "fragility": 0.25, "augmentation_robustness": 0.6599905871716598, "xie_feature_dispersion": 2.9578595858885146, "pac": 0.7326659154577781, "whitened_cosine_id": 0.8191479444503784}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 1, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687935, "raptor_stability": 0.7920674681663513, "fragility": 0.25, "augmentation_robustness": 0.7530199776918786, "xie_feature_dispersion": 2.0304992840234406, "pac": 0.772543722929115, "whitened_cosine_id": 0.8043911457061768}} +{"model": "pythia-410m", "dataset": "subj", "seed": 1, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882515, "raptor_stability": 0.8890897631645203, "fragility": 0.5, "augmentation_robustness": 0.774157157496381, "xie_feature_dispersion": 1.4231569822537715, "pac": 0.8316234603304506, "whitened_cosine_id": 0.8959898948669434}} +{"model": "pythia-410m", "dataset": "spam", "seed": 1, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882854, "raptor_stability": 0.9181560277938843, "fragility": 3.0, "augmentation_robustness": 0.8613923232327275, "xie_feature_dispersion": 13.35429844074093, "pac": 0.889774175513306, "whitened_cosine_id": 0.9216764569282532}} +{"model": "pythia-410m", "dataset": "cola", "seed": 1, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879537, "raptor_stability": 0.8038613796234131, "fragility": 0.25, "augmentation_robustness": 0.5332025416049512, "xie_feature_dispersion": 0.5611200077386973, "pac": 0.6685319606141822, "whitened_cosine_id": 0.8066012859344482}} +{"model": "pythia-410m", "dataset": "stance", "seed": 1, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.5266347984206482, "raptor_stability": 0.8334753513336182, "fragility": 0.5, "augmentation_robustness": 0.8255680326919436, "xie_feature_dispersion": 2.548552169168803, "pac": 0.8295216920127808, "whitened_cosine_id": 0.82421875}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 1, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.210307295688135, "raptor_stability": 0.8675035238265991, "fragility": 1.0, "augmentation_robustness": 0.7796142351617439, "xie_feature_dispersion": 2.10629803849922, "pac": 0.8235588794941715, "whitened_cosine_id": 0.8668701648712158}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 1, "layer": 22, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8375368714332581, "fragility": 0.25, "augmentation_robustness": 0.5707958641951811, "xie_feature_dispersion": 2.2166913426405066, "pac": 0.7041663678142196, "whitened_cosine_id": 0.8229580521583557}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8483022451400757, "fragility": 1.0, "augmentation_robustness": 0.773674431811487, "xie_feature_dispersion": 2.6155304376742743, "pac": 0.8109883384757813, "whitened_cosine_id": 0.8468784689903259}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 1, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.34505361071401985, "raptor_stability": 0.8623528480529785, "fragility": 1.5, "augmentation_robustness": 0.866225226560754, "xie_feature_dispersion": 4.689887830271909, "pac": 0.8642890373068662, "whitened_cosine_id": 0.8534741997718811}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.058822548518337014, "raptor_stability": 0.9422677159309387, "fragility": 2.0, "augmentation_robustness": 0.8626061839556527, "xie_feature_dispersion": 8.534984669917208, "pac": 0.9024369499432957, "whitened_cosine_id": 0.8359106779098511}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 1, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010858, "raptor_stability": 0.7976512312889099, "fragility": 0.25, "augmentation_robustness": 0.743646934055148, "xie_feature_dispersion": 0.8601541662036642, "pac": 0.7706490826720289, "whitened_cosine_id": 0.8195480704307556}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 1, "layer": 19, "concept": "emotion", "predictors": {"sip_eigengap": 0.40445308965852644, "raptor_stability": 0.8435872197151184, "fragility": 0.5, "augmentation_robustness": 0.7387802241177431, "xie_feature_dispersion": 2.4410346204498894, "pac": 0.7911837219164308, "whitened_cosine_id": 0.8220205903053284}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 1, "layer": 23, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.7921804785728455, "fragility": 0.25, "augmentation_robustness": 0.6227259022979126, "xie_feature_dispersion": 3.0062653097311682, "pac": 0.707453190435379, "whitened_cosine_id": 0.8225616812705994}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 1, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010583, "raptor_stability": 0.8012144565582275, "fragility": 0.25, "augmentation_robustness": 0.6679021398605163, "xie_feature_dispersion": 3.9129339183725933, "pac": 0.7345582982093719, "whitened_cosine_id": 0.8250377774238586}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 1, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010749, "raptor_stability": 0.7956711649894714, "fragility": 0.25, "augmentation_robustness": 0.7021864744022904, "xie_feature_dispersion": 2.452541305670078, "pac": 0.748928819695881, "whitened_cosine_id": 0.8224895000457764}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 1, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010877, "raptor_stability": 0.8905167579650879, "fragility": 1.0, "augmentation_robustness": 0.7827455359108311, "xie_feature_dispersion": 2.6061836065846578, "pac": 0.8366311469379595, "whitened_cosine_id": 0.8333907723426819}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 1, "layer": 13, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010649, "raptor_stability": 0.9169431924819946, "fragility": 1.5, "augmentation_robustness": 0.8734181256563801, "xie_feature_dispersion": 22.457465915951406, "pac": 0.8951806590691873, "whitened_cosine_id": 0.8557054400444031}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 1, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.855816496100951, "raptor_stability": 0.8068552017211914, "fragility": 0.25, "augmentation_robustness": 0.5148135283708677, "xie_feature_dispersion": 0.6902828208408504, "pac": 0.6608343650460295, "whitened_cosine_id": 0.8203985691070557}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 1, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.4012383857529635, "raptor_stability": 0.8363339304924011, "fragility": 0.5, "augmentation_robustness": 0.8057210469674079, "xie_feature_dispersion": 3.057908787470783, "pac": 0.8210274887299045, "whitened_cosine_id": 0.8217589259147644}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 1, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010833, "raptor_stability": 0.8418669700622559, "fragility": 1.0, "augmentation_robustness": 0.7293225446950196, "xie_feature_dispersion": 1.6761850609669111, "pac": 0.7855947573786377, "whitened_cosine_id": 0.8340412378311157}} +{"model": "gpt2", "dataset": "sst2", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.39754248593502, "raptor_stability": 0.8069735169410706, "fragility": 0.25, "augmentation_robustness": 0.5092073410955438, "xie_feature_dispersion": 1.648414982559877, "pac": 0.6580904290183072, "whitened_cosine_id": 0.8618525862693787}} +{"model": "gpt2", "dataset": "imdb", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350674, "raptor_stability": 0.8155258893966675, "fragility": 0.25, "augmentation_robustness": 0.6918368797960317, "xie_feature_dispersion": 1.679121568516715, "pac": 0.7536813845963496, "whitened_cosine_id": 0.8795225024223328}} +{"model": "gpt2", "dataset": "ag_news", "seed": 1, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.3994197076471763, "raptor_stability": 0.8592859506607056, "fragility": 0.5, "augmentation_robustness": 0.8643756652798957, "xie_feature_dispersion": 1.95723297240118, "pac": 0.8618308079703006, "whitened_cosine_id": 0.9026321768760681}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.13768748508218773, "raptor_stability": 0.9254063367843628, "fragility": 1.5, "augmentation_robustness": 0.7987703771800989, "xie_feature_dispersion": 5.392898240224899, "pac": 0.8620883569822309, "whitened_cosine_id": 0.9057890176773071}} +{"model": "gpt2", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342838, "raptor_stability": 0.7716324925422668, "fragility": 0.25, "augmentation_robustness": 0.7788427795247373, "xie_feature_dispersion": 0.5555909547173986, "pac": 0.775237636033502, "whitened_cosine_id": 0.7441529631614685}} +{"model": "gpt2", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5908930964643848, "raptor_stability": 0.8605509400367737, "fragility": 0.25, "augmentation_robustness": 0.7363054353676359, "xie_feature_dispersion": 1.3002246100849617, "pac": 0.7984281877022048, "whitened_cosine_id": 0.859701931476593}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348176, "raptor_stability": 0.8326132297515869, "fragility": 0.25, "augmentation_robustness": 0.7692316541993737, "xie_feature_dispersion": 0.8694719793739331, "pac": 0.8009224419754803, "whitened_cosine_id": 0.8273347020149231}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 1, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934648, "raptor_stability": 0.775840163230896, "fragility": 0.25, "augmentation_robustness": 0.6870212631337806, "xie_feature_dispersion": 2.3737390065905095, "pac": 0.7314307131823383, "whitened_cosine_id": 0.7987164855003357}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 1, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346262, "raptor_stability": 0.8207367062568665, "fragility": 0.25, "augmentation_robustness": 0.7990409525750108, "xie_feature_dispersion": 1.0977706807422258, "pac": 0.8098888294159386, "whitened_cosine_id": 0.7928319573402405}} +{"model": "gpt2", "dataset": "subj", "seed": 1, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351993, "raptor_stability": 0.8593925833702087, "fragility": 0.5, "augmentation_robustness": 0.6926262488173032, "xie_feature_dispersion": 1.2545052058156476, "pac": 0.776009416093756, "whitened_cosine_id": 0.9041862487792969}} +{"model": "gpt2", "dataset": "spam", "seed": 1, "layer": 12, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353376, "raptor_stability": 0.9029030203819275, "fragility": 0.25, "augmentation_robustness": 0.835250541372843, "xie_feature_dispersion": 1.4550691568045606, "pac": 0.8690767808773853, "whitened_cosine_id": 0.9343510270118713}} +{"model": "gpt2", "dataset": "cola", "seed": 1, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344225, "raptor_stability": 0.804100751876831, "fragility": 0.25, "augmentation_robustness": 0.6480747813745288, "xie_feature_dispersion": 0.35759111190860965, "pac": 0.72608776662568, "whitened_cosine_id": 0.7744043469429016}} +{"model": "gpt2", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.5721762371971069, "raptor_stability": 0.8337757587432861, "fragility": 0.5, "augmentation_robustness": 0.7933240817840598, "xie_feature_dispersion": 1.1954515513989168, "pac": 0.8135499202636729, "whitened_cosine_id": 0.8244547843933105}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 1, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934976, "raptor_stability": 0.854945182800293, "fragility": 0.25, "augmentation_robustness": 0.8063951328046506, "xie_feature_dispersion": 0.46463961416788957, "pac": 0.8306701578024718, "whitened_cosine_id": 0.8690304160118103}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 1, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688203, "raptor_stability": 0.8271149396896362, "fragility": 0.25, "augmentation_robustness": 0.5702797141411624, "xie_feature_dispersion": 1.6809287391267713, "pac": 0.6986973269153993, "whitened_cosine_id": 0.8607844114303589}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 1, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 1.210307295687949, "raptor_stability": 0.791394829750061, "fragility": 0.25, "augmentation_robustness": 0.8160921898342837, "xie_feature_dispersion": 0.5899893741394086, "pac": 0.8037435097921724, "whitened_cosine_id": 0.795337975025177}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5592522550238864, "raptor_stability": 0.8738617300987244, "fragility": 0.25, "augmentation_robustness": 0.7757794731716269, "xie_feature_dispersion": 0.9112899816215909, "pac": 0.8248206016351756, "whitened_cosine_id": 0.8513531684875488}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 1, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880974, "raptor_stability": 0.8126772046089172, "fragility": 0.25, "augmentation_robustness": 0.7615894417916816, "xie_feature_dispersion": 1.322075514358211, "pac": 0.7871333232002995, "whitened_cosine_id": 0.8305314779281616}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 1, "layer": 9, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879932, "raptor_stability": 0.7768864035606384, "fragility": 0.25, "augmentation_robustness": 0.6407154851585959, "xie_feature_dispersion": 2.304316868570724, "pac": 0.7088009443596172, "whitened_cosine_id": 0.7996089458465576}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 1, "layer": 24, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687991, "raptor_stability": 0.808860182762146, "fragility": 0.25, "augmentation_robustness": 0.779057007645201, "xie_feature_dispersion": 1.0954582686409513, "pac": 0.7939585952036735, "whitened_cosine_id": 0.8093989491462708}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 1, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882626, "raptor_stability": 0.8778501749038696, "fragility": 0.5, "augmentation_robustness": 0.7528926011129462, "xie_feature_dispersion": 1.211853663340529, "pac": 0.8153713880084079, "whitened_cosine_id": 0.8855262398719788}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 1, "layer": 16, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880065, "raptor_stability": 0.8000022768974304, "fragility": 0.25, "augmentation_robustness": 0.5320479658836585, "xie_feature_dispersion": 0.37533178694301084, "pac": 0.6660251213905445, "whitened_cosine_id": 0.8167566061019897}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 1, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.4372306810871382, "raptor_stability": 0.8121774792671204, "fragility": 0.5, "augmentation_robustness": 0.7829438977858721, "xie_feature_dispersion": 1.6812559297796246, "pac": 0.7975606885264963, "whitened_cosine_id": 0.7655757069587708}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 1, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.210307295688115, "raptor_stability": 0.8233330845832825, "fragility": 0.25, "augmentation_robustness": 0.6852795330724812, "xie_feature_dispersion": 0.620429769989345, "pac": 0.7543063088278819, "whitened_cosine_id": 0.856203019618988}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 1, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237649584, "raptor_stability": 0.8650567531585693, "fragility": 0.25, "augmentation_robustness": 0.6688278069812248, "xie_feature_dispersion": 2.1604793022510873, "pac": 0.7669422800698971, "whitened_cosine_id": 0.8860111236572266}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 1, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923765019, "raptor_stability": 0.8548077940940857, "fragility": 1.0, "augmentation_robustness": 0.8296912251707328, "xie_feature_dispersion": 2.619265146348305, "pac": 0.8422495096324092, "whitened_cosine_id": 0.8861786127090454}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.3284735817230639, "raptor_stability": 0.865305483341217, "fragility": 1.0, "augmentation_robustness": 0.8707412223835058, "xie_feature_dispersion": 3.3146556682562625, "pac": 0.8680233528623614, "whitened_cosine_id": 0.9071072340011597}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 1, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.06700207827316693, "raptor_stability": 0.9375728368759155, "fragility": 3.0, "augmentation_robustness": 0.8370451297198389, "xie_feature_dispersion": 9.643792516378186, "pac": 0.8873089832978772, "whitened_cosine_id": 0.8979226350784302}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646287, "raptor_stability": 0.8297770023345947, "fragility": 0.25, "augmentation_robustness": 0.8462807577423243, "xie_feature_dispersion": 0.5968347455995571, "pac": 0.8380288800384594, "whitened_cosine_id": 0.7946618795394897}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 1, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.6161992460964921, "raptor_stability": 0.882070004940033, "fragility": 0.25, "augmentation_robustness": 0.8630147174643309, "xie_feature_dispersion": 2.820954214278998, "pac": 0.8725423612021819, "whitened_cosine_id": 0.8536849617958069}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 1, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645619, "raptor_stability": 0.8290610313415527, "fragility": 0.25, "augmentation_robustness": 0.7926104311252984, "xie_feature_dispersion": 5.467260144092137, "pac": 0.8108357312334256, "whitened_cosine_id": 0.8316267728805542}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 1, "layer": 20, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647093, "raptor_stability": 0.7731820940971375, "fragility": 0.25, "augmentation_robustness": 0.6355369799746609, "xie_feature_dispersion": 2.999448892433556, "pac": 0.7043595370358992, "whitened_cosine_id": 0.798532247543335}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 1, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237645867, "raptor_stability": 0.819866955280304, "fragility": 0.25, "augmentation_robustness": 0.8020143361928123, "xie_feature_dispersion": 2.0378126951408726, "pac": 0.8109406457365581, "whitened_cosine_id": 0.8144164681434631}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 1, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.293872923765124, "raptor_stability": 0.8871787190437317, "fragility": 0.5, "augmentation_robustness": 0.7845579254901294, "xie_feature_dispersion": 1.6425796792755814, "pac": 0.8358683222669305, "whitened_cosine_id": 0.9134312868118286}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 1, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237639786, "raptor_stability": 0.9124530553817749, "fragility": 0.5, "augmentation_robustness": 0.907099253659993, "xie_feature_dispersion": 18.266615233549604, "pac": 0.909776154520884, "whitened_cosine_id": 0.9418519139289856}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 1, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647346, "raptor_stability": 0.8199473023414612, "fragility": 0.25, "augmentation_robustness": 0.5916327937131608, "xie_feature_dispersion": 0.5437385315920449, "pac": 0.7057900480273109, "whitened_cosine_id": 0.8188185691833496}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 1, "layer": 22, "concept": "stance", "predictors": {"sip_eigengap": 0.5082992376979554, "raptor_stability": 0.8089979290962219, "fragility": 1.0, "augmentation_robustness": 0.7725685252374112, "xie_feature_dispersion": 2.217566060904384, "pac": 0.7907832271668165, "whitened_cosine_id": 0.8215915560722351}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 1, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649438, "raptor_stability": 0.8504278063774109, "fragility": 0.25, "augmentation_robustness": 0.7801897363682817, "xie_feature_dispersion": 0.4453040461938641, "pac": 0.8153087713728463, "whitened_cosine_id": 0.8711152076721191}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 2, "layer": 1, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996196, "raptor_stability": 0.8725253343582153, "fragility": 1.0, "augmentation_robustness": 0.7320160436555694, "xie_feature_dispersion": 1.8850577574441822, "pac": 0.8022706890068924, "whitened_cosine_id": 0.8443945646286011}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998612, "raptor_stability": 0.8727290630340576, "fragility": 1.0, "augmentation_robustness": 0.8893448732634346, "xie_feature_dispersion": 1.1304006682636438, "pac": 0.8810369681487461, "whitened_cosine_id": 0.8432037830352783}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5762362270014945, "raptor_stability": 0.8451294302940369, "fragility": 1.0, "augmentation_robustness": 0.8227084099515827, "xie_feature_dispersion": 4.598590726152205, "pac": 0.8339189201228098, "whitened_cosine_id": 0.9046899676322937}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.13862003115739582, "raptor_stability": 0.9396008253097534, "fragility": 2.0, "augmentation_robustness": 0.8714397420888463, "xie_feature_dispersion": 7.460918475973291, "pac": 0.9055202836992999, "whitened_cosine_id": 0.9276723265647888}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921976638, "raptor_stability": 0.7323077321052551, "fragility": 0.25, "augmentation_robustness": 0.6404787869477748, "xie_feature_dispersion": 0.06278074668041304, "pac": 0.686393259526515, "whitened_cosine_id": 0.7365422248840332}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 2, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.6201478550392475, "raptor_stability": 0.860109269618988, "fragility": 0.5, "augmentation_robustness": 0.7869247885334385, "xie_feature_dispersion": 1.544046125551409, "pac": 0.8235170290762133, "whitened_cosine_id": 0.841076672077179}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 2, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921991476, "raptor_stability": 0.8527593612670898, "fragility": 1.0, "augmentation_robustness": 0.794283550859188, "xie_feature_dispersion": 1.5619058370885555, "pac": 0.823521456063139, "whitened_cosine_id": 0.7776172161102295}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987974, "raptor_stability": 0.8192676901817322, "fragility": 1.0, "augmentation_robustness": 0.7427685123086125, "xie_feature_dispersion": 1.857559017759172, "pac": 0.7810181012451723, "whitened_cosine_id": 0.7365460991859436}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990956, "raptor_stability": 0.8345112800598145, "fragility": 1.0, "augmentation_robustness": 0.7886622062234387, "xie_feature_dispersion": 1.342889203260779, "pac": 0.8115867431416266, "whitened_cosine_id": 0.8005088567733765}} +{"model": "pythia-70m", "dataset": "subj", "seed": 2, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.711632992200211, "raptor_stability": 0.8931885957717896, "fragility": 1.5, "augmentation_robustness": 0.8300863608388453, "xie_feature_dispersion": 2.385960236411005, "pac": 0.8616374783053175, "whitened_cosine_id": 0.9023992419242859}} +{"model": "pythia-70m", "dataset": "spam", "seed": 2, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004394, "raptor_stability": 0.9427627921104431, "fragility": 1.5, "augmentation_robustness": 0.9205893937056647, "xie_feature_dispersion": 10.434188689895425, "pac": 0.9316760929080539, "whitened_cosine_id": 0.9575247764587402}} +{"model": "pythia-70m", "dataset": "cola", "seed": 2, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921978055, "raptor_stability": 0.7819207310676575, "fragility": 1.0, "augmentation_robustness": 0.7494339269084461, "xie_feature_dispersion": 0.5849615167254145, "pac": 0.7656773289880519, "whitened_cosine_id": 0.7173330187797546}} +{"model": "pythia-70m", "dataset": "stance", "seed": 2, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.467708916223557, "raptor_stability": 0.823031485080719, "fragility": 1.0, "augmentation_robustness": 0.8142662820172029, "xie_feature_dispersion": 1.6869566033029149, "pac": 0.8186488835489609, "whitened_cosine_id": 0.8191723823547363}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 2, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998414, "raptor_stability": 0.8615893721580505, "fragility": 0.5, "augmentation_robustness": 0.8216595763224332, "xie_feature_dispersion": 0.9880768598724781, "pac": 0.8416244742402419, "whitened_cosine_id": 0.8701478838920593}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859348855, "raptor_stability": 0.8407484292984009, "fragility": 0.25, "augmentation_robustness": 0.5918794762745934, "xie_feature_dispersion": 2.033621000673696, "pac": 0.7163139527864971, "whitened_cosine_id": 0.8444841504096985}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 2, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350019, "raptor_stability": 0.8570092916488647, "fragility": 1.0, "augmentation_robustness": 0.8472864724655493, "xie_feature_dispersion": 1.332977638352146, "pac": 0.852147882057207, "whitened_cosine_id": 0.8642057180404663}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.4756762658594781, "raptor_stability": 0.8755650520324707, "fragility": 1.0, "augmentation_robustness": 0.8736053793270782, "xie_feature_dispersion": 2.6831447841823626, "pac": 0.8745852156797744, "whitened_cosine_id": 0.9056297540664673}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.10936523074513556, "raptor_stability": 0.938740074634552, "fragility": 2.0, "augmentation_robustness": 0.849824439887086, "xie_feature_dispersion": 8.126249426404561, "pac": 0.894282257260819, "whitened_cosine_id": 0.9102428555488586}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343164, "raptor_stability": 0.7734526991844177, "fragility": 0.25, "augmentation_robustness": 0.7341807784351317, "xie_feature_dispersion": 0.24746600780733177, "pac": 0.7538167388097747, "whitened_cosine_id": 0.7335423827171326}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 2, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.5898968502609128, "raptor_stability": 0.8509833216667175, "fragility": 0.25, "augmentation_robustness": 0.7503952782199734, "xie_feature_dispersion": 1.9109009411804163, "pac": 0.8006892999433455, "whitened_cosine_id": 0.8450579643249512}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 2, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347338, "raptor_stability": 0.8163250684738159, "fragility": 0.25, "augmentation_robustness": 0.720428393459381, "xie_feature_dispersion": 1.672145195874313, "pac": 0.7683767309665985, "whitened_cosine_id": 0.8197868466377258}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345715, "raptor_stability": 0.8170239925384521, "fragility": 1.0, "augmentation_robustness": 0.7672939677957877, "xie_feature_dispersion": 2.1306331599217914, "pac": 0.7921589801671199, "whitened_cosine_id": 0.7781753540039062}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 2, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345553, "raptor_stability": 0.8093597888946533, "fragility": 0.25, "augmentation_robustness": 0.7487295667365286, "xie_feature_dispersion": 1.9731796308382967, "pac": 0.779044677815591, "whitened_cosine_id": 0.7859788537025452}} +{"model": "pythia-160m", "dataset": "subj", "seed": 2, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935202, "raptor_stability": 0.8897599577903748, "fragility": 1.0, "augmentation_robustness": 0.8051559896221852, "xie_feature_dispersion": 1.9497282630608124, "pac": 0.84745797370628, "whitened_cosine_id": 0.9099186658859253}} +{"model": "pythia-160m", "dataset": "spam", "seed": 2, "layer": 8, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935196, "raptor_stability": 0.9361570477485657, "fragility": 1.5, "augmentation_robustness": 0.9125687730559409, "xie_feature_dispersion": 12.123957991903673, "pac": 0.9243629104022533, "whitened_cosine_id": 0.9451765418052673}} +{"model": "pythia-160m", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934449, "raptor_stability": 0.7818026542663574, "fragility": 0.25, "augmentation_robustness": 0.6248388717067496, "xie_feature_dispersion": 0.36909274612641374, "pac": 0.7033207629865534, "whitened_cosine_id": 0.7660055756568909}} +{"model": "pythia-160m", "dataset": "stance", "seed": 2, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.550133319968111, "raptor_stability": 0.8319461345672607, "fragility": 1.5, "augmentation_robustness": 0.8296352226027173, "xie_feature_dispersion": 2.567878497868356, "pac": 0.830790678584989, "whitened_cosine_id": 0.8248243927955627}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350152, "raptor_stability": 0.8500860929489136, "fragility": 0.5, "augmentation_robustness": 0.7916745698533383, "xie_feature_dispersion": 1.015287990938809, "pac": 0.8208803314011259, "whitened_cosine_id": 0.8699454069137573}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880925, "raptor_stability": 0.8507713079452515, "fragility": 0.25, "augmentation_robustness": 0.6548658634701003, "xie_feature_dispersion": 2.5382673428676514, "pac": 0.7528185857076759, "whitened_cosine_id": 0.8639923334121704}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881815, "raptor_stability": 0.8507294654846191, "fragility": 1.0, "augmentation_robustness": 0.8077367041653364, "xie_feature_dispersion": 2.1999878871216194, "pac": 0.8292330848249778, "whitened_cosine_id": 0.876013994216919}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 2, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.48444830765922703, "raptor_stability": 0.8797903060913086, "fragility": 1.5, "augmentation_robustness": 0.8897077200880827, "xie_feature_dispersion": 4.318369085246188, "pac": 0.8847490130896957, "whitened_cosine_id": 0.8930330276489258}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06934869454749534, "raptor_stability": 0.9469916224479675, "fragility": 2.0, "augmentation_robustness": 0.8838004807942493, "xie_feature_dispersion": 7.256347007074097, "pac": 0.9153960516211084, "whitened_cosine_id": 0.8971635699272156}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878984, "raptor_stability": 0.7757969498634338, "fragility": 0.25, "augmentation_robustness": 0.7602377504416505, "xie_feature_dispersion": 0.20512071078637148, "pac": 0.7680173501525422, "whitened_cosine_id": 0.7685407996177673}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 2, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.5276419919506735, "raptor_stability": 0.8558531403541565, "fragility": 0.25, "augmentation_robustness": 0.7811194669553142, "xie_feature_dispersion": 2.490653998171255, "pac": 0.8184863036547354, "whitened_cosine_id": 0.8436367511749268}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 2, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880245, "raptor_stability": 0.8074635863304138, "fragility": 0.5, "augmentation_robustness": 0.7107950361323764, "xie_feature_dispersion": 1.9566993340581085, "pac": 0.759129311231395, "whitened_cosine_id": 0.8125779628753662}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295687971, "raptor_stability": 0.7986416220664978, "fragility": 0.5, "augmentation_robustness": 0.6903031510581926, "xie_feature_dispersion": 1.8402433518812902, "pac": 0.7444723865623453, "whitened_cosine_id": 0.7969480156898499}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 2, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879715, "raptor_stability": 0.8084327578544617, "fragility": 0.5, "augmentation_robustness": 0.7257825239499871, "xie_feature_dispersion": 1.7043242651025499, "pac": 0.7671076409022244, "whitened_cosine_id": 0.7994747161865234}} +{"model": "pythia-410m", "dataset": "subj", "seed": 2, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688262, "raptor_stability": 0.8808537721633911, "fragility": 0.5, "augmentation_robustness": 0.7730695710492665, "xie_feature_dispersion": 1.6627162013824885, "pac": 0.8269616716063288, "whitened_cosine_id": 0.8980955481529236}} +{"model": "pythia-410m", "dataset": "spam", "seed": 2, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956871068, "raptor_stability": 0.9453094601631165, "fragility": 1.0, "augmentation_robustness": 0.931993646437433, "xie_feature_dispersion": 17.657395386814123, "pac": 0.9386515533002747, "whitened_cosine_id": 0.9302859306335449}} +{"model": "pythia-410m", "dataset": "cola", "seed": 2, "layer": 17, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879637, "raptor_stability": 0.7948347926139832, "fragility": 0.25, "augmentation_robustness": 0.6269925244354974, "xie_feature_dispersion": 0.47979217522467377, "pac": 0.7109136585247402, "whitened_cosine_id": 0.8036185503005981}} +{"model": "pythia-410m", "dataset": "stance", "seed": 2, "layer": 10, "concept": "stance", "predictors": {"sip_eigengap": 0.47549258334344463, "raptor_stability": 0.8262801170349121, "fragility": 0.5, "augmentation_robustness": 0.8202575512360268, "xie_feature_dispersion": 3.2823490583418935, "pac": 0.8232688341354695, "whitened_cosine_id": 0.8195136189460754}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 2, "layer": 18, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881415, "raptor_stability": 0.8284711837768555, "fragility": 0.5, "augmentation_robustness": 0.7694913743994085, "xie_feature_dispersion": 1.2261896328299637, "pac": 0.798981279088132, "whitened_cosine_id": 0.8577603101730347}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010692, "raptor_stability": 0.8523985147476196, "fragility": 0.25, "augmentation_robustness": 0.6344970850996317, "xie_feature_dispersion": 3.125584115559324, "pac": 0.7434477999236256, "whitened_cosine_id": 0.8270771503448486}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8408520221710205, "fragility": 1.0, "augmentation_robustness": 0.7462513872391434, "xie_feature_dispersion": 2.261174552822513, "pac": 0.793551704705082, "whitened_cosine_id": 0.8419722318649292}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.3112674068238993, "raptor_stability": 0.8652853965759277, "fragility": 1.0, "augmentation_robustness": 0.8618662021334369, "xie_feature_dispersion": 4.2348607484097185, "pac": 0.8635757993546823, "whitened_cosine_id": 0.8362071514129639}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05012914539691602, "raptor_stability": 0.9452676773071289, "fragility": 3.0, "augmentation_robustness": 0.8590266216701495, "xie_feature_dispersion": 9.524616008618372, "pac": 0.9021471494886393, "whitened_cosine_id": 0.8305992484092712}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.7967193126678467, "fragility": 0.25, "augmentation_robustness": 0.755117458917892, "xie_feature_dispersion": 0.32762786809964206, "pac": 0.7759183857928693, "whitened_cosine_id": 0.8186400532722473}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 2, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.3604267252609762, "raptor_stability": 0.8496060371398926, "fragility": 0.5, "augmentation_robustness": 0.7497415128763687, "xie_feature_dispersion": 2.6993559873766775, "pac": 0.7996737750081306, "whitened_cosine_id": 0.8220704197883606}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 2, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8090983033180237, "fragility": 0.25, "augmentation_robustness": 0.6782138616148163, "xie_feature_dispersion": 3.2068116736663326, "pac": 0.74365608246642, "whitened_cosine_id": 0.8261517286300659}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 2, "layer": 7, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010822, "raptor_stability": 0.8073044419288635, "fragility": 0.25, "augmentation_robustness": 0.6867498455562944, "xie_feature_dispersion": 4.820388126528082, "pac": 0.747027143742579, "whitened_cosine_id": 0.8241050839424133}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 2, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010454, "raptor_stability": 0.7936219573020935, "fragility": 0.25, "augmentation_robustness": 0.6887170523418908, "xie_feature_dispersion": 2.728943510383841, "pac": 0.7411695048219922, "whitened_cosine_id": 0.8242641091346741}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 2, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010878, "raptor_stability": 0.8761361837387085, "fragility": 1.0, "augmentation_robustness": 0.7502694820898129, "xie_feature_dispersion": 2.7954401496665007, "pac": 0.8132028329142607, "whitened_cosine_id": 0.829967737197876}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 2, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010256, "raptor_stability": 0.9437923431396484, "fragility": 1.0, "augmentation_robustness": 0.9235808889611689, "xie_feature_dispersion": 23.76038075342399, "pac": 0.9336866160504087, "whitened_cosine_id": 0.8749403953552246}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 2, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.8045715689659119, "fragility": 0.25, "augmentation_robustness": 0.5269196391645194, "xie_feature_dispersion": 0.5939571817383262, "pac": 0.6657456040652157, "whitened_cosine_id": 0.8205759525299072}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 2, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.33865901445727387, "raptor_stability": 0.8298296332359314, "fragility": 1.0, "augmentation_robustness": 0.817334225300112, "xie_feature_dispersion": 3.230859359898638, "pac": 0.8235819292680218, "whitened_cosine_id": 0.8170763850212097}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 2, "layer": 14, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010837, "raptor_stability": 0.8219150900840759, "fragility": 0.5, "augmentation_robustness": 0.7537574160413394, "xie_feature_dispersion": 1.8431532727005182, "pac": 0.7878362530627077, "whitened_cosine_id": 0.8218834400177002}} +{"model": "gpt2", "dataset": "sst2", "seed": 2, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350012, "raptor_stability": 0.8311133980751038, "fragility": 0.25, "augmentation_robustness": 0.5968769748584443, "xie_feature_dispersion": 2.0407834350862544, "pac": 0.713995186466774, "whitened_cosine_id": 0.8748435378074646}} +{"model": "gpt2", "dataset": "imdb", "seed": 2, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350318, "raptor_stability": 0.8567911386489868, "fragility": 0.25, "augmentation_robustness": 0.8673792464607593, "xie_feature_dispersion": 0.26606041071725683, "pac": 0.862085192554873, "whitened_cosine_id": 0.8767779469490051}} +{"model": "gpt2", "dataset": "ag_news", "seed": 2, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.3821667228940741, "raptor_stability": 0.845832109451294, "fragility": 0.5, "augmentation_robustness": 0.8470052851683942, "xie_feature_dispersion": 1.979682604320564, "pac": 0.8464186973098441, "whitened_cosine_id": 0.903249979019165}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 2, "layer": 7, "concept": "topic", "predictors": {"sip_eigengap": 0.11558697704158609, "raptor_stability": 0.932966411113739, "fragility": 1.0, "augmentation_robustness": 0.8559305238325257, "xie_feature_dispersion": 4.299506274009608, "pac": 0.8944484674731323, "whitened_cosine_id": 0.9021454453468323}} +{"model": "gpt2", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343073, "raptor_stability": 0.7622057199478149, "fragility": 0.25, "augmentation_robustness": 0.7918529895588636, "xie_feature_dispersion": 0.33360574253280906, "pac": 0.7770293547533393, "whitened_cosine_id": 0.7459770441055298}} +{"model": "gpt2", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5200547285263808, "raptor_stability": 0.85597825050354, "fragility": 0.25, "augmentation_robustness": 0.7622544507506325, "xie_feature_dispersion": 1.2683994964221992, "pac": 0.8091163506270862, "whitened_cosine_id": 0.8535709977149963}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 2, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.397542485934766, "raptor_stability": 0.8122290968894958, "fragility": 0.25, "augmentation_robustness": 0.7701636575129145, "xie_feature_dispersion": 0.6036175754728635, "pac": 0.7911963772012052, "whitened_cosine_id": 0.8273720741271973}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 2, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859344792, "raptor_stability": 0.7899907827377319, "fragility": 0.25, "augmentation_robustness": 0.775070377256843, "xie_feature_dispersion": 1.1126097391994054, "pac": 0.7825305799972875, "whitened_cosine_id": 0.7636452317237854}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346548, "raptor_stability": 0.7969914674758911, "fragility": 0.25, "augmentation_robustness": 0.7083710583755256, "xie_feature_dispersion": 1.8684634814954304, "pac": 0.7526812629257084, "whitened_cosine_id": 0.8185012936592102}} +{"model": "gpt2", "dataset": "subj", "seed": 2, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352377, "raptor_stability": 0.8654428720474243, "fragility": 0.5, "augmentation_robustness": 0.7478951378849938, "xie_feature_dispersion": 1.4494284784328508, "pac": 0.8066690049662091, "whitened_cosine_id": 0.9153006076812744}} +{"model": "gpt2", "dataset": "spam", "seed": 2, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935319, "raptor_stability": 0.9391940832138062, "fragility": 1.0, "augmentation_robustness": 0.9242307247684519, "xie_feature_dispersion": 13.972658689577166, "pac": 0.9317124039911291, "whitened_cosine_id": 0.9425793290138245}} +{"model": "gpt2", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859345233, "raptor_stability": 0.7784033417701721, "fragility": 0.25, "augmentation_robustness": 0.5640885986826307, "xie_feature_dispersion": 0.38566559649240273, "pac": 0.6712459702264014, "whitened_cosine_id": 0.7816012501716614}} +{"model": "gpt2", "dataset": "stance", "seed": 2, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.5282860197072011, "raptor_stability": 0.8207774758338928, "fragility": 0.25, "augmentation_robustness": 0.7971149286838108, "xie_feature_dispersion": 2.2792793583875905, "pac": 0.8089462022588518, "whitened_cosine_id": 0.8131434321403503}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 2, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934994, "raptor_stability": 0.808012068271637, "fragility": 0.25, "augmentation_robustness": 0.7076387264686993, "xie_feature_dispersion": 0.8358972401809576, "pac": 0.7578253973701681, "whitened_cosine_id": 0.8693022727966309}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 2, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881729, "raptor_stability": 0.8265162706375122, "fragility": 0.25, "augmentation_robustness": 0.5469683098756675, "xie_feature_dispersion": 2.1253471965524975, "pac": 0.6867422902565898, "whitened_cosine_id": 0.8618488311767578}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 2, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.35097833794424904, "raptor_stability": 0.8510255813598633, "fragility": 0.5, "augmentation_robustness": 0.868126023664805, "xie_feature_dispersion": 1.511481734340918, "pac": 0.8595758025123341, "whitened_cosine_id": 0.8767551183700562}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 2, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878433, "raptor_stability": 0.7515677809715271, "fragility": 0.25, "augmentation_robustness": 0.8471055642765235, "xie_feature_dispersion": 0.3165629770195801, "pac": 0.7993366726240253, "whitened_cosine_id": 0.7615948915481567}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5517110732147434, "raptor_stability": 0.8649501204490662, "fragility": 0.25, "augmentation_robustness": 0.7727700403781212, "xie_feature_dispersion": 0.9242099717833849, "pac": 0.8188600804135937, "whitened_cosine_id": 0.8501837849617004}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 2, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880352, "raptor_stability": 0.7788109183311462, "fragility": 0.25, "augmentation_robustness": 0.602822870481301, "xie_feature_dispersion": 2.344924657926786, "pac": 0.6908168944062236, "whitened_cosine_id": 0.8237905502319336}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 2, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295688015, "raptor_stability": 0.7836035490036011, "fragility": 0.25, "augmentation_robustness": 0.6445047638465966, "xie_feature_dispersion": 2.829344259675601, "pac": 0.7140541564250988, "whitened_cosine_id": 0.8160997629165649}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880054, "raptor_stability": 0.7852885723114014, "fragility": 0.25, "augmentation_robustness": 0.6693173469866233, "xie_feature_dispersion": 0.831323350714088, "pac": 0.7273029596490124, "whitened_cosine_id": 0.8087374567985535}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 2, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882837, "raptor_stability": 0.8820376992225647, "fragility": 0.5, "augmentation_robustness": 0.7285834645747785, "xie_feature_dispersion": 1.6491215963892647, "pac": 0.8053105818986717, "whitened_cosine_id": 0.8916261792182922}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 2, "layer": 19, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880483, "raptor_stability": 0.7810653448104858, "fragility": 0.25, "augmentation_robustness": 0.5184999598231592, "xie_feature_dispersion": 0.4506855901276054, "pac": 0.6497826523168225, "whitened_cosine_id": 0.8220202326774597}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 2, "layer": 24, "concept": "stance", "predictors": {"sip_eigengap": 0.4048263633949508, "raptor_stability": 0.8287909626960754, "fragility": 0.25, "augmentation_robustness": 0.8148653549769995, "xie_feature_dispersion": 0.8817497017545193, "pac": 0.8218281588365375, "whitened_cosine_id": 0.8191941380500793}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881538, "raptor_stability": 0.8170264959335327, "fragility": 0.25, "augmentation_robustness": 0.6658562775951427, "xie_feature_dispersion": 0.6916798290341591, "pac": 0.7414413867643377, "whitened_cosine_id": 0.8560160398483276}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923764946, "raptor_stability": 0.8504462838172913, "fragility": 0.25, "augmentation_robustness": 0.6405801888029593, "xie_feature_dispersion": 2.428287550117154, "pac": 0.7455132363101253, "whitened_cosine_id": 0.8769184350967407}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650424, "raptor_stability": 0.8790234923362732, "fragility": 1.0, "augmentation_robustness": 0.873146163657816, "xie_feature_dispersion": 2.109825225717331, "pac": 0.8760848279970446, "whitened_cosine_id": 0.8947857618331909}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.4777053541327048, "raptor_stability": 0.8827468156814575, "fragility": 0.5, "augmentation_robustness": 0.9086818995974959, "xie_feature_dispersion": 1.9664542049888671, "pac": 0.8957143576394767, "whitened_cosine_id": 0.9089540839195251}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05912307025918351, "raptor_stability": 0.9463669061660767, "fragility": 1.5, "augmentation_robustness": 0.8757474533027418, "xie_feature_dispersion": 6.768079984244763, "pac": 0.9110571797344093, "whitened_cosine_id": 0.9140043258666992}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 2, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646398, "raptor_stability": 0.8356521129608154, "fragility": 0.25, "augmentation_robustness": 0.8160727049350692, "xie_feature_dispersion": 0.23710320099138982, "pac": 0.8258624089479423, "whitened_cosine_id": 0.7763546705245972}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 2, "layer": 16, "concept": "emotion", "predictors": {"sip_eigengap": 0.48285928508127296, "raptor_stability": 0.8633195757865906, "fragility": 0.25, "augmentation_robustness": 0.8474371685693614, "xie_feature_dispersion": 2.284945012773296, "pac": 0.855378372177976, "whitened_cosine_id": 0.8482233881950378}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 2, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645508, "raptor_stability": 0.8284585475921631, "fragility": 0.25, "augmentation_robustness": 0.7660624849853699, "xie_feature_dispersion": 3.252034579020222, "pac": 0.7972605162887665, "whitened_cosine_id": 0.8060616850852966}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237642506, "raptor_stability": 0.8207340240478516, "fragility": 0.25, "augmentation_robustness": 0.733566852009586, "xie_feature_dispersion": 3.71696516386688, "pac": 0.7771504380287189, "whitened_cosine_id": 0.8012250661849976}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 2, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237644577, "raptor_stability": 0.8056316375732422, "fragility": 0.25, "augmentation_robustness": 0.7408950314947991, "xie_feature_dispersion": 2.6206447748583725, "pac": 0.7732633345340206, "whitened_cosine_id": 0.8052534461021423}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 2, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.29387292376514, "raptor_stability": 0.8994405269622803, "fragility": 0.5, "augmentation_robustness": 0.8304811500465553, "xie_feature_dispersion": 1.2535193562102473, "pac": 0.8649608385044179, "whitened_cosine_id": 0.9212391972541809}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 2, "layer": 13, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237650766, "raptor_stability": 0.9231064319610596, "fragility": 1.0, "augmentation_robustness": 0.9155044674854933, "xie_feature_dispersion": 18.157099012256424, "pac": 0.9193054497232764, "whitened_cosine_id": 0.9482038021087646}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 2, "layer": 14, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647517, "raptor_stability": 0.8107694983482361, "fragility": 0.25, "augmentation_robustness": 0.6126361282502457, "xie_feature_dispersion": 0.29378764734852275, "pac": 0.7117028132992409, "whitened_cosine_id": 0.8085699677467346}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 2, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.47687488644054804, "raptor_stability": 0.8371041417121887, "fragility": 0.25, "augmentation_robustness": 0.8437004456467965, "xie_feature_dispersion": 2.6985372907022853, "pac": 0.8404022936794926, "whitened_cosine_id": 0.8233371376991272}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 2, "layer": 14, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764935, "raptor_stability": 0.8342365026473999, "fragility": 0.25, "augmentation_robustness": 0.7913619262318612, "xie_feature_dispersion": 0.5692187925507834, "pac": 0.8127992144396305, "whitened_cosine_id": 0.8644318580627441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996696, "raptor_stability": 0.8488397598266602, "fragility": 0.25, "augmentation_robustness": 0.5708345486394, "xie_feature_dispersion": 2.9434633106744124, "pac": 0.7098371542330301, "whitened_cosine_id": 0.8510316610336304}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998183, "raptor_stability": 0.8842445015907288, "fragility": 1.0, "augmentation_robustness": 0.8984364327435839, "xie_feature_dispersion": 1.2037141877282427, "pac": 0.8913404671671563, "whitened_cosine_id": 0.8669403195381165}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 3, "layer": 1, "concept": "topic", "predictors": {"sip_eigengap": 0.6283839672205405, "raptor_stability": 0.9187504649162292, "fragility": 1.5, "augmentation_robustness": 0.9445047803748948, "xie_feature_dispersion": 3.4172576932701224, "pac": 0.931627622645562, "whitened_cosine_id": 0.9079165458679199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 3, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14101513130976387, "raptor_stability": 0.9509273767471313, "fragility": 1.5, "augmentation_robustness": 0.8868008425779641, "xie_feature_dispersion": 5.186993220693557, "pac": 0.9188641096625477, "whitened_cosine_id": 0.9339398145675659}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921976405, "raptor_stability": 0.7710468173027039, "fragility": 0.5, "augmentation_robustness": 0.7989766303576111, "xie_feature_dispersion": 0.4743571278941463, "pac": 0.7850117238301575, "whitened_cosine_id": 0.6775692105293274}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 3, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.635597069567996, "raptor_stability": 0.8650970458984375, "fragility": 1.0, "augmentation_robustness": 0.7848846813705812, "xie_feature_dispersion": 1.4784056981569556, "pac": 0.8249908636345094, "whitened_cosine_id": 0.847564160823822}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 3, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.711632992199183, "raptor_stability": 0.8229442834854126, "fragility": 0.5, "augmentation_robustness": 0.7320233370031859, "xie_feature_dispersion": 1.57080462526856, "pac": 0.7774838102442992, "whitened_cosine_id": 0.8052536249160767}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 3, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921988689, "raptor_stability": 0.8289263248443604, "fragility": 1.0, "augmentation_robustness": 0.7538693401418205, "xie_feature_dispersion": 1.8985337702191092, "pac": 0.7913978324930904, "whitened_cosine_id": 0.7646318078041077}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 3, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989895, "raptor_stability": 0.8404061198234558, "fragility": 1.0, "augmentation_robustness": 0.8369988993389674, "xie_feature_dispersion": 1.1920844617447306, "pac": 0.8387025095812116, "whitened_cosine_id": 0.7623035311698914}} +{"model": "pythia-70m", "dataset": "subj", "seed": 3, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002775, "raptor_stability": 0.899202823638916, "fragility": 1.0, "augmentation_robustness": 0.7801516491279672, "xie_feature_dispersion": 2.0439990482699595, "pac": 0.8396772363834416, "whitened_cosine_id": 0.9309227466583252}} +{"model": "pythia-70m", "dataset": "spam", "seed": 3, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004678, "raptor_stability": 0.9567713141441345, "fragility": 3.0, "augmentation_robustness": 0.9371847229705604, "xie_feature_dispersion": 6.424426785927112, "pac": 0.9469780185573475, "whitened_cosine_id": 0.9486836791038513}} +{"model": "pythia-70m", "dataset": "cola", "seed": 3, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921973576, "raptor_stability": 0.7688994407653809, "fragility": 1.0, "augmentation_robustness": 0.7138657324958716, "xie_feature_dispersion": 0.7233065123875068, "pac": 0.7413825866306263, "whitened_cosine_id": 0.6779297590255737}} +{"model": "pythia-70m", "dataset": "stance", "seed": 3, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.764248302629601, "raptor_stability": 0.8229038119316101, "fragility": 0.5, "augmentation_robustness": 0.7822202756663522, "xie_feature_dispersion": 2.080304032697393, "pac": 0.8025620437989811, "whitened_cosine_id": 0.8263651728630066}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 3, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.711632992199799, "raptor_stability": 0.8503726124763489, "fragility": 0.5, "augmentation_robustness": 0.8033091683218101, "xie_feature_dispersion": 1.2407015450743701, "pac": 0.8268408903990795, "whitened_cosine_id": 0.866639256477356}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 3, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349443, "raptor_stability": 0.8484313488006592, "fragility": 0.25, "augmentation_robustness": 0.637225872199634, "xie_feature_dispersion": 4.044131456746566, "pac": 0.7428286105001466, "whitened_cosine_id": 0.8518450260162354}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 3, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350145, "raptor_stability": 0.8780309557914734, "fragility": 1.0, "augmentation_robustness": 0.874329700092187, "xie_feature_dispersion": 1.4259597904886765, "pac": 0.8761803279418302, "whitened_cosine_id": 0.8702338933944702}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 3, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.35191923415776655, "raptor_stability": 0.8508845567703247, "fragility": 0.25, "augmentation_robustness": 0.8265529420495982, "xie_feature_dispersion": 0.9266248016633525, "pac": 0.8387187494099615, "whitened_cosine_id": 0.8987531661987305}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 3, "layer": 9, "concept": "topic", "predictors": {"sip_eigengap": 0.11774124797544638, "raptor_stability": 0.9511698484420776, "fragility": 1.5, "augmentation_robustness": 0.8745928253851561, "xie_feature_dispersion": 6.276849376401518, "pac": 0.9128813369136168, "whitened_cosine_id": 0.917269766330719}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342074, "raptor_stability": 0.78183913230896, "fragility": 0.5, "augmentation_robustness": 0.7622129099393176, "xie_feature_dispersion": 0.5886846655394216, "pac": 0.7720260211241388, "whitened_cosine_id": 0.7248976230621338}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 3, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5119239787691542, "raptor_stability": 0.8616406917572021, "fragility": 0.25, "augmentation_robustness": 0.7504844246649606, "xie_feature_dispersion": 1.6907559164308787, "pac": 0.8060625582110814, "whitened_cosine_id": 0.852891206741333}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 3, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347243, "raptor_stability": 0.8252765536308289, "fragility": 0.25, "augmentation_robustness": 0.7347966668215313, "xie_feature_dispersion": 2.064566431309704, "pac": 0.7800366102261801, "whitened_cosine_id": 0.8165638446807861}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 3, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345138, "raptor_stability": 0.7875639796257019, "fragility": 0.25, "augmentation_robustness": 0.6763370074729932, "xie_feature_dispersion": 3.2860294351071695, "pac": 0.7319504935493475, "whitened_cosine_id": 0.7773585915565491}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 3, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934679, "raptor_stability": 0.7985202670097351, "fragility": 0.25, "augmentation_robustness": 0.7440947615507673, "xie_feature_dispersion": 0.7467058167702149, "pac": 0.7713075142802512, "whitened_cosine_id": 0.7971633076667786}} +{"model": "pythia-160m", "dataset": "subj", "seed": 3, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935211, "raptor_stability": 0.8905059099197388, "fragility": 1.0, "augmentation_robustness": 0.7853861456321554, "xie_feature_dispersion": 2.317051234221485, "pac": 0.8379460277759471, "whitened_cosine_id": 0.913381040096283}} +{"model": "pythia-160m", "dataset": "spam", "seed": 3, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859350205, "raptor_stability": 0.9280425310134888, "fragility": 1.0, "augmentation_robustness": 0.9003377239515604, "xie_feature_dispersion": 13.64857545430944, "pac": 0.9141901274825246, "whitened_cosine_id": 0.9369125366210938}} +{"model": "pythia-160m", "dataset": "cola", "seed": 3, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343062, "raptor_stability": 0.7884514331817627, "fragility": 0.25, "augmentation_robustness": 0.6450635280093344, "xie_feature_dispersion": 0.17857729363649755, "pac": 0.7167574805955486, "whitened_cosine_id": 0.7649105191230774}} +{"model": "pythia-160m", "dataset": "stance", "seed": 3, "layer": 4, "concept": "stance", "predictors": {"sip_eigengap": 0.6506681129814839, "raptor_stability": 0.8327218890190125, "fragility": 0.5, "augmentation_robustness": 0.8189402967776248, "xie_feature_dispersion": 1.7873782154455453, "pac": 0.8258310928983186, "whitened_cosine_id": 0.8258171081542969}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 3, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350412, "raptor_stability": 0.8567770719528198, "fragility": 0.5, "augmentation_robustness": 0.7799380638785056, "xie_feature_dispersion": 1.305808488628754, "pac": 0.8183575679156627, "whitened_cosine_id": 0.8785260319709778}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 3, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881342, "raptor_stability": 0.8563058376312256, "fragility": 0.25, "augmentation_robustness": 0.6228553515841018, "xie_feature_dispersion": 4.914029083398948, "pac": 0.7395805946076637, "whitened_cosine_id": 0.8632848858833313}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 3, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688201, "raptor_stability": 0.8571096062660217, "fragility": 1.0, "augmentation_robustness": 0.8391733545717237, "xie_feature_dispersion": 2.1649660520641487, "pac": 0.8481414804188727, "whitened_cosine_id": 0.8794662356376648}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 3, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.3940956743493873, "raptor_stability": 0.8732040524482727, "fragility": 1.0, "augmentation_robustness": 0.8558702619852715, "xie_feature_dispersion": 3.495736654891566, "pac": 0.8645371572167722, "whitened_cosine_id": 0.894050121307373}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 3, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.0777198899969492, "raptor_stability": 0.9526127576828003, "fragility": 2.0, "augmentation_robustness": 0.8767089502212883, "xie_feature_dispersion": 7.889853603758875, "pac": 0.9146608539520442, "whitened_cosine_id": 0.8949448466300964}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 3, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879801, "raptor_stability": 0.8159152269363403, "fragility": 0.25, "augmentation_robustness": 0.7880748749868189, "xie_feature_dispersion": 0.24343282857388523, "pac": 0.8019950509615796, "whitened_cosine_id": 0.800447404384613}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 3, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.48221468653288235, "raptor_stability": 0.8573822379112244, "fragility": 0.25, "augmentation_robustness": 0.7660417601070145, "xie_feature_dispersion": 1.9953148592463423, "pac": 0.8117119990091195, "whitened_cosine_id": 0.8465678095817566}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 3, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879983, "raptor_stability": 0.8090613484382629, "fragility": 0.25, "augmentation_robustness": 0.7199613185198703, "xie_feature_dispersion": 4.138422068582218, "pac": 0.7645113334790666, "whitened_cosine_id": 0.8162167072296143}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 3, "layer": 9, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956877805, "raptor_stability": 0.8041320443153381, "fragility": 0.25, "augmentation_robustness": 0.6772706695322654, "xie_feature_dispersion": 4.201391999844916, "pac": 0.7407013569238018, "whitened_cosine_id": 0.8077095150947571}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 3, "layer": 13, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880465, "raptor_stability": 0.8054513335227966, "fragility": 0.25, "augmentation_robustness": 0.7134721107475187, "xie_feature_dispersion": 0.942681786130636, "pac": 0.7594617221351576, "whitened_cosine_id": 0.8189021944999695}} +{"model": "pythia-410m", "dataset": "subj", "seed": 3, "layer": 15, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882677, "raptor_stability": 0.908913254737854, "fragility": 1.0, "augmentation_robustness": 0.791121231421306, "xie_feature_dispersion": 1.6016867212241452, "pac": 0.85001724307958, "whitened_cosine_id": 0.9041904807090759}} +{"model": "pythia-410m", "dataset": "spam", "seed": 3, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688292, "raptor_stability": 0.9435619115829468, "fragility": 3.0, "augmentation_robustness": 0.9066470040376822, "xie_feature_dispersion": 6.763845431690352, "pac": 0.9251044578103145, "whitened_cosine_id": 0.9281833171844482}} +{"model": "pythia-410m", "dataset": "cola", "seed": 3, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956878849, "raptor_stability": 0.7881181836128235, "fragility": 0.25, "augmentation_robustness": 0.600597854138349, "xie_feature_dispersion": 0.9458123063554009, "pac": 0.6943580188755862, "whitened_cosine_id": 0.7995412349700928}} +{"model": "pythia-410m", "dataset": "stance", "seed": 3, "layer": 17, "concept": "stance", "predictors": {"sip_eigengap": 0.4936540695216237, "raptor_stability": 0.8370916247367859, "fragility": 0.5, "augmentation_robustness": 0.8332617730055567, "xie_feature_dispersion": 2.485399050185102, "pac": 0.8351766988711713, "whitened_cosine_id": 0.8208258748054504}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 3, "layer": 13, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881398, "raptor_stability": 0.8489888310432434, "fragility": 0.5, "augmentation_robustness": 0.7844172937998297, "xie_feature_dispersion": 1.6180674325869053, "pac": 0.8167030624215366, "whitened_cosine_id": 0.865872859954834}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010287, "raptor_stability": 0.8647460341453552, "fragility": 0.25, "augmentation_robustness": 0.641727222279711, "xie_feature_dispersion": 5.808094646975757, "pac": 0.7532366282125331, "whitened_cosine_id": 0.8289788365364075}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8411266803741455, "fragility": 1.0, "augmentation_robustness": 0.7503016837377494, "xie_feature_dispersion": 2.282144440990373, "pac": 0.7957141820559475, "whitened_cosine_id": 0.8417641520500183}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 3, "layer": 13, "concept": "topic", "predictors": {"sip_eigengap": 0.20340792936838478, "raptor_stability": 0.8698489665985107, "fragility": 1.0, "augmentation_robustness": 0.8437237058947922, "xie_feature_dispersion": 4.342567776713613, "pac": 0.8567863362466515, "whitened_cosine_id": 0.8350374698638916}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 3, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.059803315773717704, "raptor_stability": 0.9504701495170593, "fragility": 2.0, "augmentation_robustness": 0.8680315611599807, "xie_feature_dispersion": 8.620883539713939, "pac": 0.90925085533852, "whitened_cosine_id": 0.8365957140922546}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 3, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010883, "raptor_stability": 0.8169371485710144, "fragility": 0.25, "augmentation_robustness": 0.7710432901435903, "xie_feature_dispersion": 0.42029681121277923, "pac": 0.7939902193573023, "whitened_cosine_id": 0.8196472525596619}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 3, "layer": 11, "concept": "emotion", "predictors": {"sip_eigengap": 0.4267118062158825, "raptor_stability": 0.8550241589546204, "fragility": 0.25, "augmentation_robustness": 0.7405822372807725, "xie_feature_dispersion": 2.564638903560827, "pac": 0.7978031981176965, "whitened_cosine_id": 0.8296079635620117}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 3, "layer": 15, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010767, "raptor_stability": 0.796916663646698, "fragility": 0.25, "augmentation_robustness": 0.6606039027652186, "xie_feature_dispersion": 3.912958891643261, "pac": 0.7287602832059583, "whitened_cosine_id": 0.8257460594177246}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 3, "layer": 18, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010796, "raptor_stability": 0.8012734055519104, "fragility": 0.25, "augmentation_robustness": 0.6623961560942748, "xie_feature_dispersion": 3.563898854267082, "pac": 0.7318347808230926, "whitened_cosine_id": 0.8225507736206055}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 3, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010296, "raptor_stability": 0.7851132154464722, "fragility": 0.25, "augmentation_robustness": 0.6617407773933088, "xie_feature_dispersion": 1.0772966237421127, "pac": 0.7234269964198905, "whitened_cosine_id": 0.8228062987327576}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 3, "layer": 17, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010871, "raptor_stability": 0.8912879228591919, "fragility": 1.0, "augmentation_robustness": 0.74955767566332, "xie_feature_dispersion": 3.0211175775214865, "pac": 0.8204227992612559, "whitened_cosine_id": 0.8278633952140808}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 3, "layer": 3, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010838, "raptor_stability": 0.925938069820404, "fragility": 3.0, "augmentation_robustness": 0.9098532007735074, "xie_feature_dispersion": 10.71642677493898, "pac": 0.9178956352969557, "whitened_cosine_id": 0.8716909289360046}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 3, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010827, "raptor_stability": 0.7985167503356934, "fragility": 0.25, "augmentation_robustness": 0.5412217711508899, "xie_feature_dispersion": 1.0588280189082442, "pac": 0.6698692607432917, "whitened_cosine_id": 0.823022723197937}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 3, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.37531177284512046, "raptor_stability": 0.8395811915397644, "fragility": 1.0, "augmentation_robustness": 0.8246880782030601, "xie_feature_dispersion": 2.787120846289539, "pac": 0.8321346348714123, "whitened_cosine_id": 0.8196287155151367}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 3, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010829, "raptor_stability": 0.8403837084770203, "fragility": 1.0, "augmentation_robustness": 0.7414207905077144, "xie_feature_dispersion": 2.400769158905071, "pac": 0.7909022494923673, "whitened_cosine_id": 0.8367124795913696}} +{"model": "gpt2", "dataset": "sst2", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934796, "raptor_stability": 0.8351662158966064, "fragility": 0.25, "augmentation_robustness": 0.5635262291495843, "xie_feature_dispersion": 4.2622215814473865, "pac": 0.6993462225230953, "whitened_cosine_id": 0.8634011149406433}} +{"model": "gpt2", "dataset": "imdb", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350287, "raptor_stability": 0.8339491486549377, "fragility": 0.25, "augmentation_robustness": 0.7662356686632448, "xie_feature_dispersion": 1.304014068501824, "pac": 0.8000924086590913, "whitened_cosine_id": 0.87335205078125}} +{"model": "gpt2", "dataset": "ag_news", "seed": 3, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.3778683465799438, "raptor_stability": 0.8665527105331421, "fragility": 0.5, "augmentation_robustness": 0.8531902817642327, "xie_feature_dispersion": 2.241047464066251, "pac": 0.8598714961486874, "whitened_cosine_id": 0.9046292304992676}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 3, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.10991110313414279, "raptor_stability": 0.9330034255981445, "fragility": 1.5, "augmentation_robustness": 0.8169811795296109, "xie_feature_dispersion": 5.468890231915903, "pac": 0.8749923025638777, "whitened_cosine_id": 0.9063879251480103}} +{"model": "gpt2", "dataset": "counterfact", "seed": 3, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343766, "raptor_stability": 0.7898539900779724, "fragility": 0.25, "augmentation_robustness": 0.8079306258304046, "xie_feature_dispersion": 0.19395964260888718, "pac": 0.7988923079541885, "whitened_cosine_id": 0.7610896825790405}} +{"model": "gpt2", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6182210878004821, "raptor_stability": 0.8550314903259277, "fragility": 0.25, "augmentation_robustness": 0.7431736186782528, "xie_feature_dispersion": 1.1654170019303518, "pac": 0.7991025545020902, "whitened_cosine_id": 0.8550549149513245}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347225, "raptor_stability": 0.8043380379676819, "fragility": 0.25, "augmentation_robustness": 0.677191562860757, "xie_feature_dispersion": 1.3127141118978531, "pac": 0.7407648004142194, "whitened_cosine_id": 0.8204189538955688}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 3, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934697, "raptor_stability": 0.780476987361908, "fragility": 0.25, "augmentation_robustness": 0.6631271115518341, "xie_feature_dispersion": 3.352185591368954, "pac": 0.7218020494568711, "whitened_cosine_id": 0.8065347075462341}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 3, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346575, "raptor_stability": 0.7473269104957581, "fragility": 0.25, "augmentation_robustness": 0.6123626549338375, "xie_feature_dispersion": 0.7096033674345994, "pac": 0.6798447827147978, "whitened_cosine_id": 0.7918455004692078}} +{"model": "gpt2", "dataset": "subj", "seed": 3, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935243, "raptor_stability": 0.8725587725639343, "fragility": 0.5, "augmentation_robustness": 0.7069137856416525, "xie_feature_dispersion": 1.7904837862778602, "pac": 0.7897362791027934, "whitened_cosine_id": 0.9169808626174927}} +{"model": "gpt2", "dataset": "spam", "seed": 3, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353023, "raptor_stability": 0.9019039273262024, "fragility": 1.5, "augmentation_robustness": 0.8805267705055504, "xie_feature_dispersion": 12.885975509137046, "pac": 0.8912153489158764, "whitened_cosine_id": 0.92914217710495}} +{"model": "gpt2", "dataset": "cola", "seed": 3, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344647, "raptor_stability": 0.796906054019928, "fragility": 0.25, "augmentation_robustness": 0.6909560170275559, "xie_feature_dispersion": 0.4966128694499476, "pac": 0.743931035523742, "whitened_cosine_id": 0.7800789475440979}} +{"model": "gpt2", "dataset": "stance", "seed": 3, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.6661294213646244, "raptor_stability": 0.8302767872810364, "fragility": 0.5, "augmentation_robustness": 0.8014307017054971, "xie_feature_dispersion": 1.9080832844053377, "pac": 0.8158537444932668, "whitened_cosine_id": 0.8170424103736877}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 3, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350134, "raptor_stability": 0.8213337063789368, "fragility": 0.25, "augmentation_robustness": 0.6974710476635542, "xie_feature_dispersion": 1.3078300052773906, "pac": 0.7594023770212455, "whitened_cosine_id": 0.8702872395515442}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295687936, "raptor_stability": 0.8499755263328552, "fragility": 0.25, "augmentation_robustness": 0.6261496459413354, "xie_feature_dispersion": 4.436216156416145, "pac": 0.7380625861370953, "whitened_cosine_id": 0.8625351190567017}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 3, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.2921886382497235, "raptor_stability": 0.8477186560630798, "fragility": 1.0, "augmentation_robustness": 0.8387418545839186, "xie_feature_dispersion": 4.131659598391053, "pac": 0.8432302553234992, "whitened_cosine_id": 0.88265460729599}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 3, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879386, "raptor_stability": 0.7917901277542114, "fragility": 0.25, "augmentation_robustness": 0.789771355720532, "xie_feature_dispersion": 0.17500181221361583, "pac": 0.7907807417373718, "whitened_cosine_id": 0.7891764640808105}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6234881751702148, "raptor_stability": 0.8669646382331848, "fragility": 0.25, "augmentation_robustness": 0.7744639183310552, "xie_feature_dispersion": 0.8652954065662566, "pac": 0.82071427828212, "whitened_cosine_id": 0.8504673838615417}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 3, "layer": 6, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295687904, "raptor_stability": 0.7907098531723022, "fragility": 0.25, "augmentation_robustness": 0.6575825596573526, "xie_feature_dispersion": 3.5458093952463834, "pac": 0.7241462064148274, "whitened_cosine_id": 0.8148408532142639}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 3, "layer": 22, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879737, "raptor_stability": 0.7582568526268005, "fragility": 0.25, "augmentation_robustness": 0.6081859135950609, "xie_feature_dispersion": 0.8063341929658796, "pac": 0.6832213831109307, "whitened_cosine_id": 0.7960034012794495}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 3, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688269, "raptor_stability": 0.8890963792800903, "fragility": 1.0, "augmentation_robustness": 0.7291926988187397, "xie_feature_dispersion": 2.080419773242272, "pac": 0.809144539049415, "whitened_cosine_id": 0.8867965340614319}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 3, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879426, "raptor_stability": 0.7859232425689697, "fragility": 0.25, "augmentation_robustness": 0.5254228710879436, "xie_feature_dispersion": 0.6803530491181083, "pac": 0.6556730568284567, "whitened_cosine_id": 0.8077235221862793}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 3, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5196901382587911, "raptor_stability": 0.8363353610038757, "fragility": 0.5, "augmentation_robustness": 0.8228205239645261, "xie_feature_dispersion": 1.7693304935455851, "pac": 0.8295779424842009, "whitened_cosine_id": 0.7967984080314636}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 3, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881522, "raptor_stability": 0.811364471912384, "fragility": 0.5, "augmentation_robustness": 0.6900060500051179, "xie_feature_dispersion": 1.3697277559493137, "pac": 0.750685260958751, "whitened_cosine_id": 0.8628206849098206}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648796, "raptor_stability": 0.8676397204399109, "fragility": 0.25, "augmentation_robustness": 0.6345770721983662, "xie_feature_dispersion": 5.001598619911871, "pac": 0.7511083963191385, "whitened_cosine_id": 0.8821530938148499}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650275, "raptor_stability": 0.8786996006965637, "fragility": 1.0, "augmentation_robustness": 0.8588046228179724, "xie_feature_dispersion": 2.4507641389974713, "pac": 0.8687521117572681, "whitened_cosine_id": 0.8870015740394592}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 3, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.2775334420918106, "raptor_stability": 0.8491312265396118, "fragility": 1.5, "augmentation_robustness": 0.8575692544379112, "xie_feature_dispersion": 8.067934468405575, "pac": 0.8533502404887615, "whitened_cosine_id": 0.8975652456283569}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 3, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.07983459673425802, "raptor_stability": 0.9590606093406677, "fragility": 1.0, "augmentation_robustness": 0.9129941790024725, "xie_feature_dispersion": 5.747537131789415, "pac": 0.9360273941715701, "whitened_cosine_id": 0.9247861504554749}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 3, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646678, "raptor_stability": 0.840959370136261, "fragility": 0.25, "augmentation_robustness": 0.8017592029882017, "xie_feature_dispersion": 0.07321360576071477, "pac": 0.8213592865622313, "whitened_cosine_id": 0.7885988354682922}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 3, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.6025878875187768, "raptor_stability": 0.8772271275520325, "fragility": 0.25, "augmentation_robustness": 0.865087290926423, "xie_feature_dispersion": 1.9052135627589408, "pac": 0.8711572092392277, "whitened_cosine_id": 0.8513849377632141}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647677, "raptor_stability": 0.8638144731521606, "fragility": 2.0, "augmentation_robustness": 0.8618714559295404, "xie_feature_dispersion": 3.073243341198275, "pac": 0.8628429645408505, "whitened_cosine_id": 0.8141002058982849}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 3, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237638515, "raptor_stability": 0.8214561939239502, "fragility": 0.25, "augmentation_robustness": 0.7518367724678281, "xie_feature_dispersion": 4.329624079200656, "pac": 0.7866464831958891, "whitened_cosine_id": 0.7974085807800293}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 3, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237646187, "raptor_stability": 0.804497480392456, "fragility": 0.25, "augmentation_robustness": 0.7534120265877415, "xie_feature_dispersion": 0.582349253613317, "pac": 0.7789547534900988, "whitened_cosine_id": 0.8017151355743408}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 3, "layer": 15, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651343, "raptor_stability": 0.9011402130126953, "fragility": 0.5, "augmentation_robustness": 0.8098720947850936, "xie_feature_dispersion": 1.5822189245994793, "pac": 0.8555061538988944, "whitened_cosine_id": 0.9193786978721619}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 3, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237650004, "raptor_stability": 0.9187747240066528, "fragility": 1.0, "augmentation_robustness": 0.9072265197438765, "xie_feature_dispersion": 18.77242405320582, "pac": 0.9130006218752647, "whitened_cosine_id": 0.9447034597396851}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 3, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646827, "raptor_stability": 0.8013314008712769, "fragility": 0.25, "augmentation_robustness": 0.5714415857266395, "xie_feature_dispersion": 0.7710445833759078, "pac": 0.6863864932989582, "whitened_cosine_id": 0.8173381686210632}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 3, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.5814899203429595, "raptor_stability": 0.8490622639656067, "fragility": 0.25, "augmentation_robustness": 0.8472747423617989, "xie_feature_dispersion": 2.1435570572895988, "pac": 0.8481685031637027, "whitened_cosine_id": 0.8264880776405334}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 3, "layer": 13, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649207, "raptor_stability": 0.8474602699279785, "fragility": 0.25, "augmentation_robustness": 0.765850633291224, "xie_feature_dispersion": 1.245317692001698, "pac": 0.8066554516096013, "whitened_cosine_id": 0.8748792409896851}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 4, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996045, "raptor_stability": 0.8566904067993164, "fragility": 1.0, "augmentation_robustness": 0.6395377546477263, "xie_feature_dispersion": 1.5854208982087719, "pac": 0.7481140807235214, "whitened_cosine_id": 0.8359927535057068}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 4, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.711632992199886, "raptor_stability": 0.9050891995429993, "fragility": 1.0, "augmentation_robustness": 0.9251332642220863, "xie_feature_dispersion": 1.2126650382704005, "pac": 0.9151112318825427, "whitened_cosine_id": 0.850266695022583}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 4, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5924204103995422, "raptor_stability": 0.8544853329658508, "fragility": 1.0, "augmentation_robustness": 0.8372965090354867, "xie_feature_dispersion": 4.656184075098706, "pac": 0.8458909210006688, "whitened_cosine_id": 0.9082167744636536}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.1519243074719573, "raptor_stability": 0.9499274492263794, "fragility": 1.5, "augmentation_robustness": 0.8830180866818871, "xie_feature_dispersion": 5.282104233636098, "pac": 0.9164727679541332, "whitened_cosine_id": 0.9344291687011719}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 4, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921975906, "raptor_stability": 0.7707024216651917, "fragility": 1.0, "augmentation_robustness": 0.7329255811603577, "xie_feature_dispersion": 0.5102383939416054, "pac": 0.7518140014127747, "whitened_cosine_id": 0.7052313089370728}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 4, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.6400948489658904, "raptor_stability": 0.8515534996986389, "fragility": 0.5, "augmentation_robustness": 0.7338919283811642, "xie_feature_dispersion": 1.307093856011509, "pac": 0.7927227140399016, "whitened_cosine_id": 0.8422322273254395}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 4, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.71163299219924, "raptor_stability": 0.8237500786781311, "fragility": 0.5, "augmentation_robustness": 0.7327459917351864, "xie_feature_dispersion": 2.5240514742338784, "pac": 0.7782480352066588, "whitened_cosine_id": 0.803077220916748}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 4, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987739, "raptor_stability": 0.8263923525810242, "fragility": 1.0, "augmentation_robustness": 0.7109290902303631, "xie_feature_dispersion": 2.0494666093180176, "pac": 0.7686607214056936, "whitened_cosine_id": 0.7474713325500488}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 4, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990436, "raptor_stability": 0.8242563605308533, "fragility": 0.5, "augmentation_robustness": 0.7821271817029374, "xie_feature_dispersion": 1.5098181981749643, "pac": 0.8031917711168953, "whitened_cosine_id": 0.7899174094200134}} +{"model": "pythia-70m", "dataset": "subj", "seed": 4, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002564, "raptor_stability": 0.8824465274810791, "fragility": 1.0, "augmentation_robustness": 0.8066508753882752, "xie_feature_dispersion": 1.925210310106005, "pac": 0.8445487014346771, "whitened_cosine_id": 0.9232692122459412}} +{"model": "pythia-70m", "dataset": "spam", "seed": 4, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004774, "raptor_stability": 0.9476951360702515, "fragility": 3.0, "augmentation_robustness": 0.9265937443188577, "xie_feature_dispersion": 6.387885559509479, "pac": 0.9371444401945546, "whitened_cosine_id": 0.9508467316627502}} +{"model": "pythia-70m", "dataset": "cola", "seed": 4, "layer": 5, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977023, "raptor_stability": 0.7777998447418213, "fragility": 0.25, "augmentation_robustness": 0.6579850681228125, "xie_feature_dispersion": 0.250601434273502, "pac": 0.7178924564323169, "whitened_cosine_id": 0.7004658579826355}} +{"model": "pythia-70m", "dataset": "stance", "seed": 4, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.7333740478467312, "raptor_stability": 0.8212578296661377, "fragility": 1.0, "augmentation_robustness": 0.7504520187465166, "xie_feature_dispersion": 1.7128280566290308, "pac": 0.7858549242063271, "whitened_cosine_id": 0.8275880217552185}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 4, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997524, "raptor_stability": 0.8540244102478027, "fragility": 1.0, "augmentation_robustness": 0.7819111270390232, "xie_feature_dispersion": 1.4981388024521016, "pac": 0.817967768643413, "whitened_cosine_id": 0.8604907989501953}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 4, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349261, "raptor_stability": 0.842431366443634, "fragility": 0.25, "augmentation_robustness": 0.670070029768046, "xie_feature_dispersion": 2.8701281427080465, "pac": 0.75625069810584, "whitened_cosine_id": 0.8559513688087463}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 4, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350782, "raptor_stability": 0.8713650703430176, "fragility": 0.5, "augmentation_robustness": 0.8521992774428392, "xie_feature_dispersion": 1.7964561061229958, "pac": 0.8617821738929283, "whitened_cosine_id": 0.8889275789260864}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.3656440339329693, "raptor_stability": 0.8521515130996704, "fragility": 1.5, "augmentation_robustness": 0.8214046671297884, "xie_feature_dispersion": 6.173757088063873, "pac": 0.8367780901147295, "whitened_cosine_id": 0.9043599367141724}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 4, "layer": 9, "concept": "topic", "predictors": {"sip_eigengap": 0.12756426041857358, "raptor_stability": 0.9445372819900513, "fragility": 1.5, "augmentation_robustness": 0.8622565473270213, "xie_feature_dispersion": 6.343488803218173, "pac": 0.9033969146585363, "whitened_cosine_id": 0.9183123707771301}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 4, "layer": 7, "concept": "truth", "predictors": {"sip_eigengap": 1.397542485934335, "raptor_stability": 0.7996639609336853, "fragility": 0.25, "augmentation_robustness": 0.7613788429632672, "xie_feature_dispersion": 0.5275399816201919, "pac": 0.7805214019484763, "whitened_cosine_id": 0.753498375415802}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.7058740799382045, "raptor_stability": 0.8810455799102783, "fragility": 1.0, "augmentation_robustness": 0.8152160353634006, "xie_feature_dispersion": 2.0356355439733544, "pac": 0.8481308076368395, "whitened_cosine_id": 0.8614587187767029}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 4, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348062, "raptor_stability": 0.8045304417610168, "fragility": 0.5, "augmentation_robustness": 0.7049071379157302, "xie_feature_dispersion": 2.0681692958689024, "pac": 0.7547187898383736, "whitened_cosine_id": 0.8201311230659485}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346577, "raptor_stability": 0.8003606200218201, "fragility": 0.25, "augmentation_robustness": 0.6612670699122942, "xie_feature_dispersion": 2.4099178377906756, "pac": 0.7308138449670571, "whitened_cosine_id": 0.8072301745414734}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 4, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345884, "raptor_stability": 0.8317489624023438, "fragility": 1.0, "augmentation_robustness": 0.772983120611079, "xie_feature_dispersion": 2.3238338563599275, "pac": 0.8023660415067113, "whitened_cosine_id": 0.8050597906112671}} +{"model": "pythia-160m", "dataset": "subj", "seed": 4, "layer": 5, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351835, "raptor_stability": 0.8953569531440735, "fragility": 1.0, "augmentation_robustness": 0.8059484746527689, "xie_feature_dispersion": 1.8870415060935894, "pac": 0.8506527138984212, "whitened_cosine_id": 0.9113974571228027}} +{"model": "pythia-160m", "dataset": "spam", "seed": 4, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352754, "raptor_stability": 0.9212907552719116, "fragility": 1.5, "augmentation_robustness": 0.8624898624867177, "xie_feature_dispersion": 12.624355801662276, "pac": 0.8918903088793146, "whitened_cosine_id": 0.9403319358825684}} +{"model": "pythia-160m", "dataset": "cola", "seed": 4, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934378, "raptor_stability": 0.7925094366073608, "fragility": 0.25, "augmentation_robustness": 0.6106269635092136, "xie_feature_dispersion": 0.1983095485933765, "pac": 0.7015682000582872, "whitened_cosine_id": 0.7796781063079834}} +{"model": "pythia-160m", "dataset": "stance", "seed": 4, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.591622106050008, "raptor_stability": 0.8417823314666748, "fragility": 1.0, "augmentation_robustness": 0.8225818656918379, "xie_feature_dispersion": 1.8092532859980797, "pac": 0.8321820985792563, "whitened_cosine_id": 0.8301091194152832}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 4, "layer": 3, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349856, "raptor_stability": 0.8758466839790344, "fragility": 1.0, "augmentation_robustness": 0.835043699708102, "xie_feature_dispersion": 1.5364396930791746, "pac": 0.8554451918435682, "whitened_cosine_id": 0.8725274801254272}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 4, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880225, "raptor_stability": 0.8483951687812805, "fragility": 0.25, "augmentation_robustness": 0.643421731590692, "xie_feature_dispersion": 3.4588103868670306, "pac": 0.7459084501859863, "whitened_cosine_id": 0.8561255931854248}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 4, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882035, "raptor_stability": 0.8837658166885376, "fragility": 1.0, "augmentation_robustness": 0.8596997806079171, "xie_feature_dispersion": 1.9038436246678319, "pac": 0.8717327986482273, "whitened_cosine_id": 0.894847571849823}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.3547804650921568, "raptor_stability": 0.8753689527511597, "fragility": 1.0, "augmentation_robustness": 0.8555244488391097, "xie_feature_dispersion": 3.4329927120747055, "pac": 0.8654467007951347, "whitened_cosine_id": 0.8951209783554077}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 4, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.10955555398923562, "raptor_stability": 0.9442237615585327, "fragility": 3.0, "augmentation_robustness": 0.8375015283715399, "xie_feature_dispersion": 10.914315920866802, "pac": 0.8908626449650363, "whitened_cosine_id": 0.887626588344574}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 4, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878207, "raptor_stability": 0.7893005013465881, "fragility": 0.25, "augmentation_robustness": 0.7388001024408721, "xie_feature_dispersion": 0.4902099764637905, "pac": 0.7640503018937301, "whitened_cosine_id": 0.7745464444160461}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.5474574770136091, "raptor_stability": 0.8616097569465637, "fragility": 0.25, "augmentation_robustness": 0.7404739041427045, "xie_feature_dispersion": 1.968628707347913, "pac": 0.8010418305446341, "whitened_cosine_id": 0.8528909087181091}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 4, "layer": 15, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295688051, "raptor_stability": 0.8146539330482483, "fragility": 0.25, "augmentation_robustness": 0.7026104422955255, "xie_feature_dispersion": 4.365462087324579, "pac": 0.7586321876718869, "whitened_cosine_id": 0.8268682956695557}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 4, "layer": 15, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878458, "raptor_stability": 0.8015760779380798, "fragility": 0.25, "augmentation_robustness": 0.6575946714155477, "xie_feature_dispersion": 3.8919081093582863, "pac": 0.7295853746768137, "whitened_cosine_id": 0.814882755279541}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 4, "layer": 16, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956878596, "raptor_stability": 0.7986729145050049, "fragility": 0.25, "augmentation_robustness": 0.700139464938641, "xie_feature_dispersion": 3.5880837975111466, "pac": 0.7494061897218229, "whitened_cosine_id": 0.8099539279937744}} +{"model": "pythia-410m", "dataset": "subj", "seed": 4, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882532, "raptor_stability": 0.8692127466201782, "fragility": 0.5, "augmentation_robustness": 0.7900862434288312, "xie_feature_dispersion": 1.4726101086077301, "pac": 0.8296494950245047, "whitened_cosine_id": 0.8998460173606873}} +{"model": "pythia-410m", "dataset": "spam", "seed": 4, "layer": 24, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688298, "raptor_stability": 0.9033204317092896, "fragility": 3.0, "augmentation_robustness": 0.8355633491173182, "xie_feature_dispersion": 12.679622904881825, "pac": 0.8694418904133039, "whitened_cosine_id": 0.9067055583000183}} +{"model": "pythia-410m", "dataset": "cola", "seed": 4, "layer": 17, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880077, "raptor_stability": 0.8151479363441467, "fragility": 0.25, "augmentation_robustness": 0.631437331569998, "xie_feature_dispersion": 0.15645623015703475, "pac": 0.7232926339570724, "whitened_cosine_id": 0.8250035643577576}} +{"model": "pythia-410m", "dataset": "stance", "seed": 4, "layer": 16, "concept": "stance", "predictors": {"sip_eigengap": 0.5699871301215584, "raptor_stability": 0.8456130027770996, "fragility": 0.5, "augmentation_robustness": 0.8244991343863916, "xie_feature_dispersion": 2.7721624801456297, "pac": 0.8350560685817456, "whitened_cosine_id": 0.8285333514213562}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 4, "layer": 11, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881264, "raptor_stability": 0.8462154269218445, "fragility": 0.5, "augmentation_robustness": 0.7743294062255062, "xie_feature_dispersion": 1.1628324109274197, "pac": 0.8102724165736753, "whitened_cosine_id": 0.8581516146659851}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 4, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010715, "raptor_stability": 0.8520708680152893, "fragility": 0.25, "augmentation_robustness": 0.6412965254676084, "xie_feature_dispersion": 4.357957129868121, "pac": 0.7466836967414489, "whitened_cosine_id": 0.8315754532814026}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 4, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010869, "raptor_stability": 0.8517562747001648, "fragility": 0.5, "augmentation_robustness": 0.771865951099836, "xie_feature_dispersion": 2.184545736254185, "pac": 0.8118111129000004, "whitened_cosine_id": 0.8343337774276733}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 4, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.24047302056829442, "raptor_stability": 0.8453668355941772, "fragility": 1.5, "augmentation_robustness": 0.8458315674904345, "xie_feature_dispersion": 5.873911972981273, "pac": 0.845599201542306, "whitened_cosine_id": 0.8292579054832458}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 4, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06567607847107426, "raptor_stability": 0.948825478553772, "fragility": 3.0, "augmentation_robustness": 0.853398753358697, "xie_feature_dispersion": 9.539501225119675, "pac": 0.9011121159562345, "whitened_cosine_id": 0.8352555632591248}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 4, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.7958840727806091, "fragility": 0.25, "augmentation_robustness": 0.7423103383870259, "xie_feature_dispersion": 0.5601797812591205, "pac": 0.7690972055838174, "whitened_cosine_id": 0.8237647414207458}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 4, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.01721014234208262, "raptor_stability": 0.8599122166633606, "fragility": 0.25, "augmentation_robustness": 0.7417171978776979, "xie_feature_dispersion": 2.3728522318398584, "pac": 0.8008147072705292, "whitened_cosine_id": 0.8342825770378113}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 4, "layer": 20, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.8056564331054688, "fragility": 0.25, "augmentation_robustness": 0.6716115591135512, "xie_feature_dispersion": 2.820489046498124, "pac": 0.73863399610951, "whitened_cosine_id": 0.8278409242630005}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010146, "raptor_stability": 0.7977588176727295, "fragility": 0.25, "augmentation_robustness": 0.64006899512257, "xie_feature_dispersion": 4.84100547742313, "pac": 0.7189139063976497, "whitened_cosine_id": 0.8262239098548889}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 4, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010499, "raptor_stability": 0.8048697710037231, "fragility": 0.25, "augmentation_robustness": 0.6973612069523063, "xie_feature_dispersion": 4.496528530562788, "pac": 0.7511154889780147, "whitened_cosine_id": 0.8300395607948303}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 4, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.8694700002670288, "fragility": 1.0, "augmentation_robustness": 0.7901078358260162, "xie_feature_dispersion": 2.50098235434014, "pac": 0.8297889180465226, "whitened_cosine_id": 0.8374378681182861}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 4, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010415, "raptor_stability": 0.9186072945594788, "fragility": 1.0, "augmentation_robustness": 0.8992452095528303, "xie_feature_dispersion": 23.667968510878133, "pac": 0.9089262520561545, "whitened_cosine_id": 0.869879961013794}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 4, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010873, "raptor_stability": 0.8157230615615845, "fragility": 0.25, "augmentation_robustness": 0.5471678545861385, "xie_feature_dispersion": 0.31676967447098947, "pac": 0.6814454580738615, "whitened_cosine_id": 0.8299345970153809}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.3335229215667947, "raptor_stability": 0.824205219745636, "fragility": 0.5, "augmentation_robustness": 0.7978037532307928, "xie_feature_dispersion": 2.0733724889181753, "pac": 0.8110044864882144, "whitened_cosine_id": 0.8242607712745667}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 4, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.855816496101083, "raptor_stability": 0.8407758474349976, "fragility": 1.0, "augmentation_robustness": 0.7768042275797109, "xie_feature_dispersion": 1.8641614757755678, "pac": 0.8087900375073542, "whitened_cosine_id": 0.8235855102539062}} +{"model": "gpt2", "dataset": "sst2", "seed": 4, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934977, "raptor_stability": 0.8240580558776855, "fragility": 0.25, "augmentation_robustness": 0.5955483532445387, "xie_feature_dispersion": 2.990446112484106, "pac": 0.7098032045611121, "whitened_cosine_id": 0.8673197031021118}} +{"model": "gpt2", "dataset": "imdb", "seed": 4, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859351042, "raptor_stability": 0.8240647912025452, "fragility": 0.25, "augmentation_robustness": 0.7103925323300748, "xie_feature_dispersion": 1.6507650137935161, "pac": 0.76722866176631, "whitened_cosine_id": 0.892749547958374}} +{"model": "gpt2", "dataset": "ag_news", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.37928405790056124, "raptor_stability": 0.8549799919128418, "fragility": 0.5, "augmentation_robustness": 0.8421595137240991, "xie_feature_dispersion": 2.2799844708419164, "pac": 0.8485697528184705, "whitened_cosine_id": 0.9051476716995239}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 4, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.15303585153401286, "raptor_stability": 0.9340819120407104, "fragility": 1.5, "augmentation_robustness": 0.8178668816980438, "xie_feature_dispersion": 6.187412713082707, "pac": 0.8759743968693772, "whitened_cosine_id": 0.907498836517334}} +{"model": "gpt2", "dataset": "counterfact", "seed": 4, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341972, "raptor_stability": 0.778740644454956, "fragility": 0.25, "augmentation_robustness": 0.7971651296687343, "xie_feature_dispersion": 0.14506327750165868, "pac": 0.7879528870618452, "whitened_cosine_id": 0.7351542115211487}} +{"model": "gpt2", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5149693378247766, "raptor_stability": 0.8657316565513611, "fragility": 0.25, "augmentation_robustness": 0.7431909261678031, "xie_feature_dispersion": 1.1678519242893954, "pac": 0.8044612913595821, "whitened_cosine_id": 0.8591601848602295}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 4, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347685, "raptor_stability": 0.8219725489616394, "fragility": 0.25, "augmentation_robustness": 0.7928502271629654, "xie_feature_dispersion": 0.8049332910819382, "pac": 0.8074113880623024, "whitened_cosine_id": 0.8097414970397949}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346366, "raptor_stability": 0.7918288111686707, "fragility": 0.25, "augmentation_robustness": 0.6646703017429912, "xie_feature_dispersion": 3.311280523732777, "pac": 0.7282495564558309, "whitened_cosine_id": 0.8104947209358215}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 4, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859344243, "raptor_stability": 0.8132399916648865, "fragility": 0.25, "augmentation_robustness": 0.7322499286526112, "xie_feature_dispersion": 2.8837150977712067, "pac": 0.7727449601587488, "whitened_cosine_id": 0.8200110197067261}} +{"model": "gpt2", "dataset": "subj", "seed": 4, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.39754248593522, "raptor_stability": 0.8405840992927551, "fragility": 0.5, "augmentation_robustness": 0.7275234344331813, "xie_feature_dispersion": 1.7854217365687841, "pac": 0.7840537668629682, "whitened_cosine_id": 0.9135460257530212}} +{"model": "gpt2", "dataset": "spam", "seed": 4, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353234, "raptor_stability": 0.9081653356552124, "fragility": 1.5, "augmentation_robustness": 0.8591723776230531, "xie_feature_dispersion": 12.538518032179876, "pac": 0.8836688566391327, "whitened_cosine_id": 0.9366497993469238}} +{"model": "gpt2", "dataset": "cola", "seed": 4, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344865, "raptor_stability": 0.7840872406959534, "fragility": 0.25, "augmentation_robustness": 0.5048661738238992, "xie_feature_dispersion": 0.16631055680345583, "pac": 0.6444767072599262, "whitened_cosine_id": 0.7860410809516907}} +{"model": "gpt2", "dataset": "stance", "seed": 4, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5528580887739023, "raptor_stability": 0.8414139747619629, "fragility": 0.25, "augmentation_robustness": 0.8322638497812136, "xie_feature_dispersion": 0.9522759018719159, "pac": 0.8368389122715882, "whitened_cosine_id": 0.8301456570625305}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 4, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349983, "raptor_stability": 0.8202458024024963, "fragility": 0.25, "augmentation_robustness": 0.7289728446571213, "xie_feature_dispersion": 0.931895464785709, "pac": 0.7746093235298088, "whitened_cosine_id": 0.8736944198608398}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 4, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881493, "raptor_stability": 0.8379729986190796, "fragility": 0.25, "augmentation_robustness": 0.6139063338760717, "xie_feature_dispersion": 3.178141409790626, "pac": 0.7259396662475757, "whitened_cosine_id": 0.8596592545509338}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 4, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.2485401486377617, "raptor_stability": 0.8904624581336975, "fragility": 0.25, "augmentation_robustness": 0.9007411868297208, "xie_feature_dispersion": 1.0727326936501331, "pac": 0.8956018224817092, "whitened_cosine_id": 0.892510175704956}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878993, "raptor_stability": 0.7903531193733215, "fragility": 0.25, "augmentation_robustness": 0.79632543598189, "xie_feature_dispersion": 0.15365171233992977, "pac": 0.7933392776776058, "whitened_cosine_id": 0.7768749594688416}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5850775560898813, "raptor_stability": 0.8753568530082703, "fragility": 0.25, "augmentation_robustness": 0.7625841583432605, "xie_feature_dispersion": 0.8499569074700448, "pac": 0.8189705056757655, "whitened_cosine_id": 0.8553555011749268}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 4, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880672, "raptor_stability": 0.7783800959587097, "fragility": 0.25, "augmentation_robustness": 0.6250450756800867, "xie_feature_dispersion": 2.9465184575878856, "pac": 0.7017125858193982, "whitened_cosine_id": 0.8258277177810669}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 4, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956877166, "raptor_stability": 0.7933056354522705, "fragility": 0.25, "augmentation_robustness": 0.6933130878602948, "xie_feature_dispersion": 2.7324461419877974, "pac": 0.7433093616562827, "whitened_cosine_id": 0.8186708092689514}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 4, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882608, "raptor_stability": 0.8612777590751648, "fragility": 0.5, "augmentation_robustness": 0.7114055951912336, "xie_feature_dispersion": 1.9179527386531523, "pac": 0.7863416771331992, "whitened_cosine_id": 0.8883224129676819}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956883079, "raptor_stability": 0.9192178249359131, "fragility": 1.0, "augmentation_robustness": 0.894155653371751, "xie_feature_dispersion": 13.303624175400875, "pac": 0.906686739153832, "whitened_cosine_id": 0.9148821830749512}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 4, "layer": 13, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.210307295688004, "raptor_stability": 0.8070831894874573, "fragility": 0.25, "augmentation_robustness": 0.5376584999703106, "xie_feature_dispersion": 0.1971499159632617, "pac": 0.6723708447288839, "whitened_cosine_id": 0.816491961479187}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 4, "layer": 15, "concept": "stance", "predictors": {"sip_eigengap": 0.4591650118237964, "raptor_stability": 0.818568766117096, "fragility": 0.5, "augmentation_robustness": 0.7938712936502795, "xie_feature_dispersion": 1.9692841257546414, "pac": 0.8062200298836877, "whitened_cosine_id": 0.7844472527503967}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 4, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881198, "raptor_stability": 0.812563955783844, "fragility": 0.25, "augmentation_robustness": 0.7084456019576522, "xie_feature_dispersion": 0.9841060606862964, "pac": 0.7605047788707481, "whitened_cosine_id": 0.8456441760063171}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 4, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237646625, "raptor_stability": 0.8596814870834351, "fragility": 0.25, "augmentation_robustness": 0.6898037113021047, "xie_feature_dispersion": 3.3502512176055483, "pac": 0.7747425991927699, "whitened_cosine_id": 0.8781615495681763}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 4, "layer": 20, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650564, "raptor_stability": 0.8608689308166504, "fragility": 0.5, "augmentation_robustness": 0.792066215002771, "xie_feature_dispersion": 2.5567349907892805, "pac": 0.8264675729097106, "whitened_cosine_id": 0.8956750631332397}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 4, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.35362597314605054, "raptor_stability": 0.8578630685806274, "fragility": 1.0, "augmentation_robustness": 0.8552613951921951, "xie_feature_dispersion": 3.846365008247924, "pac": 0.8565622318864112, "whitened_cosine_id": 0.9092753529548645}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 4, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.08672486871948791, "raptor_stability": 0.9325566291809082, "fragility": 3.0, "augmentation_robustness": 0.7997953113506687, "xie_feature_dispersion": 11.842901742955636, "pac": 0.8661759702657885, "whitened_cosine_id": 0.894465446472168}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.293872923764685, "raptor_stability": 0.8403885364532471, "fragility": 0.25, "augmentation_robustness": 0.8397827650909445, "xie_feature_dispersion": 0.14763852549031706, "pac": 0.8400856507720957, "whitened_cosine_id": 0.7999392747879028}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.6393559906046417, "raptor_stability": 0.8851574063301086, "fragility": 0.25, "augmentation_robustness": 0.8701497898578472, "xie_feature_dispersion": 1.6371745621378626, "pac": 0.8776535980939779, "whitened_cosine_id": 0.8543353080749512}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 4, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647368, "raptor_stability": 0.7992756366729736, "fragility": 0.25, "augmentation_robustness": 0.6877395999249556, "xie_feature_dispersion": 4.044282793013871, "pac": 0.7435076182989646, "whitened_cosine_id": 0.8199916481971741}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.293872923764363, "raptor_stability": 0.8176515698432922, "fragility": 0.25, "augmentation_robustness": 0.7400420269154885, "xie_feature_dispersion": 3.976714100479099, "pac": 0.7788467983793903, "whitened_cosine_id": 0.793383777141571}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 4, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647522, "raptor_stability": 0.8136802911758423, "fragility": 0.25, "augmentation_robustness": 0.7463283551572726, "xie_feature_dispersion": 3.432685813095943, "pac": 0.7800043231665574, "whitened_cosine_id": 0.8169722557067871}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 4, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651185, "raptor_stability": 0.886184573173523, "fragility": 0.5, "augmentation_robustness": 0.8108833430892144, "xie_feature_dispersion": 0.9492959086809997, "pac": 0.8485339581313687, "whitened_cosine_id": 0.919140100479126}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923764812, "raptor_stability": 0.9170556664466858, "fragility": 0.5, "augmentation_robustness": 0.9123323230914238, "xie_feature_dispersion": 18.31873452250279, "pac": 0.9146939947690548, "whitened_cosine_id": 0.9489960670471191}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 4, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646698, "raptor_stability": 0.8118268251419067, "fragility": 0.25, "augmentation_robustness": 0.6088240065593917, "xie_feature_dispersion": 0.15540364667380704, "pac": 0.7103254158506492, "whitened_cosine_id": 0.8048716187477112}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.5271400318991657, "raptor_stability": 0.8213255405426025, "fragility": 0.5, "augmentation_robustness": 0.7692249112172099, "xie_feature_dispersion": 2.6328433172283368, "pac": 0.7952752258799063, "whitened_cosine_id": 0.826519787311554}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 4, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649123, "raptor_stability": 0.8499941825866699, "fragility": 0.25, "augmentation_robustness": 0.8088275899418353, "xie_feature_dispersion": 0.7460673348935266, "pac": 0.8294108862642526, "whitened_cosine_id": 0.8643565773963928}} diff --git a/repro_bundle/results_A/results/stats_ranking.jsonl b/repro_bundle/results_A/results/stats_ranking.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..f40c693159227d6224410f1b2a1fc1a2bdf8292d --- /dev/null +++ b/repro_bundle/results_A/results/stats_ranking.jsonl @@ -0,0 +1,7 @@ +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17628948336103584, "p": 0.00011810508174001726, "kendall": 0.11986019460492477, "ci95": [0.0864486183837455, 0.2656858828070762], "loco_mean": 0.06306440054904434, "n": 472}, {"predictor": "pac", "spearman": 0.11790738396920045, "p": 0.010354502513996207, "kendall": 0.07946596854652109, "ci95": [0.03199899524267362, 0.19739551469136], "loco_mean": 0.03027311003896434, "n": 472}, {"predictor": "fragility", "spearman": 0.10714381169126685, "p": 0.019896547808716226, "kendall": 0.0819652122104107, "ci95": [0.02184423176267955, 0.19486119226176976], "loco_mean": -0.08367198887077822, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.05468516346193012, "p": 0.23570006318676667, "kendall": 0.0401513209128409, "ci95": [-0.03209005896781303, 0.1425023272515047], "loco_mean": -0.012901634988281898, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.000509027245282521, "p": 0.9911998467763682, "kendall": 0.0014395350322827368, "ci95": [-0.08621837109724974, 0.09490063671198525], "loco_mean": 0.10342197319983981, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.062397712100437104, "p": 0.17594389124505538, "kendall": -0.04115892744258281, "ci95": [-0.14459614732709314, 0.02266644621275267], "loco_mean": -0.06425062206029349, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.07663916907355152, "p": 0.0963002843104417, "kendall": -0.05037133000022321, "ci95": [-0.16246440320558697, 0.010007628269806309], "loco_mean": -0.09396010191341637, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9449489763429444, "p": 3.6596555462199364e-230, "kendall": 0.7988412681276762, "ci95": [0.9320857918652855, 0.9546405882585741], "loco_mean": 0.714095456119887, "n": 472}, {"predictor": "pac", "spearman": 0.8386192401102935, "p": 4.389790032967171e-126, "kendall": 0.6566807009967972, "ci95": [0.8041701035011758, 0.8686758317302297], "loco_mean": 0.7336833473426215, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7690200093623065, "p": 2.168995457934309e-93, "kendall": 0.5761092518622476, "ci95": [0.7221527504365517, 0.8086111694029999], "loco_mean": 0.17490397542579705, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7108589060252217, "p": 7.570490911324105e-74, "kendall": 0.5244521213429774, "ci95": [0.6647373626280134, 0.7590285725368467], "loco_mean": 0.6466917724774698, "n": 472}, {"predictor": "fragility", "spearman": 0.6402369351573364, "p": 8.412937618018045e-56, "kendall": 0.5176148935000774, "ci95": [0.5762098860164254, 0.6965181301409472], "loco_mean": 0.34843798881555416, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4577071159234756, "p": 8.112397111719808e-26, "kendall": 0.30139623592068804, "ci95": [0.3719011054200997, 0.5362764648442968], "loco_mean": -0.15587895297039037, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.07233275965780582, "p": 0.11656426348042145, "kendall": -0.045983526648003925, "ci95": [-0.1676638264409657, 0.0276888325777201], "loco_mean": 0.27295969447731966, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9663956649462877, "p": 1.9769716421360586e-279, "kendall": 0.844200942819101, "ci95": [0.9582833668806262, 0.972249511018557], "loco_mean": 0.822013981169579, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8142104839116457, "p": 4.6349247787002345e-113, "kendall": 0.6190938860700277, "ci95": [0.7784315483263554, 0.8457827980307192], "loco_mean": 0.2507519638805218, "n": 472}, {"predictor": "pac", "spearman": 0.7989523437605205, "p": 7.370681520368353e-106, "kendall": 0.6131202993990427, "ci95": [0.7548202970707872, 0.8362192541519101], "loco_mean": 0.7092535567689743, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6528079273309736, "p": 1.1636128383807618e-58, "kendall": 0.47211126704811257, "ci95": [0.594467447577123, 0.7075880242965953], "loco_mean": 0.5938518661656708, "n": 472}, {"predictor": "fragility", "spearman": 0.6434621748707247, "p": 1.6002788335451959e-56, "kendall": 0.5162034040583596, "ci95": [0.5777199855374996, 0.6972436181916808], "loco_mean": 0.30639625747257265, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.46410427354367867, "p": 1.377354979546913e-26, "kendall": 0.31332541653172113, "ci95": [0.38047007067527516, 0.5387551616620547], "loco_mean": -0.07487413986761712, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.05687906812360867, "p": 0.21740724040384907, "kendall": -0.03140843113445152, "ci95": [-0.15064310621609026, 0.04683266118594591], "loco_mean": 0.24522656676053, "n": 472}]} +{"target": "excess_rotation", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.045730142358639335, "p": 0.3214915785134481, "kendall": 0.029813955162114502, "ci95": [-0.04291475943890026, 0.13280723679010745], "loco_mean": 0.11571837660547339, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.019548618802250367, "p": 0.6718436438563953, "kendall": 0.01587965653174074, "ci95": [-0.06592842183087877, 0.11021518205534078], "loco_mean": 0.11096384112403068, "n": 472}, {"predictor": "pac", "spearman": -0.028921926535337233, "p": 0.5307837018089083, "kendall": -0.022059088128396125, "ci95": [-0.11461587366461326, 0.05697485566708844], "loco_mean": 0.0566288049921826, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": -0.11740393878758819, "p": 0.010688017745394272, "kendall": -0.08521357371621864, "ci95": [-0.20901249520866028, -0.030594774861803075], "loco_mean": -0.17614889459054547, "n": 472}, {"predictor": "fragility", "spearman": -0.13634890441214695, "p": 0.0029947554168688454, "kendall": -0.0995317208632813, "ci95": [-0.2216850177242833, -0.05109443013239468], "loco_mean": 0.01926842737612711, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.19517738512287972, "p": 1.9527049508461923e-05, "kendall": -0.12812623700025189, "ci95": [-0.28507442645385095, -0.11649292054466626], "loco_mean": -0.11715629043368607, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.24497186857439832, "p": 7.030580997726293e-08, "kendall": -0.1661088920076289, "ci95": [-0.3367444423080759, -0.15504169609386367], "loco_mean": -0.11500811181491921, "n": 472}]} +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17628948336103584, "p": 0.00011810508174001726, "kendall": 0.11986019460492477, "ci95": [0.0864486183837455, 0.2656858828070762], "loco_mean": 0.06306440054904434, "n": 472}, {"predictor": "pac", "spearman": 0.11790738396920045, "p": 0.010354502513996207, "kendall": 0.07946596854652109, "ci95": [0.03199899524267362, 0.19739551469136], "loco_mean": 0.03027311003896434, "n": 472}, {"predictor": "fragility", "spearman": 0.10714381169126685, "p": 0.019896547808716226, "kendall": 0.0819652122104107, "ci95": [0.02184423176267955, 0.19486119226176976], "loco_mean": -0.08367198887077822, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.05468516346193012, "p": 0.23570006318676667, "kendall": 0.0401513209128409, "ci95": [-0.03209005896781303, 0.1425023272515047], "loco_mean": -0.012901634988281898, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.000509027245282521, "p": 0.9911998467763682, "kendall": 0.0014395350322827368, "ci95": [-0.08621837109724974, 0.09490063671198525], "loco_mean": 0.10342197319983981, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.062397712100437104, "p": 0.17594389124505538, "kendall": -0.04115892744258281, "ci95": [-0.14459614732709314, 0.02266644621275267], "loco_mean": -0.06425062206029349, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.07663916907355152, "p": 0.0963002843104417, "kendall": -0.05037133000022321, "ci95": [-0.16246440320558697, 0.010007628269806309], "loco_mean": -0.09396010191341637, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9449489763429444, "p": 3.6596555462199364e-230, "kendall": 0.7988412681276762, "ci95": [0.9320857918652855, 0.9546405882585741], "loco_mean": 0.714095456119887, "n": 472}, {"predictor": "pac", "spearman": 0.8386192401102935, "p": 4.389790032967171e-126, "kendall": 0.6566807009967972, "ci95": [0.8041701035011758, 0.8686758317302297], "loco_mean": 0.7336833473426215, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7690200093623065, "p": 2.168995457934309e-93, "kendall": 0.5761092518622476, "ci95": [0.7221527504365517, 0.8086111694029999], "loco_mean": 0.17490397542579705, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7108589060252217, "p": 7.570490911324105e-74, "kendall": 0.5244521213429774, "ci95": [0.6647373626280134, 0.7590285725368467], "loco_mean": 0.6466917724774698, "n": 472}, {"predictor": "fragility", "spearman": 0.6402369351573364, "p": 8.412937618018045e-56, "kendall": 0.5176148935000774, "ci95": [0.5762098860164254, 0.6965181301409472], "loco_mean": 0.34843798881555416, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4577071159234756, "p": 8.112397111719808e-26, "kendall": 0.30139623592068804, "ci95": [0.3719011054200997, 0.5362764648442968], "loco_mean": -0.15587895297039037, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.07233275965780582, "p": 0.11656426348042145, "kendall": -0.045983526648003925, "ci95": [-0.1676638264409657, 0.0276888325777201], "loco_mean": 0.27295969447731966, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9663956649462877, "p": 1.9769716421360586e-279, "kendall": 0.844200942819101, "ci95": [0.9582833668806262, 0.972249511018557], "loco_mean": 0.822013981169579, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8142104839116457, "p": 4.6349247787002345e-113, "kendall": 0.6190938860700277, "ci95": [0.7784315483263554, 0.8457827980307192], "loco_mean": 0.2507519638805218, "n": 472}, {"predictor": "pac", "spearman": 0.7989523437605205, "p": 7.370681520368353e-106, "kendall": 0.6131202993990427, "ci95": [0.7548202970707872, 0.8362192541519101], "loco_mean": 0.7092535567689743, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6528079273309736, "p": 1.1636128383807618e-58, "kendall": 0.47211126704811257, "ci95": [0.594467447577123, 0.7075880242965953], "loco_mean": 0.5938518661656708, "n": 472}, {"predictor": "fragility", "spearman": 0.6434621748707247, "p": 1.6002788335451959e-56, "kendall": 0.5162034040583596, "ci95": [0.5777199855374996, 0.6972436181916808], "loco_mean": 0.30639625747257265, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.46410427354367867, "p": 1.377354979546913e-26, "kendall": 0.31332541653172113, "ci95": [0.38047007067527516, 0.5387551616620547], "loco_mean": -0.07487413986761712, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.05687906812360867, "p": 0.21740724040384907, "kendall": -0.03140843113445152, "ci95": [-0.15064310621609026, 0.04683266118594591], "loco_mean": 0.24522656676053, "n": 472}]} diff --git a/repro_bundle/results_final/results/audit.jsonl b/repro_bundle/results_final/results/audit.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..d3786c463a8c5eb6e91db212df4e338c35270412 --- /dev/null +++ b/repro_bundle/results_final/results/audit.jsonl @@ -0,0 +1,162 @@ +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": 0.15384615384615385, "n": 208} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 1, "flip_rate": 0.10124999999999995, "n": 800} +{"dataset": "imdb", "seed": 1, "flip_rate": 0.38749999999999996, "n": 800} +{"dataset": "ag_news", "seed": 1, "flip_rate": 0.20750000000000002, "n": 800} +{"dataset": "dbpedia", "seed": 1, "flip_rate": 0.5137499999999999, "n": 800} +{"dataset": "counterfact", "seed": 1, "flip_rate": 0.03249999999999997, "n": 800} +{"dataset": "emotion", "seed": 1, "flip_rate": 0.07625000000000004, "n": 800} +{"dataset": "tweet_hate", "seed": 1, "flip_rate": 0.26, "n": 800} +{"dataset": "tweet_irony", "seed": 1, "flip_rate": 0.23124999999999996, "n": 800} +{"dataset": "tweet_offensive", "seed": 1, "flip_rate": 0.23375, "n": 800} +{"dataset": "subj", "seed": 1, "flip_rate": 0.08250000000000002, "n": 800} +{"dataset": "spam", "seed": 1, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "cola", "seed": 1, "flip_rate": 0.13624999999999998, "n": 800} +{"dataset": "stance", "seed": 1, "flip_rate": 0.1826923076923077, "n": 208} +{"dataset": "amazon_cf", "seed": 1, "flip_rate": 0.11375000000000002, "n": 800} +{"dataset": "sst2", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "imdb", "seed": 2, "flip_rate": 0.35624999999999996, "n": 800} +{"dataset": "ag_news", "seed": 2, "flip_rate": 0.20625000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 2, "flip_rate": 0.5037499999999999, "n": 800} +{"dataset": "counterfact", "seed": 2, "flip_rate": 0.04125000000000001, "n": 800} +{"dataset": "emotion", "seed": 2, "flip_rate": 0.05874999999999997, "n": 800} +{"dataset": "tweet_hate", "seed": 2, "flip_rate": 0.22750000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 2, "flip_rate": 0.22624999999999995, "n": 800} +{"dataset": "tweet_offensive", "seed": 2, "flip_rate": 0.23875000000000002, "n": 800} +{"dataset": "subj", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "spam", "seed": 2, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 2, "flip_rate": 0.10875000000000001, "n": 800} +{"dataset": "stance", "seed": 2, "flip_rate": 0.14903846153846156, "n": 208} +{"dataset": "amazon_cf", "seed": 2, "flip_rate": 0.08750000000000002, "n": 800} +{"dataset": "sst2", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "imdb", "seed": 3, "flip_rate": 0.3425, "n": 800} +{"dataset": "ag_news", "seed": 3, "flip_rate": 0.19874999999999998, "n": 800} +{"dataset": "dbpedia", "seed": 3, "flip_rate": 0.47875, "n": 800} +{"dataset": "counterfact", "seed": 3, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 3, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 3, "flip_rate": 0.25125, "n": 800} +{"dataset": "tweet_irony", "seed": 3, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "tweet_offensive", "seed": 3, "flip_rate": 0.2025, "n": 800} +{"dataset": "subj", "seed": 3, "flip_rate": 0.10375000000000001, "n": 800} +{"dataset": "spam", "seed": 3, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "cola", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "stance", "seed": 3, "flip_rate": 0.16826923076923073, "n": 208} +{"dataset": "amazon_cf", "seed": 3, "flip_rate": 0.09875, "n": 800} +{"dataset": "sst2", "seed": 4, "flip_rate": 0.12124999999999997, "n": 800} +{"dataset": "imdb", "seed": 4, "flip_rate": 0.36250000000000004, "n": 800} +{"dataset": "ag_news", "seed": 4, "flip_rate": 0.17500000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 4, "flip_rate": 0.48624999999999996, "n": 800} +{"dataset": "counterfact", "seed": 4, "flip_rate": 0.03874999999999995, "n": 800} +{"dataset": "emotion", "seed": 4, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 4, "flip_rate": 0.24750000000000005, "n": 800} +{"dataset": "tweet_irony", "seed": 4, "flip_rate": 0.24624999999999997, "n": 800} +{"dataset": "tweet_offensive", "seed": 4, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "subj", "seed": 4, "flip_rate": 0.08125000000000004, "n": 800} +{"dataset": "spam", "seed": 4, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 4, "flip_rate": 0.11499999999999999, "n": 800} +{"dataset": "stance", "seed": 4, "flip_rate": 0.1875, "n": 208} +{"dataset": "amazon_cf", "seed": 4, "flip_rate": 0.13, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": 0.15384615384615385, "n": 208} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 1, "flip_rate": 0.10124999999999995, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": 0.15384615384615385, "n": 208} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 1, "flip_rate": 0.10124999999999995, "n": 800} +{"dataset": "imdb", "seed": 1, "flip_rate": 0.38749999999999996, "n": 800} +{"dataset": "ag_news", "seed": 1, "flip_rate": 0.20750000000000002, "n": 800} +{"dataset": "dbpedia", "seed": 1, "flip_rate": 0.5137499999999999, "n": 800} +{"dataset": "counterfact", "seed": 1, "flip_rate": 0.03249999999999997, "n": 800} +{"dataset": "emotion", "seed": 1, "flip_rate": 0.07625000000000004, "n": 800} +{"dataset": "tweet_hate", "seed": 1, "flip_rate": 0.26, "n": 800} +{"dataset": "tweet_irony", "seed": 1, "flip_rate": 0.23124999999999996, "n": 800} +{"dataset": "tweet_offensive", "seed": 1, "flip_rate": 0.23375, "n": 800} +{"dataset": "subj", "seed": 1, "flip_rate": 0.08250000000000002, "n": 800} +{"dataset": "spam", "seed": 1, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "cola", "seed": 1, "flip_rate": 0.13624999999999998, "n": 800} +{"dataset": "stance", "seed": 1, "flip_rate": 0.1826923076923077, "n": 208} +{"dataset": "amazon_cf", "seed": 1, "flip_rate": 0.11375000000000002, "n": 800} +{"dataset": "sst2", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "imdb", "seed": 2, "flip_rate": 0.35624999999999996, "n": 800} +{"dataset": "ag_news", "seed": 2, "flip_rate": 0.20625000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 2, "flip_rate": 0.5037499999999999, "n": 800} +{"dataset": "counterfact", "seed": 2, "flip_rate": 0.04125000000000001, "n": 800} +{"dataset": "emotion", "seed": 2, "flip_rate": 0.05874999999999997, "n": 800} +{"dataset": "tweet_hate", "seed": 2, "flip_rate": 0.22750000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 2, "flip_rate": 0.22624999999999995, "n": 800} +{"dataset": "tweet_offensive", "seed": 2, "flip_rate": 0.23875000000000002, "n": 800} +{"dataset": "subj", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "spam", "seed": 2, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 2, "flip_rate": 0.10875000000000001, "n": 800} +{"dataset": "stance", "seed": 2, "flip_rate": 0.14903846153846156, "n": 208} +{"dataset": "amazon_cf", "seed": 2, "flip_rate": 0.08750000000000002, "n": 800} +{"dataset": "sst2", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "imdb", "seed": 3, "flip_rate": 0.3425, "n": 800} +{"dataset": "ag_news", "seed": 3, "flip_rate": 0.19874999999999998, "n": 800} +{"dataset": "dbpedia", "seed": 3, "flip_rate": 0.47875, "n": 800} +{"dataset": "counterfact", "seed": 3, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 3, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 3, "flip_rate": 0.25125, "n": 800} +{"dataset": "tweet_irony", "seed": 3, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "tweet_offensive", "seed": 3, "flip_rate": 0.2025, "n": 800} +{"dataset": "subj", "seed": 3, "flip_rate": 0.10375000000000001, "n": 800} +{"dataset": "spam", "seed": 3, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "cola", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "stance", "seed": 3, "flip_rate": 0.16826923076923073, "n": 208} +{"dataset": "amazon_cf", "seed": 3, "flip_rate": 0.09875, "n": 800} +{"dataset": "sst2", "seed": 4, "flip_rate": 0.12124999999999997, "n": 800} +{"dataset": "imdb", "seed": 4, "flip_rate": 0.36250000000000004, "n": 800} +{"dataset": "ag_news", "seed": 4, "flip_rate": 0.17500000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 4, "flip_rate": 0.48624999999999996, "n": 800} +{"dataset": "counterfact", "seed": 4, "flip_rate": 0.03874999999999995, "n": 800} +{"dataset": "emotion", "seed": 4, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 4, "flip_rate": 0.24750000000000005, "n": 800} +{"dataset": "tweet_irony", "seed": 4, "flip_rate": 0.24624999999999997, "n": 800} +{"dataset": "tweet_offensive", "seed": 4, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "subj", "seed": 4, "flip_rate": 0.08125000000000004, "n": 800} +{"dataset": "spam", "seed": 4, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 4, "flip_rate": 0.11499999999999999, "n": 800} +{"dataset": "stance", "seed": 4, "flip_rate": 0.1875, "n": 208} +{"dataset": "amazon_cf", "seed": 4, "flip_rate": 0.13, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} diff --git a/repro_bundle/results_final/results/eval.jsonl b/repro_bundle/results_final/results/eval.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..ed81278f9fce138c9634f27c12c4b40138fefafc --- /dev/null +++ b/repro_bundle/results_final/results/eval.jsonl @@ -0,0 +1,1437 @@ +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.78875, "drop": 0.0}, "paraphrase": {"acc": 0.7645429362880887, "drop": 0.02420706371191128, "rotation": {"subspace_principal_angle": 1.2173428997633269, "mean_class_cosine": 0.34613981140180783}}, "domain": {"acc": 0.66375, "drop": 0.125, "rotation": {"subspace_principal_angle": 1.2327061371099897, "mean_class_cosine": 0.3316860024743622}}}, "iid_acc": 0.78875, "selectivity": 0.29624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.287801427158086, "mean_class_cosine": 0.2792326694699942}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5415512465373962, "drop": -0.005301246537396165, "rotation": {"subspace_principal_angle": 1.4826873586765374, "mean_class_cosine": 0.8742178904437565}}, "domain": {"acc": 0.50125, "drop": 0.03500000000000003, "rotation": {"subspace_principal_angle": 1.4901649913072454, "mean_class_cosine": 0.6563624615789627}}}, "iid_acc": 0.53625, "selectivity": 0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.9335911744705346, "mean_class_cosine": 0.9103338651934845}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.7562326869806094, "drop": 0.017517313019390635}, "domain": {"acc": 0.7725, "drop": 0.0012500000000000844}}, "iid_acc": 0.77375, "selectivity": 0.28125000000000006} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8075, "drop": 0.0}, "paraphrase": {"acc": 0.7965116279069767, "drop": 0.01098837209302328, "rotation": {"subspace_principal_angle": 0.8035537849788047, "mean_class_cosine": 0.6941529859280129}}, "domain": {"acc": 0.6875, "drop": 0.12, "rotation": {"subspace_principal_angle": 1.1686325991529884, "mean_class_cosine": 0.3914103543132651}}, "length": {"acc": 0.8, "drop": 0.007499999999999951, "rotation": {"subspace_principal_angle": 0.7897497491232404, "mean_class_cosine": 0.7040230601402793}}}, "iid_acc": 0.8075, "selectivity": 0.30374999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7674349618290983, "mean_class_cosine": 0.7196939193951748}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6414728682170543, "drop": 0.01602713178294568, "rotation": {"subspace_principal_angle": 1.288879288748167, "mean_class_cosine": 0.8485185221141581}}, "domain": {"acc": 0.6625, "drop": -0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.4368459632860788, "mean_class_cosine": 0.5607042188856695}}, "length": {"acc": 0.6621212121212121, "drop": -0.004621212121212137, "rotation": {"subspace_principal_angle": 1.2230547063204023, "mean_class_cosine": 0.9166318400068196}}}, "iid_acc": 0.6575, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.8046269051289581, "mean_class_cosine": 0.8820641372607203}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.7926356589147286, "drop": -0.005135658914728669}, "domain": {"acc": 0.69625, "drop": 0.09124999999999994}, "length": {"acc": 0.7878787878787878, "drop": -0.00037878787878786735}}, "iid_acc": 0.7875, "selectivity": 0.28374999999999995} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8307692307692308, "drop": 0.015480769230769131, "rotation": {"subspace_principal_angle": 1.1461706135196499, "mean_class_cosine": 0.5254429176325506}}}, "iid_acc": 0.84625, "selectivity": 0.58, "iid_split_rotation": {"subspace_principal_angle": 1.1688799127850784, "mean_class_cosine": 0.5261001194927581}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8, "drop": 0.0}, "paraphrase": {"acc": 0.796923076923077, "drop": 0.003076923076923088, "rotation": {"subspace_principal_angle": 1.3242841060180541, "mean_class_cosine": 0.9641264652586653}}}, "iid_acc": 0.8, "selectivity": 0.5337500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3981863180782648, "mean_class_cosine": 0.9370201686458131}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.8415384615384616, "drop": 0.003461538461538405}}, "iid_acc": 0.845, "selectivity": 0.57875} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.8927680798004988, "drop": 0.04223192019950128, "rotation": {"subspace_principal_angle": 1.2106415683121374, "mean_class_cosine": 0.7066695308643387}}}, "iid_acc": 0.935, "selectivity": 0.8525, "iid_split_rotation": {"subspace_principal_angle": 1.1511700616225666, "mean_class_cosine": 0.7422879247769008}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5311720698254364, "drop": 0.02382793017456364, "rotation": {"subspace_principal_angle": 1.373149046572336, "mean_class_cosine": 0.8538155049730057}}}, "iid_acc": 0.555, "selectivity": 0.47250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3861255031014492, "mean_class_cosine": 0.8730732976173553}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.85785536159601, "drop": 0.03214463840399007}}, "iid_acc": 0.89, "selectivity": 0.8075} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5019607843137255, "drop": 0.01678921568627456, "rotation": {"subspace_principal_angle": 1.5573365448958074, "mean_class_cosine": -0.013459375494237761}}}, "iid_acc": 0.51875, "selectivity": 0.011250000000000093, "iid_split_rotation": {"subspace_principal_angle": 1.5530648223509733, "mean_class_cosine": -0.017730575309227073}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.0015604575163399304, "rotation": {"subspace_principal_angle": 1.5316164005556767, "mean_class_cosine": 0.04651835530761543}}}, "iid_acc": 0.50875, "selectivity": 0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.4740729716187824, "mean_class_cosine": 0.03841895244711871}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.5202614379084968, "drop": -0.010261437908496762}}, "iid_acc": 0.51, "selectivity": 0.0025000000000000577} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.61, "drop": 0.0}, "paraphrase": {"acc": 0.5486018641810919, "drop": 0.06139813581890807, "rotation": {"subspace_principal_angle": 1.4117139037388426, "mean_class_cosine": 0.3559915861697644}}}, "iid_acc": 0.61, "selectivity": 0.37375, "iid_split_rotation": {"subspace_principal_angle": 1.3714632310508308, "mean_class_cosine": 0.37969570224814214}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.275, "drop": 0.0}, "paraphrase": {"acc": 0.2583222370173103, "drop": 0.016677762982689748, "rotation": {"subspace_principal_angle": 1.5666753155926543, "mean_class_cosine": 0.5349177629069118}}}, "iid_acc": 0.275, "selectivity": 0.038750000000000034, "iid_split_rotation": {"subspace_principal_angle": 1.5082299186645924, "mean_class_cosine": 0.6046278432817893}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.507323568575233, "drop": -0.009823568575232977}}, "iid_acc": 0.4975, "selectivity": 0.26125} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.70875, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.0009274193548387544, "rotation": {"subspace_principal_angle": 1.3772445961922097, "mean_class_cosine": 0.19234551093090788}}}, "iid_acc": 0.70875, "selectivity": 0.18625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3584465776147807, "mean_class_cosine": 0.2107574495845495}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.5825, "drop": 0.0}, "paraphrase": {"acc": 0.5942275042444821, "drop": -0.011727504244482101, "rotation": {"subspace_principal_angle": 1.2205673531066674, "mean_class_cosine": 0.9079473240997651}}}, "iid_acc": 0.5825, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1442505653422113, "mean_class_cosine": 0.940086831782734}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.7045840407470289, "drop": 0.007915959252971172}}, "iid_acc": 0.7125, "selectivity": 0.19000000000000006} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.028266129032258003, "rotation": {"subspace_principal_angle": 1.4282418254454519, "mean_class_cosine": 0.142072164699308}}}, "iid_acc": 0.66375, "selectivity": 0.16374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3468259132949723, "mean_class_cosine": 0.2221026092043146}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5838709677419355, "drop": 0.016129032258064502, "rotation": {"subspace_principal_angle": 1.3876894491282326, "mean_class_cosine": 0.9792973864192993}}}, "iid_acc": 0.6, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.0920941984969763, "mean_class_cosine": 0.9685148140397846}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66125, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.019314516129032233}}, "iid_acc": 0.66125, "selectivity": 0.16125} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7600644122383253, "drop": -0.006314412238325295, "rotation": {"subspace_principal_angle": 1.115598956625283, "mean_class_cosine": 0.43963959517656737}}}, "iid_acc": 0.75375, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1918132960864172, "mean_class_cosine": 0.369975854351391}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.007481884057971011, "rotation": {"subspace_principal_angle": 1.431073468315769, "mean_class_cosine": 0.8636351111958729}}}, "iid_acc": 0.52875, "selectivity": -0.1037499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9984226130449826, "mean_class_cosine": 0.8215813059697017}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7342995169082126, "drop": -0.009299516908212624}}, "iid_acc": 0.725, "selectivity": 0.09250000000000003} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.879286694101509, "drop": 0.02321330589849102, "rotation": {"subspace_principal_angle": 0.9139246510391348, "mean_class_cosine": 0.6106425040490346}}}, "iid_acc": 0.9025, "selectivity": 0.44499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.884850806935217, "mean_class_cosine": 0.6334049571887388}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8491083676268861, "drop": 0.012141632373113831, "rotation": {"subspace_principal_angle": 0.783728326709098, "mean_class_cosine": 0.9410916640925551}}}, "iid_acc": 0.86125, "selectivity": 0.40374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.9780082924965329, "mean_class_cosine": 0.9403009956281769}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8710562414266118, "drop": 0.025193758573388236}}, "iid_acc": 0.89625, "selectivity": 0.43875} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9862778730703259, "drop": 0.007472126929674139, "rotation": {"subspace_principal_angle": 0.8999811261795105, "mean_class_cosine": 0.6216247525314251}}}, "iid_acc": 0.99375, "selectivity": 0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.8781549302342212, "mean_class_cosine": 0.6385721258735755}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64375, "drop": 0.0}, "paraphrase": {"acc": 0.4854202401372213, "drop": 0.15832975986277875, "rotation": {"subspace_principal_angle": 0.9223293102070813, "mean_class_cosine": 0.9921756731272653}}}, "iid_acc": 0.64375, "selectivity": -0.2024999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3103002492542979, "mean_class_cosine": 0.9993471541844559}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.97875, "drop": 0.0}, "paraphrase": {"acc": 0.9845626072041166, "drop": -0.005812607204116582}}, "iid_acc": 0.97875, "selectivity": 0.13250000000000006} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6826086956521739, "drop": -0.032608695652173836, "rotation": {"subspace_principal_angle": 1.5134217389057267, "mean_class_cosine": 0.05734311504400581}}}, "iid_acc": 0.65, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4857097607143221, "mean_class_cosine": 0.0849839360330566}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5260869565217391, "drop": 0.0001630434782609047, "rotation": {"subspace_principal_angle": 1.5681344894004097, "mean_class_cosine": 0.0027859134562919843}}}, "iid_acc": 0.52625, "selectivity": -0.10375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3461131864613545, "mean_class_cosine": -0.37599024476007736}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7202898550724638, "drop": -0.014039855072463747}}, "iid_acc": 0.70625, "selectivity": 0.07625000000000004} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6306818181818182, "drop": 0.023164335664335622, "rotation": {"subspace_principal_angle": 1.3319284714979758, "mean_class_cosine": 0.23716017540167048}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.21634615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.372005152667984, "mean_class_cosine": 0.24328801012012044}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.5384615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.5511363636363636, "drop": -0.01267482517482521, "rotation": {"subspace_principal_angle": 1.3661942544847507, "mean_class_cosine": 0.680691975446298}}}, "iid_acc": 0.5384615384615384, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.326965973200558, "mean_class_cosine": 0.5637407887003096}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6363636363636364, "drop": -0.011363636363636354}}, "iid_acc": 0.625, "selectivity": 0.1875} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.8817663817663818, "drop": -0.005516381766381806, "rotation": {"subspace_principal_angle": 1.0257225092803133, "mean_class_cosine": 0.5184812124632806}}}, "iid_acc": 0.87625, "selectivity": 0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.022872498909312, "mean_class_cosine": 0.5209161160438877}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7236467236467237, "drop": -0.0023967236467237196, "rotation": {"subspace_principal_angle": 1.5032004808784296, "mean_class_cosine": 0.9058820512519088}}}, "iid_acc": 0.72125, "selectivity": -0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.3645997436602872, "mean_class_cosine": 0.8668774947492062}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.8148148148148148, "drop": 0.002685185185185235}}, "iid_acc": 0.8175, "selectivity": 0.03375000000000006} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8019390581717452, "drop": 0.021810941828254826, "rotation": {"subspace_principal_angle": 1.2297289568018384, "mean_class_cosine": 0.33449317002319223}}, "domain": {"acc": 0.7775, "drop": 0.04625000000000001, "rotation": {"subspace_principal_angle": 1.2534806736867017, "mean_class_cosine": 0.31201735319379614}}}, "iid_acc": 0.82375, "selectivity": 0.33375, "iid_split_rotation": {"subspace_principal_angle": 1.249974932944242, "mean_class_cosine": 0.31534615054656556}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5526315789473685, "drop": -0.022631578947368447, "rotation": {"subspace_principal_angle": 1.3043937828785666, "mean_class_cosine": 0.9883475306540368}}, "domain": {"acc": 0.50125, "drop": 0.028750000000000053, "rotation": {"subspace_principal_angle": 1.4264081497308534, "mean_class_cosine": 0.495004129731469}}}, "iid_acc": 0.53, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.1840707821435512, "mean_class_cosine": 0.9761804914155537}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8060941828254847, "drop": 0.02015581717451531}, "domain": {"acc": 0.77375, "drop": 0.05249999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33625000000000005} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8275193798449613, "drop": 0.01498062015503876, "rotation": {"subspace_principal_angle": 1.095349444903557, "mean_class_cosine": 0.4577358103130952}}, "domain": {"acc": 0.71125, "drop": 0.13124999999999998, "rotation": {"subspace_principal_angle": 1.266476857896887, "mean_class_cosine": 0.29964398320533164}}, "length": {"acc": 0.843939393939394, "drop": -0.0014393939393939625, "rotation": {"subspace_principal_angle": 1.020996775517062, "mean_class_cosine": 0.5225163309800367}}}, "iid_acc": 0.8425, "selectivity": 0.38125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0185057868759257, "mean_class_cosine": 0.5246385975150215}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6996124031007752, "drop": 0.007887596899224802, "rotation": {"subspace_principal_angle": 1.5092754395213872, "mean_class_cosine": 0.8994925840610262}}, "domain": {"acc": 0.56125, "drop": 0.14625, "rotation": {"subspace_principal_angle": 1.5628669635664512, "mean_class_cosine": 0.41448835148075136}}, "length": {"acc": 0.7257575757575757, "drop": -0.018257575757575695, "rotation": {"subspace_principal_angle": 1.4863663784751628, "mean_class_cosine": 0.9364420792767987}}}, "iid_acc": 0.7075, "selectivity": 0.24625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5618431812325368, "mean_class_cosine": 0.89092989954099}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8352713178294574, "drop": 0.018478682170542604}, "domain": {"acc": 0.71875, "drop": 0.135}, "length": {"acc": 0.853030303030303, "drop": 0.0007196969696969813}}, "iid_acc": 0.85375, "selectivity": 0.3925} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8430769230769231, "drop": -0.0005769230769230305, "rotation": {"subspace_principal_angle": 1.4488633502262789, "mean_class_cosine": 0.2627189029713985}}}, "iid_acc": 0.8425, "selectivity": 0.60375, "iid_split_rotation": {"subspace_principal_angle": 1.3920742480225834, "mean_class_cosine": 0.2804084873444433}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8784615384615385, "drop": -0.019711538461538503, "rotation": {"subspace_principal_angle": 1.5157375083147102, "mean_class_cosine": 0.9772553909019803}}}, "iid_acc": 0.85875, "selectivity": 0.62, "iid_split_rotation": {"subspace_principal_angle": 1.5565345003357298, "mean_class_cosine": 0.9660708595879833}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5030769230769231, "drop": -0.0043269230769230616}}, "iid_acc": 0.49875, "selectivity": 0.26} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.027319201995012476, "rotation": {"subspace_principal_angle": 1.371071144342142, "mean_class_cosine": 0.6983572474879793}}}, "iid_acc": 0.955, "selectivity": 0.87125, "iid_split_rotation": {"subspace_principal_angle": 1.1190162175121645, "mean_class_cosine": 0.7424385333421911}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.46, "drop": 0.0}, "paraphrase": {"acc": 0.4339152119700748, "drop": 0.0260847880299252, "rotation": {"subspace_principal_angle": 1.2702079599204308, "mean_class_cosine": 0.8441489999187578}}}, "iid_acc": 0.46, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.54813616950135, "mean_class_cosine": 0.8639196384527672}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8952618453865336, "drop": 0.025988154613466374}}, "iid_acc": 0.92125, "selectivity": 0.8375} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": 0.003088235294117614, "rotation": {"subspace_principal_angle": 1.5507981147481529, "mean_class_cosine": 0.019996879097623668}}}, "iid_acc": 0.5325, "selectivity": 0.04124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4729086760692658, "mean_class_cosine": 0.09773139915411025}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.51125, "drop": 0.0}, "paraphrase": {"acc": 0.4915032679738562, "drop": 0.0197467320261438, "rotation": {"subspace_principal_angle": 1.5063411676739642, "mean_class_cosine": 0.09376607824598783}}}, "iid_acc": 0.51125, "selectivity": 0.019999999999999962, "iid_split_rotation": {"subspace_principal_angle": 1.3815718454510815, "mean_class_cosine": -0.780613294925026}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.01531045751633986}}, "iid_acc": 0.5225, "selectivity": 0.031249999999999944} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.5872170439414115, "drop": 0.0377829560585885, "rotation": {"subspace_principal_angle": 1.415219489070654, "mean_class_cosine": 0.29709377332489545}}}, "iid_acc": 0.625, "selectivity": 0.36125, "iid_split_rotation": {"subspace_principal_angle": 1.3868652000931787, "mean_class_cosine": 0.30693642876692934}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1975, "drop": 0.0}, "paraphrase": {"acc": 0.20905459387483355, "drop": -0.01155459387483354, "rotation": {"subspace_principal_angle": 1.508152156475103, "mean_class_cosine": 0.3868276136007782}}}, "iid_acc": 0.1975, "selectivity": -0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4103810864475195, "mean_class_cosine": 0.663164345963827}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.5539280958721704, "drop": 0.014821904127829577}}, "iid_acc": 0.56875, "selectivity": 0.305} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7470288624787776, "drop": 0.0017211375212223823, "rotation": {"subspace_principal_angle": 1.218944387247129, "mean_class_cosine": 0.34463687992333025}}}, "iid_acc": 0.74875, "selectivity": 0.21000000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.161541923084499, "mean_class_cosine": 0.39792541536895626}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.6281833616298812, "drop": -0.029433361629881194, "rotation": {"subspace_principal_angle": 1.4828776907746226, "mean_class_cosine": 0.8507212382776503}}}, "iid_acc": 0.59875, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5285287072761573, "mean_class_cosine": 0.889675723801062}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7589134125636672, "drop": -0.007663412563667205}}, "iid_acc": 0.75125, "selectivity": 0.21250000000000002} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6258064516129033, "drop": 0.029193548387096757, "rotation": {"subspace_principal_angle": 1.4380753701349163, "mean_class_cosine": 0.13233165634932476}}}, "iid_acc": 0.655, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.3608689467779782, "mean_class_cosine": 0.20838887470800788}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.6129032258064516, "drop": -0.009153225806451615, "rotation": {"subspace_principal_angle": 1.3501977335705537, "mean_class_cosine": 0.8788630744675302}}}, "iid_acc": 0.60375, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3042174668679, "mean_class_cosine": 0.9016543168707795}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6274193548387097, "drop": 0.01758064516129032}}, "iid_acc": 0.645, "selectivity": 0.13624999999999998} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7584541062801933, "drop": -0.015954106280193225, "rotation": {"subspace_principal_angle": 1.2412464706999458, "mean_class_cosine": 0.3236171403840489}}}, "iid_acc": 0.7425, "selectivity": 0.125, "iid_split_rotation": {"subspace_principal_angle": 1.2363583878771438, "mean_class_cosine": 0.32823830168996404}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5797101449275363, "drop": -0.013460144927536222, "rotation": {"subspace_principal_angle": 1.4328194682708153, "mean_class_cosine": 0.8163073052539073}}}, "iid_acc": 0.56625, "selectivity": -0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.393646227569837, "mean_class_cosine": 0.7747593471553436}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7471819645732689, "drop": -0.00843196457326889}}, "iid_acc": 0.73875, "selectivity": 0.12124999999999997} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.887517146776406, "drop": 0.02623285322359392, "rotation": {"subspace_principal_angle": 1.1698183926314918, "mean_class_cosine": 0.39031889296657585}}}, "iid_acc": 0.91375, "selectivity": 0.3949999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1565211457406315, "mean_class_cosine": 0.4025265308866283}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8847736625514403, "drop": 0.007726337448559661, "rotation": {"subspace_principal_angle": 0.9750908966255823, "mean_class_cosine": 0.724907117214936}}}, "iid_acc": 0.8925, "selectivity": 0.3737499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5507156284967207, "mean_class_cosine": 0.9656213717308835}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9012345679012346, "drop": 0.013765432098765462}}, "iid_acc": 0.915, "selectivity": 0.39625} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.7143429012203775, "mean_class_cosine": 0.7555238835683384}}}, "iid_acc": 0.99125, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6854415650897067, "mean_class_cosine": 0.7741396052879256}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.93625, "drop": 0.0}, "paraphrase": {"acc": 0.9468267581475128, "drop": -0.0105767581475128, "rotation": {"subspace_principal_angle": 1.309840282431262, "mean_class_cosine": 0.963893627422916}}}, "iid_acc": 0.93625, "selectivity": 0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.94741234766098, "mean_class_cosine": 0.9835024086435691}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9775, "drop": 0.0}, "paraphrase": {"acc": 0.9828473413379074, "drop": -0.005347341337907374}}, "iid_acc": 0.9775, "selectivity": 0.11250000000000004} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.006666666666666599, "rotation": {"subspace_principal_angle": 1.5316360507140625, "mean_class_cosine": 0.0391502679566816}}}, "iid_acc": 0.66, "selectivity": 0.07250000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3743529128243746, "mean_class_cosine": 0.19518239053186837}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.47246376811594204, "drop": 0.03753623188405797, "rotation": {"subspace_principal_angle": 1.118894178326917, "mean_class_cosine": -0.9771850067325109}}}, "iid_acc": 0.51, "selectivity": -0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3803133457755477, "mean_class_cosine": 0.946898836596458}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7072463768115942, "drop": -0.0009963768115941463}}, "iid_acc": 0.70625, "selectivity": 0.11875000000000002} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6079545454545454, "drop": -0.0021853146853146876, "rotation": {"subspace_principal_angle": 1.5660532537843979, "mean_class_cosine": 0.23593286698387292}}}, "iid_acc": 0.6057692307692307, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.3587755369509562, "mean_class_cosine": 0.22622489885469554}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6022727272727273, "drop": -0.02054195804195802, "rotation": {"subspace_principal_angle": 1.425664296220229, "mean_class_cosine": 0.6549701520168614}}}, "iid_acc": 0.5817307692307693, "selectivity": 0.1538461538461539, "iid_split_rotation": {"subspace_principal_angle": 1.3242514768412414, "mean_class_cosine": 0.5857958842617208}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.0021853146853146876}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2163461538461539} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8703703703703703, "drop": 0.00462962962962965, "rotation": {"subspace_principal_angle": 1.2653845793904546, "mean_class_cosine": 0.30068589374987864}}}, "iid_acc": 0.875, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 1.1530609002795849, "mean_class_cosine": 0.4056916522664149}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.57875, "drop": 0.0}, "paraphrase": {"acc": 0.5911680911680912, "drop": -0.012418091168091228, "rotation": {"subspace_principal_angle": 1.5387855283082137, "mean_class_cosine": 0.9739607963186667}}}, "iid_acc": 0.57875, "selectivity": -0.17625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.0864399863171719, "mean_class_cosine": 0.8793845387817394}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8475783475783476, "drop": 0.006171652421652407}}, "iid_acc": 0.85375, "selectivity": 0.09875} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8213296398891967, "drop": 0.03117036011080332, "rotation": {"subspace_principal_angle": 1.2728072178283032, "mean_class_cosine": 0.2935985328813959}}, "domain": {"acc": 0.6475, "drop": 0.20500000000000007, "rotation": {"subspace_principal_angle": 1.2548682229869004, "mean_class_cosine": 0.31069877540648505}}}, "iid_acc": 0.8525, "selectivity": 0.35625, "iid_split_rotation": {"subspace_principal_angle": 1.1836995794199188, "mean_class_cosine": 0.37750157196014655}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5318559556786704, "drop": -0.006855955678670367, "rotation": {"subspace_principal_angle": 1.183174660406376, "mean_class_cosine": 0.9978979106979409}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.4296731020342406, "mean_class_cosine": 0.41513571955567125}}}, "iid_acc": 0.525, "selectivity": 0.028749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9398355095136322, "mean_class_cosine": 0.9889864954707369}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8421052631578947, "drop": 0.004144736842105257}, "domain": {"acc": 0.8225, "drop": 0.023749999999999938}}, "iid_acc": 0.84625, "selectivity": 0.3499999999999999} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8604651162790697, "drop": 0.024534883720930267, "rotation": {"subspace_principal_angle": 1.1635939988143396, "mean_class_cosine": 0.396041968116701}}, "domain": {"acc": 0.56, "drop": 0.32499999999999996, "rotation": {"subspace_principal_angle": 1.3073928416159974, "mean_class_cosine": 0.26036815053777257}}, "length": {"acc": 0.8878787878787879, "drop": -0.002878787878787925, "rotation": {"subspace_principal_angle": 1.1306543361974652, "mean_class_cosine": 0.42606792700100266}}}, "iid_acc": 0.885, "selectivity": 0.41375, "iid_split_rotation": {"subspace_principal_angle": 1.1314507236110818, "mean_class_cosine": 0.42534730729927617}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7383720930232558, "drop": 0.03412790697674417, "rotation": {"subspace_principal_angle": 0.9299498041850935, "mean_class_cosine": 0.9017395970359101}}, "domain": {"acc": 0.5575, "drop": 0.21499999999999997, "rotation": {"subspace_principal_angle": 1.4111384197278427, "mean_class_cosine": 0.27301695590829955}}, "length": {"acc": 0.7878787878787878, "drop": -0.01537878787878788, "rotation": {"subspace_principal_angle": 1.0571900651122168, "mean_class_cosine": 0.9405678373193577}}}, "iid_acc": 0.7725, "selectivity": 0.30124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.339591955524043, "mean_class_cosine": 0.8972887886451133}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8662790697674418, "drop": 0.022470930232558195}, "domain": {"acc": 0.65, "drop": 0.23875000000000002}, "length": {"acc": 0.8878787878787879, "drop": 0.000871212121212106}}, "iid_acc": 0.88875, "selectivity": 0.41750000000000004} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.88, "drop": -0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.4381874155007441, "mean_class_cosine": 0.2360974231087299}}}, "iid_acc": 0.875, "selectivity": 0.5974999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3889398949790899, "mean_class_cosine": 0.28986888876348327}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8123076923076923, "drop": 0.013942307692307754, "rotation": {"subspace_principal_angle": 1.4726936861262263, "mean_class_cosine": 0.9825256512058247}}}, "iid_acc": 0.82625, "selectivity": 0.5487500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.413087776634539, "mean_class_cosine": 0.9710401150655834}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8861538461538462, "drop": 0.005096153846153806}}, "iid_acc": 0.89125, "selectivity": 0.61375} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.9725, "drop": 0.0}, "paraphrase": {"acc": 0.9551122194513716, "drop": 0.01738778054862844, "rotation": {"subspace_principal_angle": 1.1805912863774084, "mean_class_cosine": 0.696959257792652}}}, "iid_acc": 0.9725, "selectivity": 0.875, "iid_split_rotation": {"subspace_principal_angle": 1.1819768188421824, "mean_class_cosine": 0.7412159284602584}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.4275, "drop": 0.0}, "paraphrase": {"acc": 0.42643391521197005, "drop": 0.0010660847880299418, "rotation": {"subspace_principal_angle": 1.267210976332211, "mean_class_cosine": 0.8469328938765297}}}, "iid_acc": 0.4275, "selectivity": 0.32999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2708278894045024, "mean_class_cosine": 0.8352289736541864}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.0210692019950125}}, "iid_acc": 0.94875, "selectivity": 0.85125} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5568627450980392, "drop": -0.006862745098039191, "rotation": {"subspace_principal_angle": 1.4681160564274516, "mean_class_cosine": 0.10249993503953822}}}, "iid_acc": 0.55, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.5020669180006716, "mean_class_cosine": 0.06867531169240668}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.4875816993464052, "drop": 0.00991830065359478, "rotation": {"subspace_principal_angle": 1.0583311490868843, "mean_class_cosine": 0.6025164388499016}}}, "iid_acc": 0.4975, "selectivity": -0.03250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2056755940549575, "mean_class_cosine": -0.8743607002802327}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5372549019607843, "drop": 0.0002450980392156854}}, "iid_acc": 0.5375, "selectivity": 0.007499999999999951} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6604527296937417, "drop": -0.005452729693741665, "rotation": {"subspace_principal_angle": 1.3690495088359922, "mean_class_cosine": 0.35780435447931885}}}, "iid_acc": 0.655, "selectivity": 0.39375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.418659878889111, "mean_class_cosine": 0.31040393556774787}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.18125, "drop": 0.0}, "paraphrase": {"acc": 0.1904127829560586, "drop": -0.009162782956058602, "rotation": {"subspace_principal_angle": 1.4238306284342992, "mean_class_cosine": 0.3336979016265693}}}, "iid_acc": 0.18125, "selectivity": -0.07999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3516589964423749, "mean_class_cosine": 0.6675023436977741}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5712383488681758, "drop": 0.030011651131824135}}, "iid_acc": 0.60125, "selectivity": 0.33999999999999997} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.00039685908319186947, "rotation": {"subspace_principal_angle": 1.3803752925053967, "mean_class_cosine": 0.18927233568020757}}}, "iid_acc": 0.72875, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 1.4107099285898823, "mean_class_cosine": 0.1594035006847374}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5229202037351444, "drop": -0.015420203735144411, "rotation": {"subspace_principal_angle": 1.132903800026888, "mean_class_cosine": 0.9957211105004683}}}, "iid_acc": 0.5075, "selectivity": -0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1567819921044111, "mean_class_cosine": 0.997306798497698}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7606112054329371, "drop": -0.0018612054329371075}}, "iid_acc": 0.75875, "selectivity": 0.22750000000000004} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.03556451612903222, "rotation": {"subspace_principal_angle": 1.4500907567958619, "mean_class_cosine": 0.12041267342093487}}}, "iid_acc": 0.6775, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4451716043887228, "mean_class_cosine": 0.12529455713007304}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5435483870967742, "drop": 0.002701612903225792, "rotation": {"subspace_principal_angle": 1.0905650289278745, "mean_class_cosine": 0.9919303109580393}}}, "iid_acc": 0.54625, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.435319911955165, "mean_class_cosine": 0.9822586493642307}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.02806451612903227}}, "iid_acc": 0.67, "selectivity": 0.16375000000000006} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7568438003220612, "drop": 0.014406199677938769, "rotation": {"subspace_principal_angle": 1.362823497882489, "mean_class_cosine": 0.2064768368861471}}}, "iid_acc": 0.77125, "selectivity": 0.21124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3556181289971392, "mean_class_cosine": 0.21352151995440388}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.01748188405797102, "rotation": {"subspace_principal_angle": 1.162087777749045, "mean_class_cosine": 0.4004265677277434}}}, "iid_acc": 0.51875, "selectivity": -0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5040593331311518, "mean_class_cosine": 0.6240661876988235}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.775, "drop": 0.0}, "paraphrase": {"acc": 0.7729468599033816, "drop": 0.002053140096618389}}, "iid_acc": 0.775, "selectivity": 0.21499999999999997} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9245541838134431, "drop": 0.017945816186556884, "rotation": {"subspace_principal_angle": 1.1580571429217417, "mean_class_cosine": 0.401119992002977}}}, "iid_acc": 0.9425, "selectivity": 0.44875, "iid_split_rotation": {"subspace_principal_angle": 1.0639249835392377, "mean_class_cosine": 0.48544434408974335}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5116598079561042, "drop": 0.0033401920438957955, "rotation": {"subspace_principal_angle": 1.474754823264982, "mean_class_cosine": 0.27562171118558787}}}, "iid_acc": 0.515, "selectivity": 0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6855036886417566, "mean_class_cosine": 0.9705005903895466}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.93, "drop": 0.0}, "paraphrase": {"acc": 0.934156378600823, "drop": -0.004156378600822963}}, "iid_acc": 0.93, "selectivity": 0.43625} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0006389365351628928, "rotation": {"subspace_principal_angle": 0.7922854168062247, "mean_class_cosine": 0.7022200256213631}}}, "iid_acc": 0.9925, "selectivity": 0.13250000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.7419982107629325, "mean_class_cosine": 0.7371197159686783}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9502572898799314, "drop": -0.00900728987993138, "rotation": {"subspace_principal_angle": 1.2962420205278622, "mean_class_cosine": 0.9560615750258359}}}, "iid_acc": 0.94125, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4821593641247295, "mean_class_cosine": 0.9814265656310696}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9875, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": -0.0022084048027444236}}, "iid_acc": 0.9875, "selectivity": 0.12750000000000006} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6608695652173913, "drop": 0.03413043478260869, "rotation": {"subspace_principal_angle": 1.5269251960916086, "mean_class_cosine": 0.04385705910466185}}}, "iid_acc": 0.695, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.452899790541391, "mean_class_cosine": 0.117623607019381}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.041141304347826146, "rotation": {"subspace_principal_angle": 1.2698364171990568, "mean_class_cosine": -0.7441747656964055}}}, "iid_acc": 0.52375, "selectivity": -0.06624999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.421349910061863, "mean_class_cosine": 0.7671339569758747}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7231884057971014, "drop": -0.009438405797101423}}, "iid_acc": 0.71375, "selectivity": 0.12375000000000003} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6826923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.6363636363636364, "drop": 0.046328671328671356, "rotation": {"subspace_principal_angle": 1.389965032090286, "mean_class_cosine": 0.19817848562152038}}}, "iid_acc": 0.6826923076923077, "selectivity": 0.25480769230769235, "iid_split_rotation": {"subspace_principal_angle": 1.5631105240216077, "mean_class_cosine": 0.18163197534662465}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6105769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.625, "drop": -0.014423076923076872, "rotation": {"subspace_principal_angle": 1.3818731643580437, "mean_class_cosine": 0.6330079224250492}}}, "iid_acc": 0.6105769230769231, "selectivity": 0.18269230769230776, "iid_split_rotation": {"subspace_principal_angle": 1.188799274694121, "mean_class_cosine": 0.5596495884045671}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.006993006993006978}}, "iid_acc": 0.6490384615384616, "selectivity": 0.2211538461538462} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8803418803418803, "drop": 0.0009081196581196549, "rotation": {"subspace_principal_angle": 1.336870712814003, "mean_class_cosine": 0.23179799551578698}}}, "iid_acc": 0.88125, "selectivity": 0.15749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.230199866592587, "mean_class_cosine": 0.3340493484245235}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.5811965811965812, "drop": -0.004946581196581201, "rotation": {"subspace_principal_angle": 0.6698879155775275, "mean_class_cosine": 0.9725719155523136}}}, "iid_acc": 0.57625, "selectivity": -0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9538385113441675, "mean_class_cosine": 0.9585511263219829}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8974358974358975, "drop": 0.0013141025641025816}}, "iid_acc": 0.89875, "selectivity": 0.17500000000000004} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8337950138504155, "drop": 0.027454986149584415, "rotation": {"subspace_principal_angle": 1.2672102146950088, "mean_class_cosine": 0.2989442428389649}}, "domain": {"acc": 0.88, "drop": -0.018750000000000044, "rotation": {"subspace_principal_angle": 1.401321553710003, "mean_class_cosine": 0.1686646701084955}}}, "iid_acc": 0.86125, "selectivity": 0.36749999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.2210418787662296, "mean_class_cosine": 0.3426671322986696}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5346260387811634, "drop": -0.009626038781163415, "rotation": {"subspace_principal_angle": 1.1075097075912679, "mean_class_cosine": 0.9893879557921825}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.2937330372237155, "mean_class_cosine": 0.27468238367063336}}}, "iid_acc": 0.525, "selectivity": 0.03125, "iid_split_rotation": {"subspace_principal_angle": 1.474523039419839, "mean_class_cosine": 0.9693041836230525}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8476454293628809, "drop": 0.04110457063711914}, "domain": {"acc": 0.86875, "drop": 0.020000000000000018}}, "iid_acc": 0.88875, "selectivity": 0.395} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.875968992248062, "drop": 0.012781007751938023, "rotation": {"subspace_principal_angle": 1.3733967390372437, "mean_class_cosine": 0.1961200848251179}}, "domain": {"acc": 0.5525, "drop": 0.33625000000000005, "rotation": {"subspace_principal_angle": 1.427574987743908, "mean_class_cosine": 0.14273220656435498}}, "length": {"acc": 0.8924242424242425, "drop": -0.003674242424242413, "rotation": {"subspace_principal_angle": 1.32660559111955, "mean_class_cosine": 0.24177114792102647}}}, "iid_acc": 0.88875, "selectivity": 0.3975, "iid_split_rotation": {"subspace_principal_angle": 1.3721751230402406, "mean_class_cosine": 0.19731783026248}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.84875, "drop": 0.0}, "paraphrase": {"acc": 0.8197674418604651, "drop": 0.028982558139534875, "rotation": {"subspace_principal_angle": 1.0890107964995421, "mean_class_cosine": 0.9365459180728041}}, "domain": {"acc": 0.5575, "drop": 0.29125, "rotation": {"subspace_principal_angle": 1.2433163855943603, "mean_class_cosine": 0.290612600127489}}, "length": {"acc": 0.8575757575757575, "drop": -0.008825757575757542, "rotation": {"subspace_principal_angle": 0.7717021170570981, "mean_class_cosine": 0.9578281621778026}}}, "iid_acc": 0.84875, "selectivity": 0.3575, "iid_split_rotation": {"subspace_principal_angle": 1.2334456221930967, "mean_class_cosine": 0.9323952105718878}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8875968992248062, "drop": 0.00990310077519374}, "domain": {"acc": 0.695, "drop": 0.2025}, "length": {"acc": 0.9015151515151515, "drop": -0.004015151515151527}}, "iid_acc": 0.8975, "selectivity": 0.40624999999999994} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8907692307692308, "drop": -0.00951923076923078, "rotation": {"subspace_principal_angle": 1.3982668079683969, "mean_class_cosine": 0.2412385692456759}}}, "iid_acc": 0.88125, "selectivity": 0.63, "iid_split_rotation": {"subspace_principal_angle": 1.3873607135770687, "mean_class_cosine": 0.27893385434758605}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.805, "drop": 0.0}, "paraphrase": {"acc": 0.8046153846153846, "drop": 0.00038461538461542766, "rotation": {"subspace_principal_angle": 1.1223434771672867, "mean_class_cosine": 0.9797187391070092}}}, "iid_acc": 0.805, "selectivity": 0.5537500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1951168916205284, "mean_class_cosine": 0.9636873787333077}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.9092307692307692, "drop": -0.010480769230769127}}, "iid_acc": 0.89875, "selectivity": 0.6475000000000001} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.972568578553616, "drop": 0.0024314214463839745, "rotation": {"subspace_principal_angle": 1.0887911120828875, "mean_class_cosine": 0.6783458118132375}}}, "iid_acc": 0.975, "selectivity": 0.8624999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.083853464584136, "mean_class_cosine": 0.7060030624663257}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.8125, "drop": 0.0}, "paraphrase": {"acc": 0.773067331670823, "drop": 0.03943266832917702, "rotation": {"subspace_principal_angle": 1.1185256458497195, "mean_class_cosine": 0.8994165755464111}}}, "iid_acc": 0.8125, "selectivity": 0.7, "iid_split_rotation": {"subspace_principal_angle": 0.9631112899466643, "mean_class_cosine": 0.9391589025650812}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9650872817955112, "drop": 0.011162718204488775}}, "iid_acc": 0.97625, "selectivity": 0.8637499999999999} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.5830065359477125, "drop": 0.024493464052287583, "rotation": {"subspace_principal_angle": 1.4835452240288083, "mean_class_cosine": 0.08714044135255376}}}, "iid_acc": 0.6075, "selectivity": 0.08750000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.483433918298054, "mean_class_cosine": 0.08725132314037118}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.4888888888888889, "drop": 0.01611111111111113, "rotation": {"subspace_principal_angle": 1.2048851781158827, "mean_class_cosine": 0.5321986735325194}}}, "iid_acc": 0.505, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4614710386057692, "mean_class_cosine": -0.6438340541689455}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5542483660130719, "drop": 0.03450163398692807}}, "iid_acc": 0.58875, "selectivity": 0.06874999999999998} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65875, "drop": 0.0}, "paraphrase": {"acc": 0.6085219707057257, "drop": 0.05022802929427428, "rotation": {"subspace_principal_angle": 1.4274964103041523, "mean_class_cosine": 0.2819535108490442}}}, "iid_acc": 0.65875, "selectivity": 0.3837499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3942837316412728, "mean_class_cosine": 0.30680585311258146}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.18, "drop": 0.0}, "paraphrase": {"acc": 0.19440745672436752, "drop": -0.014407456724367523, "rotation": {"subspace_principal_angle": 1.42729065004506, "mean_class_cosine": 0.3233324825686333}}}, "iid_acc": 0.18, "selectivity": -0.09500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.543456747904511, "mean_class_cosine": 0.6716440420685439}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02558089214380832}}, "iid_acc": 0.64875, "selectivity": 0.37375} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.005396859083191874, "rotation": {"subspace_principal_angle": 1.4458284299287356, "mean_class_cosine": 0.1246428806664818}}}, "iid_acc": 0.73375, "selectivity": 0.23625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4100208755322998, "mean_class_cosine": 0.16008370527990506}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5314091680814941, "drop": -0.012659168081494032, "rotation": {"subspace_principal_angle": 1.5017224503953317, "mean_class_cosine": 0.9922645891758275}}}, "iid_acc": 0.51875, "selectivity": 0.021250000000000047, "iid_split_rotation": {"subspace_principal_angle": 0.5510785135387823, "mean_class_cosine": 0.994402608644392}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7504244482173175, "drop": 0.020825551782682528}}, "iid_acc": 0.77125, "selectivity": 0.27375} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.6532258064516129, "drop": 0.031774193548387175, "rotation": {"subspace_principal_angle": 1.4862916929900118, "mean_class_cosine": 0.08440409464427018}}}, "iid_acc": 0.685, "selectivity": 0.18375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.504979130468674, "mean_class_cosine": 0.06576968766218992}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5564516129032258, "drop": 0.009798387096774275, "rotation": {"subspace_principal_angle": 1.4670137647528678, "mean_class_cosine": 0.9493974522411648}}}, "iid_acc": 0.56625, "selectivity": 0.06500000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2038765745555904, "mean_class_cosine": 0.9198294280970227}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.6935483870967742, "drop": 0.01895161290322578}}, "iid_acc": 0.7125, "selectivity": 0.21125000000000005} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7439613526570048, "drop": 0.0047886473429952225, "rotation": {"subspace_principal_angle": 1.4293803795970716, "mean_class_cosine": 0.14094506785538105}}}, "iid_acc": 0.74875, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4308280823099806, "mean_class_cosine": 0.13951166975487922}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.571658615136876, "drop": -0.019158615136876, "rotation": {"subspace_principal_angle": 1.3467680040328998, "mean_class_cosine": 0.780244054310545}}}, "iid_acc": 0.5525, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.5305723014222863, "mean_class_cosine": 0.8225959610699823}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.7616747181964574, "drop": 0.01207528180354267}}, "iid_acc": 0.77375, "selectivity": 0.19625000000000004} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.02258058984910838, "rotation": {"subspace_principal_angle": 1.1992451710106438, "mean_class_cosine": 0.3630611813014619}}}, "iid_acc": 0.95125, "selectivity": 0.44125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.197550530346397, "mean_class_cosine": 0.3646396665958699}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6598079561042524, "drop": -0.0023079561042523844, "rotation": {"subspace_principal_angle": 0.8946495725521172, "mean_class_cosine": 0.6299993713824044}}}, "iid_acc": 0.6575, "selectivity": 0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7999262041956853, "mean_class_cosine": 0.9697684559622158}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.945, "drop": 0.0}, "paraphrase": {"acc": 0.9314128943758574, "drop": 0.01358710562414256}}, "iid_acc": 0.945, "selectivity": 0.43499999999999994} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99875, "drop": 0.0}, "paraphrase": {"acc": 0.9982847341337907, "drop": 0.000465265866209319, "rotation": {"subspace_principal_angle": 0.931771543682588, "mean_class_cosine": 0.5964129401770092}}}, "iid_acc": 0.99875, "selectivity": 0.18125000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9266228730293125, "mean_class_cosine": 0.600537740322992}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6375, "drop": 0.0}, "paraphrase": {"acc": 0.4065180102915952, "drop": 0.23098198970840478, "rotation": {"subspace_principal_angle": 0.4804883746103172, "mean_class_cosine": 0.9773138541986853}}}, "iid_acc": 0.6375, "selectivity": -0.18000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0322313143428026, "mean_class_cosine": 0.9999323022857498}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9948542024013722, "drop": 0.001395797598627735}}, "iid_acc": 0.99625, "selectivity": 0.17874999999999996} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.6884057971014492, "drop": 0.04284420289855073, "rotation": {"subspace_principal_angle": 1.5103706788720144, "mean_class_cosine": 0.060388883021034304}}}, "iid_acc": 0.73125, "selectivity": 0.14374999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.4200020283035577, "mean_class_cosine": 0.1502234646254889}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.4666666666666667, "drop": 0.04083333333333328, "rotation": {"subspace_principal_angle": 0.9965627610068227, "mean_class_cosine": -0.9475923907212384}}}, "iid_acc": 0.5075, "selectivity": -0.08000000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.7465288923899847, "mean_class_cosine": 0.963542850231548}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7275362318840579, "drop": 0.039963768115942044}}, "iid_acc": 0.7675, "selectivity": 0.17999999999999994} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6647727272727273, "drop": 0.008304195804195835, "rotation": {"subspace_principal_angle": 1.4238317227425443, "mean_class_cosine": 0.16606688749587134}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.18269230769230776, "iid_split_rotation": {"subspace_principal_angle": 1.4493637206220726, "mean_class_cosine": 0.18022085560237225}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.4423076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.4090909090909091, "drop": 0.033216783216783174, "rotation": {"subspace_principal_angle": 1.314437097913778, "mean_class_cosine": 0.6492731252980364}}}, "iid_acc": 0.4423076923076923, "selectivity": -0.04807692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.0082539944993898, "mean_class_cosine": 0.81084397247279}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6761363636363636, "drop": -0.0030594405594405183}}, "iid_acc": 0.6730769230769231, "selectivity": 0.18269230769230776} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8874643874643875, "drop": 0.005035612535612466, "rotation": {"subspace_principal_angle": 1.373498225543769, "mean_class_cosine": 0.19602056818834496}}}, "iid_acc": 0.8925, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.272801169025973, "mean_class_cosine": 0.29360431510047513}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.5883190883190883, "drop": 0.0016809116809116675, "rotation": {"subspace_principal_angle": 1.0911846628602964, "mean_class_cosine": 0.9555356213618866}}}, "iid_acc": 0.59, "selectivity": -0.125, "iid_split_rotation": {"subspace_principal_angle": 1.1224788303280901, "mean_class_cosine": 0.9089747436544128}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8846153846153846, "drop": 0.01163461538461541}}, "iid_acc": 0.89625, "selectivity": 0.18125000000000002} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7853185595567868, "drop": 0.01843144044321321, "rotation": {"subspace_principal_angle": 1.3395977046860834, "mean_class_cosine": 0.2291444175563393}}, "domain": {"acc": 0.78875, "drop": 0.015000000000000013, "rotation": {"subspace_principal_angle": 1.3075865369139061, "mean_class_cosine": 0.2601811310213691}}}, "iid_acc": 0.80375, "selectivity": 0.31124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3161987442877714, "mean_class_cosine": 0.25185598358320005}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5373961218836565, "drop": -0.017396121883656468, "rotation": {"subspace_principal_angle": 1.1243174241253076, "mean_class_cosine": 0.997655368526527}}, "domain": {"acc": 0.50125, "drop": 0.018750000000000044, "rotation": {"subspace_principal_angle": 1.434291363492328, "mean_class_cosine": 0.5041835597720421}}}, "iid_acc": 0.52, "selectivity": 0.027500000000000024, "iid_split_rotation": {"subspace_principal_angle": 1.4733549164259618, "mean_class_cosine": 0.9931712323643861}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.775623268698061, "drop": 0.050626731301939065}, "domain": {"acc": 0.7475, "drop": 0.07874999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33375000000000005} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8081395348837209, "drop": 0.01561046511627906, "rotation": {"subspace_principal_angle": 1.4878379029744278, "mean_class_cosine": 0.08286330186240415}}, "domain": {"acc": 0.63875, "drop": 0.18499999999999994, "rotation": {"subspace_principal_angle": 1.454691067906504, "mean_class_cosine": 0.11584457649805602}}, "length": {"acc": 0.8227272727272728, "drop": 0.0010227272727272307, "rotation": {"subspace_principal_angle": 1.4554754679619462, "mean_class_cosine": 0.11506542197274067}}}, "iid_acc": 0.82375, "selectivity": 0.3475, "iid_split_rotation": {"subspace_principal_angle": 1.3842160089393434, "mean_class_cosine": 0.18549965489482007}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7325581395348837, "drop": 0.03494186046511627, "rotation": {"subspace_principal_angle": 1.2562269465379385, "mean_class_cosine": 0.9025478152861728}}, "domain": {"acc": 0.5575, "drop": 0.20999999999999996, "rotation": {"subspace_principal_angle": 1.3251714868629496, "mean_class_cosine": 0.29596572601612736}}, "length": {"acc": 0.7727272727272727, "drop": -0.005227272727272747, "rotation": {"subspace_principal_angle": 1.0637760788494128, "mean_class_cosine": 0.9397971855033075}}}, "iid_acc": 0.7675, "selectivity": 0.29124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.061070800271238, "mean_class_cosine": 0.9127773418749959}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.8725, "drop": 0.0}, "paraphrase": {"acc": 0.8527131782945736, "drop": 0.01978682170542645}, "domain": {"acc": 0.57875, "drop": 0.29375000000000007}, "length": {"acc": 0.8712121212121212, "drop": 0.0012878787878788378}}, "iid_acc": 0.8725, "selectivity": 0.39625000000000005} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8661538461538462, "drop": 0.0013461538461538858, "rotation": {"subspace_principal_angle": 1.2824688472167018, "mean_class_cosine": 0.4528696194346733}}}, "iid_acc": 0.8675, "selectivity": 0.64, "iid_split_rotation": {"subspace_principal_angle": 1.267228970027625, "mean_class_cosine": 0.4341122200684181}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.47875, "drop": 0.0}, "paraphrase": {"acc": 0.42615384615384616, "drop": 0.05259615384615385, "rotation": {"subspace_principal_angle": 0.5775569724987617, "mean_class_cosine": 0.9093320332006872}}}, "iid_acc": 0.47875, "selectivity": 0.25125, "iid_split_rotation": {"subspace_principal_angle": 0.7727337369433303, "mean_class_cosine": 0.8668228792038879}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.0036538461538461187}}, "iid_acc": 0.8775, "selectivity": 0.6499999999999999} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.95375, "drop": 0.0}, "paraphrase": {"acc": 0.9451371571072319, "drop": 0.008612842892768091, "rotation": {"subspace_principal_angle": 1.198927059140121, "mean_class_cosine": 0.6054966764735671}}}, "iid_acc": 0.95375, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.1388755195423386, "mean_class_cosine": 0.6482459135365446}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6084788029925187, "drop": 0.01402119700748139, "rotation": {"subspace_principal_angle": 1.493246923874091, "mean_class_cosine": 0.8677677280952006}}}, "iid_acc": 0.6225, "selectivity": 0.5375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5091321723919393, "mean_class_cosine": 0.8880006055054167}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.9351620947630923, "drop": 0.012337905236907698}}, "iid_acc": 0.9475, "selectivity": 0.8625} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.495, "drop": 0.0}, "paraphrase": {"acc": 0.4745098039215686, "drop": 0.02049019607843139, "rotation": {"subspace_principal_angle": 1.557032384565418, "mean_class_cosine": -0.013763507646052914}}}, "iid_acc": 0.495, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.5359063301369187, "mean_class_cosine": -0.03488291842098544}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.0017892156862744923, "rotation": {"subspace_principal_angle": 1.3709234192517532, "mean_class_cosine": 0.6267242100586605}}}, "iid_acc": 0.49625, "selectivity": -0.014999999999999958, "iid_split_rotation": {"subspace_principal_angle": 1.3896477426172604, "mean_class_cosine": -0.3304774205182823}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.4825, "drop": 0.0}, "paraphrase": {"acc": 0.47320261437908495, "drop": 0.009297385620915033}}, "iid_acc": 0.4825, "selectivity": -0.028749999999999998} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5805592543275633, "drop": 0.02319074567243673, "rotation": {"subspace_principal_angle": 1.3989325795834617, "mean_class_cosine": 0.365872980266819}}}, "iid_acc": 0.60375, "selectivity": 0.335, "iid_split_rotation": {"subspace_principal_angle": 1.4035880336880546, "mean_class_cosine": 0.3488005843140671}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.1775, "drop": 0.0}, "paraphrase": {"acc": 0.16511318242343542, "drop": 0.012386817576564568, "rotation": {"subspace_principal_angle": 1.5619855127249644, "mean_class_cosine": 0.3341308312243725}}}, "iid_acc": 0.1775, "selectivity": -0.09125, "iid_split_rotation": {"subspace_principal_angle": 1.3905000233838531, "mean_class_cosine": 0.6611663679873362}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5778961384820239, "drop": 0.022103861517976053}}, "iid_acc": 0.6, "selectivity": 0.33125} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.015927419354838768, "rotation": {"subspace_principal_angle": 1.5028161180968338, "mean_class_cosine": 0.06792786120606348}}}, "iid_acc": 0.69375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4722418499371575, "mean_class_cosine": 0.09839501129737169}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.533106960950764, "drop": -0.020606960950764086, "rotation": {"subspace_principal_angle": 1.024515247435546, "mean_class_cosine": 0.9955533421541893}}}, "iid_acc": 0.5125, "selectivity": -0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.2562415527805546, "mean_class_cosine": 0.9954953689417243}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7555178268251274, "drop": 0.00448217317487265}}, "iid_acc": 0.76, "selectivity": 0.24624999999999997} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.6338709677419355, "drop": 0.017379032258064475, "rotation": {"subspace_principal_angle": 1.4673214972757795, "mean_class_cosine": 0.10329027681844688}}}, "iid_acc": 0.65125, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.5077800700720831, "mean_class_cosine": 0.06297455823321907}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.535483870967742, "drop": 0.005766129032258038, "rotation": {"subspace_principal_angle": 1.440925356190957, "mean_class_cosine": 0.9961212841602167}}}, "iid_acc": 0.54125, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3284144435763523, "mean_class_cosine": 0.9886091312692893}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6209677419354839, "drop": 0.03403225806451615}}, "iid_acc": 0.655, "selectivity": 0.17000000000000004} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.750402576489533, "drop": -0.009152576489533049, "rotation": {"subspace_principal_angle": 1.4211901787085208, "mean_class_cosine": 0.14904869151211453}}}, "iid_acc": 0.74125, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.4965103140884382, "mean_class_cosine": 0.07421770808908856}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5265700483091788, "drop": 0.03967995169082128, "rotation": {"subspace_principal_angle": 1.5565645508048995, "mean_class_cosine": 0.00760410778674959}}}, "iid_acc": 0.56625, "selectivity": 0.0025000000000000577, "iid_split_rotation": {"subspace_principal_angle": 0.9594487263253761, "mean_class_cosine": -0.4124908154959152}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7536231884057971, "drop": -0.0011231884057971708}}, "iid_acc": 0.7525, "selectivity": 0.18874999999999997} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8998628257887518, "drop": 0.02138717421124825, "rotation": {"subspace_principal_angle": 1.2563517152731103, "mean_class_cosine": 0.30928836211834454}}}, "iid_acc": 0.92125, "selectivity": 0.43375, "iid_split_rotation": {"subspace_principal_angle": 1.1864007137524348, "mean_class_cosine": 0.3749989233038349}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.545, "drop": 0.0}, "paraphrase": {"acc": 0.5569272976680384, "drop": -0.01192729766803835, "rotation": {"subspace_principal_angle": 0.6735562930427222, "mean_class_cosine": 0.3885547331126563}}}, "iid_acc": 0.545, "selectivity": 0.05750000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.3659301793739206, "mean_class_cosine": 0.9497848325707625}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.9039780521262003, "drop": 0.009771947873799647}}, "iid_acc": 0.91375, "selectivity": 0.42624999999999996} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9965694682675815, "drop": -0.0015694682675815308, "rotation": {"subspace_principal_angle": 0.9204115550682747, "mean_class_cosine": 0.605492671467479}}}, "iid_acc": 0.995, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 0.964600661915144, "mean_class_cosine": 0.5697451063410308}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7753001715265866, "drop": -0.00780017152658663, "rotation": {"subspace_principal_angle": 1.5561305399812533, "mean_class_cosine": 0.9880733179203292}}}, "iid_acc": 0.7675, "selectivity": -0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.8575713187579257, "mean_class_cosine": 0.9958656438305629}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": 0.0006110634648370805}}, "iid_acc": 0.99375, "selectivity": 0.19125000000000003} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6478260869565218, "drop": 0.0096739130434782, "rotation": {"subspace_principal_angle": 1.5203803161832818, "mean_class_cosine": 0.050394655640553065}}}, "iid_acc": 0.6575, "selectivity": 0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.4388032016199581, "mean_class_cosine": 0.13161019079523278}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4623188405797101, "drop": 0.040181159420289825, "rotation": {"subspace_principal_angle": 1.5601182686089141, "mean_class_cosine": -0.9599175649315286}}}, "iid_acc": 0.5025, "selectivity": -0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3056048703674055, "mean_class_cosine": 0.9664617957303832}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.6956521739130435, "drop": 0.018097826086956537}}, "iid_acc": 0.71375, "selectivity": 0.15500000000000003} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6477272727272727, "drop": -0.008304195804195835, "rotation": {"subspace_principal_angle": 1.4577549081283396, "mean_class_cosine": 0.13478050883475684}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.24999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.5438885641755995, "mean_class_cosine": 0.12661575417990478}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.40865384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.3693181818181818, "drop": 0.03933566433566432, "rotation": {"subspace_principal_angle": 1.3024704772930913, "mean_class_cosine": 0.751592937345538}}}, "iid_acc": 0.40865384615384615, "selectivity": 0.019230769230769218, "iid_split_rotation": {"subspace_principal_angle": 1.0672377902275059, "mean_class_cosine": 0.7053655504479837}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.0021853146853146876}}, "iid_acc": 0.6442307692307693, "selectivity": 0.25480769230769235} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8831908831908832, "drop": 0.0030591168091167464, "rotation": {"subspace_principal_angle": 1.4868944039444256, "mean_class_cosine": 0.08380351910501391}}}, "iid_acc": 0.88625, "selectivity": 0.20999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3010782735174253, "mean_class_cosine": 0.26645969404516406}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.575, "drop": 0.0}, "paraphrase": {"acc": 0.5797720797720798, "drop": -0.004772079772079829, "rotation": {"subspace_principal_angle": 1.1062280403413047, "mean_class_cosine": 0.9715489933239789}}}, "iid_acc": 0.575, "selectivity": -0.10125000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.5004862855814347, "mean_class_cosine": 0.9429161811360605}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.9017094017094017, "drop": -0.0017094017094017033}}, "iid_acc": 0.9, "selectivity": 0.22375} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.8448753462603878, "drop": 0.010124653739612133, "rotation": {"subspace_principal_angle": 1.2793309538353963, "mean_class_cosine": 0.28735610225096786}}, "domain": {"acc": 0.835, "drop": 0.020000000000000018, "rotation": {"subspace_principal_angle": 1.4170648249450182, "mean_class_cosine": 0.15312668462263485}}}, "iid_acc": 0.855, "selectivity": 0.36374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2898741682012294, "mean_class_cosine": 0.27724177646677317}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5373961218836565, "drop": -0.017396121883656468, "rotation": {"subspace_principal_angle": 1.5469965306207487, "mean_class_cosine": 0.995777039241573}}, "domain": {"acc": 0.50125, "drop": 0.018750000000000044, "rotation": {"subspace_principal_angle": 1.4241429560997638, "mean_class_cosine": 0.3829223526550779}}}, "iid_acc": 0.52, "selectivity": 0.028749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9408750735467649, "mean_class_cosine": 0.9869080660276348}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8365650969529086, "drop": 0.04593490304709136}, "domain": {"acc": 0.68125, "drop": 0.20124999999999993}}, "iid_acc": 0.8825, "selectivity": 0.39124999999999993} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5241830065359477, "drop": -0.009183006535947724, "rotation": {"subspace_principal_angle": 1.534491721385843, "mean_class_cosine": 0.03629663087546754}}}, "iid_acc": 0.515, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.5355851741198854, "mean_class_cosine": 0.03520387717988177}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5215686274509804, "drop": -0.006568627450980413, "rotation": {"subspace_principal_angle": 1.4962498880827226, "mean_class_cosine": -0.003140814081629747}}}, "iid_acc": 0.515, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.3928150255004763, "mean_class_cosine": -0.22397488753911812}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.5098039215686274, "drop": -0.0035539215686274384}}, "iid_acc": 0.50625, "selectivity": 0.0012499999999999734} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6625, "drop": 0.0}, "paraphrase": {"acc": 0.6351531291611185, "drop": 0.027346870838881432, "rotation": {"subspace_principal_angle": 1.3792866566009878, "mean_class_cosine": 0.32365198757370595}}}, "iid_acc": 0.6625, "selectivity": 0.41625, "iid_split_rotation": {"subspace_principal_angle": 1.3919845995237112, "mean_class_cosine": 0.34481299862959075}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.30625, "drop": 0.0}, "paraphrase": {"acc": 0.2969374167776298, "drop": 0.009312583222370219, "rotation": {"subspace_principal_angle": 1.1990223569073335, "mean_class_cosine": 0.6137390921956974}}}, "iid_acc": 0.30625, "selectivity": 0.060000000000000026, "iid_split_rotation": {"subspace_principal_angle": 1.4316896651831825, "mean_class_cosine": 0.6279344942694856}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.64375, "drop": 0.0}, "paraphrase": {"acc": 0.6378162450066578, "drop": 0.005933754993342255}}, "iid_acc": 0.64375, "selectivity": 0.3975000000000001} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.7028862478777589, "drop": -0.012886247877758938, "rotation": {"subspace_principal_angle": 1.4427027917966888, "mean_class_cosine": 0.12774353013272075}}}, "iid_acc": 0.69, "selectivity": 0.1612499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3294573165238053, "mean_class_cosine": 0.23900304446924894}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5517826825127334, "drop": -0.0017826825127333912, "rotation": {"subspace_principal_angle": 0.8282909656023721, "mean_class_cosine": 0.9050477286017182}}}, "iid_acc": 0.55, "selectivity": 0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5117065619049865, "mean_class_cosine": 0.9515453007674062}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7623089983022071, "drop": -0.008558998302207077}}, "iid_acc": 0.75375, "selectivity": 0.22499999999999998} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5887096774193549, "drop": 0.015040322580645138, "rotation": {"subspace_principal_angle": 1.5014698349556506, "mean_class_cosine": 0.06927097278538774}}}, "iid_acc": 0.60375, "selectivity": 0.08875, "iid_split_rotation": {"subspace_principal_angle": 1.505561040213484, "mean_class_cosine": 0.0651890267481754}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5419354838709678, "drop": -0.0006854838709677846, "rotation": {"subspace_principal_angle": 1.2893002371450315, "mean_class_cosine": 0.9894024792146818}}}, "iid_acc": 0.54125, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.295158193254292, "mean_class_cosine": 0.9686636169079809}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6675, "drop": 0.0}, "paraphrase": {"acc": 0.6580645161290323, "drop": 0.009435483870967709}}, "iid_acc": 0.6675, "selectivity": 0.15249999999999997} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7053140096618358, "drop": 0.019685990338164205, "rotation": {"subspace_principal_angle": 1.48218154327834, "mean_class_cosine": 0.08849885326614075}}}, "iid_acc": 0.725, "selectivity": 0.18874999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4683469244564828, "mean_class_cosine": 0.10227028026027415}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.5942028985507246, "drop": 0.0007971014492753614, "rotation": {"subspace_principal_angle": 1.4158562736410767, "mean_class_cosine": 0.12335819756367886}}}, "iid_acc": 0.595, "selectivity": 0.05874999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6688260501385498, "mean_class_cosine": 0.15014143544853242}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7568438003220612, "drop": 0.015656199677938742}}, "iid_acc": 0.7725, "selectivity": 0.23624999999999996} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9259259259259259, "drop": 0.016574074074074074, "rotation": {"subspace_principal_angle": 1.157452401071877, "mean_class_cosine": 0.4016738775889428}}}, "iid_acc": 0.9425, "selectivity": 0.42000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.104624499684144, "mean_class_cosine": 0.4494698976663737}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.585, "drop": 0.0}, "paraphrase": {"acc": 0.5624142661179699, "drop": 0.022585733882030112, "rotation": {"subspace_principal_angle": 1.2005636922035423, "mean_class_cosine": 0.8139615746809903}}}, "iid_acc": 0.585, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 0.549247413832069, "mean_class_cosine": 0.30809866212710807}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9272976680384087, "drop": 0.003952331961591282}}, "iid_acc": 0.93125, "selectivity": 0.40875000000000006} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.6623188405797101, "drop": 0.0339311594202899, "rotation": {"subspace_principal_angle": 1.496893033147877, "mean_class_cosine": 0.07383603911862277}}}, "iid_acc": 0.69625, "selectivity": 0.11625000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4214587189180432, "mean_class_cosine": 0.1487831455638894}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.4797101449275362, "drop": 0.03028985507246379, "rotation": {"subspace_principal_angle": 1.1094284833982664, "mean_class_cosine": -0.7834732078378617}}}, "iid_acc": 0.51, "selectivity": -0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2652565110598484, "mean_class_cosine": 0.8601341288733604}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.6420289855072464, "drop": 0.050471014492753596}}, "iid_acc": 0.6925, "selectivity": 0.11250000000000004} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6704545454545454, "drop": -0.02622377622377614, "rotation": {"subspace_principal_angle": 1.4296664682007603, "mean_class_cosine": 0.16574943658385013}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2163461538461539, "iid_split_rotation": {"subspace_principal_angle": 1.551374852609326, "mean_class_cosine": 0.11936510654614986}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.4230769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.4147727272727273, "drop": 0.00830419580419578, "rotation": {"subspace_principal_angle": 1.3518141128300416, "mean_class_cosine": 0.666584729764562}}}, "iid_acc": 0.4230769230769231, "selectivity": -0.004807692307692291, "iid_split_rotation": {"subspace_principal_angle": 1.4575770760056468, "mean_class_cosine": 0.7420250665679694}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6761363636363636, "drop": -0.00786713286713292}}, "iid_acc": 0.6682692307692307, "selectivity": 0.24038461538461536} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8846153846153846, "drop": 0.0066346153846154055, "rotation": {"subspace_principal_angle": 1.4268427703533795, "mean_class_cosine": 0.1434568887054392}}}, "iid_acc": 0.89125, "selectivity": 0.20499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.368747840402915, "mean_class_cosine": 0.20067656560132463}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5826210826210826, "drop": -0.0026210826210826266, "rotation": {"subspace_principal_angle": 0.9151023657985915, "mean_class_cosine": 0.9645849728606082}}}, "iid_acc": 0.58, "selectivity": -0.10625000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.47399007231295404, "mean_class_cosine": 0.9356570924866734}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.9002849002849003, "drop": 0.0009650997150997265}}, "iid_acc": 0.90125, "selectivity": 0.21499999999999997} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8490304709141274, "drop": 0.024719529085872605, "rotation": {"subspace_principal_angle": 1.176281709220393, "mean_class_cosine": 0.384360137185048}}, "domain": {"acc": 0.73625, "drop": 0.13750000000000007, "rotation": {"subspace_principal_angle": 1.1970353549367978, "mean_class_cosine": 0.36511932310217704}}}, "iid_acc": 0.87375, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.0786643668776719, "mean_class_cosine": 0.4725059154778732}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5304709141274239, "drop": -0.004220914127423869, "rotation": {"subspace_principal_angle": 1.3152812819817528, "mean_class_cosine": 0.9992665145477913}}, "domain": {"acc": 0.50125, "drop": 0.025000000000000022, "rotation": {"subspace_principal_angle": 1.308682431264261, "mean_class_cosine": 0.5566033754658513}}}, "iid_acc": 0.52625, "selectivity": 0.04249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.391200206971303, "mean_class_cosine": 0.9964068683023172}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8130193905817175, "drop": 0.018230609418282584}, "domain": {"acc": 0.82875, "drop": 0.0025000000000000577}}, "iid_acc": 0.83125, "selectivity": 0.34750000000000003} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8643410852713178, "drop": 0.023158914728682145, "rotation": {"subspace_principal_angle": 1.1410222217173775, "mean_class_cosine": 0.4166654610472472}}, "domain": {"acc": 0.815, "drop": 0.07250000000000001, "rotation": {"subspace_principal_angle": 1.250009657687724, "mean_class_cosine": 0.3153131973834549}}, "length": {"acc": 0.8878787878787879, "drop": -0.00037878787878797837, "rotation": {"subspace_principal_angle": 1.0751426744599302, "mean_class_cosine": 0.4756067433505382}}}, "iid_acc": 0.8875, "selectivity": 0.42999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.080452581028767, "mean_class_cosine": 0.4709291585448011}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7364341085271318, "drop": -0.0014341085271317722, "rotation": {"subspace_principal_angle": 1.133371334962397, "mean_class_cosine": 0.8868509368956804}}, "domain": {"acc": 0.5575, "drop": 0.1775, "rotation": {"subspace_principal_angle": 1.4962557115172213, "mean_class_cosine": 0.27127763310435205}}, "length": {"acc": 0.7651515151515151, "drop": -0.03015151515151515, "rotation": {"subspace_principal_angle": 1.3646255328037025, "mean_class_cosine": 0.9602447183571129}}}, "iid_acc": 0.735, "selectivity": 0.27749999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.9276681297932009, "mean_class_cosine": 0.8062680171943362}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8798449612403101, "drop": 0.01765503875968988}, "domain": {"acc": 0.69625, "drop": 0.20124999999999993}, "length": {"acc": 0.9, "drop": -0.0025000000000000577}}, "iid_acc": 0.8975, "selectivity": 0.43999999999999995} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.004903846153846203, "rotation": {"subspace_principal_angle": 1.3890007649660276, "mean_class_cosine": 0.284822099506686}}}, "iid_acc": 0.87875, "selectivity": 0.60625, "iid_split_rotation": {"subspace_principal_angle": 1.5702965678160945, "mean_class_cosine": 0.3020514335534847}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.5661538461538461, "drop": -0.009903846153846096, "rotation": {"subspace_principal_angle": 1.553740809284586, "mean_class_cosine": 0.9736297672635466}}}, "iid_acc": 0.55625, "selectivity": 0.28375, "iid_split_rotation": {"subspace_principal_angle": 1.4574124850080719, "mean_class_cosine": 0.9616324023244914}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8861538461538462, "drop": -0.001153846153846172}}, "iid_acc": 0.885, "selectivity": 0.6125} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9476309226932669, "drop": 0.02361907730673307, "rotation": {"subspace_principal_angle": 1.0954764445356109, "mean_class_cosine": 0.7159555553963199}}}, "iid_acc": 0.97125, "selectivity": 0.8775, "iid_split_rotation": {"subspace_principal_angle": 1.0349871771004895, "mean_class_cosine": 0.757327529417333}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.26375, "drop": 0.0}, "paraphrase": {"acc": 0.29177057356608477, "drop": -0.028020573566084783, "rotation": {"subspace_principal_angle": 1.4115196982796947, "mean_class_cosine": 0.8384434499541353}}}, "iid_acc": 0.26375, "selectivity": 0.16999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2413128275371268, "mean_class_cosine": 0.7859751886800755}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.912718204488778, "drop": 0.03603179551122193}}, "iid_acc": 0.94875, "selectivity": 0.855} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.5882352941176471, "drop": 0.02926470588235297, "rotation": {"subspace_principal_angle": 1.360311035504444, "mean_class_cosine": 0.20893450516523}}}, "iid_acc": 0.6175, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4063133781244486, "mean_class_cosine": 0.1637422801353008}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.4925, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.005539215686274523, "rotation": {"subspace_principal_angle": 1.441946064731949, "mean_class_cosine": 0.6865636818700251}}}, "iid_acc": 0.4925, "selectivity": -0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5463324868354524, "mean_class_cosine": -0.9630196496317205}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5581699346405229, "drop": -0.009419934640522931}}, "iid_acc": 0.54875, "selectivity": 0.012499999999999956} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02183089214380829, "rotation": {"subspace_principal_angle": 1.25177621884401, "mean_class_cosine": 0.5310790447646092}}}, "iid_acc": 0.645, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2608266828416124, "mean_class_cosine": 0.45773127447897305}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1825, "drop": 0.0}, "paraphrase": {"acc": 0.17709720372836218, "drop": 0.005402796271637811, "rotation": {"subspace_principal_angle": 1.4219864569031055, "mean_class_cosine": 0.3343326799788812}}}, "iid_acc": 0.1825, "selectivity": -0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4213554573473413, "mean_class_cosine": 0.6639945990861887}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.521970705725699, "drop": 0.00427929427430096}}, "iid_acc": 0.52625, "selectivity": 0.27} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7707979626485568, "drop": -0.01704796264855679, "rotation": {"subspace_principal_angle": 1.3829256711520628, "mean_class_cosine": 0.18676744322680136}}}, "iid_acc": 0.75375, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3348303902625251, "mean_class_cosine": 0.23378226387175424}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.5212224108658744, "drop": -0.00497241086587441, "rotation": {"subspace_principal_angle": 1.478236251419889, "mean_class_cosine": 0.9994667861891016}}}, "iid_acc": 0.51625, "selectivity": -0.0050000000000000044, "iid_split_rotation": {"subspace_principal_angle": 1.3448819076659233, "mean_class_cosine": 0.9996533797319738}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6960950764006791, "drop": 0.008904923599320824}}, "iid_acc": 0.705, "selectivity": 0.18374999999999997} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67625, "drop": 0.0}, "paraphrase": {"acc": 0.632258064516129, "drop": 0.043991935483871036, "rotation": {"subspace_principal_angle": 1.4188129223223702, "mean_class_cosine": 0.15139897022933}}}, "iid_acc": 0.67625, "selectivity": 0.1875, "iid_split_rotation": {"subspace_principal_angle": 1.381661689857851, "mean_class_cosine": 0.18800903417603418}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5338709677419354, "drop": 0.002379032258064573, "rotation": {"subspace_principal_angle": 1.2619016165915666, "mean_class_cosine": 0.9985457827336248}}}, "iid_acc": 0.53625, "selectivity": 0.04749999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2970420520156083, "mean_class_cosine": 0.9948538875393387}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67125, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.035766129032258065}}, "iid_acc": 0.67125, "selectivity": 0.1825} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7681159420289855, "drop": -0.008115942028985468, "rotation": {"subspace_principal_angle": 1.310808643208806, "mean_class_cosine": 0.2570686494136881}}}, "iid_acc": 0.76, "selectivity": 0.15375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3820982126149526, "mean_class_cosine": 0.18758027789963122}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.47125, "drop": 0.0}, "paraphrase": {"acc": 0.49597423510466987, "drop": -0.024724235104669867, "rotation": {"subspace_principal_angle": 1.452719223490081, "mean_class_cosine": 0.2128800431912587}}}, "iid_acc": 0.47125, "selectivity": -0.13499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.9161871593522571, "mean_class_cosine": -0.32128070846271867}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6843800322061192, "drop": 0.010619967793880747}}, "iid_acc": 0.695, "selectivity": 0.08875} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.94, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.011330589849108286, "rotation": {"subspace_principal_angle": 1.137125759529054, "mean_class_cosine": 0.4202044066034193}}}, "iid_acc": 0.94, "selectivity": 0.47624999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.0643181542191005, "mean_class_cosine": 0.4851005702018669}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5253772290809328, "drop": 0.008372770919067185, "rotation": {"subspace_principal_angle": 0.516621762104431, "mean_class_cosine": 0.21743596239076726}}}, "iid_acc": 0.53375, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.8840737623848459, "mean_class_cosine": 0.976674523046531}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9368998628257887, "drop": 0.0006001371742112571}}, "iid_acc": 0.9375, "selectivity": 0.47375} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.6634680250005783, "mean_class_cosine": 0.7878611804971933}}}, "iid_acc": 0.99125, "selectivity": 0.11874999999999991, "iid_split_rotation": {"subspace_principal_angle": 0.6348168891373653, "mean_class_cosine": 0.8051803002223438}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9777015437392796, "drop": -0.006451543739279697, "rotation": {"subspace_principal_angle": 1.3391381172924783, "mean_class_cosine": 0.9776428455730666}}}, "iid_acc": 0.97125, "selectivity": 0.0987499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4227835246448142, "mean_class_cosine": 0.9908198389945848}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": 0.0002915951972555231}}, "iid_acc": 0.99, "selectivity": 0.11749999999999994} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7, "drop": 0.02750000000000008, "rotation": {"subspace_principal_angle": 1.472522322965299, "mean_class_cosine": 0.09811589541574661}}}, "iid_acc": 0.7275, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.3920524003697496, "mean_class_cosine": 0.17779365245715506}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.03989130434782606, "rotation": {"subspace_principal_angle": 1.4319943336187273, "mean_class_cosine": -0.9921843120457909}}}, "iid_acc": 0.5225, "selectivity": -0.08374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.0962721333543264, "mean_class_cosine": 0.9947347244035363}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.717391304347826, "drop": -0.0036413043478260576}}, "iid_acc": 0.71375, "selectivity": 0.10750000000000004} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.6153846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.6136363636363636, "drop": 0.0017482517482517723, "rotation": {"subspace_principal_angle": 1.4739338903736143, "mean_class_cosine": 0.13152150958787132}}}, "iid_acc": 0.6153846153846154, "selectivity": 0.17307692307692313, "iid_split_rotation": {"subspace_principal_angle": 1.45720032842477, "mean_class_cosine": 0.19720918409507146}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.41346153846153844, "drop": 0.0}, "paraphrase": {"acc": 0.4147727272727273, "drop": -0.001311188811188857, "rotation": {"subspace_principal_angle": 1.550847759851561, "mean_class_cosine": 0.7251546533887714}}}, "iid_acc": 0.41346153846153844, "selectivity": -0.028846153846153855, "iid_split_rotation": {"subspace_principal_angle": 0.8940696505827914, "mean_class_cosine": 0.8831178368398586}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": -0.017045454545454586}}, "iid_acc": 0.625, "selectivity": 0.1826923076923077} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.8888888888888888, "drop": 0.0011111111111111738, "rotation": {"subspace_principal_angle": 1.346328125428775, "mean_class_cosine": 0.22258793643339736}}}, "iid_acc": 0.89, "selectivity": 0.135, "iid_split_rotation": {"subspace_principal_angle": 1.2354156575580886, "mean_class_cosine": 0.3291286539962054}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5598290598290598, "drop": 0.007670940170940166, "rotation": {"subspace_principal_angle": 1.089834172363605, "mean_class_cosine": 0.9822532101639228}}}, "iid_acc": 0.5675, "selectivity": -0.1875, "iid_split_rotation": {"subspace_principal_angle": 0.9991972798605081, "mean_class_cosine": 0.8996231510474297}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8532763532763533, "drop": 0.00672364672364667}}, "iid_acc": 0.86, "selectivity": 0.10499999999999998} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7649513212795549, "drop": -0.0074513212795549455, "rotation": {"subspace_principal_angle": 1.260155827346341, "mean_class_cosine": 0.30566854295453516}}, "domain": {"acc": 0.72625, "drop": 0.03125, "rotation": {"subspace_principal_angle": 1.2669695481971335, "mean_class_cosine": 0.2991738951377597}}}, "iid_acc": 0.7575, "selectivity": 0.24874999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3245794968247167, "mean_class_cosine": 0.24373663709592575}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5521557719054242, "drop": -0.017155771905424166, "rotation": {"subspace_principal_angle": 1.4523160206390673, "mean_class_cosine": 0.9818329782722122}}, "domain": {"acc": 0.495, "drop": 0.040000000000000036, "rotation": {"subspace_principal_angle": 1.4451431888735655, "mean_class_cosine": 0.5803352104264222}}}, "iid_acc": 0.535, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3059026276224222, "mean_class_cosine": 0.9628009093282439}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.741307371349096, "drop": -0.008807371349095927}, "domain": {"acc": 0.75125, "drop": -0.018749999999999933}}, "iid_acc": 0.7325, "selectivity": 0.22375} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.8020408163265306, "drop": 0.015459183673469434, "rotation": {"subspace_principal_angle": 0.9261920049968357, "mean_class_cosine": 0.6008822051011515}}, "domain": {"acc": 0.68375, "drop": 0.13375000000000004, "rotation": {"subspace_principal_angle": 1.1716370438461978, "mean_class_cosine": 0.38864385309548416}}, "length": {"acc": 0.8115501519756839, "drop": 0.0059498480243160845, "rotation": {"subspace_principal_angle": 0.8775406092988467, "mean_class_cosine": 0.6390447633833902}}}, "iid_acc": 0.8175, "selectivity": 0.31625000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.957610160630722, "mean_class_cosine": 0.5754760826890111}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6877551020408164, "drop": 0.0034948979591836737, "rotation": {"subspace_principal_angle": 1.2806469825658773, "mean_class_cosine": 0.8910748524092273}}, "domain": {"acc": 0.56875, "drop": 0.12250000000000005, "rotation": {"subspace_principal_angle": 1.438815881455066, "mean_class_cosine": 0.41273195854934175}}, "length": {"acc": 0.6975683890577508, "drop": -0.0063183890577507285, "rotation": {"subspace_principal_angle": 1.5481828698411682, "mean_class_cosine": 0.9192162272707554}}}, "iid_acc": 0.69125, "selectivity": 0.19000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3565470315186383, "mean_class_cosine": 0.8817941693849485}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.78875, "drop": 0.0}, "paraphrase": {"acc": 0.7755102040816326, "drop": 0.013239795918367325}, "domain": {"acc": 0.69125, "drop": 0.09749999999999992}, "length": {"acc": 0.7781155015197568, "drop": 0.010634498480243115}}, "iid_acc": 0.78875, "selectivity": 0.2875} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8454258675078864, "drop": 0.008324132492113612, "rotation": {"subspace_principal_angle": 1.4872713948684544, "mean_class_cosine": 0.5396680342819231}}}, "iid_acc": 0.85375, "selectivity": 0.62375, "iid_split_rotation": {"subspace_principal_angle": 1.1001196196694507, "mean_class_cosine": 0.5629295021677052}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.78, "drop": 0.0}, "paraphrase": {"acc": 0.7744479495268138, "drop": 0.0055520504731861875, "rotation": {"subspace_principal_angle": 1.3957068228612415, "mean_class_cosine": 0.9585242100327607}}}, "iid_acc": 0.78, "selectivity": 0.55, "iid_split_rotation": {"subspace_principal_angle": 1.0534755137152318, "mean_class_cosine": 0.9472616950218877}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8249211356466877, "drop": 0.006328864353312369}}, "iid_acc": 0.83125, "selectivity": 0.6012500000000001} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.9177377892030848, "drop": 0.0022622107969152427, "rotation": {"subspace_principal_angle": 1.5481489232177876, "mean_class_cosine": 0.7602277235959066}}}, "iid_acc": 0.92, "selectivity": 0.8500000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.0913499627869498, "mean_class_cosine": 0.8064226222572987}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.81375, "drop": 0.0}, "paraphrase": {"acc": 0.7583547557840618, "drop": 0.05539524421593822, "rotation": {"subspace_principal_angle": 1.434488039184557, "mean_class_cosine": 0.8699474461477923}}}, "iid_acc": 0.81375, "selectivity": 0.7437499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4670162663468989, "mean_class_cosine": 0.9406629826516945}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8174807197943444, "drop": 0.03626928020565556}}, "iid_acc": 0.85375, "selectivity": 0.78375} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.53875, "drop": 0.0}, "paraphrase": {"acc": 0.5297157622739018, "drop": 0.009034237726098127, "rotation": {"subspace_principal_angle": 1.506602262713435, "mean_class_cosine": 0.06414998384863341}}}, "iid_acc": 0.53875, "selectivity": 0.062499999999999944, "iid_split_rotation": {"subspace_principal_angle": 1.5141486385951506, "mean_class_cosine": 0.056617396360705646}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.47674418604651164, "drop": 0.022005813953488385, "rotation": {"subspace_principal_angle": 1.5018907330856117, "mean_class_cosine": 0.3385630136055}}}, "iid_acc": 0.49875, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5223535908671477, "mean_class_cosine": 0.3337844099038709}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5335917312661499, "drop": 0.017658268733850124}}, "iid_acc": 0.55125, "selectivity": 0.07500000000000001} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5115020297699594, "drop": 0.028497970230040637, "rotation": {"subspace_principal_angle": 1.3837974822380827, "mean_class_cosine": 0.2853919153689364}}}, "iid_acc": 0.54, "selectivity": 0.3175, "iid_split_rotation": {"subspace_principal_angle": 1.4430652239408488, "mean_class_cosine": 0.32302299850041155}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.2075, "drop": 0.0}, "paraphrase": {"acc": 0.21380243572395127, "drop": -0.006302435723951283, "rotation": {"subspace_principal_angle": 1.4121359208224795, "mean_class_cosine": 0.761742118163654}}}, "iid_acc": 0.2075, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4351405486147384, "mean_class_cosine": 0.6192669179592563}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.4438430311231394, "drop": 0.03115696887686059}}, "iid_acc": 0.475, "selectivity": 0.25249999999999995} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7043918918918919, "drop": 0.021858108108108065, "rotation": {"subspace_principal_angle": 1.2557206455227148, "mean_class_cosine": 0.3098884278198152}}}, "iid_acc": 0.72625, "selectivity": 0.20999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2635908058672536, "mean_class_cosine": 0.30239617255824447}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5506756756756757, "drop": -0.02317567567567569, "rotation": {"subspace_principal_angle": 1.563519056112211, "mean_class_cosine": 0.9678685291584537}}}, "iid_acc": 0.5275, "selectivity": 0.011249999999999982, "iid_split_rotation": {"subspace_principal_angle": 1.4948991382902788, "mean_class_cosine": 0.9835336511500095}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7010135135135135, "drop": 0.005236486486486558}}, "iid_acc": 0.70625, "selectivity": 0.19000000000000006} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6715447154471544, "drop": 0.01595528455284556, "rotation": {"subspace_principal_angle": 1.3527346728817384, "mean_class_cosine": 0.21633758729886746}}}, "iid_acc": 0.6875, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.396125798434941, "mean_class_cosine": 0.17378368868184255}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.59125, "drop": 0.0}, "paraphrase": {"acc": 0.583739837398374, "drop": 0.007510162601626069, "rotation": {"subspace_principal_angle": 1.533612671019165, "mean_class_cosine": 0.9008068153198737}}}, "iid_acc": 0.59125, "selectivity": 0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.260460650130829, "mean_class_cosine": 0.890537070105323}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6520325203252032, "drop": 0.007967479674796829}}, "iid_acc": 0.66, "selectivity": 0.19125000000000003} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7079934747145188, "drop": 0.013256525285481158, "rotation": {"subspace_principal_angle": 1.4022375641064082, "mean_class_cosine": 0.16776171230922843}}}, "iid_acc": 0.72125, "selectivity": 0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4265826075864037, "mean_class_cosine": 0.14371435563850823}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5432300163132137, "drop": -0.005730016313213682, "rotation": {"subspace_principal_angle": 1.1709151024773818, "mean_class_cosine": 0.6775048474955148}}}, "iid_acc": 0.5375, "selectivity": -0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1992395749899722, "mean_class_cosine": 0.7778731026109806}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7259380097879282, "drop": 0.0003119902120717333}}, "iid_acc": 0.72625, "selectivity": 0.14749999999999996} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8446866485013624, "drop": 0.041563351498637546, "rotation": {"subspace_principal_angle": 0.9763955114091522, "mean_class_cosine": 0.5600124400720721}}}, "iid_acc": 0.88625, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 0.901229402467838, "mean_class_cosine": 0.620646474717384}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.84, "drop": 0.0}, "paraphrase": {"acc": 0.8079019073569482, "drop": 0.032098092643051745, "rotation": {"subspace_principal_angle": 1.5194268659444015, "mean_class_cosine": 0.9313227017091655}}}, "iid_acc": 0.84, "selectivity": 0.36, "iid_split_rotation": {"subspace_principal_angle": 1.4080821249653852, "mean_class_cosine": 0.9589713457056335}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8337874659400545, "drop": 0.037462534059945485}}, "iid_acc": 0.87125, "selectivity": 0.39125} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9818481848184818, "drop": 0.006901815181518223, "rotation": {"subspace_principal_angle": 0.66528493719011, "mean_class_cosine": 0.7867409300908236}}}, "iid_acc": 0.98875, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6314500555079575, "mean_class_cosine": 0.8071723665064887}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8861386138613861, "drop": 0.0026113861386138915, "rotation": {"subspace_principal_angle": 1.4453777964256416, "mean_class_cosine": 0.9697563041315995}}}, "iid_acc": 0.88875, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9974066490876098, "mean_class_cosine": 0.9768283657309265}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98, "drop": 0.0}, "paraphrase": {"acc": 0.9752475247524752, "drop": 0.004752475247524757}}, "iid_acc": 0.98, "selectivity": 0.11749999999999994} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.703328509406657, "drop": -0.013328509406657085, "rotation": {"subspace_principal_angle": 1.4989704535588646, "mean_class_cosine": 0.07176413141027983}}}, "iid_acc": 0.69, "selectivity": 0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.5357515116941332, "mean_class_cosine": 0.03503764222349777}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.54375, "drop": 0.0}, "paraphrase": {"acc": 0.5470332850940666, "drop": -0.0032832850940666125, "rotation": {"subspace_principal_angle": 1.5166062020639395, "mean_class_cosine": 0.39412823203921477}}}, "iid_acc": 0.54375, "selectivity": -0.14, "iid_split_rotation": {"subspace_principal_angle": 1.5290148158529324, "mean_class_cosine": -0.0372845302255322}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": -0.009247467438494916}}, "iid_acc": 0.71, "selectivity": 0.026249999999999996} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5961538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.6647058823529411, "drop": -0.068552036199095, "rotation": {"subspace_principal_angle": 1.4788344627937355, "mean_class_cosine": 0.29399670434967345}}}, "iid_acc": 0.5961538461538461, "selectivity": 0.15384615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.3726634159821796, "mean_class_cosine": 0.2274413627536632}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5048076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": -0.03636877828054297, "rotation": {"subspace_principal_angle": 0.913621189609579, "mean_class_cosine": 0.6670655608930605}}}, "iid_acc": 0.5048076923076923, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.0837588978590438, "mean_class_cosine": 0.5902856300722443}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6009615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6235294117647059, "drop": -0.02256787330316745}}, "iid_acc": 0.6009615384615384, "selectivity": 0.15865384615384615} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": -0.0021985190409027355, "rotation": {"subspace_principal_angle": 0.9713008818157643, "mean_class_cosine": 0.564225974312156}}}, "iid_acc": 0.89625, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 0.9979017449204138, "mean_class_cosine": 0.5420667359536389}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7362482369534555, "drop": -0.009998236953455586, "rotation": {"subspace_principal_angle": 1.3253351141042944, "mean_class_cosine": 0.8910372103240486}}}, "iid_acc": 0.72625, "selectivity": -0.030000000000000027, "iid_split_rotation": {"subspace_principal_angle": 1.5153708369050465, "mean_class_cosine": 0.9019435549852257}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.785, "drop": 0.0}, "paraphrase": {"acc": 0.7870239774330042, "drop": -0.002023977433004198}}, "iid_acc": 0.785, "selectivity": 0.028750000000000053} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.8, "drop": 0.0}, "paraphrase": {"acc": 0.8052851182197497, "drop": -0.0052851182197496405, "rotation": {"subspace_principal_angle": 1.289669049843915, "mean_class_cosine": 0.2774388483889756}}, "domain": {"acc": 0.71, "drop": 0.09000000000000008, "rotation": {"subspace_principal_angle": 1.2978069882976488, "mean_class_cosine": 0.2696112780665089}}}, "iid_acc": 0.8, "selectivity": 0.2875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2350581884793985, "mean_class_cosine": 0.32946618568105485}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5549374130737135, "drop": 0.005062586926286583, "rotation": {"subspace_principal_angle": 1.3268922292489094, "mean_class_cosine": 0.9883969165833375}}, "domain": {"acc": 0.495, "drop": 0.06500000000000006, "rotation": {"subspace_principal_angle": 1.131095069142722, "mean_class_cosine": 0.7133167140261951}}}, "iid_acc": 0.56, "selectivity": 0.0475000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2611138196752891, "mean_class_cosine": 0.9895181201439}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.815, "drop": 0.0}, "paraphrase": {"acc": 0.780250347705146, "drop": 0.03474965229485394}, "domain": {"acc": 0.76625, "drop": 0.04874999999999996}}, "iid_acc": 0.815, "selectivity": 0.3025} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.0027295918367347127, "rotation": {"subspace_principal_angle": 1.1073395630387064, "mean_class_cosine": 0.4470428901807506}}, "domain": {"acc": 0.71375, "drop": 0.14, "rotation": {"subspace_principal_angle": 1.310897035728456, "mean_class_cosine": 0.2569832264845267}}, "length": {"acc": 0.8556231003039514, "drop": -0.0018731003039513805, "rotation": {"subspace_principal_angle": 1.106355203321873, "mean_class_cosine": 0.44792319554846216}}}, "iid_acc": 0.85375, "selectivity": 0.3375, "iid_split_rotation": {"subspace_principal_angle": 1.1106698792892344, "mean_class_cosine": 0.4440614070193354}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6795918367346939, "drop": 0.002908163265306074, "rotation": {"subspace_principal_angle": 1.0125111213870783, "mean_class_cosine": 0.9329519454684184}}, "domain": {"acc": 0.56625, "drop": 0.11624999999999996, "rotation": {"subspace_principal_angle": 1.1283485572455334, "mean_class_cosine": 0.5142940873907857}}, "length": {"acc": 0.6960486322188449, "drop": -0.013548632218844947, "rotation": {"subspace_principal_angle": 1.3998779208397523, "mean_class_cosine": 0.9308749733569011}}}, "iid_acc": 0.6825, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 0.8902790332464676, "mean_class_cosine": 0.8762482918205301}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.000229591836734655}, "domain": {"acc": 0.7775, "drop": 0.07374999999999998}, "length": {"acc": 0.8541033434650456, "drop": -0.0028533434650456213}}, "iid_acc": 0.85125, "selectivity": 0.33499999999999996} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8470031545741324, "drop": 0.0154968454258676, "rotation": {"subspace_principal_angle": 1.3265156223459154, "mean_class_cosine": 0.357128289219289}}}, "iid_acc": 0.8625, "selectivity": 0.6375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.347641540682651, "mean_class_cosine": 0.38870803168480306}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6876971608832808, "drop": -0.00019716088328081138, "rotation": {"subspace_principal_angle": 1.49331931022075, "mean_class_cosine": 0.9421571273887641}}}, "iid_acc": 0.6875, "selectivity": 0.4625, "iid_split_rotation": {"subspace_principal_angle": 0.8842746339269641, "mean_class_cosine": 0.9544218315516564}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8517350157728707, "drop": 0.010764984227129348}}, "iid_acc": 0.8625, "selectivity": 0.6375000000000001} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9280205655526992, "drop": 0.01447943444730082, "rotation": {"subspace_principal_angle": 1.543953519865052, "mean_class_cosine": 0.6356132585262798}}}, "iid_acc": 0.9425, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.1199023344741124, "mean_class_cosine": 0.6818253733919893}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8791773778920309, "drop": -0.007927377892030907, "rotation": {"subspace_principal_angle": 1.347264998699133, "mean_class_cosine": 0.909582314970496}}}, "iid_acc": 0.87125, "selectivity": 0.7975, "iid_split_rotation": {"subspace_principal_angle": 1.4779951280492787, "mean_class_cosine": 0.9618009806772534}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94375, "drop": 0.0}, "paraphrase": {"acc": 0.922879177377892, "drop": 0.020870822622107932}}, "iid_acc": 0.94375, "selectivity": 0.87} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.49483204134366926, "drop": 0.010167958656330744, "rotation": {"subspace_principal_angle": 1.524398385950912, "mean_class_cosine": -0.046381295294995115}}}, "iid_acc": 0.505, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5107404805094091, "mean_class_cosine": 0.06001975217823446}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.4909560723514212, "drop": 0.015293927648578787, "rotation": {"subspace_principal_angle": 1.5668349405587232, "mean_class_cosine": -0.03841414374121909}}}, "iid_acc": 0.50625, "selectivity": 0.023749999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.5352103349454524, "mean_class_cosine": -0.02088599099775313}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5167958656330749, "drop": 0.0007041343669250688}}, "iid_acc": 0.5175, "selectivity": 0.034999999999999976} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.5764546684709067, "drop": 0.026045331529093385, "rotation": {"subspace_principal_angle": 1.5428367399341, "mean_class_cosine": 0.37101745351040355}}}, "iid_acc": 0.6025, "selectivity": 0.33875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4285270805830252, "mean_class_cosine": 0.3347078246464954}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.21, "drop": 0.0}, "paraphrase": {"acc": 0.19079837618403248, "drop": 0.019201623815967517, "rotation": {"subspace_principal_angle": 1.4456365520979793, "mean_class_cosine": 0.6187678845759252}}}, "iid_acc": 0.21, "selectivity": -0.05374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5049088148460505, "mean_class_cosine": 0.472889652110318}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.496617050067659, "drop": 0.009632949932340984}}, "iid_acc": 0.50625, "selectivity": 0.2425} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7331081081081081, "drop": 0.009391891891891935, "rotation": {"subspace_principal_angle": 1.3068079885037773, "mean_class_cosine": 0.2609327871606552}}}, "iid_acc": 0.7425, "selectivity": 0.23625000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.2782285316613433, "mean_class_cosine": 0.28841185363838734}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5929054054054054, "drop": -0.020405405405405364, "rotation": {"subspace_principal_angle": 1.4497936937308118, "mean_class_cosine": 0.9158504322082752}}}, "iid_acc": 0.5725, "selectivity": 0.06625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3799098692242713, "mean_class_cosine": 0.8877578773800847}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7364864864864865, "drop": -0.0027364864864865}}, "iid_acc": 0.73375, "selectivity": 0.22750000000000004} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.67875, "drop": 0.0}, "paraphrase": {"acc": 0.6764227642276422, "drop": 0.0023272357723577164, "rotation": {"subspace_principal_angle": 1.3819592931432023, "mean_class_cosine": 0.18771672963966035}}}, "iid_acc": 0.67875, "selectivity": 0.19374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.416763883086912, "mean_class_cosine": 0.15342407040448366}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.5788617886178862, "drop": -0.010111788617886197, "rotation": {"subspace_principal_angle": 0.8085931323629197, "mean_class_cosine": 0.9884789144089932}}}, "iid_acc": 0.56875, "selectivity": 0.08374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3601702468050045, "mean_class_cosine": 0.9887573457930535}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.6455284552845528, "drop": 0.03947154471544723}}, "iid_acc": 0.685, "selectivity": 0.20000000000000007} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.732463295269168, "drop": -0.003713295269167971, "rotation": {"subspace_principal_angle": 1.4514696427072735, "mean_class_cosine": 0.11904370635094103}}}, "iid_acc": 0.72875, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4175736807895711, "mean_class_cosine": 0.1526238101323753}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.4893964110929853, "drop": -0.011896411092985337, "rotation": {"subspace_principal_angle": 1.3731868842609583, "mean_class_cosine": 0.8675782704012913}}}, "iid_acc": 0.4775, "selectivity": -0.10625000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4543529756472604, "mean_class_cosine": 0.7758454557886542}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7259380097879282, "drop": 0.006561990212071822}}, "iid_acc": 0.7325, "selectivity": 0.14875000000000005} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.8991825613079019, "drop": 0.019567438692098094, "rotation": {"subspace_principal_angle": 1.17550589332711, "mean_class_cosine": 0.38507624166782645}}}, "iid_acc": 0.91875, "selectivity": 0.41625, "iid_split_rotation": {"subspace_principal_angle": 1.1563275504871824, "mean_class_cosine": 0.40270374204802895}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.6771117166212534, "drop": 0.11038828337874662, "rotation": {"subspace_principal_angle": 0.9740972022042724, "mean_class_cosine": 0.9590361076960369}}}, "iid_acc": 0.7875, "selectivity": 0.28500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4700765532293367, "mean_class_cosine": 0.8827661242087979}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.888283378746594, "drop": 0.012966621253405974}}, "iid_acc": 0.90125, "selectivity": 0.39875000000000005} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9867986798679867, "drop": 0.003201320132013241, "rotation": {"subspace_principal_angle": 0.7435989691909886, "mean_class_cosine": 0.7360370400174403}}}, "iid_acc": 0.99, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.7520555858872627, "mean_class_cosine": 0.7302861569945356}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9422442244224423, "drop": -0.0009942244224422359, "rotation": {"subspace_principal_angle": 1.441143445268482, "mean_class_cosine": 0.9558286303461359}}}, "iid_acc": 0.94125, "selectivity": 0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5613733244607428, "mean_class_cosine": 0.973426291781849}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98125, "drop": 0.0}, "paraphrase": {"acc": 0.9818481848184818, "drop": -0.0005981848184818395}}, "iid_acc": 0.98125, "selectivity": 0.125} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.662807525325615, "drop": 0.023442474674384983, "rotation": {"subspace_principal_angle": 1.5311195648072153, "mean_class_cosine": 0.03966635264677357}}}, "iid_acc": 0.68625, "selectivity": 0.07874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.46679518399481, "mean_class_cosine": 0.10381376065362727}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5151953690303908, "drop": 0.01980463096960927, "rotation": {"subspace_principal_angle": 1.1930287896465346, "mean_class_cosine": 0.911295872781087}}}, "iid_acc": 0.535, "selectivity": -0.07250000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.6904458872144239, "mean_class_cosine": -0.9709984265270024}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": -0.007997467438494832}}, "iid_acc": 0.71125, "selectivity": 0.10375000000000001} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6294117647058823, "drop": -0.004411764705882337, "rotation": {"subspace_principal_angle": 1.5554803313899561, "mean_class_cosine": 0.31267859264684567}}}, "iid_acc": 0.625, "selectivity": 0.1875, "iid_split_rotation": {"subspace_principal_angle": 1.48134181230664, "mean_class_cosine": 0.16283662737976473}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5647058823529412, "drop": -0.01182126696832575, "rotation": {"subspace_principal_angle": 1.0131680189558905, "mean_class_cosine": 0.6680893935892259}}}, "iid_acc": 0.5528846153846154, "selectivity": 0.11538461538461542, "iid_split_rotation": {"subspace_principal_angle": 1.4160182659695404, "mean_class_cosine": 0.5680667839120414}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6346153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.6411764705882353, "drop": -0.0065610859728507664}}, "iid_acc": 0.6346153846153846, "selectivity": 0.19711538461538458} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8899858956276445, "drop": 0.00876410437235553, "rotation": {"subspace_principal_angle": 1.2045865557395712, "mean_class_cosine": 0.3580791088837819}}}, "iid_acc": 0.89875, "selectivity": 0.16000000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2021828264056524, "mean_class_cosine": 0.3603224132960571}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.61625, "drop": 0.0}, "paraphrase": {"acc": 0.5994358251057827, "drop": 0.016814174894217215, "rotation": {"subspace_principal_angle": 1.1806325446572814, "mean_class_cosine": 0.8922635781825321}}}, "iid_acc": 0.61625, "selectivity": -0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3837719103631978, "mean_class_cosine": 0.7905423303610484}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.835, "drop": 0.0}, "paraphrase": {"acc": 0.8349788434414669, "drop": 2.1156558533075298e-05}}, "iid_acc": 0.835, "selectivity": 0.09624999999999995} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8247566063977747, "drop": 0.012743393602225317, "rotation": {"subspace_principal_angle": 1.2798965624456102, "mean_class_cosine": 0.28681430291560384}}, "domain": {"acc": 0.805, "drop": 0.03249999999999997, "rotation": {"subspace_principal_angle": 1.2851613063016167, "mean_class_cosine": 0.28176679938187554}}}, "iid_acc": 0.8375, "selectivity": 0.33625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1765812661272652, "mean_class_cosine": 0.3840835740297509}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.56884561891516, "drop": -0.012595618915159923, "rotation": {"subspace_principal_angle": 1.2901370394335034, "mean_class_cosine": 0.9977925223579588}}, "domain": {"acc": 0.495, "drop": 0.06125000000000003, "rotation": {"subspace_principal_angle": 1.4599392249622891, "mean_class_cosine": 0.6915007566423281}}}, "iid_acc": 0.55625, "selectivity": 0.05500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4453053092946215, "mean_class_cosine": 0.998233597525252}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.8122392211404729, "drop": -0.008489221140472902}, "domain": {"acc": 0.79625, "drop": 0.007499999999999951}}, "iid_acc": 0.80375, "selectivity": 0.3025} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8816326530612245, "drop": 0.0021173469387755217, "rotation": {"subspace_principal_angle": 1.190715066843778, "mean_class_cosine": 0.3709959314311648}}, "domain": {"acc": 0.60625, "drop": 0.2775000000000001, "rotation": {"subspace_principal_angle": 1.4172475398253837, "mean_class_cosine": 0.1529461220229834}}, "length": {"acc": 0.8890577507598785, "drop": -0.005307750759878438, "rotation": {"subspace_principal_angle": 1.2010441268654077, "mean_class_cosine": 0.3613843900828707}}}, "iid_acc": 0.88375, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2079982748541473, "mean_class_cosine": 0.35489153878867186}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.79, "drop": 0.0}, "paraphrase": {"acc": 0.7816326530612245, "drop": 0.0083673469387755, "rotation": {"subspace_principal_angle": 1.3105422369786384, "mean_class_cosine": 0.9556516177420615}}, "domain": {"acc": 0.56625, "drop": 0.22375, "rotation": {"subspace_principal_angle": 1.5060453332017627, "mean_class_cosine": 0.35625753388954995}}, "length": {"acc": 0.8069908814589666, "drop": -0.016990881458966545, "rotation": {"subspace_principal_angle": 1.5564351284094222, "mean_class_cosine": 0.9479993237414271}}}, "iid_acc": 0.79, "selectivity": 0.29500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3575732701756293, "mean_class_cosine": 0.922075109814884}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8877551020408163, "drop": 0.004744897959183647}, "domain": {"acc": 0.6275, "drop": 0.265}, "length": {"acc": 0.8890577507598785, "drop": 0.0034422492401214866}}, "iid_acc": 0.8925, "selectivity": 0.39749999999999996} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8485804416403786, "drop": 0.018919558359621447, "rotation": {"subspace_principal_angle": 1.4399059560351106, "mean_class_cosine": 0.2556332542951133}}}, "iid_acc": 0.8675, "selectivity": 0.64875, "iid_split_rotation": {"subspace_principal_angle": 1.3788930048065602, "mean_class_cosine": 0.2888745735494024}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.7996845425867508, "drop": 0.011565457413249214, "rotation": {"subspace_principal_angle": 1.5548543754884978, "mean_class_cosine": 0.9775272980914157}}}, "iid_acc": 0.81125, "selectivity": 0.5925, "iid_split_rotation": {"subspace_principal_angle": 1.3348767857160242, "mean_class_cosine": 0.979268775558342}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8596214511041009, "drop": 0.020378548895899073}}, "iid_acc": 0.88, "selectivity": 0.66125} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.9525, "drop": 0.0}, "paraphrase": {"acc": 0.9434447300771208, "drop": 0.009055269922879194, "rotation": {"subspace_principal_angle": 1.134461645570175, "mean_class_cosine": 0.6354679842524055}}}, "iid_acc": 0.9525, "selectivity": 0.875, "iid_split_rotation": {"subspace_principal_angle": 1.1124246218092984, "mean_class_cosine": 0.678012477701743}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7712082262210797, "drop": 0.03254177377892031, "rotation": {"subspace_principal_angle": 1.5560406223970316, "mean_class_cosine": 0.8899947966184091}}}, "iid_acc": 0.80375, "selectivity": 0.72625, "iid_split_rotation": {"subspace_principal_angle": 1.503910501910838, "mean_class_cosine": 0.9534530359529768}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9511568123393316, "drop": 0.005093187660668463}}, "iid_acc": 0.95625, "selectivity": 0.87875} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5258397932816538, "drop": 0.015410206718346253, "rotation": {"subspace_principal_angle": 1.496424070388461, "mean_class_cosine": 0.07430371365823527}}}, "iid_acc": 0.54125, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4808569146777435, "mean_class_cosine": 0.08981820636557197}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5025839793281653, "drop": 0.012416020671834671, "rotation": {"subspace_principal_angle": 1.5496948617311233, "mean_class_cosine": 0.9662275396228005}}}, "iid_acc": 0.515, "selectivity": -0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.1782486440521842, "mean_class_cosine": 0.8742419568492337}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5012919896640827, "drop": 0.01870801033591729}}, "iid_acc": 0.52, "selectivity": -0.0012499999999999734} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.6425, "drop": 0.0}, "paraphrase": {"acc": 0.5872801082543978, "drop": 0.055219891745602157, "rotation": {"subspace_principal_angle": 1.36919244552489, "mean_class_cosine": 0.29038444079442377}}}, "iid_acc": 0.6425, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3882673901543217, "mean_class_cosine": 0.28989125551170686}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.32125, "drop": 0.0}, "paraphrase": {"acc": 0.3071718538565629, "drop": 0.014078146143437054, "rotation": {"subspace_principal_angle": 1.3170533049461681, "mean_class_cosine": 0.6594180780518364}}}, "iid_acc": 0.32125, "selectivity": 0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3815192780699195, "mean_class_cosine": 0.5713567606372191}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5453315290933695, "drop": 0.019668470906630486}}, "iid_acc": 0.565, "selectivity": 0.30999999999999994} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7162162162162162, "drop": 0.017533783783783785, "rotation": {"subspace_principal_angle": 1.3990618518124023, "mean_class_cosine": 0.1708915658468793}}}, "iid_acc": 0.73375, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.3739741567595274, "mean_class_cosine": 0.19555384795811886}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.527027027027027, "drop": -0.014527027027027017, "rotation": {"subspace_principal_angle": 1.2081420951650945, "mean_class_cosine": 0.994931982276282}}}, "iid_acc": 0.5125, "selectivity": -0.007500000000000062, "iid_split_rotation": {"subspace_principal_angle": 1.4812482829852658, "mean_class_cosine": 0.9985002258751091}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7297297297297297, "drop": 0.02902027027027032}}, "iid_acc": 0.75875, "selectivity": 0.23875000000000002} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.656910569105691, "drop": 0.02058943089430898, "rotation": {"subspace_principal_angle": 1.4402611653004675, "mean_class_cosine": 0.13016476977575708}}}, "iid_acc": 0.6775, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.4323787501656056, "mean_class_cosine": 0.1379759996513438}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5495934959349593, "drop": -0.0033434959349593063, "rotation": {"subspace_principal_angle": 1.1049873185467238, "mean_class_cosine": 0.9918135119807125}}}, "iid_acc": 0.54625, "selectivity": 0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2358796490004342, "mean_class_cosine": 0.9310981395287434}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6731707317073171, "drop": 0.00932926829268288}}, "iid_acc": 0.6825, "selectivity": 0.17125} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7308319738988581, "drop": 0.011668026101141926, "rotation": {"subspace_principal_angle": 1.4422948230704635, "mean_class_cosine": 0.12814814582491812}}}, "iid_acc": 0.7425, "selectivity": 0.17500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4341137765391143, "mean_class_cosine": 0.13625736099808608}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.46875, "drop": 0.0}, "paraphrase": {"acc": 0.499184339314845, "drop": -0.030434339314845016, "rotation": {"subspace_principal_angle": 0.5863090291944898, "mean_class_cosine": 0.9879298243670935}}}, "iid_acc": 0.46875, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.1523867074914143, "mean_class_cosine": 0.9911941124762024}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7487765089722676, "drop": 0.003723491027732395}}, "iid_acc": 0.7525, "selectivity": 0.18499999999999994} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9209809264305178, "drop": 0.016519073569482234, "rotation": {"subspace_principal_angle": 1.1107561694947945, "mean_class_cosine": 0.44398408965660313}}}, "iid_acc": 0.9375, "selectivity": 0.43125, "iid_split_rotation": {"subspace_principal_angle": 1.134035745754033, "mean_class_cosine": 0.42300636563354255}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5926430517711172, "drop": -0.025143051771117197, "rotation": {"subspace_principal_angle": 0.950772040164655, "mean_class_cosine": 0.9618998317960137}}}, "iid_acc": 0.5675, "selectivity": 0.06125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2078832510638606, "mean_class_cosine": 0.4895900188006161}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9073569482288828, "drop": 0.010143051771117184}}, "iid_acc": 0.9175, "selectivity": 0.41125} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9587458745874587, "drop": 0.0337541254125413, "rotation": {"subspace_principal_angle": 0.9486144857930342, "mean_class_cosine": 0.5828095295275909}}}, "iid_acc": 0.9925, "selectivity": 0.16375000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.9074513638691858, "mean_class_cosine": 0.6157559117610208}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.14026402640264027, "drop": 0.5609859735973598, "rotation": {"subspace_principal_angle": 1.1024334053734877, "mean_class_cosine": 0.5757723487754411}}}, "iid_acc": 0.70125, "selectivity": -0.12749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2565568409948566, "mean_class_cosine": 0.9967803523356993}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.976897689768977, "drop": 0.01685231023102307}}, "iid_acc": 0.99375, "selectivity": 0.16500000000000004} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6613603473227206, "drop": 0.026139652677279357, "rotation": {"subspace_principal_angle": 1.5419209125002344, "mean_class_cosine": 0.0288714017921598}}}, "iid_acc": 0.6875, "selectivity": 0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4675689654827935, "mean_class_cosine": 0.10304412909555002}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5050651230101303, "drop": 0.021184876989869705, "rotation": {"subspace_principal_angle": 1.5385757961576287, "mean_class_cosine": 0.48096021437512015}}}, "iid_acc": 0.52625, "selectivity": -0.05249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1269562907021227, "mean_class_cosine": -0.8913469139771255}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.723589001447178, "drop": -0.006089001447177944}}, "iid_acc": 0.7175, "selectivity": 0.13875000000000004} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": 0.014253393665158431, "rotation": {"subspace_principal_angle": 1.380845212313945, "mean_class_cosine": 0.22774336814303378}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.3028846153846154, "iid_split_rotation": {"subspace_principal_angle": 1.4371716903391298, "mean_class_cosine": 0.16228231461949003}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.47115384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.4411764705882353, "drop": 0.029977375565610864, "rotation": {"subspace_principal_angle": 1.077660750569373, "mean_class_cosine": 0.5488122950260849}}}, "iid_acc": 0.47115384615384615, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.3611080124878545, "mean_class_cosine": 0.7221103135486793}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": -0.014592760180995423}}, "iid_acc": 0.6442307692307693, "selectivity": 0.27403846153846156} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8885754583921015, "drop": 0.006424541607898515, "rotation": {"subspace_principal_angle": 1.1803089373724376, "mean_class_cosine": 0.38063916084124166}}}, "iid_acc": 0.895, "selectivity": 0.16500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.198391794536694, "mean_class_cosine": 0.36385619559330473}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.8208744710860366, "drop": -0.017124471086036652, "rotation": {"subspace_principal_angle": 1.4701060816070168, "mean_class_cosine": 0.8930227036508337}}}, "iid_acc": 0.80375, "selectivity": 0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.1336871573868261, "mean_class_cosine": 0.8902119338649208}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8716502115655853, "drop": 0.005849788434414638}}, "iid_acc": 0.8775, "selectivity": 0.14749999999999996} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8344923504867872, "drop": 0.03300764951321289, "rotation": {"subspace_principal_angle": 1.32725054312888, "mean_class_cosine": 0.24114527926927334}}, "domain": {"acc": 0.855, "drop": 0.012500000000000067, "rotation": {"subspace_principal_angle": 1.4511968441379492, "mean_class_cosine": 0.11931456062079314}}}, "iid_acc": 0.8675, "selectivity": 0.38125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3310106874774308, "mean_class_cosine": 0.23749440429619095}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5827538247566064, "drop": -0.02025382475660642, "rotation": {"subspace_principal_angle": 0.7190626485497094, "mean_class_cosine": 0.9647786618667977}}, "domain": {"acc": 0.495, "drop": 0.0675, "rotation": {"subspace_principal_angle": 1.1078756653311121, "mean_class_cosine": 0.28125598087122977}}}, "iid_acc": 0.5625, "selectivity": 0.07624999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.8988412922378591, "mean_class_cosine": 0.9806205060335059}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.827538247566064, "drop": 0.027461752433936004}, "domain": {"acc": 0.84875, "drop": 0.006249999999999978}}, "iid_acc": 0.855, "selectivity": 0.36874999999999997} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.8959183673469387, "drop": 0.009081632653061278, "rotation": {"subspace_principal_angle": 1.329136111358117, "mean_class_cosine": 0.2393149283948771}}, "domain": {"acc": 0.615, "drop": 0.29000000000000004, "rotation": {"subspace_principal_angle": 1.4430058206254768, "mean_class_cosine": 0.12744297808753202}}, "length": {"acc": 0.9072948328267477, "drop": -0.0022948328267476947, "rotation": {"subspace_principal_angle": 1.3188768354543379, "mean_class_cosine": 0.2492633213424405}}}, "iid_acc": 0.905, "selectivity": 0.43625, "iid_split_rotation": {"subspace_principal_angle": 1.2936213190349724, "mean_class_cosine": 0.2736395752811731}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.84375, "drop": 0.0}, "paraphrase": {"acc": 0.8448979591836735, "drop": -0.001147959183673497, "rotation": {"subspace_principal_angle": 1.2791968807329361, "mean_class_cosine": 0.9478558905818975}}, "domain": {"acc": 0.56625, "drop": 0.27749999999999997, "rotation": {"subspace_principal_angle": 1.5624247261675832, "mean_class_cosine": 0.3303097598518465}}, "length": {"acc": 0.8556231003039514, "drop": -0.01187310030395139, "rotation": {"subspace_principal_angle": 1.3538731412108984, "mean_class_cosine": 0.9594238444634923}}}, "iid_acc": 0.84375, "selectivity": 0.375, "iid_split_rotation": {"subspace_principal_angle": 1.0790921789981887, "mean_class_cosine": 0.9412243630037906}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9142857142857143, "drop": 0.003214285714285725}, "domain": {"acc": 0.8325, "drop": 0.08499999999999996}, "length": {"acc": 0.9133738601823708, "drop": 0.004126139817629215}}, "iid_acc": 0.9175, "selectivity": 0.44875} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8533123028391167, "drop": 0.016687697160883253, "rotation": {"subspace_principal_angle": 1.351979607774418, "mean_class_cosine": 0.3225426038590604}}}, "iid_acc": 0.87, "selectivity": 0.61625, "iid_split_rotation": {"subspace_principal_angle": 1.3508402767430103, "mean_class_cosine": 0.3455994411857771}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.81875, "drop": 0.0}, "paraphrase": {"acc": 0.804416403785489, "drop": 0.014333596214511024, "rotation": {"subspace_principal_angle": 1.2203955464014318, "mean_class_cosine": 0.9625472880536573}}}, "iid_acc": 0.81875, "selectivity": 0.565, "iid_split_rotation": {"subspace_principal_angle": 1.38855296445251, "mean_class_cosine": 0.9564765506858832}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8659305993690851, "drop": 0.017819400630914917}}, "iid_acc": 0.88375, "selectivity": 0.6300000000000001} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.96875, "drop": 0.0}, "paraphrase": {"acc": 0.9640102827763496, "drop": 0.004739717223650408, "rotation": {"subspace_principal_angle": 1.0857448901953828, "mean_class_cosine": 0.6613254867828217}}}, "iid_acc": 0.96875, "selectivity": 0.89375, "iid_split_rotation": {"subspace_principal_angle": 1.076414289934142, "mean_class_cosine": 0.7082374564146615}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6838046272493573, "drop": 0.026195372750642654, "rotation": {"subspace_principal_angle": 1.5350687578200197, "mean_class_cosine": 0.8594207995140349}}}, "iid_acc": 0.71, "selectivity": 0.635, "iid_split_rotation": {"subspace_principal_angle": 1.0895700706545426, "mean_class_cosine": 0.9341302332271046}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9665809768637532, "drop": 0.0009190231362468104}}, "iid_acc": 0.9675, "selectivity": 0.8925000000000001} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.60625, "drop": 0.0}, "paraphrase": {"acc": 0.5930232558139535, "drop": 0.013226744186046413, "rotation": {"subspace_principal_angle": 1.4617197666146766, "mean_class_cosine": 0.10886039552142644}}}, "iid_acc": 0.60625, "selectivity": 0.11499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.466798042788668, "mean_class_cosine": 0.10381091730611129}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": 0.005998062015503902, "rotation": {"subspace_principal_angle": 1.5323338067846937, "mean_class_cosine": 0.9209037611997781}}}, "iid_acc": 0.51375, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.7482139187600273, "mean_class_cosine": 0.7359681788398151}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.57875, "drop": 0.0}, "paraphrase": {"acc": 0.5671834625322998, "drop": 0.0115665374677002}}, "iid_acc": 0.57875, "selectivity": 0.08749999999999997} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.6125, "drop": 0.0}, "paraphrase": {"acc": 0.5926928281461434, "drop": 0.01980717185385661, "rotation": {"subspace_principal_angle": 1.445337558644254, "mean_class_cosine": 0.2135394258196436}}}, "iid_acc": 0.6125, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.42444703173012, "mean_class_cosine": 0.21657072763653504}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.2075, "drop": 0.0}, "paraphrase": {"acc": 0.21109607577807848, "drop": -0.0035960757780784947, "rotation": {"subspace_principal_angle": 1.2059558141673055, "mean_class_cosine": 0.8778793379671459}}}, "iid_acc": 0.2075, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4627224949895972, "mean_class_cosine": 0.5073250273765634}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.6102841677943166, "drop": 0.007215832205683448}}, "iid_acc": 0.6175, "selectivity": 0.395} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7094594594594594, "drop": 0.018040540540540606, "rotation": {"subspace_principal_angle": 1.439408162778464, "mean_class_cosine": 0.13101046781611214}}}, "iid_acc": 0.7275, "selectivity": 0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.440617717267834, "mean_class_cosine": 0.12981124295188987}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5726351351351351, "drop": -0.04263513513513506, "rotation": {"subspace_principal_angle": 1.1956589180448767, "mean_class_cosine": 0.9074769300513319}}}, "iid_acc": 0.53, "selectivity": 0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.4496764494101746, "mean_class_cosine": 0.9668956137286093}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.76875, "drop": 0.0}, "paraphrase": {"acc": 0.7635135135135135, "drop": 0.005236486486486558}}, "iid_acc": 0.76875, "selectivity": 0.26375000000000004} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6634146341463415, "drop": 0.019085365853658498, "rotation": {"subspace_principal_angle": 1.4585106839043807, "mean_class_cosine": 0.11204984079812053}}}, "iid_acc": 0.6825, "selectivity": 0.2025, "iid_split_rotation": {"subspace_principal_angle": 1.4658854601898588, "mean_class_cosine": 0.10471852591428352}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5560975609756098, "drop": -0.004847560975609788, "rotation": {"subspace_principal_angle": 0.7177846273187343, "mean_class_cosine": 0.9945223665026368}}}, "iid_acc": 0.55125, "selectivity": 0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2365217302923708, "mean_class_cosine": 0.9628480684104785}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6731707317073171, "drop": 0.0343292682926829}}, "iid_acc": 0.7075, "selectivity": 0.22750000000000004} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.7210440456769984, "drop": -0.008544045676998402, "rotation": {"subspace_principal_angle": 1.5017342011669776, "mean_class_cosine": 0.06900723919555485}}}, "iid_acc": 0.7125, "selectivity": 0.12375000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4884858123978102, "mean_class_cosine": 0.08221760330215533}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.47125, "drop": 0.0}, "paraphrase": {"acc": 0.5024469820554649, "drop": -0.031196982055464895, "rotation": {"subspace_principal_angle": 0.9746986945864435, "mean_class_cosine": 0.9741912343048922}}}, "iid_acc": 0.47125, "selectivity": -0.1175, "iid_split_rotation": {"subspace_principal_angle": 0.721596565453364, "mean_class_cosine": 0.9782502269601113}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.76625, "drop": 0.0}, "paraphrase": {"acc": 0.7846655791190864, "drop": -0.018415579119086423}}, "iid_acc": 0.76625, "selectivity": 0.1775} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.94375, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.024131471389645776, "rotation": {"subspace_principal_angle": 1.1654703724010276, "mean_class_cosine": 0.39431832489873625}}}, "iid_acc": 0.94375, "selectivity": 0.46499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1557667157993363, "mean_class_cosine": 0.40321702771416507}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.67625, "drop": 0.0}, "paraphrase": {"acc": 0.6743869209809265, "drop": 0.0018630790190735658, "rotation": {"subspace_principal_angle": 1.141242730507284, "mean_class_cosine": 0.9694783192528957}}}, "iid_acc": 0.67625, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 1.4347294745233075, "mean_class_cosine": 0.7429243512254556}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.9318801089918256, "drop": 0.015619891008174402}}, "iid_acc": 0.9475, "selectivity": 0.46875} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9900990099009901, "drop": 0.0024009900990099586, "rotation": {"subspace_principal_angle": 0.9638933203750397, "mean_class_cosine": 0.5703262722196243}}}, "iid_acc": 0.9925, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9412550884957132, "mean_class_cosine": 0.5887740038848737}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.43399339933993397, "drop": 0.223506600660066, "rotation": {"subspace_principal_angle": 1.0823516673301268, "mean_class_cosine": 0.9745159076170016}}}, "iid_acc": 0.6575, "selectivity": -0.13875000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.4071181854905652, "mean_class_cosine": 0.9998877741118155}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9884488448844885, "drop": 0.00655115511551152}}, "iid_acc": 0.995, "selectivity": 0.19874999999999998} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.6772793053545586, "drop": 0.03522069464544142, "rotation": {"subspace_principal_angle": 1.5464207554561848, "mean_class_cosine": 0.02437315754440208}}}, "iid_acc": 0.7125, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 1.4523838978678731, "mean_class_cosine": 0.11813590281951224}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5123010130246021, "drop": 0.016448986975397983, "rotation": {"subspace_principal_angle": 1.5271447554808601, "mean_class_cosine": 0.5274519249488747}}}, "iid_acc": 0.52875, "selectivity": -0.043749999999999956, "iid_split_rotation": {"subspace_principal_angle": 0.7321498396093691, "mean_class_cosine": -0.8691134472727473}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7337192474674384, "drop": 0.007530752532561524}}, "iid_acc": 0.74125, "selectivity": 0.16874999999999996} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6634615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6764705882352942, "drop": -0.013009049773755721, "rotation": {"subspace_principal_angle": 1.4059071468679976, "mean_class_cosine": 0.20250091764866876}}}, "iid_acc": 0.6634615384615384, "selectivity": 0.3125, "iid_split_rotation": {"subspace_principal_angle": 1.4463777868534364, "mean_class_cosine": 0.14772015722986645}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.4951923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.4411764705882353, "drop": 0.05401583710407243, "rotation": {"subspace_principal_angle": 1.155152788254596, "mean_class_cosine": 0.6509669876420695}}}, "iid_acc": 0.4951923076923077, "selectivity": 0.14423076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.3178705376329367, "mean_class_cosine": 0.6023058035884701}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6705882352941176, "drop": -0.026357466063348323}}, "iid_acc": 0.6442307692307693, "selectivity": 0.29326923076923084} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": 0.01030148095909722, "rotation": {"subspace_principal_angle": 1.3664411608119795, "mean_class_cosine": 0.20293578579819216}}}, "iid_acc": 0.90875, "selectivity": 0.2024999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3058033713210209, "mean_class_cosine": 0.2619024695236912}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.7249647390691114, "drop": -0.0012147390691114035, "rotation": {"subspace_principal_angle": 1.3860419944865392, "mean_class_cosine": 0.9342763772250853}}}, "iid_acc": 0.72375, "selectivity": 0.01749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5390421123713445, "mean_class_cosine": 0.8051652349687315}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.9097320169252469, "drop": -0.003482016925246856}}, "iid_acc": 0.90625, "selectivity": 0.19999999999999996} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.78125, "drop": 0.0}, "paraphrase": {"acc": 0.7858136300417247, "drop": -0.0045636300417246645, "rotation": {"subspace_principal_angle": 1.325361507791648, "mean_class_cosine": 0.24297813599017176}}, "domain": {"acc": 0.78, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.4795657935318307, "mean_class_cosine": 0.09110403380754725}}}, "iid_acc": 0.78125, "selectivity": 0.26875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3436786283410562, "mean_class_cosine": 0.22517018028571784}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.55875, "drop": 0.0}, "paraphrase": {"acc": 0.5744089012517385, "drop": -0.015658901251738522, "rotation": {"subspace_principal_angle": 0.6230331668549962, "mean_class_cosine": 0.9910755485040825}}, "domain": {"acc": 0.495, "drop": 0.06374999999999997, "rotation": {"subspace_principal_angle": 1.4491819836983844, "mean_class_cosine": 0.5529294755790912}}}, "iid_acc": 0.55875, "selectivity": 0.04625000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4276392742434922, "mean_class_cosine": 0.9917849416289807}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.7746870653685675, "drop": 0.047812934631432547}, "domain": {"acc": 0.80625, "drop": 0.016249999999999987}}, "iid_acc": 0.8225, "selectivity": 0.31000000000000005} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.8551020408163266, "drop": -0.00010204081632658735, "rotation": {"subspace_principal_angle": 1.3692001923291004, "mean_class_cosine": 0.20023339510507437}}, "domain": {"acc": 0.6, "drop": 0.255, "rotation": {"subspace_principal_angle": 1.468428009163586, "mean_class_cosine": 0.10218962037267862}}, "length": {"acc": 0.8556231003039514, "drop": -0.0006231003039514071, "rotation": {"subspace_principal_angle": 1.3923528233365947, "mean_class_cosine": 0.17749800787298398}}}, "iid_acc": 0.855, "selectivity": 0.37374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3944016978212828, "mean_class_cosine": 0.17548129599139065}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7510204081632653, "drop": 0.0014795918367346284, "rotation": {"subspace_principal_angle": 1.3529173793192326, "mean_class_cosine": 0.9403316707188716}}, "domain": {"acc": 0.56625, "drop": 0.18624999999999992, "rotation": {"subspace_principal_angle": 1.553813034487346, "mean_class_cosine": 0.4365397449834313}}, "length": {"acc": 0.7613981762917933, "drop": -0.008898176291793347, "rotation": {"subspace_principal_angle": 1.484045792925862, "mean_class_cosine": 0.9449458799225073}}}, "iid_acc": 0.7525, "selectivity": 0.27124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.102117298311522, "mean_class_cosine": 0.9232016388495949}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.02272959183673473}, "domain": {"acc": 0.67125, "drop": 0.2025}, "length": {"acc": 0.8723404255319149, "drop": 0.0014095744680850952}}, "iid_acc": 0.87375, "selectivity": 0.3925} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8280757097791798, "drop": 0.024424290220820266, "rotation": {"subspace_principal_angle": 1.4366178069899613, "mean_class_cosine": 0.2843840411694384}}}, "iid_acc": 0.8525, "selectivity": 0.6175, "iid_split_rotation": {"subspace_principal_angle": 1.3589778090630555, "mean_class_cosine": 0.3305635161629984}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5930599369085173, "drop": 0.018190063091482633, "rotation": {"subspace_principal_angle": 0.6501933782850093, "mean_class_cosine": 0.9238960788618794}}}, "iid_acc": 0.61125, "selectivity": 0.37625, "iid_split_rotation": {"subspace_principal_angle": 0.576594648277471, "mean_class_cosine": 0.9553403579368684}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.861198738170347, "drop": 0.028801261829653035}}, "iid_acc": 0.89, "selectivity": 0.655} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9383033419023136, "drop": 0.007946658097686465, "rotation": {"subspace_principal_angle": 1.2309707985530614, "mean_class_cosine": 0.5687671045365533}}}, "iid_acc": 0.94625, "selectivity": 0.85875, "iid_split_rotation": {"subspace_principal_angle": 1.1443944465717892, "mean_class_cosine": 0.6294410202468741}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7095115681233933, "drop": 0.03423843187660669, "rotation": {"subspace_principal_angle": 1.3105209680737453, "mean_class_cosine": 0.8754616020973991}}}, "iid_acc": 0.74375, "selectivity": 0.65625, "iid_split_rotation": {"subspace_principal_angle": 1.3306223102176358, "mean_class_cosine": 0.927772151836221}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94, "drop": 0.0}, "paraphrase": {"acc": 0.9203084832904884, "drop": 0.019691516709511525}}, "iid_acc": 0.94, "selectivity": 0.8524999999999999} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5116279069767442, "drop": 0.019622093023255793, "rotation": {"subspace_principal_angle": 1.4907936041998628, "mean_class_cosine": 0.07991740785625166}}}, "iid_acc": 0.53125, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.494326896717623, "mean_class_cosine": 0.07639492509245417}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": -0.01025193798449614, "rotation": {"subspace_principal_angle": 1.4876221863427175, "mean_class_cosine": 0.8807712318567114}}}, "iid_acc": 0.4975, "selectivity": 0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.2199667575281368, "mean_class_cosine": 0.8996775514646336}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.47375, "drop": 0.0}, "paraphrase": {"acc": 0.5064599483204134, "drop": -0.03270994832041341}}, "iid_acc": 0.47375, "selectivity": -0.00874999999999998} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6275, "drop": 0.0}, "paraphrase": {"acc": 0.5723951285520974, "drop": 0.05510487144790255, "rotation": {"subspace_principal_angle": 1.4343786178569309, "mean_class_cosine": 0.2763953601626957}}}, "iid_acc": 0.6275, "selectivity": 0.37625, "iid_split_rotation": {"subspace_principal_angle": 1.3550101339445768, "mean_class_cosine": 0.31016139289436095}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.24125, "drop": 0.0}, "paraphrase": {"acc": 0.2449255751014885, "drop": -0.0036755751014884985, "rotation": {"subspace_principal_angle": 1.2918205373490343, "mean_class_cosine": 0.7766803935782208}}}, "iid_acc": 0.24125, "selectivity": -0.009999999999999981, "iid_split_rotation": {"subspace_principal_angle": 1.5684032048848013, "mean_class_cosine": 0.5981109246694539}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.60875, "drop": 0.0}, "paraphrase": {"acc": 0.5683355886332883, "drop": 0.040414411366711755}}, "iid_acc": 0.60875, "selectivity": 0.35750000000000004} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7398648648648649, "drop": 0.00013513513513507824, "rotation": {"subspace_principal_angle": 1.2673314142482994, "mean_class_cosine": 0.2988285834755996}}}, "iid_acc": 0.74, "selectivity": 0.24, "iid_split_rotation": {"subspace_principal_angle": 1.296110954087143, "mean_class_cosine": 0.27124411828692846}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.6199324324324325, "drop": -0.01993243243243248, "rotation": {"subspace_principal_angle": 1.5658191148065415, "mean_class_cosine": 0.9653627770010756}}}, "iid_acc": 0.6, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.684621553461914, "mean_class_cosine": 0.9768155693913088}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7246621621621622, "drop": 0.010337837837837816}}, "iid_acc": 0.735, "selectivity": 0.235} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6308943089430894, "drop": 0.014105691056910619, "rotation": {"subspace_principal_angle": 1.4810005396497288, "mean_class_cosine": 0.08967516097444077}}}, "iid_acc": 0.645, "selectivity": 0.16875, "iid_split_rotation": {"subspace_principal_angle": 1.4383531851019704, "mean_class_cosine": 0.13205627952425741}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.551219512195122, "drop": -0.008719512195122014, "rotation": {"subspace_principal_angle": 1.5478870220336094, "mean_class_cosine": 0.9974275968935058}}}, "iid_acc": 0.5425, "selectivity": 0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3106732220928752, "mean_class_cosine": 0.9809909328157596}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6861788617886179, "drop": -0.011178861788617822}}, "iid_acc": 0.675, "selectivity": 0.19875000000000004} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.7225, "drop": 0.0}, "paraphrase": {"acc": 0.7177814029363785, "drop": 0.0047185970636215435, "rotation": {"subspace_principal_angle": 1.3869179562010008, "mean_class_cosine": 0.18284392784898534}}}, "iid_acc": 0.7225, "selectivity": 0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3863158843348609, "mean_class_cosine": 0.18343581678343168}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.46492659053833607, "drop": 0.040073409461663934, "rotation": {"subspace_principal_angle": 1.3638991068226356, "mean_class_cosine": 0.5772026508098606}}}, "iid_acc": 0.505, "selectivity": -0.12, "iid_split_rotation": {"subspace_principal_angle": 1.0426083209720547, "mean_class_cosine": 0.9895063250474988}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7210440456769984, "drop": 0.003955954323001554}}, "iid_acc": 0.725, "selectivity": 0.09999999999999998} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8950953678474114, "drop": 0.013654632152588553, "rotation": {"subspace_principal_angle": 1.1926328016497498, "mean_class_cosine": 0.36921437562940024}}}, "iid_acc": 0.90875, "selectivity": 0.43124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2197817002343454, "mean_class_cosine": 0.34385074329564713}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5994550408719346, "drop": -0.019455040871934615, "rotation": {"subspace_principal_angle": 1.0711301307338577, "mean_class_cosine": 0.9676646451280264}}}, "iid_acc": 0.58, "selectivity": 0.10249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4692738400107856, "mean_class_cosine": 0.5961110196205139}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.9059945504087193, "drop": -0.003494550408719377}}, "iid_acc": 0.9025, "selectivity": 0.425} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9851485148514851, "drop": 0.0073514851485149135, "rotation": {"subspace_principal_angle": 0.9650258351359023, "mean_class_cosine": 0.5693956383770296}}}, "iid_acc": 0.9925, "selectivity": 0.15250000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9566139505713168, "mean_class_cosine": 0.5762905161188008}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.7838283828382838, "drop": -0.10132838283828383, "rotation": {"subspace_principal_angle": 0.7837036206433732, "mean_class_cosine": 0.9616722641195252}}}, "iid_acc": 0.6825, "selectivity": -0.15749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2673965735923087, "mean_class_cosine": 0.9865395355198125}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9875, "drop": 0.0}, "paraphrase": {"acc": 0.9834983498349835, "drop": 0.004001650165016524}}, "iid_acc": 0.9875, "selectivity": 0.14750000000000008} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.6801736613603473, "drop": 0.009826338639652654, "rotation": {"subspace_principal_angle": 1.446957515536241, "mean_class_cosine": 0.12352252083060832}}}, "iid_acc": 0.69, "selectivity": 0.0724999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.482921903323322, "mean_class_cosine": 0.08776137400044447}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5354558610709117, "drop": 0.022044138929088297, "rotation": {"subspace_principal_angle": 1.0513666721812371, "mean_class_cosine": 0.686624560357272}}}, "iid_acc": 0.5575, "selectivity": -0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.6796913400547621, "mean_class_cosine": 0.5471521743827539}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7308248914616498, "drop": -0.01207489146164975}}, "iid_acc": 0.71875, "selectivity": 0.10124999999999995} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": 0.013744343891402777, "rotation": {"subspace_principal_angle": 1.4301231189422874, "mean_class_cosine": 0.21099799551640833}}}, "iid_acc": 0.6490384615384616, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.480503372414954, "mean_class_cosine": 0.14932363173924582}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5288461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": -0.0005656108597285048, "rotation": {"subspace_principal_angle": 1.2751790898102846, "mean_class_cosine": 0.6777353047064866}}}, "iid_acc": 0.5288461538461539, "selectivity": 0.13942307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.437011284720561, "mean_class_cosine": 0.6376580778974325}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": -0.010294117647058787}}, "iid_acc": 0.625, "selectivity": 0.23557692307692307} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8899858956276445, "drop": 0.011264104372355477, "rotation": {"subspace_principal_angle": 1.2097432263622687, "mean_class_cosine": 0.3532596311031677}}}, "iid_acc": 0.90125, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.2597303970258766, "mean_class_cosine": 0.306073583648639}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5077574047954866, "drop": 0.017242595204513433, "rotation": {"subspace_principal_angle": 1.5668153896789387, "mean_class_cosine": 0.821628018438601}}}, "iid_acc": 0.525, "selectivity": -0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.244997779584516, "mean_class_cosine": 0.9282425176629551}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8561354019746121, "drop": 0.01886459802538787}}, "iid_acc": 0.875, "selectivity": 0.12749999999999995} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.7927677329624478, "drop": 0.030982267037552136, "rotation": {"subspace_principal_angle": 1.3441679611646846, "mean_class_cosine": 0.22469338685922752}}, "domain": {"acc": 0.68875, "drop": 0.135, "rotation": {"subspace_principal_angle": 1.4619007929505858, "mean_class_cosine": 0.10868044323365433}}}, "iid_acc": 0.82375, "selectivity": 0.2925, "iid_split_rotation": {"subspace_principal_angle": 1.3382551814356736, "mean_class_cosine": 0.23045101264055012}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.55875, "drop": 0.0}, "paraphrase": {"acc": 0.5744089012517385, "drop": -0.015658901251738522, "rotation": {"subspace_principal_angle": 1.1096712314062809, "mean_class_cosine": 0.9710046862155046}}, "domain": {"acc": 0.495, "drop": 0.06374999999999997, "rotation": {"subspace_principal_angle": 1.3394837230563983, "mean_class_cosine": 0.4067141442483526}}}, "iid_acc": 0.55875, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.0571042364623964, "mean_class_cosine": 0.9732840432471883}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.7955493741307371, "drop": 0.05820062586926289}, "domain": {"acc": 0.8325, "drop": 0.02124999999999999}}, "iid_acc": 0.85375, "selectivity": 0.3225} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5297157622739018, "drop": 0.02403423772609814, "rotation": {"subspace_principal_angle": 1.4849996721834149, "mean_class_cosine": 0.08569143420695843}}}, "iid_acc": 0.55375, "selectivity": 0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4079692233979924, "mean_class_cosine": 0.1621085598267091}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": -0.009001937984496111, "rotation": {"subspace_principal_angle": 1.5312281927446767, "mean_class_cosine": 0.9121731321951317}}}, "iid_acc": 0.49875, "selectivity": -0.013749999999999929, "iid_split_rotation": {"subspace_principal_angle": 1.414487810468904, "mean_class_cosine": 0.9131370420534566}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5129198966408268, "drop": -0.004169896640826787}}, "iid_acc": 0.50875, "selectivity": -0.00374999999999992} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6224627875507442, "drop": 0.03503721244925573, "rotation": {"subspace_principal_angle": 1.377806750699231, "mean_class_cosine": 0.3147008679395653}}}, "iid_acc": 0.6575, "selectivity": 0.40375, "iid_split_rotation": {"subspace_principal_angle": 1.3221669271577434, "mean_class_cosine": 0.32518311136584716}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.27375, "drop": 0.0}, "paraphrase": {"acc": 0.2665764546684709, "drop": 0.007173545331529085, "rotation": {"subspace_principal_angle": 1.3088731708244634, "mean_class_cosine": 0.746477870474282}}}, "iid_acc": 0.27375, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4761219391324176, "mean_class_cosine": 0.6090428280673176}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5737483085250338, "drop": 0.025001691474966226}}, "iid_acc": 0.59875, "selectivity": 0.34500000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.722972972972973, "drop": 0.018277027027026937, "rotation": {"subspace_principal_angle": 1.3556838479482727, "mean_class_cosine": 0.21345731612896818}}}, "iid_acc": 0.74125, "selectivity": 0.24, "iid_split_rotation": {"subspace_principal_angle": 1.3660213343678467, "mean_class_cosine": 0.20334685861759932}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5827702702702703, "drop": 0.00597972972972971, "rotation": {"subspace_principal_angle": 0.7662797422733312, "mean_class_cosine": 0.8372309786422542}}}, "iid_acc": 0.58875, "selectivity": 0.08750000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.19745934756960443, "mean_class_cosine": 0.9910109223557726}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7466216216216216, "drop": -0.007871621621621583}}, "iid_acc": 0.73875, "selectivity": 0.23750000000000004} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.65625, "drop": 0.0}, "paraphrase": {"acc": 0.6113821138211382, "drop": 0.04486788617886184, "rotation": {"subspace_principal_angle": 1.5244430780993927, "mean_class_cosine": 0.046336651197513255}}}, "iid_acc": 0.65625, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 1.48060525921707, "mean_class_cosine": 0.09006884183258959}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5560975609756098, "drop": -0.01609756097560977, "rotation": {"subspace_principal_angle": 1.4447759012617603, "mean_class_cosine": 0.9966733139859822}}}, "iid_acc": 0.54, "selectivity": 0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5053011307485271, "mean_class_cosine": 0.9733537861349781}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.6682926829268293, "drop": 0.024207317073170698}}, "iid_acc": 0.6925, "selectivity": 0.17625000000000002} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7047308319738989, "drop": 0.014019168026101148, "rotation": {"subspace_principal_angle": 1.4273253943610407, "mean_class_cosine": 0.14297923999787693}}}, "iid_acc": 0.71875, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.410754254834671, "mean_class_cosine": 0.15935974106023298}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.49375, "drop": 0.0}, "paraphrase": {"acc": 0.46492659053833607, "drop": 0.02882340946166395, "rotation": {"subspace_principal_angle": 1.2475152228835678, "mean_class_cosine": -0.2873781232568553}}}, "iid_acc": 0.49375, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 0.3595110102120592, "mean_class_cosine": 0.9955684555352899}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7373572593800979, "drop": 0.005142740619902164}}, "iid_acc": 0.7425, "selectivity": 0.15000000000000002} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.01163147138964582, "rotation": {"subspace_principal_angle": 1.1462155217150516, "mean_class_cosine": 0.4119388427182844}}}, "iid_acc": 0.93125, "selectivity": 0.4525, "iid_split_rotation": {"subspace_principal_angle": 1.2036153547134156, "mean_class_cosine": 0.3589857417663745}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.57375, "drop": 0.0}, "paraphrase": {"acc": 0.5885558583106267, "drop": -0.014805858310626752, "rotation": {"subspace_principal_angle": 0.6949506609048397, "mean_class_cosine": 0.9687824706589729}}}, "iid_acc": 0.57375, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4852957312103852, "mean_class_cosine": 0.7009380316990886}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.9155313351498637, "drop": -0.00428133514986373}}, "iid_acc": 0.91125, "selectivity": 0.4325} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6555716353111433, "drop": 0.0544283646888567, "rotation": {"subspace_principal_angle": 1.530036517597367, "mean_class_cosine": 0.04074852400167777}}}, "iid_acc": 0.71, "selectivity": 0.12249999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4332650590425584, "mean_class_cosine": 0.1370981137331397}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5180897250361794, "drop": 0.008160274963820546, "rotation": {"subspace_principal_angle": 1.3868122762825261, "mean_class_cosine": 0.5674558871637044}}}, "iid_acc": 0.52625, "selectivity": -0.06125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3698230368243622, "mean_class_cosine": -0.902684531716938}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": 0.03450253256150515}}, "iid_acc": 0.75375, "selectivity": 0.16625} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": -0.004977375565610842, "rotation": {"subspace_principal_angle": 1.4918296501472632, "mean_class_cosine": 0.14481202131126455}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.2548076923076923, "iid_split_rotation": {"subspace_principal_angle": 1.4905675320918412, "mean_class_cosine": 0.12376594768867437}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.5576923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.5235294117647059, "drop": 0.0341628959276018, "rotation": {"subspace_principal_angle": 1.1976344371529, "mean_class_cosine": 0.7296312684418567}}}, "iid_acc": 0.5576923076923077, "selectivity": 0.15865384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.5298301184380643, "mean_class_cosine": 0.5946390438469394}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.6298076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": -0.005486425339366496}}, "iid_acc": 0.6298076923076923, "selectivity": 0.23076923076923073} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.9012693935119888, "drop": -0.0012693935119887367, "rotation": {"subspace_principal_angle": 1.3718118378240132, "mean_class_cosine": 0.1976739601069373}}}, "iid_acc": 0.9, "selectivity": 0.21999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.320617301683332, "mean_class_cosine": 0.24757741494334223}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7475317348377997, "drop": -0.015031734837799626, "rotation": {"subspace_principal_angle": 0.7004143627777456, "mean_class_cosine": 0.8791149950634916}}}, "iid_acc": 0.7325, "selectivity": 0.05249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.298216893295834, "mean_class_cosine": 0.687969391669262}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.9055007052186178, "drop": -0.0005007052186177807}}, "iid_acc": 0.905, "selectivity": 0.22499999999999998} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8150208623087621, "drop": 0.031229137691237807, "rotation": {"subspace_principal_angle": 1.2085181784579044, "mean_class_cosine": 0.3544054290791189}}, "domain": {"acc": 0.84375, "drop": 0.0024999999999999467, "rotation": {"subspace_principal_angle": 1.2026486146003328, "mean_class_cosine": 0.35988787403384076}}}, "iid_acc": 0.84625, "selectivity": 0.3499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1571656029830548, "mean_class_cosine": 0.40193650581631374}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5660639777468707, "drop": -0.014813977746870655, "rotation": {"subspace_principal_angle": 1.245440017394346, "mean_class_cosine": 0.9990012115847782}}, "domain": {"acc": 0.495, "drop": 0.05625000000000002, "rotation": {"subspace_principal_angle": 1.122537065823632, "mean_class_cosine": 0.8260255526449155}}}, "iid_acc": 0.55125, "selectivity": 0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.503465091230446, "mean_class_cosine": 0.9992316721985006}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.79375, "drop": 0.0}, "paraphrase": {"acc": 0.8108484005563282, "drop": -0.017098400556328275}, "domain": {"acc": 0.81875, "drop": -0.025000000000000022}}, "iid_acc": 0.79375, "selectivity": 0.29749999999999993} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.889795918367347, "drop": 0.0089540816326531, "rotation": {"subspace_principal_angle": 1.1767967051679535, "mean_class_cosine": 0.38388465061115085}}, "domain": {"acc": 0.62875, "drop": 0.27, "rotation": {"subspace_principal_angle": 1.383856999934436, "mean_class_cosine": 0.18585242109581535}}, "length": {"acc": 0.9057750759878419, "drop": -0.007025075987841856, "rotation": {"subspace_principal_angle": 1.1410061611227862, "mean_class_cosine": 0.41668006103594923}}}, "iid_acc": 0.89875, "selectivity": 0.37875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2153568944914745, "mean_class_cosine": 0.3480023638189218}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.67875, "drop": 0.0}, "paraphrase": {"acc": 0.6959183673469388, "drop": -0.01716836734693883, "rotation": {"subspace_principal_angle": 1.4280844473025907, "mean_class_cosine": 0.9628713037828026}}, "domain": {"acc": 0.56625, "drop": 0.11249999999999993, "rotation": {"subspace_principal_angle": 1.3603869108648967, "mean_class_cosine": 0.44643897304207514}}, "length": {"acc": 0.6990881458966566, "drop": -0.020338145896656612, "rotation": {"subspace_principal_angle": 1.3643554774870585, "mean_class_cosine": 0.9296532611805729}}}, "iid_acc": 0.67875, "selectivity": 0.15874999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4371116325199274, "mean_class_cosine": 0.8728452884225864}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8959183673469387, "drop": 0.012831632653061198}, "domain": {"acc": 0.7025, "drop": 0.20624999999999993}, "length": {"acc": 0.9103343465045592, "drop": -0.0015843465045592975}}, "iid_acc": 0.90875, "selectivity": 0.38874999999999993} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8690851735015773, "drop": 0.013414826498422627, "rotation": {"subspace_principal_angle": 1.3839824797303943, "mean_class_cosine": 0.31779551037707243}}}, "iid_acc": 0.8825, "selectivity": 0.6425, "iid_split_rotation": {"subspace_principal_angle": 1.3316581856764766, "mean_class_cosine": 0.3540098873003071}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.64625, "drop": 0.0}, "paraphrase": {"acc": 0.6403785488958991, "drop": 0.005871451104100922, "rotation": {"subspace_principal_angle": 0.9413392447354468, "mean_class_cosine": 0.9542917678949543}}}, "iid_acc": 0.64625, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 1.3936985587921362, "mean_class_cosine": 0.957153329410988}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8706624605678234, "drop": 0.014337539432176638}}, "iid_acc": 0.885, "selectivity": 0.645} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9588688946015425, "drop": -0.00011889460154246301, "rotation": {"subspace_principal_angle": 1.1552531570580125, "mean_class_cosine": 0.6468507558692805}}}, "iid_acc": 0.95875, "selectivity": 0.8825, "iid_split_rotation": {"subspace_principal_angle": 1.135852999209662, "mean_class_cosine": 0.6934419744150427}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.6349614395886889, "drop": 0.055038560411311055, "rotation": {"subspace_principal_angle": 0.9810281506834297, "mean_class_cosine": 0.8971151689162156}}}, "iid_acc": 0.69, "selectivity": 0.6137499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9002976361653219, "mean_class_cosine": 0.9196584919183077}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9588688946015425, "drop": -0.00011889460154246301}}, "iid_acc": 0.95875, "selectivity": 0.8825} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.6253229974160207, "drop": -0.01782299741602067, "rotation": {"subspace_principal_angle": 1.2853027316760957, "mean_class_cosine": 0.28163110133345426}}}, "iid_acc": 0.6075, "selectivity": 0.12625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3656019976170561, "mean_class_cosine": 0.20375741617602003}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.50375, "drop": 0.0}, "paraphrase": {"acc": 0.5258397932816538, "drop": -0.022089793281653725, "rotation": {"subspace_principal_angle": 1.415941389813765, "mean_class_cosine": 0.989806795103265}}}, "iid_acc": 0.50375, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.1847539410134869, "mean_class_cosine": 0.9708123793493897}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5658914728682171, "drop": -0.008391472868217065}}, "iid_acc": 0.5575, "selectivity": 0.07624999999999998} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5520974289580515, "drop": 0.04665257104194853, "rotation": {"subspace_principal_angle": 1.2698836003378577, "mean_class_cosine": 0.41354171313299243}}}, "iid_acc": 0.59875, "selectivity": 0.36125, "iid_split_rotation": {"subspace_principal_angle": 1.279435072679175, "mean_class_cosine": 0.4311997525753071}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.1775, "drop": 0.0}, "paraphrase": {"acc": 0.18132611637347767, "drop": -0.0038261163734776837, "rotation": {"subspace_principal_angle": 1.2291192051808308, "mean_class_cosine": 0.9975774044324508}}}, "iid_acc": 0.1775, "selectivity": -0.06, "iid_split_rotation": {"subspace_principal_angle": 1.4905226932759024, "mean_class_cosine": 0.6650981545648424}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5128552097428958, "drop": 0.013394790257104217}}, "iid_acc": 0.52625, "selectivity": 0.28875} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7398648648648649, "drop": 0.0038851351351351093, "rotation": {"subspace_principal_angle": 1.2773857388810559, "mean_class_cosine": 0.289218730636968}}}, "iid_acc": 0.74375, "selectivity": 0.23625000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.283587337214898, "mean_class_cosine": 0.2832766461897106}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5202702702702703, "drop": -0.012770270270270334, "rotation": {"subspace_principal_angle": 1.416822487851161, "mean_class_cosine": 0.9990366010707538}}}, "iid_acc": 0.5075, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.172452398513104, "mean_class_cosine": 0.9997365426935431}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7212837837837838, "drop": 0.009966216216216184}}, "iid_acc": 0.73125, "selectivity": 0.22375} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6260162601626016, "drop": 0.04773373983739837, "rotation": {"subspace_principal_angle": 1.4713221046099612, "mean_class_cosine": 0.09931025175584188}}}, "iid_acc": 0.67375, "selectivity": 0.18624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4688407807398294, "mean_class_cosine": 0.10177900098741591}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5495934959349593, "drop": -0.007093495934959337, "rotation": {"subspace_principal_angle": 1.1126001122879392, "mean_class_cosine": 0.9966307622920645}}}, "iid_acc": 0.5425, "selectivity": 0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2227358373040218, "mean_class_cosine": 0.9830887042288201}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6325203252032521, "drop": 0.044979674796747915}}, "iid_acc": 0.6775, "selectivity": 0.19} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.734094616639478, "drop": 0.0009053833605220385, "rotation": {"subspace_principal_angle": 1.3384619923865202, "mean_class_cosine": 0.23024976330131475}}}, "iid_acc": 0.735, "selectivity": 0.14874999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3807433555450965, "mean_class_cosine": 0.1889109127031547}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.46375, "drop": 0.0}, "paraphrase": {"acc": 0.4926590538336052, "drop": -0.028909053833605203, "rotation": {"subspace_principal_angle": 1.240506779082087, "mean_class_cosine": 0.9979825325820296}}}, "iid_acc": 0.46375, "selectivity": -0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4130621801133338, "mean_class_cosine": 0.9984335729440443}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.755, "drop": 0.0}, "paraphrase": {"acc": 0.7504078303425775, "drop": 0.004592169657422485}}, "iid_acc": 0.755, "selectivity": 0.16874999999999996} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.93, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.010381471389645847, "rotation": {"subspace_principal_angle": 1.1842664562934508, "mean_class_cosine": 0.37697657827864284}}}, "iid_acc": 0.93, "selectivity": 0.42125, "iid_split_rotation": {"subspace_principal_angle": 1.1417564076798083, "mean_class_cosine": 0.4159979297955525}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5722070844686649, "drop": -0.014707084468664866, "rotation": {"subspace_principal_angle": 0.9890820205173421, "mean_class_cosine": 0.9786448762488624}}}, "iid_acc": 0.5575, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4310370860587216, "mean_class_cosine": 0.6707511189916638}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9291553133514986, "drop": 0.0020946866485014315}}, "iid_acc": 0.93125, "selectivity": 0.4225} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9900990099009901, "drop": 0.0024009900990099586, "rotation": {"subspace_principal_angle": 0.9546523855786807, "mean_class_cosine": 0.5778924853874827}}}, "iid_acc": 0.9925, "selectivity": 0.15500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.9113621026036162, "mean_class_cosine": 0.6126697953735774}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.6815181518151815, "drop": -0.017768151815181543, "rotation": {"subspace_principal_angle": 0.7769812943933525, "mean_class_cosine": 0.9999668724105906}}}, "iid_acc": 0.66375, "selectivity": -0.17375000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.4808794246553239, "mean_class_cosine": 0.999987060235844}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9867986798679867, "drop": 0.005701320132013299}}, "iid_acc": 0.9925, "selectivity": 0.15500000000000003} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.6975397973950795, "drop": 0.047460202604920454, "rotation": {"subspace_principal_angle": 1.4955638990865465, "mean_class_cosine": 0.07516147955849134}}}, "iid_acc": 0.745, "selectivity": 0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.378258975516904, "mean_class_cosine": 0.19134997405306095}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.53875, "drop": 0.0}, "paraphrase": {"acc": 0.5340086830680174, "drop": 0.004741316931982542, "rotation": {"subspace_principal_angle": 1.021497555457415, "mean_class_cosine": 0.9732173089300729}}}, "iid_acc": 0.53875, "selectivity": -0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5200003895350536, "mean_class_cosine": -0.9852001616272563}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7264833574529667, "drop": -0.016483357452966696}}, "iid_acc": 0.71, "selectivity": 0.08624999999999994} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6634615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6764705882352942, "drop": -0.013009049773755721, "rotation": {"subspace_principal_angle": 1.411172607075842, "mean_class_cosine": 0.18933079531066535}}}, "iid_acc": 0.6634615384615384, "selectivity": 0.2740384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.4720863809909137, "mean_class_cosine": 0.12696001888551056}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.5764705882352941, "drop": 0.0052601809954752055, "rotation": {"subspace_principal_angle": 1.0998469877442718, "mean_class_cosine": 0.6836666330824704}}}, "iid_acc": 0.5817307692307693, "selectivity": 0.19230769230769235, "iid_split_rotation": {"subspace_principal_angle": 1.540737942845596, "mean_class_cosine": 0.6936262779934396}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6470588235294118, "drop": 0.006787330316742057}}, "iid_acc": 0.6538461538461539, "selectivity": 0.2644230769230769} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": -0.0009485190409027622, "rotation": {"subspace_principal_angle": 1.2279252167452932, "mean_class_cosine": 0.3361924663049618}}}, "iid_acc": 0.8975, "selectivity": 0.15499999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.2122444869762357, "mean_class_cosine": 0.3509185358458786}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.6234132581100141, "drop": -0.005913258110014086, "rotation": {"subspace_principal_angle": 0.8729067617905422, "mean_class_cosine": 0.7807952771130344}}}, "iid_acc": 0.6175, "selectivity": -0.125, "iid_split_rotation": {"subspace_principal_angle": 0.5477116558377627, "mean_class_cosine": 0.21905717301122568}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.9026798307475318, "drop": 0.003570169252468225}}, "iid_acc": 0.90625, "selectivity": 0.16374999999999995} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7503467406380028, "drop": -0.007846740638002725, "rotation": {"subspace_principal_angle": 1.1359120026803335, "mean_class_cosine": 0.4213054955523352}}, "domain": {"acc": 0.75, "drop": -0.007499999999999951, "rotation": {"subspace_principal_angle": 1.2165320665064534, "mean_class_cosine": 0.34690040746322265}}}, "iid_acc": 0.7425, "selectivity": 0.18375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1365226365220666, "mean_class_cosine": 0.4207516217391289}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6324549237170597, "drop": -0.009954923717059638, "rotation": {"subspace_principal_angle": 1.4631222901212197, "mean_class_cosine": 0.8231564279245471}}, "domain": {"acc": 0.5025, "drop": 0.1200000000000001, "rotation": {"subspace_principal_angle": 1.3032516443365474, "mean_class_cosine": 0.7093055206790797}}}, "iid_acc": 0.6225, "selectivity": 0.06375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.0549484337600161, "mean_class_cosine": 0.8394584109810561}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.7475, "drop": 0.0}, "paraphrase": {"acc": 0.7517337031900139, "drop": -0.0042337031900138156}, "domain": {"acc": 0.7575, "drop": -0.009999999999999898}}, "iid_acc": 0.7475, "selectivity": 0.18875000000000008} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.7747572815533981, "drop": 0.02774271844660192, "rotation": {"subspace_principal_angle": 1.0704211059136206, "mean_class_cosine": 0.4797547921494105}}, "domain": {"acc": 0.5775, "drop": 0.22499999999999998, "rotation": {"subspace_principal_angle": 1.3307188056276382, "mean_class_cosine": 0.23777792496643144}}, "length": {"acc": 0.8054298642533937, "drop": -0.002929864253393699, "rotation": {"subspace_principal_angle": 1.0251786604711886, "mean_class_cosine": 0.5189461749096541}}}, "iid_acc": 0.8025, "selectivity": 0.26, "iid_split_rotation": {"subspace_principal_angle": 1.1020990971387528, "mean_class_cosine": 0.4517243926605984}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.76125, "drop": 0.0}, "paraphrase": {"acc": 0.7475728155339806, "drop": 0.013677184466019399, "rotation": {"subspace_principal_angle": 1.3667128970253954, "mean_class_cosine": 0.8708735415100642}}, "domain": {"acc": 0.6175, "drop": 0.14374999999999993, "rotation": {"subspace_principal_angle": 1.3448216837274964, "mean_class_cosine": 0.549993690079513}}, "length": {"acc": 0.751131221719457, "drop": 0.010118778280542973, "rotation": {"subspace_principal_angle": 1.5333491169148545, "mean_class_cosine": 0.8840723107383672}}}, "iid_acc": 0.76125, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.4498931415198635, "mean_class_cosine": 0.8774451765117859}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.79, "drop": 0.0}, "paraphrase": {"acc": 0.7708737864077669, "drop": 0.0191262135922331}, "domain": {"acc": 0.6, "drop": 0.19000000000000006}, "length": {"acc": 0.799396681749623, "drop": -0.009396681749622937}}, "iid_acc": 0.79, "selectivity": 0.24750000000000005} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8440944881889764, "drop": 0.014655511811023647, "rotation": {"subspace_principal_angle": 1.420395850894294, "mean_class_cosine": 0.3177960201739077}}}, "iid_acc": 0.85875, "selectivity": 0.6012500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.457737166249678, "mean_class_cosine": 0.2964930332276186}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.83, "drop": 0.0}, "paraphrase": {"acc": 0.8125984251968504, "drop": 0.017401574803149522, "rotation": {"subspace_principal_angle": 1.4786133211134285, "mean_class_cosine": 0.9691458832694383}}}, "iid_acc": 0.83, "selectivity": 0.5725, "iid_split_rotation": {"subspace_principal_angle": 1.5646208839230873, "mean_class_cosine": 0.9660745167896441}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.84251968503937, "drop": 0.02748031496062997}}, "iid_acc": 0.87, "selectivity": 0.6125} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9193954659949622, "drop": 0.019354534005037727, "rotation": {"subspace_principal_angle": 1.2541440798612775, "mean_class_cosine": 0.6719234888555385}}}, "iid_acc": 0.93875, "selectivity": 0.8875, "iid_split_rotation": {"subspace_principal_angle": 1.3380418703494907, "mean_class_cosine": 0.7071156351962342}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.84125, "drop": 0.0}, "paraphrase": {"acc": 0.8312342569269522, "drop": 0.010015743073047867, "rotation": {"subspace_principal_angle": 1.4201631598283133, "mean_class_cosine": 0.8910464593425866}}}, "iid_acc": 0.84125, "selectivity": 0.79, "iid_split_rotation": {"subspace_principal_angle": 1.5685103886275429, "mean_class_cosine": 0.9592733462262945}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.9017632241813602, "drop": 0.01698677581863972}}, "iid_acc": 0.91875, "selectivity": 0.8674999999999999} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5058670143415906, "drop": 0.012882985658409396, "rotation": {"subspace_principal_angle": 1.4825642540594317, "mean_class_cosine": 0.08811763766369723}}}, "iid_acc": 0.51875, "selectivity": 0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.514014896298524, "mean_class_cosine": -0.056750923621074825}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.49375, "drop": 0.0}, "paraphrase": {"acc": 0.4980443285528031, "drop": -0.004294328552803095, "rotation": {"subspace_principal_angle": 1.5186826580706572, "mean_class_cosine": -0.08658206386049154}}}, "iid_acc": 0.49375, "selectivity": -0.0012499999999999734, "iid_split_rotation": {"subspace_principal_angle": 1.5356626656582177, "mean_class_cosine": 0.14967252329717184}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.516297262059974, "drop": -0.001297262059973936}}, "iid_acc": 0.515, "selectivity": 0.020000000000000018} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5391766268260292, "drop": 0.02707337317397085, "rotation": {"subspace_principal_angle": 1.3218665508999698, "mean_class_cosine": 0.40140129389793905}}}, "iid_acc": 0.56625, "selectivity": 0.30500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4544563315042423, "mean_class_cosine": 0.3616549205752137}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.2425, "drop": 0.0}, "paraphrase": {"acc": 0.1952191235059761, "drop": 0.047280876494023905, "rotation": {"subspace_principal_angle": 1.5288782648781811, "mean_class_cosine": 0.49201628880852194}}}, "iid_acc": 0.2425, "selectivity": -0.01874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4051891188356136, "mean_class_cosine": 0.5694708832697903}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.4675, "drop": 0.0}, "paraphrase": {"acc": 0.44754316069057104, "drop": 0.019956839309428986}}, "iid_acc": 0.4675, "selectivity": 0.20625000000000004} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7249190938511327, "drop": 0.001330906148867217, "rotation": {"subspace_principal_angle": 1.1511849070923794, "mean_class_cosine": 0.40740561391162183}}}, "iid_acc": 0.72625, "selectivity": 0.16374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1756361449389974, "mean_class_cosine": 0.3849560311735334}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.63915857605178, "drop": -0.03165857605177991, "rotation": {"subspace_principal_angle": 0.8931518264214847, "mean_class_cosine": 0.793022261839432}}}, "iid_acc": 0.6075, "selectivity": 0.04500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.0684485107486943, "mean_class_cosine": 0.6944069959278446}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.7313915857605178, "drop": -0.015141585760517717}}, "iid_acc": 0.71625, "selectivity": 0.15375000000000005} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6494345718901454, "drop": 0.020565428109854622, "rotation": {"subspace_principal_angle": 1.326914373237704, "mean_class_cosine": 0.2414715148282628}}}, "iid_acc": 0.67, "selectivity": 0.19250000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.297386449224978, "mean_class_cosine": 0.27001622040531514}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.6187399030694669, "drop": 0.007510096930533061, "rotation": {"subspace_principal_angle": 1.1476051561016243, "mean_class_cosine": 0.909033474781661}}}, "iid_acc": 0.62625, "selectivity": 0.14875, "iid_split_rotation": {"subspace_principal_angle": 1.4081586003059547, "mean_class_cosine": 0.8941316972733946}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6625, "drop": 0.0}, "paraphrase": {"acc": 0.6235864297253635, "drop": 0.03891357027463649}}, "iid_acc": 0.6625, "selectivity": 0.185} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7405582922824302, "drop": -0.011808292282430188, "rotation": {"subspace_principal_angle": 1.1923774338443787, "mean_class_cosine": 0.36945168819099083}}}, "iid_acc": 0.72875, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2706314343701106, "mean_class_cosine": 0.29567773039327916}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.54375, "drop": 0.0}, "paraphrase": {"acc": 0.5500821018062397, "drop": -0.006332101806239776, "rotation": {"subspace_principal_angle": 1.539326603180651, "mean_class_cosine": 0.8691103147755386}}}, "iid_acc": 0.54375, "selectivity": -0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3002024261544294, "mean_class_cosine": 0.7081293912566515}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7274220032840722, "drop": -0.00867200328407225}}, "iid_acc": 0.71875, "selectivity": 0.12375000000000003} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8363384188626907, "drop": 0.05491158113730932, "rotation": {"subspace_principal_angle": 0.9678914596829231, "mean_class_cosine": 0.5670375780110272}}}, "iid_acc": 0.89125, "selectivity": 0.435, "iid_split_rotation": {"subspace_principal_angle": 0.9419088602911864, "mean_class_cosine": 0.5882454358575304}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.02100728155339804, "rotation": {"subspace_principal_angle": 1.2291895890809326, "mean_class_cosine": 0.9098769842554898}}}, "iid_acc": 0.84625, "selectivity": 0.38999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3955051971962031, "mean_class_cosine": 0.9616189776316897}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8626907073509015, "drop": 0.016059292649098555}}, "iid_acc": 0.87875, "selectivity": 0.42250000000000004} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0034102141680395492, "rotation": {"subspace_principal_angle": 0.8491797480583015, "mean_class_cosine": 0.6605991630042665}}}, "iid_acc": 0.99, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.7965198420896995, "mean_class_cosine": 0.69919899770492}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.5420098846787479, "drop": 0.1567401153212521, "rotation": {"subspace_principal_angle": 1.4157293467420775, "mean_class_cosine": 0.9954481825771943}}}, "iid_acc": 0.69875, "selectivity": -0.13750000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.4943111294129614, "mean_class_cosine": 0.9994355209688002}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.98625, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": -0.0005704283360791207}}, "iid_acc": 0.98625, "selectivity": 0.1499999999999999} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6816269284712483, "drop": -0.006626928471248239, "rotation": {"subspace_principal_angle": 1.493369684917862, "mean_class_cosine": 0.07734930442811239}}}, "iid_acc": 0.675, "selectivity": 0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4475486480920703, "mean_class_cosine": 0.12293589375679194}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.517531556802244, "drop": -0.005031556802244097, "rotation": {"subspace_principal_angle": 1.407007437734732, "mean_class_cosine": 0.1636007713519359}}}, "iid_acc": 0.5125, "selectivity": -0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4687784126101222, "mean_class_cosine": 0.14581035402784961}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.7110799438990182, "drop": -0.009829943899018145}}, "iid_acc": 0.70125, "selectivity": 0.04125000000000001} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6214689265536724, "drop": 0.01795415036940451, "rotation": {"subspace_principal_angle": 1.4099267955197121, "mean_class_cosine": 0.21703022459825547}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.19230769230769224, "iid_split_rotation": {"subspace_principal_angle": 1.4925996250553264, "mean_class_cosine": 0.20125792019838662}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.5769230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.519774011299435, "drop": 0.05714906562364186, "rotation": {"subspace_principal_angle": 1.5438703377286263, "mean_class_cosine": 0.6081265892996393}}}, "iid_acc": 0.5769230769230769, "selectivity": 0.12980769230769224, "iid_split_rotation": {"subspace_principal_angle": 1.367531053578908, "mean_class_cosine": 0.5070016095884643}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6101694915254238, "drop": -0.004400260756193042}}, "iid_acc": 0.6057692307692307, "selectivity": 0.1586538461538461} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8849315068493151, "drop": 0.017568493150684872, "rotation": {"subspace_principal_angle": 1.171653684236408, "mean_class_cosine": 0.3886285207848383}}}, "iid_acc": 0.9025, "selectivity": 0.13249999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1155850933836575, "mean_class_cosine": 0.43965204674130126}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.6, "drop": 0.0025000000000000577, "rotation": {"subspace_principal_angle": 1.1040454036261176, "mean_class_cosine": 0.827992833300117}}}, "iid_acc": 0.6025, "selectivity": -0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.126081986855619, "mean_class_cosine": 0.6844596631580138}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8479452054794521, "drop": 0.01955479452054798}}, "iid_acc": 0.8675, "selectivity": 0.09750000000000003} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.76625, "drop": 0.0}, "paraphrase": {"acc": 0.7503467406380028, "drop": 0.015903259361997213, "rotation": {"subspace_principal_angle": 1.2612626710199533, "mean_class_cosine": 0.3046144880272891}}, "domain": {"acc": 0.72875, "drop": 0.03749999999999998, "rotation": {"subspace_principal_angle": 1.3058544057767159, "mean_class_cosine": 0.2618532161180114}}}, "iid_acc": 0.76625, "selectivity": 0.2875, "iid_split_rotation": {"subspace_principal_angle": 1.2715880951662406, "mean_class_cosine": 0.294763708893907}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.49930651872399445, "drop": 0.0031934812760054943, "rotation": {"subspace_principal_angle": 1.1711657362498487, "mean_class_cosine": 0.9874129557052536}}, "domain": {"acc": 0.49875, "drop": 0.00374999999999992, "rotation": {"subspace_principal_angle": 1.5016858906817072, "mean_class_cosine": 0.3434093433034723}}}, "iid_acc": 0.5025, "selectivity": 0.023749999999999938, "iid_split_rotation": {"subspace_principal_angle": 1.2751956782954972, "mean_class_cosine": 0.972007963226671}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.765, "drop": 0.0}, "paraphrase": {"acc": 0.7600554785020804, "drop": 0.004944521497919574}, "domain": {"acc": 0.72375, "drop": 0.04125000000000001}}, "iid_acc": 0.765, "selectivity": 0.28625} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.829126213592233, "drop": 0.02212378640776691, "rotation": {"subspace_principal_angle": 1.1254573688815634, "mean_class_cosine": 0.43076380259456154}}, "domain": {"acc": 0.53625, "drop": 0.31499999999999995, "rotation": {"subspace_principal_angle": 1.3011899623729255, "mean_class_cosine": 0.2663520415198698}}, "length": {"acc": 0.8597285067873304, "drop": -0.00847850678733042, "rotation": {"subspace_principal_angle": 1.0844876622709578, "mean_class_cosine": 0.4673657026245072}}}, "iid_acc": 0.85125, "selectivity": 0.36624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1313772551394048, "mean_class_cosine": 0.4254137973371145}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7320388349514563, "drop": -0.0007888349514563187, "rotation": {"subspace_principal_angle": 1.3889114115978614, "mean_class_cosine": 0.9101913863100519}}, "domain": {"acc": 0.58, "drop": 0.15125, "rotation": {"subspace_principal_angle": 1.2835193898835013, "mean_class_cosine": 0.5072723519659894}}, "length": {"acc": 0.7450980392156863, "drop": -0.013848039215686336, "rotation": {"subspace_principal_angle": 1.474745277649587, "mean_class_cosine": 0.9218320832330005}}}, "iid_acc": 0.73125, "selectivity": 0.24624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4693373559237952, "mean_class_cosine": 0.8923141074617422}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.01725728155339812}, "domain": {"acc": 0.5775, "drop": 0.265}, "length": {"acc": 0.8476621417797888, "drop": -0.005162141779788798}}, "iid_acc": 0.8425, "selectivity": 0.35750000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8677165354330708, "drop": -0.00021653543307076362, "rotation": {"subspace_principal_angle": 1.3138102642722838, "mean_class_cosine": 0.4178222215166151}}}, "iid_acc": 0.8675, "selectivity": 0.6100000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.269543032505647, "mean_class_cosine": 0.41884051584646115}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.59375, "drop": 0.0}, "paraphrase": {"acc": 0.5748031496062992, "drop": 0.01894685039370081, "rotation": {"subspace_principal_angle": 1.462647268843256, "mean_class_cosine": 0.9428914975925955}}}, "iid_acc": 0.59375, "selectivity": 0.33625, "iid_split_rotation": {"subspace_principal_angle": 1.5667150391990796, "mean_class_cosine": 0.7928532264394269}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8566929133858268, "drop": 0.01455708661417321}}, "iid_acc": 0.87125, "selectivity": 0.61375} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.965, "drop": 0.0}, "paraphrase": {"acc": 0.9319899244332494, "drop": 0.03301007556675062, "rotation": {"subspace_principal_angle": 1.2176081199133377, "mean_class_cosine": 0.6442873004051297}}}, "iid_acc": 0.965, "selectivity": 0.9037499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1385695404879772, "mean_class_cosine": 0.7109378375596532}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.6801007556675063, "drop": 0.04114924433249367, "rotation": {"subspace_principal_angle": 1.5679042974829251, "mean_class_cosine": 0.8621074777728631}}}, "iid_acc": 0.72125, "selectivity": 0.6599999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2212236304536004, "mean_class_cosine": 0.9460677922131192}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9395465994962217, "drop": 0.02170340050377839}}, "iid_acc": 0.96125, "selectivity": 0.9} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5071707953063885, "drop": 0.02907920469361147, "rotation": {"subspace_principal_angle": 1.4838566858436246, "mean_class_cosine": 0.08683016010539553}}}, "iid_acc": 0.53625, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.465396612413356, "mean_class_cosine": 0.10520467342535929}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.49674054758800523, "drop": -0.021740547588005255, "rotation": {"subspace_principal_angle": 1.5469172283064518, "mean_class_cosine": -0.5397626183074191}}}, "iid_acc": 0.475, "selectivity": -0.03375000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.408080647568939, "mean_class_cosine": 0.4860859905388361}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4876140808344198, "drop": 0.01488591916558013}}, "iid_acc": 0.5025, "selectivity": -0.006250000000000089} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5524568393094289, "drop": 0.058793160690571056, "rotation": {"subspace_principal_angle": 1.3850534905313705, "mean_class_cosine": 0.28991503737744245}}}, "iid_acc": 0.61125, "selectivity": 0.37749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.419189771224309, "mean_class_cosine": 0.29718954050314617}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.0925, "drop": 0.0}, "paraphrase": {"acc": 0.1049136786188579, "drop": -0.012413678618857907, "rotation": {"subspace_principal_angle": 1.540744396600694, "mean_class_cosine": 0.06532846024974738}}}, "iid_acc": 0.0925, "selectivity": -0.14125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4223489889510221, "mean_class_cosine": 0.5690159019187561}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5232403718459495, "drop": 0.011759628154050517}}, "iid_acc": 0.535, "selectivity": 0.30125} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": -0.0007322006472492149, "rotation": {"subspace_principal_angle": 1.3881466203882136, "mean_class_cosine": 0.18163583883336032}}}, "iid_acc": 0.73875, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3959757692399888, "mean_class_cosine": 0.17393143305355055}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5906148867313916, "drop": -0.0393648867313916, "rotation": {"subspace_principal_angle": 1.4683816158112597, "mean_class_cosine": 0.9483792840431817}}}, "iid_acc": 0.55125, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.3971642089849379, "mean_class_cosine": 0.9161271445117759}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7540453074433657, "drop": -0.0027953074433657576}}, "iid_acc": 0.75125, "selectivity": 0.20875} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6397415185783522, "drop": 0.044008481421647816, "rotation": {"subspace_principal_angle": 1.3758917619051532, "mean_class_cosine": 0.19367290768765433}}}, "iid_acc": 0.68375, "selectivity": 0.18374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.345477898881277, "mean_class_cosine": 0.22341675246913884}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5993537964458805, "drop": 0.011896203554119467, "rotation": {"subspace_principal_angle": 1.4573232614008496, "mean_class_cosine": 0.870794766106819}}}, "iid_acc": 0.61125, "selectivity": 0.11124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.385445998953203, "mean_class_cosine": 0.8535614192902857}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.6397415185783522, "drop": 0.046508481421647874}}, "iid_acc": 0.68625, "selectivity": 0.18625000000000003} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7405582922824302, "drop": -0.013058292282430162, "rotation": {"subspace_principal_angle": 1.3879649771162461, "mean_class_cosine": 0.1818144576289447}}}, "iid_acc": 0.7275, "selectivity": 0.1312500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4079457648518725, "mean_class_cosine": 0.16213170804038407}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.50375, "drop": 0.0}, "paraphrase": {"acc": 0.5320197044334976, "drop": -0.028269704433497522, "rotation": {"subspace_principal_angle": 1.5607306280375703, "mean_class_cosine": 0.9876176460724118}}}, "iid_acc": 0.50375, "selectivity": -0.09249999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.166861107011101, "mean_class_cosine": 0.9302660401353273}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7422003284072249, "drop": 0.010299671592775006}}, "iid_acc": 0.7525, "selectivity": 0.15625} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8793342579750347, "drop": 0.020665742024965295, "rotation": {"subspace_principal_angle": 1.1063584135999445, "mean_class_cosine": 0.4479203253262932}}}, "iid_acc": 0.9, "selectivity": 0.46125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0249583822746835, "mean_class_cosine": 0.5191344576441722}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.5728155339805825, "drop": 0.14468446601941753, "rotation": {"subspace_principal_angle": 1.284951624986405, "mean_class_cosine": 0.8542882000899992}}}, "iid_acc": 0.7175, "selectivity": 0.27875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0904185384770684, "mean_class_cosine": 0.9172822881817656}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8904299583911235, "drop": 0.00832004160887656}}, "iid_acc": 0.89875, "selectivity": 0.4600000000000001} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9967051070840197, "drop": -0.00045510708401974576, "rotation": {"subspace_principal_angle": 0.8815453404218065, "mean_class_cosine": 0.6359593299455631}}}, "iid_acc": 0.99625, "selectivity": 0.16874999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.829603647498299, "mean_class_cosine": 0.6751681879987259}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.5634266886326195, "drop": 0.12157331136738059, "rotation": {"subspace_principal_angle": 1.1147774126986099, "mean_class_cosine": 0.9975604664853437}}}, "iid_acc": 0.685, "selectivity": -0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.351613895560496, "mean_class_cosine": 0.9997571774292637}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9950576606260296, "drop": -0.005057660626029636}}, "iid_acc": 0.99, "selectivity": 0.16249999999999998} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6633941093969145, "drop": 0.01910589060308554, "rotation": {"subspace_principal_angle": 1.5456837867774578, "mean_class_cosine": 0.02510990060669071}}}, "iid_acc": 0.6825, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4420552406446043, "mean_class_cosine": 0.12838574922234897}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.46875, "drop": 0.0}, "paraphrase": {"acc": 0.4950911640953717, "drop": -0.026341164095371683, "rotation": {"subspace_principal_angle": 0.8450834965822174, "mean_class_cosine": 0.8344612624127071}}}, "iid_acc": 0.46875, "selectivity": -0.14375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.447600013734126, "mean_class_cosine": -0.6148431774148773}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7237026647966339, "drop": -0.009952664796633925}}, "iid_acc": 0.71375, "selectivity": 0.10124999999999995} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.615819209039548, "drop": 0.00918079096045199, "rotation": {"subspace_principal_angle": 1.344146750268554, "mean_class_cosine": 0.24255494243559642}}}, "iid_acc": 0.625, "selectivity": 0.16346153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.4369180768778205, "mean_class_cosine": 0.21822584946678156}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5423728813559322, "drop": 0.0201271186440678, "rotation": {"subspace_principal_angle": 1.5227921354385472, "mean_class_cosine": 0.6841529664773013}}}, "iid_acc": 0.5625, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.2545192468784838, "mean_class_cosine": 0.5447114834286542}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.632768361581921, "drop": -0.02699913081269023}}, "iid_acc": 0.6057692307692307, "selectivity": 0.14423076923076916} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8945205479452055, "drop": 0.004229452054794525, "rotation": {"subspace_principal_angle": 1.248973078003772, "mean_class_cosine": 0.3162967291391359}}}, "iid_acc": 0.89875, "selectivity": 0.15250000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.2122862189044603, "mean_class_cosine": 0.35087945751268557}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5767123287671233, "drop": -0.009212328767123301, "rotation": {"subspace_principal_angle": 1.21580262592107, "mean_class_cosine": 0.8869377958216829}}}, "iid_acc": 0.5675, "selectivity": -0.17874999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.839488040678943, "mean_class_cosine": 0.7546602597663057}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8767123287671232, "drop": 0.012037671232876801}}, "iid_acc": 0.88875, "selectivity": 0.14250000000000007} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7739251040221914, "drop": 0.029824895977808574, "rotation": {"subspace_principal_angle": 1.2659109438320224, "mean_class_cosine": 0.30018384612544047}}, "domain": {"acc": 0.76875, "drop": 0.03499999999999992, "rotation": {"subspace_principal_angle": 1.34258356435988, "mean_class_cosine": 0.22623698719981744}}}, "iid_acc": 0.80375, "selectivity": 0.32499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2506624679779867, "mean_class_cosine": 0.3146936213275413}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5034674063800277, "drop": -0.004717406380027711, "rotation": {"subspace_principal_angle": 1.239282670172901, "mean_class_cosine": 0.9986892573730854}}, "domain": {"acc": 0.5, "drop": -0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.3097283989710364, "mean_class_cosine": 0.1603346495077691}}}, "iid_acc": 0.49875, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.420934784951417, "mean_class_cosine": 0.9934847558506945}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.8030513176144244, "drop": 0.019448682385575622}, "domain": {"acc": 0.81625, "drop": 0.006249999999999978}}, "iid_acc": 0.8225, "selectivity": 0.34375} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.8524271844660194, "drop": 0.02382281553398058, "rotation": {"subspace_principal_angle": 1.220154188841714, "mean_class_cosine": 0.34350094358966443}}, "domain": {"acc": 0.4975, "drop": 0.37875, "rotation": {"subspace_principal_angle": 1.438605482929702, "mean_class_cosine": 0.13180618706503908}}, "length": {"acc": 0.8808446455505279, "drop": -0.004594645550527909, "rotation": {"subspace_principal_angle": 1.1893943688634518, "mean_class_cosine": 0.37222205321854485}}}, "iid_acc": 0.87625, "selectivity": 0.38625, "iid_split_rotation": {"subspace_principal_angle": 1.1874268592484474, "mean_class_cosine": 0.3740474632573129}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.829126213592233, "drop": 0.013373786407766985, "rotation": {"subspace_principal_angle": 1.525376087221384, "mean_class_cosine": 0.942182847156327}}, "domain": {"acc": 0.57875, "drop": 0.26375000000000004, "rotation": {"subspace_principal_angle": 1.436822367907557, "mean_class_cosine": 0.27136585793969137}}, "length": {"acc": 0.8536953242835595, "drop": -0.011195324283559516, "rotation": {"subspace_principal_angle": 1.4983468510851918, "mean_class_cosine": 0.9461442600007871}}}, "iid_acc": 0.8425, "selectivity": 0.35250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.287421066876229, "mean_class_cosine": 0.910695802907292}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8640776699029126, "drop": 0.015922330097087434}, "domain": {"acc": 0.505, "drop": 0.375}, "length": {"acc": 0.8883861236802413, "drop": -0.008386123680241275}}, "iid_acc": 0.88, "selectivity": 0.39} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8582677165354331, "drop": 0.015482283464566926, "rotation": {"subspace_principal_angle": 1.2827164275902705, "mean_class_cosine": 0.4163674814452723}}}, "iid_acc": 0.87375, "selectivity": 0.6000000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.26135377780834, "mean_class_cosine": 0.4442699111544142}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7975, "drop": 0.0}, "paraphrase": {"acc": 0.7968503937007874, "drop": 0.0006496062992126239, "rotation": {"subspace_principal_angle": 1.424694638770457, "mean_class_cosine": 0.9469278805103363}}}, "iid_acc": 0.7975, "selectivity": 0.5237499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4837220741637616, "mean_class_cosine": 0.9518191309197193}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8598425196850393, "drop": 0.017657480314960616}}, "iid_acc": 0.8775, "selectivity": 0.60375} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9672544080604534, "drop": 0.00024559193954665215, "rotation": {"subspace_principal_angle": 1.1145296779462641, "mean_class_cosine": 0.7020277810816539}}}, "iid_acc": 0.9675, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.5624189489347664, "mean_class_cosine": 0.7403838209375735}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.44125, "drop": 0.0}, "paraphrase": {"acc": 0.4256926952141058, "drop": 0.015557304785894177, "rotation": {"subspace_principal_angle": 1.033444992470805, "mean_class_cosine": 0.7765287771414837}}}, "iid_acc": 0.44125, "selectivity": 0.38, "iid_split_rotation": {"subspace_principal_angle": 1.2853261455962783, "mean_class_cosine": 0.8790932160427343}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.96375, "drop": 0.0}, "paraphrase": {"acc": 0.9571788413098237, "drop": 0.006571158690176326}}, "iid_acc": 0.96375, "selectivity": 0.9025} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5397653194263363, "drop": 0.013984680573663644, "rotation": {"subspace_principal_angle": 1.5274443365773367, "mean_class_cosine": 0.0433384122408627}}}, "iid_acc": 0.55375, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5117604681674472, "mean_class_cosine": 0.05900157231956833}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.5071707953063885, "drop": -0.029670795306388553, "rotation": {"subspace_principal_angle": 0.8759468408895317, "mean_class_cosine": -0.6781424224487453}}}, "iid_acc": 0.4775, "selectivity": -0.006250000000000033, "iid_split_rotation": {"subspace_principal_angle": 1.0844620330806314, "mean_class_cosine": -0.4769926683134672}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.49282920469361147, "drop": 0.02717079530638855}}, "iid_acc": 0.52, "selectivity": 0.036250000000000004} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.62875, "drop": 0.0}, "paraphrase": {"acc": 0.5803452855245684, "drop": 0.0484047144754316, "rotation": {"subspace_principal_angle": 1.4145413502440674, "mean_class_cosine": 0.27663015100707133}}}, "iid_acc": 0.62875, "selectivity": 0.37125, "iid_split_rotation": {"subspace_principal_angle": 1.4392990909748595, "mean_class_cosine": 0.2769217490323135}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.09, "drop": 0.0}, "paraphrase": {"acc": 0.07436918990703852, "drop": 0.01563081009296148, "rotation": {"subspace_principal_angle": 1.3350767969885171, "mean_class_cosine": -0.06735755908616899}}}, "iid_acc": 0.09, "selectivity": -0.1675, "iid_split_rotation": {"subspace_principal_angle": 1.2649936899279066, "mean_class_cosine": 0.6127869206852473}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5391766268260292, "drop": 0.01332337317397081}}, "iid_acc": 0.5525, "selectivity": 0.295} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": 0.0030177993527508162, "rotation": {"subspace_principal_angle": 1.3800650372385295, "mean_class_cosine": 0.18957697385454825}}}, "iid_acc": 0.7425, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3745352340470285, "mean_class_cosine": 0.19500357265711427}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.64125, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.025416666666666643, "rotation": {"subspace_principal_angle": 1.2209077466423848, "mean_class_cosine": 0.8479639619526143}}}, "iid_acc": 0.64125, "selectivity": 0.09124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.563353933733748, "mean_class_cosine": 0.8055681849044787}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7572815533980582, "drop": 0.0027184466019417597}}, "iid_acc": 0.76, "selectivity": 0.20999999999999996} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.65375, "drop": 0.0}, "paraphrase": {"acc": 0.6155088852988692, "drop": 0.038241114701130896, "rotation": {"subspace_principal_angle": 1.4613251262739329, "mean_class_cosine": 0.10925268204629429}}}, "iid_acc": 0.65375, "selectivity": 0.1462500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.467586611767328, "mean_class_cosine": 0.10302657672993193}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.6125, "drop": 0.0}, "paraphrase": {"acc": 0.5912762520193862, "drop": 0.021223747980613883, "rotation": {"subspace_principal_angle": 1.2485040409587074, "mean_class_cosine": 0.866292700063821}}}, "iid_acc": 0.6125, "selectivity": 0.1050000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5556750038869702, "mean_class_cosine": 0.8382504448687405}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6510500807754442, "drop": 0.03269991922455573}}, "iid_acc": 0.68375, "selectivity": 0.17625000000000002} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.722495894909688, "drop": -0.0024958949096880456, "rotation": {"subspace_principal_angle": 1.4020686555705464, "mean_class_cosine": 0.1679282246076349}}}, "iid_acc": 0.72, "selectivity": 0.13374999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3616761916383724, "mean_class_cosine": 0.20759928424855742}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.56375, "drop": 0.0}, "paraphrase": {"acc": 0.5369458128078818, "drop": 0.02680418719211819, "rotation": {"subspace_principal_angle": 1.525739498161781, "mean_class_cosine": 0.8790506170094416}}}, "iid_acc": 0.56375, "selectivity": -0.022500000000000075, "iid_split_rotation": {"subspace_principal_angle": 1.355838296904698, "mean_class_cosine": 0.7623898156919341}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7504105090311987, "drop": 0.0008394909688013152}}, "iid_acc": 0.75125, "selectivity": 0.16499999999999992} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.897364771151179, "drop": 0.02388522884882105, "rotation": {"subspace_principal_angle": 1.2009122822029357, "mean_class_cosine": 0.3615073211363401}}}, "iid_acc": 0.92125, "selectivity": 0.45375, "iid_split_rotation": {"subspace_principal_angle": 1.1365147935104427, "mean_class_cosine": 0.42075873671714054}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7253814147018031, "drop": 0.008368585298196951, "rotation": {"subspace_principal_angle": 1.382859019749326, "mean_class_cosine": 0.9518475699667834}}}, "iid_acc": 0.73375, "selectivity": 0.26625, "iid_split_rotation": {"subspace_principal_angle": 0.8342063363329182, "mean_class_cosine": 0.8834390160996848}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.92625, "drop": 0.0}, "paraphrase": {"acc": 0.9140083217753121, "drop": 0.012241678224687913}}, "iid_acc": 0.92625, "selectivity": 0.45875} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.0007372322899505956, "rotation": {"subspace_principal_angle": 0.9022113956514256, "mean_class_cosine": 0.6198762040410312}}}, "iid_acc": 0.9925, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 0.7803136404078311, "mean_class_cosine": 0.7106929262056456}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.4744645799011532, "drop": 0.21553542009884674, "rotation": {"subspace_principal_angle": 1.5013958950641693, "mean_class_cosine": 0.997742196187771}}}, "iid_acc": 0.69, "selectivity": -0.1362500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3648075725684445, "mean_class_cosine": 0.9999469425446367}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": -0.00011532125205926391}}, "iid_acc": 0.99, "selectivity": 0.16374999999999995} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6900420757363254, "drop": 0.019957924263674598, "rotation": {"subspace_principal_angle": 1.4053917466501165, "mean_class_cosine": 0.16465140281778454}}}, "iid_acc": 0.71, "selectivity": 0.10249999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.4198215714680307, "mean_class_cosine": 0.15040187120207954}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.511921458625526, "drop": -0.03442145862552598, "rotation": {"subspace_principal_angle": 1.088765737718087, "mean_class_cosine": -0.46901391020350736}}}, "iid_acc": 0.4775, "selectivity": -0.13000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4869971047307045, "mean_class_cosine": 0.19635370929712698}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7223001402524544, "drop": -0.012300140252454406}}, "iid_acc": 0.71, "selectivity": 0.10249999999999992} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.672316384180791, "drop": -0.018470230334637128, "rotation": {"subspace_principal_angle": 1.4095901786238862, "mean_class_cosine": 0.19654959439481243}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.27403846153846156, "iid_split_rotation": {"subspace_principal_angle": 1.5567476235509987, "mean_class_cosine": 0.20024194309517288}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.423728813559322, "drop": 0.03780964797913955, "rotation": {"subspace_principal_angle": 1.318674137405094, "mean_class_cosine": 0.5319901746483242}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.08173076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.1328152149633164, "mean_class_cosine": 0.6908347513167752}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6201923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": -0.01822577140373749}}, "iid_acc": 0.6201923076923077, "selectivity": 0.24038461538461542} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8780821917808219, "drop": 0.02066780821917813, "rotation": {"subspace_principal_angle": 1.4022671366129482, "mean_class_cosine": 0.16773255884346316}}}, "iid_acc": 0.89875, "selectivity": 0.17500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3156344568426237, "mean_class_cosine": 0.25240204094053126}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.584931506849315, "drop": 0.017568493150684983, "rotation": {"subspace_principal_angle": 1.2909886826134447, "mean_class_cosine": 0.8167709778669217}}}, "iid_acc": 0.6025, "selectivity": -0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.181444552949185, "mean_class_cosine": 0.8742409810564655}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.8972602739726028, "drop": 0.013989726027397231}}, "iid_acc": 0.91125, "selectivity": 0.1875} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.7988904299583911, "drop": 0.047359570041608845, "rotation": {"subspace_principal_angle": 1.307852975470961, "mean_class_cosine": 0.25992385942235496}}, "domain": {"acc": 0.865, "drop": -0.018750000000000044, "rotation": {"subspace_principal_angle": 1.4101720618034508, "mean_class_cosine": 0.15993446696213712}}}, "iid_acc": 0.84625, "selectivity": 0.3374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2803570242694347, "mean_class_cosine": 0.2863731564533277}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5034674063800277, "drop": -0.004717406380027711, "rotation": {"subspace_principal_angle": 1.4967653281587188, "mean_class_cosine": 0.9956396889374676}}, "domain": {"acc": 0.5, "drop": -0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5195064626946009, "mean_class_cosine": 0.17206983639586906}}}, "iid_acc": 0.49875, "selectivity": -0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.53305740516774, "mean_class_cosine": 0.9726130574764482}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.812760055478502, "drop": 0.04223994452149793}, "domain": {"acc": 0.84375, "drop": 0.011249999999999982}}, "iid_acc": 0.855, "selectivity": 0.34624999999999995} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8815533980582524, "drop": 0.020946601941747578, "rotation": {"subspace_principal_angle": 1.3231535337909477, "mean_class_cosine": 0.24511934689015158}}, "domain": {"acc": 0.56, "drop": 0.3424999999999999, "rotation": {"subspace_principal_angle": 1.4299398396010543, "mean_class_cosine": 0.14039117067809936}}, "length": {"acc": 0.9140271493212669, "drop": -0.011527149321266972, "rotation": {"subspace_principal_angle": 1.2714271775831436, "mean_class_cosine": 0.2949174731196204}}}, "iid_acc": 0.9025, "selectivity": 0.43749999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3312504683770023, "mean_class_cosine": 0.23726147696520083}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8524271844660194, "drop": 0.03507281553398056, "rotation": {"subspace_principal_angle": 1.1382054241911463, "mean_class_cosine": 0.9322577478917091}}, "domain": {"acc": 0.57875, "drop": 0.30874999999999997, "rotation": {"subspace_principal_angle": 1.5632845119627006, "mean_class_cosine": 0.34081581535856803}}, "length": {"acc": 0.8883861236802413, "drop": -0.0008861236802413242, "rotation": {"subspace_principal_angle": 1.098044803005494, "mean_class_cosine": 0.9434252502763861}}}, "iid_acc": 0.8875, "selectivity": 0.42249999999999993, "iid_split_rotation": {"subspace_principal_angle": 0.7127743899001, "mean_class_cosine": 0.9390001647602704}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.90375, "drop": 0.0}, "paraphrase": {"acc": 0.8990291262135922, "drop": 0.004720873786407842}, "domain": {"acc": 0.73125, "drop": 0.1725000000000001}, "length": {"acc": 0.9155354449472096, "drop": -0.011785444947209567}}, "iid_acc": 0.90375, "selectivity": 0.43875000000000003} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8866141732283465, "drop": 0.005885826771653502, "rotation": {"subspace_principal_angle": 1.3590153860798901, "mean_class_cosine": 0.3384795168112319}}}, "iid_acc": 0.8925, "selectivity": 0.67375, "iid_split_rotation": {"subspace_principal_angle": 1.5137713950823037, "mean_class_cosine": 0.34878144427948704}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6661417322834645, "drop": 0.021358267716535484, "rotation": {"subspace_principal_angle": 1.2049572219735756, "mean_class_cosine": 0.9516923360218306}}}, "iid_acc": 0.6875, "selectivity": 0.46875, "iid_split_rotation": {"subspace_principal_angle": 1.161004465491632, "mean_class_cosine": 0.8826822088496902}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8771653543307086, "drop": 0.017834645669291382}}, "iid_acc": 0.895, "selectivity": 0.67625} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.964735516372796, "drop": 0.010264483627204002, "rotation": {"subspace_principal_angle": 1.110589965959208, "mean_class_cosine": 0.671095672439944}}}, "iid_acc": 0.975, "selectivity": 0.92375, "iid_split_rotation": {"subspace_principal_angle": 1.0922410354486654, "mean_class_cosine": 0.7234153961945159}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.84375, "drop": 0.0}, "paraphrase": {"acc": 0.8488664987405542, "drop": -0.005116498740554198, "rotation": {"subspace_principal_angle": 1.4098145943553153, "mean_class_cosine": 0.8856785823869348}}}, "iid_acc": 0.84375, "selectivity": 0.7925, "iid_split_rotation": {"subspace_principal_angle": 0.8542184898946922, "mean_class_cosine": 0.9572548062279118}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.9596977329974811, "drop": 0.01530226700251891}}, "iid_acc": 0.975, "selectivity": 0.92375} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5736636245110821, "drop": 0.006336375488917856, "rotation": {"subspace_principal_angle": 1.4824473599970098, "mean_class_cosine": 0.08823407641471295}}}, "iid_acc": 0.58, "selectivity": 0.05499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4950406885229706, "mean_class_cosine": 0.07568319984431088}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.5058670143415906, "drop": -0.02961701434159064, "rotation": {"subspace_principal_angle": 1.1493594930263484, "mean_class_cosine": -0.4424939882567903}}}, "iid_acc": 0.47625, "selectivity": -0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.551473168662097, "mean_class_cosine": 0.01234111003206595}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5867014341590613, "drop": -0.00045143415906123696}}, "iid_acc": 0.58625, "selectivity": 0.06125000000000003} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6148738379814077, "drop": 0.03512616201859231, "rotation": {"subspace_principal_angle": 1.4073412354609784, "mean_class_cosine": 0.26349298302241686}}}, "iid_acc": 0.65, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.413824435567323, "mean_class_cosine": 0.24889219745848057}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.095, "drop": 0.0}, "paraphrase": {"acc": 0.10092961487383798, "drop": -0.005929614873837974, "rotation": {"subspace_principal_angle": 1.331943855465791, "mean_class_cosine": 0.03732221069944747}}}, "iid_acc": 0.095, "selectivity": -0.165, "iid_split_rotation": {"subspace_principal_angle": 1.5577947437159871, "mean_class_cosine": 0.5674187824840483}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.64, "drop": 0.0}, "paraphrase": {"acc": 0.6347941567065073, "drop": 0.005205843293492718}}, "iid_acc": 0.64, "selectivity": 0.38} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.77625, "drop": 0.0}, "paraphrase": {"acc": 0.7427184466019418, "drop": 0.033531553398058245, "rotation": {"subspace_principal_angle": 1.4430864610974254, "mean_class_cosine": 0.12736299475185067}}}, "iid_acc": 0.77625, "selectivity": 0.22499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4358406552905756, "mean_class_cosine": 0.13454638570974817}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5598705501618123, "drop": -0.02237055016181233, "rotation": {"subspace_principal_angle": 1.1405278653750983, "mean_class_cosine": 0.804713745342145}}}, "iid_acc": 0.5375, "selectivity": -0.01375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.509364684034334, "mean_class_cosine": 0.9432691877610848}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.7925, "drop": 0.0}, "paraphrase": {"acc": 0.7944983818770227, "drop": -0.0019983818770227257}}, "iid_acc": 0.7925, "selectivity": 0.24124999999999996} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.6494345718901454, "drop": 0.03681542810985461, "rotation": {"subspace_principal_angle": 1.4705871216508313, "mean_class_cosine": 0.10004157445035826}}}, "iid_acc": 0.68625, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4090091592636642, "mean_class_cosine": 0.16108229179329214}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5282714054927302, "drop": 0.00922859450726976, "rotation": {"subspace_principal_angle": 0.8130735690110286, "mean_class_cosine": 0.9877317674481327}}}, "iid_acc": 0.5375, "selectivity": 0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 0.6815839977232387, "mean_class_cosine": 0.9939458887778236}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6591276252019386, "drop": 0.04587237479806139}}, "iid_acc": 0.705, "selectivity": 0.18999999999999995} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6962233169129721, "drop": 0.011276683087027894, "rotation": {"subspace_principal_angle": 1.49979082124605, "mean_class_cosine": 0.07094585487697073}}}, "iid_acc": 0.7075, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.482620368933285, "mean_class_cosine": 0.08806174093377613}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.4875, "drop": 0.0}, "paraphrase": {"acc": 0.49589490968801314, "drop": -0.008394909688013152, "rotation": {"subspace_principal_angle": 1.524068667642953, "mean_class_cosine": 0.9875864078448118}}}, "iid_acc": 0.4875, "selectivity": -0.07500000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8171205192766501, "mean_class_cosine": 0.770949459806412}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7684729064039408, "drop": -0.004722906403940796}}, "iid_acc": 0.76375, "selectivity": 0.20125000000000004} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.93375, "drop": 0.0}, "paraphrase": {"acc": 0.9181692094313454, "drop": 0.015580790568654579, "rotation": {"subspace_principal_angle": 1.245296425269521, "mean_class_cosine": 0.3197824779830509}}}, "iid_acc": 0.93375, "selectivity": 0.46624999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.1642335870118337, "mean_class_cosine": 0.3954545965394525}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.855755894590846, "drop": -0.004505894590846049, "rotation": {"subspace_principal_angle": 0.9569293407079221, "mean_class_cosine": 0.948995043172376}}}, "iid_acc": 0.85125, "selectivity": 0.3837499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5155592875644858, "mean_class_cosine": 0.9503525832749367}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.9292649098474342, "drop": 0.003235090152565845}}, "iid_acc": 0.9325, "selectivity": 0.46499999999999997} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.001987232289950569, "rotation": {"subspace_principal_angle": 0.9228456065715872, "mean_class_cosine": 0.6035537376948626}}}, "iid_acc": 0.99375, "selectivity": 0.19000000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.836801876997636, "mean_class_cosine": 0.6698408686733577}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.42174629324546953, "drop": 0.2632537067545305, "rotation": {"subspace_principal_angle": 1.1528917670379706, "mean_class_cosine": 0.9973529147197886}}}, "iid_acc": 0.685, "selectivity": -0.11874999999999991, "iid_split_rotation": {"subspace_principal_angle": 1.1220283287635184, "mean_class_cosine": 0.999929258322932}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.0007372322899505956}}, "iid_acc": 0.9925, "selectivity": 0.18875000000000008} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.699859747545582, "drop": 0.048890252454418026, "rotation": {"subspace_principal_angle": 1.4762812316903078, "mean_class_cosine": 0.09437443909422669}}}, "iid_acc": 0.74875, "selectivity": 0.16500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4435800079379149, "mean_class_cosine": 0.12687345177932846}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.5063113604488079, "drop": -0.03131136044880789, "rotation": {"subspace_principal_angle": 1.4269418061859824, "mean_class_cosine": -0.10748417912085814}}}, "iid_acc": 0.475, "selectivity": -0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4676928130838114, "mean_class_cosine": 0.13776060857811906}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.726507713884993, "drop": 0.004742286115006933}}, "iid_acc": 0.73125, "selectivity": 0.14749999999999996} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.6923076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6836158192090396, "drop": 0.008691873098652714, "rotation": {"subspace_principal_angle": 1.414178459512195, "mean_class_cosine": 0.1826323602732557}}}, "iid_acc": 0.6923076923076923, "selectivity": 0.2932692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.4350751526080228, "mean_class_cosine": 0.1805559724526645}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.4855769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.4180790960451977, "drop": 0.06749782703172535, "rotation": {"subspace_principal_angle": 1.2602978332522063, "mean_class_cosine": 0.6807601457540157}}}, "iid_acc": 0.4855769230769231, "selectivity": 0.08653846153846151, "iid_split_rotation": {"subspace_principal_angle": 1.1362894639616432, "mean_class_cosine": 0.6888773765938021}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6836158192090396, "drop": -0.010538896132116449}}, "iid_acc": 0.6730769230769231, "selectivity": 0.27403846153846156} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8945205479452055, "drop": 0.01297945205479445, "rotation": {"subspace_principal_angle": 1.411004019350323, "mean_class_cosine": 0.15911316342225837}}}, "iid_acc": 0.9075, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.3881521587997803, "mean_class_cosine": 0.1816303925457004}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.647945205479452, "drop": 0.020804794520547953, "rotation": {"subspace_principal_angle": 1.1198503241914632, "mean_class_cosine": 0.8163975702558504}}}, "iid_acc": 0.66875, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.401866352414971, "mean_class_cosine": 0.9059315149970117}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9125, "drop": 0.0}, "paraphrase": {"acc": 0.9082191780821918, "drop": 0.0042808219178082085}}, "iid_acc": 0.9125, "selectivity": 0.21875} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.78375, "drop": 0.0}, "paraphrase": {"acc": 0.7961165048543689, "drop": -0.012366504854368965, "rotation": {"subspace_principal_angle": 1.2928467517215603, "mean_class_cosine": 0.2743844969854279}}, "domain": {"acc": 0.725, "drop": 0.05874999999999997, "rotation": {"subspace_principal_angle": 1.3363808809411806, "mean_class_cosine": 0.23227445847260966}}}, "iid_acc": 0.78375, "selectivity": 0.26625, "iid_split_rotation": {"subspace_principal_angle": 1.252105311514529, "mean_class_cosine": 0.3133237568177669}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.007765256588072145, "rotation": {"subspace_principal_angle": 0.5795075771494037, "mean_class_cosine": 0.9984633717830367}}, "domain": {"acc": 0.5, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5279705890862962, "mean_class_cosine": 0.23923966769590235}}}, "iid_acc": 0.50125, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.2020721809074764, "mean_class_cosine": 0.9962696884890198}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.805, "drop": 0.0}, "paraphrase": {"acc": 0.7364771151178918, "drop": 0.06852288488210823}, "domain": {"acc": 0.69625, "drop": 0.10875000000000001}}, "iid_acc": 0.805, "selectivity": 0.2875000000000001} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8271844660194175, "drop": 0.035315533980582514, "rotation": {"subspace_principal_angle": 1.1721917031350069, "mean_class_cosine": 0.38813273689356925}}, "domain": {"acc": 0.59, "drop": 0.2725000000000001, "rotation": {"subspace_principal_angle": 1.3310637787422674, "mean_class_cosine": 0.23744283169580327}}, "length": {"acc": 0.8748114630467572, "drop": -0.01231146304675712, "rotation": {"subspace_principal_angle": 1.0758385777551667, "mean_class_cosine": 0.4749944713239639}}}, "iid_acc": 0.8625, "selectivity": 0.33125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.16650891029775, "mean_class_cosine": 0.3933637231840712}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5805825242718446, "drop": -0.014332524271844616, "rotation": {"subspace_principal_angle": 1.390314107154939, "mean_class_cosine": 0.9207722901404116}}, "domain": {"acc": 0.57875, "drop": -0.012499999999999956, "rotation": {"subspace_principal_angle": 1.0701854814469178, "mean_class_cosine": 0.8833474305608653}}, "length": {"acc": 0.5641025641025641, "drop": 0.002147435897435934, "rotation": {"subspace_principal_angle": 0.9369329512012231, "mean_class_cosine": 0.9177680143320355}}}, "iid_acc": 0.56625, "selectivity": 0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.9871644210171315, "mean_class_cosine": 0.4073480943439788}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.006007281553398136}, "domain": {"acc": 0.75375, "drop": 0.07750000000000001}, "length": {"acc": 0.8355957767722474, "drop": -0.004345776772247345}}, "iid_acc": 0.83125, "selectivity": 0.30000000000000004} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.8725, "drop": 0.0}, "paraphrase": {"acc": 0.8614173228346457, "drop": 0.011082677165354382, "rotation": {"subspace_principal_angle": 1.3840540885530221, "mean_class_cosine": 0.2934795135845933}}}, "iid_acc": 0.8725, "selectivity": 0.59125, "iid_split_rotation": {"subspace_principal_angle": 1.4099364101494178, "mean_class_cosine": 0.29058275145224166}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.5925, "drop": 0.0}, "paraphrase": {"acc": 0.5669291338582677, "drop": 0.025570866141732318, "rotation": {"subspace_principal_angle": 0.998890821047691, "mean_class_cosine": 0.94716741691378}}}, "iid_acc": 0.5925, "selectivity": 0.31125, "iid_split_rotation": {"subspace_principal_angle": 0.9488762179483594, "mean_class_cosine": 0.7707278199904818}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8866141732283465, "drop": -0.0016141732283464494}}, "iid_acc": 0.885, "selectivity": 0.60375} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9370277078085643, "drop": 0.02422229219143579, "rotation": {"subspace_principal_angle": 1.2359971876297358, "mean_class_cosine": 0.6039952840126505}}}, "iid_acc": 0.96125, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.2037218056275953, "mean_class_cosine": 0.6646147092871253}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.2575, "drop": 0.0}, "paraphrase": {"acc": 0.24181360201511334, "drop": 0.015686397984886663, "rotation": {"subspace_principal_angle": 1.3871046502979878, "mean_class_cosine": 0.6602864355874835}}}, "iid_acc": 0.2575, "selectivity": 0.165, "iid_split_rotation": {"subspace_principal_angle": 1.5631281003472528, "mean_class_cosine": 0.7824016096532195}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9168765743073047, "drop": 0.024373425692695294}}, "iid_acc": 0.94125, "selectivity": 0.84875} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5580182529335072, "drop": -0.006768252933507135, "rotation": {"subspace_principal_angle": 1.4094629345382517, "mean_class_cosine": 0.16063442581070056}}}, "iid_acc": 0.55125, "selectivity": 0.05375000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4680870349990278, "mean_class_cosine": 0.10252880356925093}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.4915254237288136, "drop": -0.015275423728813575, "rotation": {"subspace_principal_angle": 1.4399498453238624, "mean_class_cosine": -0.9176241029118328}}}, "iid_acc": 0.47625, "selectivity": -0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4611313378914106, "mean_class_cosine": 0.7185000764227502}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5032594524119948, "drop": 0.009240547588005188}}, "iid_acc": 0.5125, "selectivity": 0.014999999999999958} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6325, "drop": 0.0}, "paraphrase": {"acc": 0.6228419654714475, "drop": 0.009658034528552406, "rotation": {"subspace_principal_angle": 1.348983488807089, "mean_class_cosine": 0.32809803411682364}}}, "iid_acc": 0.6325, "selectivity": 0.39249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3929305981197904, "mean_class_cosine": 0.31905211151253626}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.17, "drop": 0.0}, "paraphrase": {"acc": 0.17795484727755645, "drop": -0.007954847277556437, "rotation": {"subspace_principal_angle": 1.4531704709040898, "mean_class_cosine": 0.4279120214634104}}}, "iid_acc": 0.17, "selectivity": -0.06999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.5500505520500567, "mean_class_cosine": 0.6189009860707383}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.602921646746348, "drop": -0.004171646746347957}}, "iid_acc": 0.59875, "selectivity": 0.35875} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7168284789644013, "drop": 0.0044215210355986745, "rotation": {"subspace_principal_angle": 1.378307022373732, "mean_class_cosine": 0.19130281479130162}}}, "iid_acc": 0.72125, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4153437530917778, "mean_class_cosine": 0.1548272315027917}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6585760517799353, "drop": -0.036076051779935225, "rotation": {"subspace_principal_angle": 1.0555425497782547, "mean_class_cosine": 0.88892568318339}}}, "iid_acc": 0.6225, "selectivity": 0.06375000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9396491512167356, "mean_class_cosine": 0.32052389745129145}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": -0.008232200647249277}}, "iid_acc": 0.73125, "selectivity": 0.1725} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6316639741518578, "drop": 0.017086025848142228, "rotation": {"subspace_principal_angle": 1.3946146188040096, "mean_class_cosine": 0.17527167497740181}}}, "iid_acc": 0.64875, "selectivity": 0.16125000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4780475383542893, "mean_class_cosine": 0.09261586955669032}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.568659127625202, "drop": 9.087237479799004e-05, "rotation": {"subspace_principal_angle": 1.042022638358511, "mean_class_cosine": 0.9908204267164294}}}, "iid_acc": 0.56875, "selectivity": 0.08124999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.4747132575045126, "mean_class_cosine": 0.9899105548993554}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6381260096930533, "drop": 0.011873990306946691}}, "iid_acc": 0.65, "selectivity": 0.16250000000000003} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.7241379310344828, "drop": -0.016637931034482745, "rotation": {"subspace_principal_angle": 1.4200343797614488, "mean_class_cosine": 0.15019148021114892}}}, "iid_acc": 0.7075, "selectivity": 0.13124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.382096362448142, "mean_class_cosine": 0.18758209522431973}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5024630541871922, "drop": -0.01246305418719218, "rotation": {"subspace_principal_angle": 1.534087120470578, "mean_class_cosine": 0.9835950818873114}}}, "iid_acc": 0.49, "selectivity": -0.08625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0584655028541943, "mean_class_cosine": 0.7127138345764553}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7323481116584565, "drop": 0.00015188834154356012}}, "iid_acc": 0.7325, "selectivity": 0.15625} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8918169209431346, "drop": 0.0031830790568654344, "rotation": {"subspace_principal_angle": 1.2670757577041036, "mean_class_cosine": 0.29907254848700277}}}, "iid_acc": 0.895, "selectivity": 0.435, "iid_split_rotation": {"subspace_principal_angle": 1.1636722638264427, "mean_class_cosine": 0.3959701014331015}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.64125, "drop": 0.0}, "paraphrase": {"acc": 0.6310679611650486, "drop": 0.01018203883495139, "rotation": {"subspace_principal_angle": 1.2408104805968017, "mean_class_cosine": 0.9660598087989238}}}, "iid_acc": 0.64125, "selectivity": 0.18124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1355294988893803, "mean_class_cosine": 0.9278898529249071}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.9195561719833565, "drop": -0.01830617198335649}}, "iid_acc": 0.90125, "selectivity": 0.44125} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915, "rotation": {"subspace_principal_angle": 0.9875978535885324, "mean_class_cosine": 0.5506965324007853}}}, "iid_acc": 0.9925, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 0.858018142885935, "mean_class_cosine": 0.6539381215477612}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.7232289950576606, "drop": -0.03822899505766053, "rotation": {"subspace_principal_angle": 0.5872782478443009, "mean_class_cosine": 0.9996739997061842}}}, "iid_acc": 0.685, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5575352393582083, "mean_class_cosine": 0.999883744660425}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915}}, "iid_acc": 0.9925, "selectivity": 0.1975} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.6465638148667602, "drop": 0.017186185133239773, "rotation": {"subspace_principal_angle": 1.519163670668585, "mean_class_cosine": 0.05160971766617552}}}, "iid_acc": 0.66375, "selectivity": 0.0774999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4801296825592345, "mean_class_cosine": 0.0905424753256214}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.5077138849929874, "drop": -0.03021388499298744, "rotation": {"subspace_principal_angle": 1.43231424996757, "mean_class_cosine": -0.6739732724428789}}}, "iid_acc": 0.4775, "selectivity": -0.10875000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.2395084599326798, "mean_class_cosine": 0.2320322894243369}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6830294530154277, "drop": 0.011970546984572228}}, "iid_acc": 0.695, "selectivity": 0.1087499999999999} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": 0.0058126901347240745, "rotation": {"subspace_principal_angle": 1.429726953139992, "mean_class_cosine": 0.17310900825401718}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.4510331206422105, "mean_class_cosine": 0.16038164209334843}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.47115384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.423728813559322, "drop": 0.04742503259452413, "rotation": {"subspace_principal_angle": 1.1889023495577067, "mean_class_cosine": 0.5428679561508623}}}, "iid_acc": 0.47115384615384615, "selectivity": 0.08653846153846151, "iid_split_rotation": {"subspace_principal_angle": 1.560544322206414, "mean_class_cosine": 0.7853248916815403}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.027243589743589758}}, "iid_acc": 0.6394230769230769, "selectivity": 0.25480769230769224} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8835616438356164, "drop": 0.005188356164383623, "rotation": {"subspace_principal_angle": 1.4311921432320234, "mean_class_cosine": 0.13915115995570887}}}, "iid_acc": 0.88875, "selectivity": 0.21500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.3924741674351642, "mean_class_cosine": 0.17737858927148076}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.59125, "drop": 0.0}, "paraphrase": {"acc": 0.589041095890411, "drop": 0.0022089041095890716, "rotation": {"subspace_principal_angle": 1.3624580504069597, "mean_class_cosine": 0.7634106223781487}}}, "iid_acc": 0.59125, "selectivity": -0.0824999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5428225535260213, "mean_class_cosine": 0.8607299633579588}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8958904109589041, "drop": 0.010359589041095907}}, "iid_acc": 0.90625, "selectivity": 0.23250000000000004} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8224687933425797, "drop": 0.015031206657420304, "rotation": {"subspace_principal_angle": 1.33328453850938, "mean_class_cosine": 0.23528499856171853}}, "domain": {"acc": 0.66, "drop": 0.1775, "rotation": {"subspace_principal_angle": 1.451106758229533, "mean_class_cosine": 0.1194040025165285}}}, "iid_acc": 0.8375, "selectivity": 0.31625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2770908389865925, "mean_class_cosine": 0.28950101482596285}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.007765256588072145, "rotation": {"subspace_principal_angle": 1.3142819239703394, "mean_class_cosine": 0.9932995750720417}}, "domain": {"acc": 0.5, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5693408077430426, "mean_class_cosine": 0.210032160652116}}}, "iid_acc": 0.50125, "selectivity": -0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.0510108236306581, "mean_class_cosine": 0.9827971130549489}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8377253814147018, "drop": 0.008524618585298183}, "domain": {"acc": 0.83625, "drop": 0.009999999999999898}}, "iid_acc": 0.84625, "selectivity": 0.32499999999999996} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8803149606299212, "drop": 0.0009350393700787718, "rotation": {"subspace_principal_angle": 1.3448528768477728, "mean_class_cosine": 0.3368814094412147}}}, "iid_acc": 0.88125, "selectivity": 0.6575, "iid_split_rotation": {"subspace_principal_angle": 1.4525237248303589, "mean_class_cosine": 0.271959475761732}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5181102362204725, "drop": 0.024389763779527507, "rotation": {"subspace_principal_angle": 1.0916569684946897, "mean_class_cosine": 0.9392101199033003}}}, "iid_acc": 0.5425, "selectivity": 0.31875, "iid_split_rotation": {"subspace_principal_angle": 1.5156324848496125, "mean_class_cosine": 0.7159087166162079}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8724409448818897, "drop": 0.022559055118110294}}, "iid_acc": 0.895, "selectivity": 0.67125} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5384615384615384, "drop": -0.01221153846153844, "rotation": {"subspace_principal_angle": 1.4851659522030354, "mean_class_cosine": 0.08552576462723123}}}, "iid_acc": 0.52625, "selectivity": 0.03249999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.491288991251595, "mean_class_cosine": 0.07942359551958941}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.4771838331160365, "drop": -0.0009338331160365088, "rotation": {"subspace_principal_angle": 1.3966119018215262, "mean_class_cosine": -0.17561134987197125}}}, "iid_acc": 0.47625, "selectivity": -0.017500000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.5601173526903696, "mean_class_cosine": 0.2908977808248975}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5149934810951761, "drop": -0.014993481095176064}}, "iid_acc": 0.5, "selectivity": 0.006249999999999978} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.6162018592297477, "drop": 0.05254814077025227, "rotation": {"subspace_principal_angle": 1.3864574345230025, "mean_class_cosine": 0.33562554763173674}}}, "iid_acc": 0.66875, "selectivity": 0.40499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3371373671231837, "mean_class_cosine": 0.3302011789609844}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.185, "drop": 0.0}, "paraphrase": {"acc": 0.20185922974767595, "drop": -0.016859229747675952, "rotation": {"subspace_principal_angle": 1.5059232736861428, "mean_class_cosine": 0.4846447217345316}}}, "iid_acc": 0.185, "selectivity": -0.07874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1753986125618363, "mean_class_cosine": 0.6615083924419292}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6254980079681275, "drop": 0.019501992031872528}}, "iid_acc": 0.645, "selectivity": 0.38125000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7184466019417476, "drop": 0.022803398058252378, "rotation": {"subspace_principal_angle": 1.5041417980246978, "mean_class_cosine": 0.06660518398513274}}}, "iid_acc": 0.74125, "selectivity": 0.22124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4864371370902933, "mean_class_cosine": 0.08425916865336666}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5485436893203883, "drop": 0.00020631067961163208, "rotation": {"subspace_principal_angle": 1.3708104020371576, "mean_class_cosine": 0.9683596041166356}}}, "iid_acc": 0.54875, "selectivity": 0.028749999999999942, "iid_split_rotation": {"subspace_principal_angle": 1.515659944278173, "mean_class_cosine": 0.9203876968475591}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.77875, "drop": 0.0}, "paraphrase": {"acc": 0.7880258899676376, "drop": -0.009275889967637507}}, "iid_acc": 0.77875, "selectivity": 0.25875000000000004} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6155088852988692, "drop": 0.03949111470113087, "rotation": {"subspace_principal_angle": 1.5045399067665934, "mean_class_cosine": 0.06620795401218435}}}, "iid_acc": 0.655, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4647655973683582, "mean_class_cosine": 0.10583216572254955}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5201938610662359, "drop": 0.017306138933764093, "rotation": {"subspace_principal_angle": 1.402582557195184, "mean_class_cosine": 0.9741348383409788}}}, "iid_acc": 0.5375, "selectivity": 0.018749999999999933, "iid_split_rotation": {"subspace_principal_angle": 1.518487153242731, "mean_class_cosine": 0.9849570196369872}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6720516962843296, "drop": 0.021698303715670386}}, "iid_acc": 0.69375, "selectivity": 0.17499999999999993} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.7027914614121511, "drop": -0.00904146141215112, "rotation": {"subspace_principal_angle": 1.4423830102309243, "mean_class_cosine": 0.12806068526617434}}}, "iid_acc": 0.69375, "selectivity": 0.16125, "iid_split_rotation": {"subspace_principal_angle": 1.380371124548562, "mean_class_cosine": 0.1892764282980448}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.545, "drop": 0.0}, "paraphrase": {"acc": 0.5500821018062397, "drop": -0.005082101806239692, "rotation": {"subspace_principal_angle": 1.0329294851079376, "mean_class_cosine": 0.9281244210082812}}}, "iid_acc": 0.545, "selectivity": 0.012500000000000067, "iid_split_rotation": {"subspace_principal_angle": 1.1987031561182855, "mean_class_cosine": 0.701876474533382}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7323481116584565, "drop": -0.0010981116584565243}}, "iid_acc": 0.73125, "selectivity": 0.19874999999999998} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.9292649098474342, "drop": 0.005735090152565903, "rotation": {"subspace_principal_angle": 1.229191428617486, "mean_class_cosine": 0.3349996872547975}}}, "iid_acc": 0.935, "selectivity": 0.43875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0937898079150548, "mean_class_cosine": 0.4591219079682709}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6907073509015257, "drop": 0.000542649098474346, "rotation": {"subspace_principal_angle": 1.2452578371926764, "mean_class_cosine": 0.9675450369885805}}}, "iid_acc": 0.69125, "selectivity": 0.195, "iid_split_rotation": {"subspace_principal_angle": 1.2562849442009068, "mean_class_cosine": 0.9533786218456732}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9251040221914009, "drop": -0.0026040221914008788}}, "iid_acc": 0.9225, "selectivity": 0.42624999999999996} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.7, "drop": 0.0}, "paraphrase": {"acc": 0.6690042075736325, "drop": 0.030995792426367408, "rotation": {"subspace_principal_angle": 1.4968125785391955, "mean_class_cosine": 0.0739162738788901}}}, "iid_acc": 0.7, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.424774943726807, "mean_class_cosine": 0.14550301874390104}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.48625, "drop": 0.0}, "paraphrase": {"acc": 0.5133239831697055, "drop": -0.027073983169705496, "rotation": {"subspace_principal_angle": 1.473176890758434, "mean_class_cosine": -0.2335794158894724}}}, "iid_acc": 0.48625, "selectivity": -0.06749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2575303716090405, "mean_class_cosine": 0.3250973513251493}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.6975, "drop": 0.0}, "paraphrase": {"acc": 0.6760168302945302, "drop": 0.021483169705469818}}, "iid_acc": 0.6975, "selectivity": 0.14375000000000004} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": 0.010620382442416365, "rotation": {"subspace_principal_angle": 1.4206287662194153, "mean_class_cosine": 0.19628077809764047}}}, "iid_acc": 0.6490384615384616, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.455372039968678, "mean_class_cosine": 0.19590134511595567}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.39903846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.3785310734463277, "drop": 0.02050738809213387, "rotation": {"subspace_principal_angle": 0.9199382488566391, "mean_class_cosine": 0.9264117531113749}}}, "iid_acc": 0.39903846153846156, "selectivity": 0.009615384615384637, "iid_split_rotation": {"subspace_principal_angle": 1.366185399901852, "mean_class_cosine": 0.35869369801616585}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6271186440677966, "drop": 0.04115058670143412}}, "iid_acc": 0.6682692307692307, "selectivity": 0.2788461538461538} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8876712328767123, "drop": 0.01357876712328765, "rotation": {"subspace_principal_angle": 1.431717330740105, "mean_class_cosine": 0.13863106275045062}}}, "iid_acc": 0.90125, "selectivity": 0.21999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4181086678178447, "mean_class_cosine": 0.15209506900442057}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5794520547945206, "drop": 0.006797945205479494, "rotation": {"subspace_principal_angle": 0.8966790530917582, "mean_class_cosine": 0.766361309893145}}}, "iid_acc": 0.58625, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.7483984334756586, "mean_class_cosine": 0.8336256970933262}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8890410958904109, "drop": 0.018458904109589058}}, "iid_acc": 0.9075, "selectivity": 0.22624999999999995} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.835, "drop": 0.0}, "paraphrase": {"acc": 0.7877947295423023, "drop": 0.04720527045769762, "rotation": {"subspace_principal_angle": 1.2702602526749749, "mean_class_cosine": 0.29603229534182074}}, "domain": {"acc": 0.78375, "drop": 0.05125000000000002, "rotation": {"subspace_principal_angle": 1.2958495701467163, "mean_class_cosine": 0.2714956938278416}}}, "iid_acc": 0.835, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 1.1803611728107277, "mean_class_cosine": 0.38059085697680345}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.001515256588072167, "rotation": {"subspace_principal_angle": 1.2132834342045293, "mean_class_cosine": 0.9993075605412058}}, "domain": {"acc": 0.5, "drop": 0.007499999999999951, "rotation": {"subspace_principal_angle": 1.3982125329965656, "mean_class_cosine": 0.036308404316974255}}}, "iid_acc": 0.5075, "selectivity": 0.0012499999999999734, "iid_split_rotation": {"subspace_principal_angle": 1.4113229091197326, "mean_class_cosine": 0.9975285200002872}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.7877947295423023, "drop": 0.029705270457697663}, "domain": {"acc": 0.855, "drop": -0.03749999999999998}}, "iid_acc": 0.8175, "selectivity": 0.31125} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8893203883495145, "drop": 0.013179611650485423, "rotation": {"subspace_principal_angle": 1.067178302018369, "mean_class_cosine": 0.48259750926799183}}, "domain": {"acc": 0.5675, "drop": 0.33499999999999996, "rotation": {"subspace_principal_angle": 1.2593770522331844, "mean_class_cosine": 0.3064099515537772}}, "length": {"acc": 0.9095022624434389, "drop": -0.0070022624434389336, "rotation": {"subspace_principal_angle": 1.0045870056146216, "mean_class_cosine": 0.536436803133892}}}, "iid_acc": 0.9025, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.0653451564032532, "mean_class_cosine": 0.4842022438803039}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.7611650485436893, "drop": -0.06491504854368924, "rotation": {"subspace_principal_angle": 1.4069108146369018, "mean_class_cosine": 0.9395668351834272}}, "domain": {"acc": 0.57875, "drop": 0.11750000000000005, "rotation": {"subspace_principal_angle": 1.2932791063335667, "mean_class_cosine": 0.34457030192799165}}, "length": {"acc": 0.7134238310708899, "drop": -0.017173831070889878, "rotation": {"subspace_principal_angle": 1.5504433084962168, "mean_class_cosine": 0.9246357737712342}}}, "iid_acc": 0.69625, "selectivity": 0.18125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3250489130035334, "mean_class_cosine": 0.7575530733718268}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8699029126213592, "drop": 0.030097087378640808}, "domain": {"acc": 0.515, "drop": 0.385}, "length": {"acc": 0.9034690799396682, "drop": -0.0034690799396681626}}, "iid_acc": 0.9, "selectivity": 0.385} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8881889763779528, "drop": 0.006811023622047219, "rotation": {"subspace_principal_angle": 1.2507965877486584, "mean_class_cosine": 0.4519091647049873}}}, "iid_acc": 0.895, "selectivity": 0.635, "iid_split_rotation": {"subspace_principal_angle": 1.2538562323332731, "mean_class_cosine": 0.4436394901742749}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.3725, "drop": 0.0}, "paraphrase": {"acc": 0.36220472440944884, "drop": 0.010295275590551156, "rotation": {"subspace_principal_angle": 1.1149135781556565, "mean_class_cosine": 0.9771677078579994}}}, "iid_acc": 0.3725, "selectivity": 0.11249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1275037300836335, "mean_class_cosine": 0.5054561634322786}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8771653543307086, "drop": 0.009084645669291347}}, "iid_acc": 0.88625, "selectivity": 0.62625} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9571788413098237, "drop": 0.010321158690176357, "rotation": {"subspace_principal_angle": 1.1895811374243113, "mean_class_cosine": 0.6930412270179037}}}, "iid_acc": 0.9675, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.1204246248516365, "mean_class_cosine": 0.734504756933933}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.28125, "drop": 0.0}, "paraphrase": {"acc": 0.3047858942065491, "drop": -0.023535894206549113, "rotation": {"subspace_principal_angle": 1.566917995749759, "mean_class_cosine": 0.7334608100976151}}}, "iid_acc": 0.28125, "selectivity": 0.22, "iid_split_rotation": {"subspace_principal_angle": 1.5646555108195546, "mean_class_cosine": 0.8084161094056437}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9445843828715366, "drop": 0.010415617128463395}}, "iid_acc": 0.955, "selectivity": 0.8937499999999999} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.60625, "drop": 0.0}, "paraphrase": {"acc": 0.6375488917861799, "drop": -0.03129889178617995, "rotation": {"subspace_principal_angle": 1.2380910520249748, "mean_class_cosine": 0.3266011440007731}}}, "iid_acc": 0.60625, "selectivity": 0.08499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2935899170666836, "mean_class_cosine": 0.2736697785737916}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.4915254237288136, "drop": -0.014025423728813602, "rotation": {"subspace_principal_angle": 1.481666093252446, "mean_class_cosine": -0.9406980138297385}}}, "iid_acc": 0.4775, "selectivity": -0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2591046492876659, "mean_class_cosine": 0.7431471438900024}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5019556714471969, "drop": 0.011794328552803157}}, "iid_acc": 0.51375, "selectivity": -0.007499999999999951} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.5962815405046481, "drop": -0.006281540504648131, "rotation": {"subspace_principal_angle": 1.3616817945475932, "mean_class_cosine": 0.34929473551176465}}}, "iid_acc": 0.59, "selectivity": 0.33499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3738625012283205, "mean_class_cosine": 0.31435666695467485}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.07625, "drop": 0.0}, "paraphrase": {"acc": 0.07702523240371846, "drop": -0.0007752324037184621, "rotation": {"subspace_principal_angle": 1.4014772128715691, "mean_class_cosine": -0.26880588026423574}}}, "iid_acc": 0.07625, "selectivity": -0.17875000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.2922523705095328, "mean_class_cosine": 0.5906582844419805}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.49125, "drop": 0.0}, "paraphrase": {"acc": 0.4913678618857902, "drop": -0.00011786188579016033}}, "iid_acc": 0.49125, "selectivity": 0.23625000000000002} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.765, "drop": 0.0}, "paraphrase": {"acc": 0.7686084142394822, "drop": -0.003608414239482216, "rotation": {"subspace_principal_angle": 1.267703414249137, "mean_class_cosine": 0.2984735606750531}}}, "iid_acc": 0.765, "selectivity": 0.21250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.297520243254995, "mean_class_cosine": 0.26988739363466807}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5355987055016181, "drop": 0.0019012944983818336, "rotation": {"subspace_principal_angle": 1.4289319500576079, "mean_class_cosine": 0.9995585763899515}}}, "iid_acc": 0.5375, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.095765507578661, "mean_class_cosine": 0.9992212510316336}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6828478964401294, "drop": 0.012152103559870508}}, "iid_acc": 0.695, "selectivity": 0.14249999999999996} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6865912762520194, "drop": 0.007158723747980544, "rotation": {"subspace_principal_angle": 1.3189864885611238, "mean_class_cosine": 0.2491571278556936}}}, "iid_acc": 0.69375, "selectivity": 0.19374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3741028601612726, "mean_class_cosine": 0.19542762781810974}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5137318255250404, "drop": 0.015018174474959678, "rotation": {"subspace_principal_angle": 1.410712709591301, "mean_class_cosine": 0.9996250431090286}}}, "iid_acc": 0.52875, "selectivity": 0.028750000000000053, "iid_split_rotation": {"subspace_principal_angle": 0.9117211516634655, "mean_class_cosine": 0.9997105404220764}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.592891760904685, "drop": 0.018358239095314977}}, "iid_acc": 0.61125, "selectivity": 0.11124999999999996} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.6995073891625616, "drop": 0.016742610837438443, "rotation": {"subspace_principal_angle": 1.398637777187281, "mean_class_cosine": 0.17130938688822317}}}, "iid_acc": 0.71625, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4203666415350267, "mean_class_cosine": 0.14986297902276266}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.4975369458128079, "drop": -0.0012869458128078604, "rotation": {"subspace_principal_angle": 1.5334169157311857, "mean_class_cosine": 0.9992809010722258}}}, "iid_acc": 0.49625, "selectivity": -0.08625, "iid_split_rotation": {"subspace_principal_angle": 1.523586797466333, "mean_class_cosine": 0.9461897010476042}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6765188834154351, "drop": 0.01473111658456494}}, "iid_acc": 0.69125, "selectivity": 0.10875000000000001} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.9275, "drop": 0.0}, "paraphrase": {"acc": 0.9181692094313454, "drop": 0.009330790568654601, "rotation": {"subspace_principal_angle": 1.0867093987105656, "mean_class_cosine": 0.4654003933713694}}}, "iid_acc": 0.9275, "selectivity": 0.45875, "iid_split_rotation": {"subspace_principal_angle": 1.0076905925780175, "mean_class_cosine": 0.533814981080676}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5312066574202496, "drop": 4.33425797503606e-05, "rotation": {"subspace_principal_angle": 0.9422006877595401, "mean_class_cosine": 0.9720210108018141}}}, "iid_acc": 0.53125, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 0.9637054547388073, "mean_class_cosine": 0.8521486536674154}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.8987517337031901, "drop": 0.016248266296809977}}, "iid_acc": 0.915, "selectivity": 0.44625000000000004} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9967051070840197, "drop": -0.0017051070840197191, "rotation": {"subspace_principal_angle": 1.004243983718731, "mean_class_cosine": 0.5367262616365928}}}, "iid_acc": 0.995, "selectivity": 0.1725, "iid_split_rotation": {"subspace_principal_angle": 0.8818699720917821, "mean_class_cosine": 0.6357087708516111}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.7166392092257001, "drop": -0.021639209225700173, "rotation": {"subspace_principal_angle": 1.3903038602930748, "mean_class_cosine": 0.999942812375732}}}, "iid_acc": 0.695, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2573644298943418, "mean_class_cosine": 0.9999792008680795}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915}}, "iid_acc": 0.9925, "selectivity": 0.17000000000000004} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6451612903225806, "drop": 0.03733870967741937, "rotation": {"subspace_principal_angle": 1.534693889353481, "mean_class_cosine": 0.03609459538397165}}}, "iid_acc": 0.6825, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.402974256327293, "mean_class_cosine": 0.16703541534575633}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.47375, "drop": 0.0}, "paraphrase": {"acc": 0.49789621318373073, "drop": -0.024146213183730725, "rotation": {"subspace_principal_angle": 1.4387989610109857, "mean_class_cosine": -0.9029123388757152}}}, "iid_acc": 0.47375, "selectivity": -0.12749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5531483539033937, "mean_class_cosine": -0.16026020693951765}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.7124824684431977, "drop": -0.011232468443197696}}, "iid_acc": 0.70125, "selectivity": 0.10000000000000009} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6779661016949152, "drop": -0.009696870925684498, "rotation": {"subspace_principal_angle": 1.3728662084585508, "mean_class_cosine": 0.23801349724373286}}}, "iid_acc": 0.6682692307692307, "selectivity": 0.23076923076923073, "iid_split_rotation": {"subspace_principal_angle": 1.3852570180024775, "mean_class_cosine": 0.240802390739838}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.49038461538461536, "drop": 0.0}, "paraphrase": {"acc": 0.4124293785310734, "drop": 0.07795523685354194, "rotation": {"subspace_principal_angle": 1.093285198742622, "mean_class_cosine": 0.5340692801147456}}}, "iid_acc": 0.49038461538461536, "selectivity": 0.05288461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.1837899565316878, "mean_class_cosine": 0.4155974505600208}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.6298076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6497175141242938, "drop": -0.019909821816601503}}, "iid_acc": 0.6298076923076923, "selectivity": 0.1923076923076923} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.873972602739726, "drop": 0.026027397260274032, "rotation": {"subspace_principal_angle": 1.2965339594699503, "mean_class_cosine": 0.2708369468728177}}}, "iid_acc": 0.9, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.2862423961352993, "mean_class_cosine": 0.28072934769702657}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.575, "drop": 0.0}, "paraphrase": {"acc": 0.5671232876712329, "drop": 0.007876712328767077, "rotation": {"subspace_principal_angle": 1.0126162968840653, "mean_class_cosine": 0.687860663967105}}}, "iid_acc": 0.575, "selectivity": -0.15375000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.5178918149259684, "mean_class_cosine": 0.7969393733558711}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8575342465753425, "drop": 0.003715753424657464}}, "iid_acc": 0.86125, "selectivity": 0.13249999999999995} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7098591549295775, "drop": 0.04639084507042246, "rotation": {"subspace_principal_angle": 1.285187527092102, "mean_class_cosine": 0.2817416408844383}}, "domain": {"acc": 0.71375, "drop": 0.04249999999999998, "rotation": {"subspace_principal_angle": 1.3127911813868958, "mean_class_cosine": 0.25515223416644334}}}, "iid_acc": 0.75625, "selectivity": 0.22124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2283794043097545, "mean_class_cosine": 0.33576468083719035}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5253521126760563, "drop": -0.026602112676056278, "rotation": {"subspace_principal_angle": 1.214807789184398, "mean_class_cosine": 0.9831397387284473}}, "domain": {"acc": 0.4875, "drop": 0.011250000000000038, "rotation": {"subspace_principal_angle": 1.5537344695259054, "mean_class_cosine": 0.34478575371993236}}}, "iid_acc": 0.49875, "selectivity": -0.036250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1073273476869623, "mean_class_cosine": 0.9951775418509261}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7126760563380282, "drop": 0.018573943661971737}, "domain": {"acc": 0.7225, "drop": 0.008749999999999925}}, "iid_acc": 0.73125, "selectivity": 0.19624999999999992} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.82, "drop": 0.0}, "paraphrase": {"acc": 0.8269961977186312, "drop": -0.006996197718631247, "rotation": {"subspace_principal_angle": 1.0415904118081376, "mean_class_cosine": 0.5048480397255118}}, "domain": {"acc": 0.59375, "drop": 0.22624999999999995, "rotation": {"subspace_principal_angle": 1.2834570977197297, "mean_class_cosine": 0.28340154844279125}}, "length": {"acc": 0.8264840182648402, "drop": -0.006484018264840241, "rotation": {"subspace_principal_angle": 0.999883755381926, "mean_class_cosine": 0.5404001186906672}}}, "iid_acc": 0.82, "selectivity": 0.3549999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9949686563230938, "mean_class_cosine": 0.5445291790193694}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6863117870722434, "drop": 0.018688212927756576, "rotation": {"subspace_principal_angle": 1.3283778591084485, "mean_class_cosine": 0.9327609071818541}}, "domain": {"acc": 0.565, "drop": 0.14, "rotation": {"subspace_principal_angle": 1.5278891311014122, "mean_class_cosine": 0.3806568220146318}}, "length": {"acc": 0.7062404870624048, "drop": -0.0012404870624048714, "rotation": {"subspace_principal_angle": 1.3125076791898616, "mean_class_cosine": 0.9017445724833599}}}, "iid_acc": 0.705, "selectivity": 0.23999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4606505742125795, "mean_class_cosine": 0.9081384351848174}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.7975, "drop": 0.0}, "paraphrase": {"acc": 0.7984790874524715, "drop": -0.0009790874524715498}, "domain": {"acc": 0.645, "drop": 0.15249999999999997}, "length": {"acc": 0.7975646879756468, "drop": -6.46879756468488e-05}}, "iid_acc": 0.7975, "selectivity": 0.33249999999999996} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8611544461778471, "drop": 0.0013455538221529562, "rotation": {"subspace_principal_angle": 0.9939269073783296, "mean_class_cosine": 0.6453866739528615}}}, "iid_acc": 0.8625, "selectivity": 0.5362500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4271388898728552, "mean_class_cosine": 0.6524369913147846}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7316692667706708, "drop": 0.00458073322932917, "rotation": {"subspace_principal_angle": 1.3724063354724922, "mean_class_cosine": 0.9469017405917457}}}, "iid_acc": 0.73625, "selectivity": 0.41, "iid_split_rotation": {"subspace_principal_angle": 1.0603683518688838, "mean_class_cosine": 0.950801823774931}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.81625, "drop": 0.0}, "paraphrase": {"acc": 0.8081123244929798, "drop": 0.00813767550702027}}, "iid_acc": 0.81625, "selectivity": 0.49000000000000005} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9160671462829736, "drop": 0.021432853717026412, "rotation": {"subspace_principal_angle": 1.1847135524986785, "mean_class_cosine": 0.706584730242114}}}, "iid_acc": 0.9375, "selectivity": 0.87625, "iid_split_rotation": {"subspace_principal_angle": 1.11150260490418, "mean_class_cosine": 0.7556327352536992}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.4460431654676259, "drop": 0.05520683453237407, "rotation": {"subspace_principal_angle": 1.199717400716062, "mean_class_cosine": 0.8611479821783318}}}, "iid_acc": 0.50125, "selectivity": 0.43999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4104637914876992, "mean_class_cosine": 0.8005163008883979}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8465227817745803, "drop": 0.03222721822541974}}, "iid_acc": 0.87875, "selectivity": 0.8175} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.515032679738562, "drop": 0.009967320261437984, "rotation": {"subspace_principal_angle": 1.566988880756832, "mean_class_cosine": 0.003807436838872146}}}, "iid_acc": 0.525, "selectivity": 0.03750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.515059781737158, "mean_class_cosine": 0.055707691363518934}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5189542483660131, "drop": 0.00729575163398688, "rotation": {"subspace_principal_angle": 1.565436854816222, "mean_class_cosine": 0.11469077846809743}}}, "iid_acc": 0.52625, "selectivity": 0.03875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.519484137603478, "mean_class_cosine": 0.056727235397258546}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5124183006535947, "drop": 0.023831699346405277}}, "iid_acc": 0.53625, "selectivity": 0.048750000000000016} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5485175202156334, "drop": 0.023982479784366628, "rotation": {"subspace_principal_angle": 1.4303692527758758, "mean_class_cosine": 0.35644155459058985}}}, "iid_acc": 0.5725, "selectivity": 0.2575, "iid_split_rotation": {"subspace_principal_angle": 1.4149642770924389, "mean_class_cosine": 0.3618340196725102}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.28125, "drop": 0.0}, "paraphrase": {"acc": 0.28975741239892183, "drop": -0.008507412398921832, "rotation": {"subspace_principal_angle": 1.5249290421568074, "mean_class_cosine": 0.644382717112205}}}, "iid_acc": 0.28125, "selectivity": -0.03375, "iid_split_rotation": {"subspace_principal_angle": 1.5382442039631525, "mean_class_cosine": 0.5744283723940081}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5053908355795148, "drop": 0.009609164420485206}}, "iid_acc": 0.515, "selectivity": 0.2} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.7345575959933222, "drop": -0.01080759599332215, "rotation": {"subspace_principal_angle": 1.2514538362385088, "mean_class_cosine": 0.31394236141299214}}}, "iid_acc": 0.72375, "selectivity": 0.22499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3629405243520136, "mean_class_cosine": 0.20636233074920893}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.6060100166944908, "drop": -0.0297600166944908, "rotation": {"subspace_principal_angle": 1.4254734942092195, "mean_class_cosine": 0.930851220197922}}}, "iid_acc": 0.57625, "selectivity": 0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5027930622930314, "mean_class_cosine": 0.8705083526841852}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7312186978297162, "drop": 0.003781302170283829}}, "iid_acc": 0.735, "selectivity": 0.23624999999999996} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6237623762376238, "drop": 0.058737623762376234, "rotation": {"subspace_principal_angle": 1.3701462267495714, "mean_class_cosine": 0.19930643009504329}}}, "iid_acc": 0.6825, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3138345624755305, "mean_class_cosine": 0.25414324935789645}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.6072607260726073, "drop": -0.007260726072607304, "rotation": {"subspace_principal_angle": 1.420901469694498, "mean_class_cosine": 0.9343142755680263}}}, "iid_acc": 0.6, "selectivity": 0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2228313385316762, "mean_class_cosine": 0.8726714354786228}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.66625, "drop": 0.0}, "paraphrase": {"acc": 0.6254125412541254, "drop": 0.04083745874587463}}, "iid_acc": 0.66625, "selectivity": 0.17625000000000002} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7648902821316614, "drop": -0.03364028213166148, "rotation": {"subspace_principal_angle": 1.1748420004300262, "mean_class_cosine": 0.38568885340073633}}}, "iid_acc": 0.73125, "selectivity": 0.10624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1987833070239917, "mean_class_cosine": 0.3634914913995175}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5094043887147336, "drop": 0.020595611285266435, "rotation": {"subspace_principal_angle": 1.3931515700293393, "mean_class_cosine": 0.8341203196581115}}}, "iid_acc": 0.53, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.5409782777316359, "mean_class_cosine": 0.7395207496379319}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7523510971786834, "drop": -0.026101097178683474}}, "iid_acc": 0.72625, "selectivity": 0.10124999999999995} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.8688981868898187, "drop": 0.046101813110181356, "rotation": {"subspace_principal_angle": 1.0974938475903255, "mean_class_cosine": 0.45582819608799463}}}, "iid_acc": 0.915, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 0.9804764137968498, "mean_class_cosine": 0.5566268231619027}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.8125, "drop": 0.0}, "paraphrase": {"acc": 0.6443514644351465, "drop": 0.16814853556485354, "rotation": {"subspace_principal_angle": 1.3274674971100646, "mean_class_cosine": 0.8179461473320861}}}, "iid_acc": 0.8125, "selectivity": 0.25125, "iid_split_rotation": {"subspace_principal_angle": 1.311349815093557, "mean_class_cosine": 0.9540198561738099}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8521617852161785, "drop": 0.04783821478382155}}, "iid_acc": 0.9, "selectivity": 0.33875} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9852700490998363, "drop": 0.0059799509001636775, "rotation": {"subspace_principal_angle": 0.6333378965948789, "mean_class_cosine": 0.8060565063456264}}}, "iid_acc": 0.99125, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 0.6062508845786333, "mean_class_cosine": 0.8217899986183517}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8772504091653028, "drop": 0.0027495908346971687, "rotation": {"subspace_principal_angle": 1.13183836457679, "mean_class_cosine": 0.9746417417399029}}}, "iid_acc": 0.88, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.0774674925239829, "mean_class_cosine": 0.989696915490704}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9885433715220949, "drop": 0.00020662847790509087}}, "iid_acc": 0.98875, "selectivity": 0.11750000000000005} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.7, "drop": -0.022499999999999964, "rotation": {"subspace_principal_angle": 1.493482753933245, "mean_class_cosine": 0.0772365736671137}}}, "iid_acc": 0.6775, "selectivity": 0.012499999999999956, "iid_split_rotation": {"subspace_principal_angle": 1.5233694038561607, "mean_class_cosine": 0.0474091452723575}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.523943661971831, "drop": 0.003556338028169015, "rotation": {"subspace_principal_angle": 1.364642797562692, "mean_class_cosine": 0.26221087675633176}}}, "iid_acc": 0.5275, "selectivity": -0.13750000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.5645617307273036, "mean_class_cosine": 0.0046736400727034775}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7492957746478873, "drop": -0.019295774647887343}}, "iid_acc": 0.73, "selectivity": 0.06499999999999995} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5722543352601156, "drop": -0.01936971987550018, "rotation": {"subspace_principal_angle": 1.4113114437701857, "mean_class_cosine": 0.18732090600899934}}}, "iid_acc": 0.5528846153846154, "selectivity": 0.1586538461538462, "iid_split_rotation": {"subspace_principal_angle": 1.4113877250233597, "mean_class_cosine": 0.1900533891006572}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.41346153846153844, "drop": 0.0}, "paraphrase": {"acc": 0.44508670520231214, "drop": -0.0316251667407737, "rotation": {"subspace_principal_angle": 1.4614748229703205, "mean_class_cosine": 0.41743847989205585}}}, "iid_acc": 0.41346153846153844, "selectivity": 0.019230769230769218, "iid_split_rotation": {"subspace_principal_angle": 1.1472146976181432, "mean_class_cosine": 0.7289310712969407}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.5336538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": -0.044380835927078754}}, "iid_acc": 0.5336538461538461, "selectivity": 0.13942307692307693} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8848821081830791, "drop": -0.009882108183079108, "rotation": {"subspace_principal_angle": 1.2416497520853838, "mean_class_cosine": 0.32323553402514815}}}, "iid_acc": 0.875, "selectivity": 0.11124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1798700863789469, "mean_class_cosine": 0.3810449400671245}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5423023578363384, "drop": 0.022697642163661547, "rotation": {"subspace_principal_angle": 1.0521118048612201, "mean_class_cosine": 0.46745139889316134}}}, "iid_acc": 0.565, "selectivity": -0.1987500000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8706941393205756, "mean_class_cosine": 0.8476985688696614}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.84875, "drop": 0.0}, "paraphrase": {"acc": 0.855755894590846, "drop": -0.007005894590845996}}, "iid_acc": 0.84875, "selectivity": 0.08499999999999996} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.81375, "drop": 0.0}, "paraphrase": {"acc": 0.7830985915492957, "drop": 0.030651408450704243, "rotation": {"subspace_principal_angle": 1.2474741652442436, "mean_class_cosine": 0.3177183323280257}}, "domain": {"acc": 0.7875, "drop": 0.026249999999999996, "rotation": {"subspace_principal_angle": 1.1724916667604595, "mean_class_cosine": 0.38785627191420236}}}, "iid_acc": 0.81375, "selectivity": 0.27125, "iid_split_rotation": {"subspace_principal_angle": 1.2460595658666687, "mean_class_cosine": 0.31905931609907595}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5450704225352113, "drop": -0.018820422535211323, "rotation": {"subspace_principal_angle": 1.217800297410841, "mean_class_cosine": 0.9938492827921086}}, "domain": {"acc": 0.4875, "drop": 0.03875000000000001, "rotation": {"subspace_principal_angle": 1.4265907543600744, "mean_class_cosine": 0.524717137763018}}}, "iid_acc": 0.52625, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.0792601449175692, "mean_class_cosine": 0.9975233270077284}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.7887323943661971, "drop": 0.03376760563380288}, "domain": {"acc": 0.78875, "drop": 0.03375000000000006}}, "iid_acc": 0.8225, "selectivity": 0.28} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8269961977186312, "drop": 0.02675380228136881, "rotation": {"subspace_principal_angle": 1.0856620278217086, "mean_class_cosine": 0.46632716618126924}}, "domain": {"acc": 0.59125, "drop": 0.26249999999999996, "rotation": {"subspace_principal_angle": 1.2546499825836788, "mean_class_cosine": 0.3109062073482294}}, "length": {"acc": 0.8508371385083714, "drop": 0.0029128614916286155, "rotation": {"subspace_principal_angle": 1.0562705758313071, "mean_class_cosine": 0.492122058178101}}}, "iid_acc": 0.85375, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.04571866168401, "mean_class_cosine": 0.5012802086419837}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.6844106463878327, "drop": -0.03316064638783267, "rotation": {"subspace_principal_angle": 1.2624327988110398, "mean_class_cosine": 0.9443739768139503}}, "domain": {"acc": 0.55625, "drop": 0.09499999999999997, "rotation": {"subspace_principal_angle": 1.2485837708508047, "mean_class_cosine": 0.5140204613064764}}, "length": {"acc": 0.6727549467275494, "drop": -0.021504946727549434, "rotation": {"subspace_principal_angle": 1.3927565265280373, "mean_class_cosine": 0.891203465902359}}}, "iid_acc": 0.65125, "selectivity": 0.17375000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4689242704183265, "mean_class_cosine": 0.9409811928992509}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8136882129277566, "drop": 0.023811787072243407}, "domain": {"acc": 0.6275, "drop": 0.21000000000000008}, "length": {"acc": 0.8340943683409436, "drop": 0.003405631659056385}}, "iid_acc": 0.8375, "selectivity": 0.36000000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8533541341653667, "drop": 0.0066458658346333355, "rotation": {"subspace_principal_angle": 1.3730733013635024, "mean_class_cosine": 0.2667006488076803}}}, "iid_acc": 0.86, "selectivity": 0.59125, "iid_split_rotation": {"subspace_principal_angle": 1.4307610404327311, "mean_class_cosine": 0.277445881216753}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": -0.006445007800311986, "rotation": {"subspace_principal_angle": 1.4820823089446629, "mean_class_cosine": 0.9745838181982655}}}, "iid_acc": 0.86875, "selectivity": 0.6000000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5380315983443016, "mean_class_cosine": 0.9691609768158005}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.40875, "drop": 0.0}, "paraphrase": {"acc": 0.38221528861154447, "drop": 0.026534711388455534}}, "iid_acc": 0.40875, "selectivity": 0.14} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.935251798561151, "drop": 0.012248201438848971, "rotation": {"subspace_principal_angle": 1.1046415070501514, "mean_class_cosine": 0.6973108185521933}}}, "iid_acc": 0.9475, "selectivity": 0.87375, "iid_split_rotation": {"subspace_principal_angle": 1.0670730680867, "mean_class_cosine": 0.747139744826814}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4628297362110312, "drop": 0.03967026378896876, "rotation": {"subspace_principal_angle": 1.5435774572835343, "mean_class_cosine": 0.8716859158620075}}}, "iid_acc": 0.5025, "selectivity": 0.42874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.389055684030504, "mean_class_cosine": 0.7772996017727697}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.92625, "drop": 0.0}, "paraphrase": {"acc": 0.9160671462829736, "drop": 0.01018285371702643}}, "iid_acc": 0.92625, "selectivity": 0.8525} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5176470588235295, "drop": 0.018602941176470544, "rotation": {"subspace_principal_angle": 1.5270884311030197, "mean_class_cosine": 0.04369398057174439}}}, "iid_acc": 0.53625, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.476543368793087, "mean_class_cosine": 0.094113468729472}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.5045751633986928, "drop": 0.011674836601307192, "rotation": {"subspace_principal_angle": 1.5294742379185886, "mean_class_cosine": 0.18570440239474292}}}, "iid_acc": 0.51625, "selectivity": 0.04249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.444417852756359, "mean_class_cosine": 0.20646522396999623}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5333333333333333, "drop": 0.0029166666666666785}}, "iid_acc": 0.53625, "selectivity": 0.0625} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.5431266846361186, "drop": 0.0518733153638814, "rotation": {"subspace_principal_angle": 1.467102949540591, "mean_class_cosine": 0.29891759428990633}}}, "iid_acc": 0.595, "selectivity": 0.31375, "iid_split_rotation": {"subspace_principal_angle": 1.5242732907156675, "mean_class_cosine": 0.27641104971513886}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.1925, "drop": 0.0}, "paraphrase": {"acc": 0.20754716981132076, "drop": -0.01504716981132076, "rotation": {"subspace_principal_angle": 1.3006810428081945, "mean_class_cosine": 0.5938365988587205}}}, "iid_acc": 0.1925, "selectivity": -0.08875, "iid_split_rotation": {"subspace_principal_angle": 1.5008582164851936, "mean_class_cosine": 0.6730802808549522}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5309973045822103, "drop": 0.017752695417789677}}, "iid_acc": 0.54875, "selectivity": 0.26749999999999996} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7429048414023373, "drop": -0.002904841402337266, "rotation": {"subspace_principal_angle": 1.321041649041896, "mean_class_cosine": 0.24716625603971074}}}, "iid_acc": 0.74, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3832981091452556, "mean_class_cosine": 0.18640154565031142}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5976627712854758, "drop": -0.03766277128547579, "rotation": {"subspace_principal_angle": 1.3172594721297197, "mean_class_cosine": 0.9641971900851679}}}, "iid_acc": 0.56, "selectivity": 0.0525000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1295344939397884, "mean_class_cosine": 0.9754037717574164}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7378964941569283, "drop": -0.009146494156928253}}, "iid_acc": 0.72875, "selectivity": 0.22125000000000006} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6551155115511551, "drop": 0.01988448844884494, "rotation": {"subspace_principal_angle": 1.4103931207701927, "mean_class_cosine": 0.1597162496412831}}}, "iid_acc": 0.675, "selectivity": 0.18500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.451975006440924, "mean_class_cosine": 0.1185419210711762}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5396039603960396, "drop": 0.0016460396039603697, "rotation": {"subspace_principal_angle": 1.5446898330353456, "mean_class_cosine": 0.9934586731916171}}}, "iid_acc": 0.54125, "selectivity": 0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3541190216750216, "mean_class_cosine": 0.9952103287013051}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.6975, "drop": 0.0}, "paraphrase": {"acc": 0.6617161716171617, "drop": 0.035783828382838334}}, "iid_acc": 0.6975, "selectivity": 0.20750000000000002} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7523510971786834, "drop": -0.00735109717868343, "rotation": {"subspace_principal_angle": 1.3838768695601789, "mean_class_cosine": 0.18583289760856309}}}, "iid_acc": 0.745, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.411496556459018, "mean_class_cosine": 0.1586268817831123}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.485, "drop": 0.0}, "paraphrase": {"acc": 0.5219435736677116, "drop": -0.036943573667711616, "rotation": {"subspace_principal_angle": 1.4598677944473912, "mean_class_cosine": 0.7189138058825407}}}, "iid_acc": 0.485, "selectivity": -0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.257226644633361, "mean_class_cosine": 0.8721290560249453}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7507836990595611, "drop": 0.005466300940438873}}, "iid_acc": 0.75625, "selectivity": 0.17374999999999996} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9121338912133892, "drop": 0.010366108786610817, "rotation": {"subspace_principal_angle": 1.190236906942168, "mean_class_cosine": 0.3714399247796692}}}, "iid_acc": 0.9225, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1180814498000364, "mean_class_cosine": 0.43740853126331475}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.8019525801952581, "drop": 0.0642974198047419, "rotation": {"subspace_principal_angle": 1.316257361185483, "mean_class_cosine": 0.9428379882047717}}}, "iid_acc": 0.86625, "selectivity": 0.33124999999999993, "iid_split_rotation": {"subspace_principal_angle": 0.8984797392010688, "mean_class_cosine": 0.9658337076254448}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.9107391910739191, "drop": 0.009260808926080943}}, "iid_acc": 0.92, "selectivity": 0.385} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9950900163666121, "drop": 0.0011599836333878732, "rotation": {"subspace_principal_angle": 0.8576615569303944, "mean_class_cosine": 0.6542078543716676}}}, "iid_acc": 0.99625, "selectivity": 0.15000000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9012913552860777, "mean_class_cosine": 0.6205978969084885}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.68125, "drop": 0.0}, "paraphrase": {"acc": 0.5548281505728314, "drop": 0.1264218494271686, "rotation": {"subspace_principal_angle": 0.6861186089833221, "mean_class_cosine": 0.9976598565566228}}}, "iid_acc": 0.68125, "selectivity": -0.16499999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.6195594346223908, "mean_class_cosine": 0.9998339526246938}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": -0.0009533551554827824}}, "iid_acc": 0.9925, "selectivity": 0.1462500000000001} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6915492957746479, "drop": -0.004049295774647854, "rotation": {"subspace_principal_angle": 1.4821835383859001, "mean_class_cosine": 0.08849686598665059}}}, "iid_acc": 0.6875, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5422652050607666, "mean_class_cosine": 0.0285272510510537}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5140845070422535, "drop": 0.03966549295774646, "rotation": {"subspace_principal_angle": 1.411758063021396, "mean_class_cosine": -0.05543427344626703}}}, "iid_acc": 0.55375, "selectivity": -0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1826722547195974, "mean_class_cosine": -0.8190873052106897}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7450704225352113, "drop": -0.01757042253521124}}, "iid_acc": 0.7275, "selectivity": 0.08875} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5673076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.5549132947976878, "drop": 0.012394397510004485, "rotation": {"subspace_principal_angle": 1.435539235192927, "mean_class_cosine": 0.18347039301699317}}}, "iid_acc": 0.5673076923076923, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.3743463171910144, "mean_class_cosine": 0.24089153525080556}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.4326923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.41040462427745666, "drop": 0.022287683414851045, "rotation": {"subspace_principal_angle": 1.3988658583948614, "mean_class_cosine": -0.27577383548106055}}}, "iid_acc": 0.4326923076923077, "selectivity": 0.04326923076923078, "iid_split_rotation": {"subspace_principal_angle": 1.0687827569428319, "mean_class_cosine": 0.8344718355538371}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5953757225433526, "drop": -0.04249110715873716}}, "iid_acc": 0.5528846153846154, "selectivity": 0.1634615384615385} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.9029126213592233, "drop": -0.006662621359223353, "rotation": {"subspace_principal_angle": 1.20453886521433, "mean_class_cosine": 0.3581236367027224}}}, "iid_acc": 0.89625, "selectivity": 0.17374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2309341256728386, "mean_class_cosine": 0.3333571784399262}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5298196948682385, "drop": 0.022680305131761447, "rotation": {"subspace_principal_angle": 1.310916058135534, "mean_class_cosine": -0.07981176620814859}}}, "iid_acc": 0.5525, "selectivity": -0.17000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1896244566676593, "mean_class_cosine": 0.9512750131760109}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.8575, "drop": 0.0}, "paraphrase": {"acc": 0.8640776699029126, "drop": -0.006577669902912531}}, "iid_acc": 0.8575, "selectivity": 0.135} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8056338028169014, "drop": 0.048116197183098564, "rotation": {"subspace_principal_angle": 1.2927332146049364, "mean_class_cosine": 0.2744936747860658}}, "domain": {"acc": 0.82375, "drop": 0.030000000000000027, "rotation": {"subspace_principal_angle": 1.2846999556990648, "mean_class_cosine": 0.28220942740249133}}}, "iid_acc": 0.85375, "selectivity": 0.32125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.220104644031074, "mean_class_cosine": 0.3435474732924775}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.532394366197183, "drop": -0.012394366197183038, "rotation": {"subspace_principal_angle": 0.365026570963859, "mean_class_cosine": 0.9987360435705452}}, "domain": {"acc": 0.4875, "drop": 0.03250000000000003, "rotation": {"subspace_principal_angle": 1.5449752153763945, "mean_class_cosine": 0.4557918147259487}}}, "iid_acc": 0.52, "selectivity": -0.012499999999999956, "iid_split_rotation": {"subspace_principal_angle": 0.34709256477040135, "mean_class_cosine": 0.9995894414380515}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.8225352112676056, "drop": 0.022464788732394325}, "domain": {"acc": 0.8325, "drop": 0.012499999999999956}}, "iid_acc": 0.845, "selectivity": 0.3125} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8688212927756654, "drop": 0.016178707224334565, "rotation": {"subspace_principal_angle": 1.2184200366683278, "mean_class_cosine": 0.34512905932789906}}, "domain": {"acc": 0.7325, "drop": 0.15249999999999997, "rotation": {"subspace_principal_angle": 1.3710819230300715, "mean_class_cosine": 0.1983894193717613}}, "length": {"acc": 0.8919330289193302, "drop": -0.006933028919330231, "rotation": {"subspace_principal_angle": 1.1635770628342312, "mean_class_cosine": 0.3960575192257396}}}, "iid_acc": 0.885, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 1.178058815472859, "mean_class_cosine": 0.38271893642492194}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7585551330798479, "drop": -0.02605513307984786, "rotation": {"subspace_principal_angle": 1.0808603554768208, "mean_class_cosine": 0.9645911983303304}}, "domain": {"acc": 0.5575, "drop": 0.17500000000000004, "rotation": {"subspace_principal_angle": 1.4332455458492723, "mean_class_cosine": 0.4266662959732691}}, "length": {"acc": 0.7595129375951294, "drop": -0.02701293759512935, "rotation": {"subspace_principal_angle": 1.1028359956060956, "mean_class_cosine": 0.9361757660470457}}}, "iid_acc": 0.7325, "selectivity": 0.25375000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.340774604922017, "mean_class_cosine": 0.9544915504856384}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.876425855513308, "drop": 0.01732414448669206}, "domain": {"acc": 0.57875, "drop": 0.31500000000000006}, "length": {"acc": 0.898021308980213, "drop": -0.0042713089802129955}}, "iid_acc": 0.89375, "selectivity": 0.41500000000000004} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": 0.0023049921996879386, "rotation": {"subspace_principal_angle": 1.2939750508328909, "mean_class_cosine": 0.3697587563743038}}}, "iid_acc": 0.8775, "selectivity": 0.60625, "iid_split_rotation": {"subspace_principal_angle": 1.3239667655285696, "mean_class_cosine": 0.3832240775461434}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.57375, "drop": 0.0}, "paraphrase": {"acc": 0.5772230889235569, "drop": -0.0034730889235569107, "rotation": {"subspace_principal_angle": 1.4461389808941905, "mean_class_cosine": 0.8942227001502406}}}, "iid_acc": 0.57375, "selectivity": 0.3025, "iid_split_rotation": {"subspace_principal_angle": 1.5670258105099684, "mean_class_cosine": 0.9555720866412488}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": 0.0010549921996879652}}, "iid_acc": 0.87625, "selectivity": 0.605} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.9625, "drop": 0.0}, "paraphrase": {"acc": 0.9568345323741008, "drop": 0.0056654676258992565, "rotation": {"subspace_principal_angle": 1.0889403294719116, "mean_class_cosine": 0.7097651658028382}}}, "iid_acc": 0.9625, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.0742515810195217, "mean_class_cosine": 0.7520191017637144}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.45083932853717024, "drop": 0.02416067146282974, "rotation": {"subspace_principal_angle": 1.1366627849078705, "mean_class_cosine": 0.861011502467378}}}, "iid_acc": 0.475, "selectivity": 0.41874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5498782331536536, "mean_class_cosine": 0.7592574844685773}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9400479616306955, "drop": 0.014952038369304477}}, "iid_acc": 0.955, "selectivity": 0.8987499999999999} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5830065359477125, "drop": 0.01699346405228752, "rotation": {"subspace_principal_angle": 1.4156934488627833, "mean_class_cosine": 0.1544817430571851}}}, "iid_acc": 0.6, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3875062308714112, "mean_class_cosine": 0.18226553873635268}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5254901960784314, "drop": 0.012009803921568585, "rotation": {"subspace_principal_angle": 0.8046612760829073, "mean_class_cosine": 0.9020520136987834}}}, "iid_acc": 0.5375, "selectivity": 0.03249999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.5910745370085485, "mean_class_cosine": -0.921862888991628}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.5825, "drop": 0.0}, "paraphrase": {"acc": 0.550326797385621, "drop": 0.03217320261437906}}, "iid_acc": 0.5825, "selectivity": 0.07750000000000001} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5876010781671159, "drop": 0.011148921832884073, "rotation": {"subspace_principal_angle": 1.4649394534970999, "mean_class_cosine": 0.27348013078523753}}}, "iid_acc": 0.59875, "selectivity": 0.3075, "iid_split_rotation": {"subspace_principal_angle": 1.5210318775010727, "mean_class_cosine": 0.26014522379767274}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.18375, "drop": 0.0}, "paraphrase": {"acc": 0.1725067385444744, "drop": 0.011243261455525594, "rotation": {"subspace_principal_angle": 1.2577695007550802, "mean_class_cosine": 0.6357274521759256}}}, "iid_acc": 0.18375, "selectivity": -0.10750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4122075752855408, "mean_class_cosine": 0.6982530139834328}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.59375, "drop": 0.0}, "paraphrase": {"acc": 0.5592991913746631, "drop": 0.03445080862533689}}, "iid_acc": 0.59375, "selectivity": 0.3025} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7161936560934892, "drop": 0.0025563439065108273, "rotation": {"subspace_principal_angle": 1.4142647213981934, "mean_class_cosine": 0.1558931614185028}}}, "iid_acc": 0.71875, "selectivity": 0.23375, "iid_split_rotation": {"subspace_principal_angle": 1.4440239039597567, "mean_class_cosine": 0.12643313043512422}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5442404006677797, "drop": -0.014240400667779651, "rotation": {"subspace_principal_angle": 1.438598695547982, "mean_class_cosine": 0.9982834430354565}}}, "iid_acc": 0.53, "selectivity": 0.04500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.312217021230763, "mean_class_cosine": 0.9973052099136428}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7712854757929883, "drop": -0.0075354757929883}}, "iid_acc": 0.76375, "selectivity": 0.27875000000000005} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6567656765676567, "drop": 0.02323432343234333, "rotation": {"subspace_principal_angle": 1.4737611346239565, "mean_class_cosine": 0.09688298605763825}}}, "iid_acc": 0.68, "selectivity": 0.18875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4500535476739749, "mean_class_cosine": 0.12044961172303534}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5379537953795379, "drop": 0.002046204620462122, "rotation": {"subspace_principal_angle": 0.9674520996825667, "mean_class_cosine": 0.9975306978126972}}}, "iid_acc": 0.54, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 0.6210793518048389, "mean_class_cosine": 0.9983833459358811}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.6914191419141914, "drop": 0.041080858085808636}}, "iid_acc": 0.7325, "selectivity": 0.24125000000000002} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7304075235109718, "drop": -0.016657523510971828, "rotation": {"subspace_principal_angle": 1.4688894465039684, "mean_class_cosine": 0.10173058782242872}}}, "iid_acc": 0.71375, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.4124632331913725, "mean_class_cosine": 0.15767237056399389}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.5266457680250783, "drop": -0.030395768025078318, "rotation": {"subspace_principal_angle": 0.7632925636553622, "mean_class_cosine": 0.9245850252469596}}}, "iid_acc": 0.49625, "selectivity": -0.03999999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.8486480460237737, "mean_class_cosine": 0.7937846466717617}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7570532915360502, "drop": -0.0058032915360501924}}, "iid_acc": 0.75125, "selectivity": 0.21499999999999997} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9316596931659693, "drop": 0.0070903068340306685, "rotation": {"subspace_principal_angle": 1.1500093845203234, "mean_class_cosine": 0.40847887501442204}}}, "iid_acc": 0.93875, "selectivity": 0.4149999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.0038823721844266, "mean_class_cosine": 0.5370313385967125}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.62375, "drop": 0.0}, "paraphrase": {"acc": 0.6192468619246861, "drop": 0.004503138075313884, "rotation": {"subspace_principal_angle": 1.486642282369967, "mean_class_cosine": 0.9620795603326425}}}, "iid_acc": 0.62375, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4105065066746125, "mean_class_cosine": 0.9378064089839906}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9386331938633193, "drop": 0.00011680613668063611}}, "iid_acc": 0.93875, "selectivity": 0.4149999999999999} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9819967266775778, "drop": 0.009253273322422206, "rotation": {"subspace_principal_angle": 0.7667082993155465, "mean_class_cosine": 0.7201982449314277}}}, "iid_acc": 0.99125, "selectivity": 0.13374999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.7510140573282249, "mean_class_cosine": 0.7309972720101955}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.939443535188216, "drop": -0.006943535188215999, "rotation": {"subspace_principal_angle": 0.996635177095526, "mean_class_cosine": 0.953156648540542}}}, "iid_acc": 0.9325, "selectivity": 0.07499999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9887430183481719, "mean_class_cosine": 0.9831308468715932}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9819967266775778, "drop": 0.00675327332242226}}, "iid_acc": 0.98875, "selectivity": 0.13124999999999998} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6873239436619718, "drop": 0.02267605633802816, "rotation": {"subspace_principal_angle": 1.4852509416429518, "mean_class_cosine": 0.08544108628330754}}}, "iid_acc": 0.71, "selectivity": 0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4922855379226647, "mean_class_cosine": 0.0784301577092764}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5154929577464789, "drop": 0.03325704225352111, "rotation": {"subspace_principal_angle": 0.5600755724824341, "mean_class_cosine": -0.860265927958924}}}, "iid_acc": 0.54875, "selectivity": -0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.2092432879483297, "mean_class_cosine": 0.6330568261731295}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7464788732394366, "drop": -0.006478873239436633}}, "iid_acc": 0.74, "selectivity": 0.15125} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6184971098265896, "drop": -0.012727879057358837, "rotation": {"subspace_principal_angle": 1.4119061291333692, "mean_class_cosine": 0.1704373905622454}}}, "iid_acc": 0.6057692307692307, "selectivity": 0.24999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4297364412104765, "mean_class_cosine": 0.20248169581710895}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.4567307692307692, "drop": 0.0}, "paraphrase": {"acc": 0.4508670520231214, "drop": 0.005863717207647834, "rotation": {"subspace_principal_angle": 1.3366057576399157, "mean_class_cosine": 0.7129666544008023}}}, "iid_acc": 0.4567307692307692, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.417385145003071, "mean_class_cosine": 0.826148295807795}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.5480769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.5433526011560693, "drop": 0.004724321920853813}}, "iid_acc": 0.5480769230769231, "selectivity": 0.19230769230769235} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8932038834951457, "drop": 0.0017961165048543393, "rotation": {"subspace_principal_angle": 1.311372754664119, "mean_class_cosine": 0.2565234550971107}}}, "iid_acc": 0.895, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.2262739349985579, "mean_class_cosine": 0.337747173335497}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5436893203883495, "drop": 0.022560679611650536, "rotation": {"subspace_principal_angle": 1.2898034193621133, "mean_class_cosine": -0.3258881311565935}}}, "iid_acc": 0.56625, "selectivity": -0.1825, "iid_split_rotation": {"subspace_principal_angle": 1.2264589211138428, "mean_class_cosine": 0.9728987695098255}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.90375, "drop": 0.0}, "paraphrase": {"acc": 0.9042995839112344, "drop": -0.0005495839112343859}}, "iid_acc": 0.90375, "selectivity": 0.15500000000000003} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8323943661971831, "drop": 0.02135563380281691, "rotation": {"subspace_principal_angle": 1.2914523895644616, "mean_class_cosine": 0.27572507644251076}}, "domain": {"acc": 0.87125, "drop": -0.01749999999999996, "rotation": {"subspace_principal_angle": 1.4039511655723063, "mean_class_cosine": 0.16607215124792452}}}, "iid_acc": 0.85375, "selectivity": 0.33125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.250730833076561, "mean_class_cosine": 0.3146287288985242}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5352112676056338, "drop": -0.011461267605633707, "rotation": {"subspace_principal_angle": 0.9170707486812013, "mean_class_cosine": 0.9954078358133513}}, "domain": {"acc": 0.4875, "drop": 0.03625000000000006, "rotation": {"subspace_principal_angle": 1.5523343990205853, "mean_class_cosine": 0.22749359327828106}}}, "iid_acc": 0.52375, "selectivity": 0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.4480914990441032, "mean_class_cosine": 0.9983051052181077}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8436619718309859, "drop": 0.033838028169014045}, "domain": {"acc": 0.82, "drop": 0.057499999999999996}}, "iid_acc": 0.8775, "selectivity": 0.355} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8840304182509505, "drop": 0.014719581749049526, "rotation": {"subspace_principal_angle": 1.3732084918619123, "mean_class_cosine": 0.19630467274275237}}, "domain": {"acc": 0.655, "drop": 0.24375000000000002, "rotation": {"subspace_principal_angle": 1.4452895556036265, "mean_class_cosine": 0.1251775345482053}}, "length": {"acc": 0.8995433789954338, "drop": -0.0007933789954337467, "rotation": {"subspace_principal_angle": 1.335992218473073, "mean_class_cosine": 0.2326524735797551}}}, "iid_acc": 0.89875, "selectivity": 0.39875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3151033264444465, "mean_class_cosine": 0.2529159390133714}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.8365019011406845, "drop": 0.032248098859315566, "rotation": {"subspace_principal_angle": 1.0533503571330138, "mean_class_cosine": 0.9658970994045344}}, "domain": {"acc": 0.5575, "drop": 0.31125, "rotation": {"subspace_principal_angle": 1.3501859541358534, "mean_class_cosine": 0.2666305404277387}}, "length": {"acc": 0.8751902587519026, "drop": -0.006440258751902572, "rotation": {"subspace_principal_angle": 1.4766030164776944, "mean_class_cosine": 0.9618498962901108}}}, "iid_acc": 0.86875, "selectivity": 0.36875, "iid_split_rotation": {"subspace_principal_angle": 0.8666123632485124, "mean_class_cosine": 0.9567597495837794}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.8954372623574145, "drop": 0.023312737642585457}, "domain": {"acc": 0.83125, "drop": 0.08749999999999991}, "length": {"acc": 0.923896499238965, "drop": -0.005146499238965041}}, "iid_acc": 0.91875, "selectivity": 0.41874999999999996} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8829953198127926, "drop": -0.0004953198127926051, "rotation": {"subspace_principal_angle": 1.300584248001103, "mean_class_cosine": 0.33037107823356726}}}, "iid_acc": 0.8825, "selectivity": 0.5887499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3544227368327597, "mean_class_cosine": 0.32442597332270096}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7410296411856474, "drop": -0.013529641185647368, "rotation": {"subspace_principal_angle": 1.0031403826052068, "mean_class_cosine": 0.9483063492551265}}}, "iid_acc": 0.7275, "selectivity": 0.43375, "iid_split_rotation": {"subspace_principal_angle": 0.835980512632536, "mean_class_cosine": 0.9567132219152211}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8939157566302652, "drop": 0.00358424336973473}}, "iid_acc": 0.8975, "selectivity": 0.60375} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.97375, "drop": 0.0}, "paraphrase": {"acc": 0.9592326139088729, "drop": 0.014517386091127071, "rotation": {"subspace_principal_angle": 1.0461394619384272, "mean_class_cosine": 0.7060277954956268}}}, "iid_acc": 0.97375, "selectivity": 0.90375, "iid_split_rotation": {"subspace_principal_angle": 1.0572533041916146, "mean_class_cosine": 0.7322189822775439}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6546762589928058, "drop": 0.005323741007194238, "rotation": {"subspace_principal_angle": 1.3147848288233726, "mean_class_cosine": 0.8625758379093228}}}, "iid_acc": 0.66, "selectivity": 0.5900000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2749765898173022, "mean_class_cosine": 0.8458260421792687}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9616306954436451, "drop": 0.009619304556354846}}, "iid_acc": 0.97125, "selectivity": 0.9012499999999999} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.62875, "drop": 0.0}, "paraphrase": {"acc": 0.6065359477124183, "drop": 0.022214052287581776, "rotation": {"subspace_principal_angle": 1.4471225946917146, "mean_class_cosine": 0.12335870420709219}}}, "iid_acc": 0.62875, "selectivity": 0.14375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.396757975212198, "mean_class_cosine": 0.1731610965171537}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5202614379084968, "drop": 0.019738562091503264, "rotation": {"subspace_principal_angle": 1.017105332000735, "mean_class_cosine": 0.8150921314497699}}}, "iid_acc": 0.54, "selectivity": 0.05500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3938330496938112, "mean_class_cosine": -0.5401181987718008}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.63125, "drop": 0.0}, "paraphrase": {"acc": 0.6143790849673203, "drop": 0.01687091503267968}}, "iid_acc": 0.63125, "selectivity": 0.14625} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6266846361185984, "drop": 0.028315363881401656, "rotation": {"subspace_principal_angle": 1.4639056921866045, "mean_class_cosine": 0.2607364800995969}}}, "iid_acc": 0.655, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5392183501899661, "mean_class_cosine": 0.23109273289109802}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.1925, "drop": 0.0}, "paraphrase": {"acc": 0.18867924528301888, "drop": 0.003820754716981123, "rotation": {"subspace_principal_angle": 1.3025058812372758, "mean_class_cosine": 0.609760025746434}}}, "iid_acc": 0.1925, "selectivity": -0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.380643945283782, "mean_class_cosine": 0.7278668025302554}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6145552560646901, "drop": 0.03419474393530997}}, "iid_acc": 0.64875, "selectivity": 0.38250000000000006} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.775, "drop": 0.0}, "paraphrase": {"acc": 0.7696160267111853, "drop": 0.005383973288814681, "rotation": {"subspace_principal_angle": 1.4245590947156432, "mean_class_cosine": 0.145716567257158}}}, "iid_acc": 0.775, "selectivity": 0.28875, "iid_split_rotation": {"subspace_principal_angle": 1.471068816739365, "mean_class_cosine": 0.09956228431408685}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5425709515859767, "drop": -0.017570951585976657, "rotation": {"subspace_principal_angle": 1.4049876290028978, "mean_class_cosine": 0.8302182537519393}}}, "iid_acc": 0.525, "selectivity": 0.03875000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.628667819566516, "mean_class_cosine": 0.9886859497074707}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.7863105175292153, "drop": 0.001189482470784653}}, "iid_acc": 0.7875, "selectivity": 0.30124999999999996} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": 0.00708333333333333, "rotation": {"subspace_principal_angle": 1.4652126246929733, "mean_class_cosine": 0.10538763834081566}}}, "iid_acc": 0.67375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.453567266755897, "mean_class_cosine": 0.1169607381052836}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5412541254125413, "drop": 0.00874587458745879, "rotation": {"subspace_principal_angle": 1.435416177004221, "mean_class_cosine": -0.8400641400418787}}}, "iid_acc": 0.55, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.353882512217704, "mean_class_cosine": 0.9873913376016163}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.6897689768976898, "drop": 0.026481023102310264}}, "iid_acc": 0.71625, "selectivity": 0.22250000000000003} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7476489028213166, "drop": -0.02014890282131654, "rotation": {"subspace_principal_angle": 1.5031766373236868, "mean_class_cosine": 0.06756817028714371}}}, "iid_acc": 0.7275, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.4985421261579637, "mean_class_cosine": 0.07219134782882808}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5297805642633229, "drop": -0.016030564263322833, "rotation": {"subspace_principal_angle": 1.4269512459642844, "mean_class_cosine": 0.6191468418063922}}}, "iid_acc": 0.51375, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 0.6941036459644455, "mean_class_cosine": 0.6890861724335152}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7586206896551724, "drop": -0.008620689655172376}}, "iid_acc": 0.75, "selectivity": 0.23624999999999996} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9400278940027894, "drop": 0.01122210599721063, "rotation": {"subspace_principal_angle": 1.212406540346009, "mean_class_cosine": 0.3507667834923325}}}, "iid_acc": 0.95125, "selectivity": 0.41875000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.1491505155466, "mean_class_cosine": 0.40926267218064927}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.785, "drop": 0.0}, "paraphrase": {"acc": 0.796373779637378, "drop": -0.011373779637377956, "rotation": {"subspace_principal_angle": 0.5776320485859313, "mean_class_cosine": 0.9642980239049406}}}, "iid_acc": 0.785, "selectivity": 0.25250000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.8609996534230742, "mean_class_cosine": 0.9712015306482873}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9372384937238494, "drop": 0.014011506276150665}}, "iid_acc": 0.95125, "selectivity": 0.41875000000000007} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.9975, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.004046644844517222, "rotation": {"subspace_principal_angle": 0.8761164033837602, "mean_class_cosine": 0.6401395707229799}}}, "iid_acc": 0.9975, "selectivity": 0.1825000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8963642783199154, "mean_class_cosine": 0.6244538122618211}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.4075286415711948, "drop": 0.3112213584288052, "rotation": {"subspace_principal_angle": 1.3896444565022879, "mean_class_cosine": 0.8607515326377927}}}, "iid_acc": 0.71875, "selectivity": -0.09624999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1374176741267867, "mean_class_cosine": 0.9949908284390802}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9918166939443536, "drop": 0.004433306055646402}}, "iid_acc": 0.99625, "selectivity": 0.18125000000000002} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7112676056338029, "drop": 0.02498239436619709, "rotation": {"subspace_principal_angle": 1.512652494056396, "mean_class_cosine": 0.05811107708207547}}}, "iid_acc": 0.73625, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.5022867064091796, "mean_class_cosine": 0.06845604053333303}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5338028169014084, "drop": 0.03119718309859154, "rotation": {"subspace_principal_angle": 1.2680709408982833, "mean_class_cosine": -0.35755644118795693}}}, "iid_acc": 0.565, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.547048981725239, "mean_class_cosine": 0.09058745031393456}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.78125, "drop": 0.0}, "paraphrase": {"acc": 0.7507042253521127, "drop": 0.030545774647887325}}, "iid_acc": 0.78125, "selectivity": 0.19125000000000003} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5961538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5953757225433526, "drop": 0.000778123610493564, "rotation": {"subspace_principal_angle": 1.4398635945392528, "mean_class_cosine": 0.13929059880991962}}}, "iid_acc": 0.5961538461538461, "selectivity": 0.22115384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.4596191174582307, "mean_class_cosine": 0.18093919697525562}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5192307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.5491329479768786, "drop": -0.029902178746109342, "rotation": {"subspace_principal_angle": 1.0835579282901093, "mean_class_cosine": 0.6041012733975983}}}, "iid_acc": 0.5192307692307693, "selectivity": 0.14423076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.0998389268494577, "mean_class_cosine": 0.6208614359967286}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5721153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": -0.0059192974655403185}}, "iid_acc": 0.5721153846153846, "selectivity": 0.19711538461538458} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.9056865464632455, "drop": -0.0006865464632455076, "rotation": {"subspace_principal_angle": 1.3934785192501797, "mean_class_cosine": 0.17639007447692165}}}, "iid_acc": 0.905, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3193210248676257, "mean_class_cosine": 0.24883312785983533}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.57, "drop": 0.0}, "paraphrase": {"acc": 0.5533980582524272, "drop": 0.01660194174757279, "rotation": {"subspace_principal_angle": 1.5354546576355417, "mean_class_cosine": 0.1563657422847014}}}, "iid_acc": 0.57, "selectivity": -0.14250000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.445383158458667, "mean_class_cosine": 0.9556115381599443}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9084604715672677, "drop": 0.006539528432732311}}, "iid_acc": 0.915, "selectivity": 0.2025} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.776056338028169, "drop": 0.033943661971831074, "rotation": {"subspace_principal_angle": 1.3558532015708853, "mean_class_cosine": 0.2132918626421837}}, "domain": {"acc": 0.8, "drop": 0.010000000000000009, "rotation": {"subspace_principal_angle": 1.3213395199914533, "mean_class_cosine": 0.24687761614550957}}}, "iid_acc": 0.81, "selectivity": 0.30000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2949820598326707, "mean_class_cosine": 0.27233051788958473}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5309859154929577, "drop": -0.002235915492957652, "rotation": {"subspace_principal_angle": 0.949671245424435, "mean_class_cosine": 0.9981407854560767}}, "domain": {"acc": 0.4875, "drop": 0.041250000000000064, "rotation": {"subspace_principal_angle": 1.5209675895490495, "mean_class_cosine": 0.5347820929835523}}}, "iid_acc": 0.52875, "selectivity": 0.018750000000000044, "iid_split_rotation": {"subspace_principal_angle": 0.5977677192699846, "mean_class_cosine": 0.9995599104407848}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.825, "drop": 0.0}, "paraphrase": {"acc": 0.7295774647887324, "drop": 0.09542253521126753}, "domain": {"acc": 0.73125, "drop": 0.09375}}, "iid_acc": 0.825, "selectivity": 0.31499999999999995} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.825, "drop": 0.0}, "paraphrase": {"acc": 0.8022813688212928, "drop": 0.022718631178707205, "rotation": {"subspace_principal_angle": 1.3701838535641038, "mean_class_cosine": 0.19926955803809265}}, "domain": {"acc": 0.56, "drop": 0.2649999999999999, "rotation": {"subspace_principal_angle": 1.3671422264532063, "mean_class_cosine": 0.2022492580826582}}, "length": {"acc": 0.8356164383561644, "drop": -0.010616438356164437, "rotation": {"subspace_principal_angle": 1.3466907976398639, "mean_class_cosine": 0.22223434808814393}}}, "iid_acc": 0.825, "selectivity": 0.32499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3129495009920322, "mean_class_cosine": 0.25499915160174697}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.6406844106463878, "drop": -0.04568441064638784, "rotation": {"subspace_principal_angle": 1.5652743111758125, "mean_class_cosine": 0.9561201033378162}}, "domain": {"acc": 0.5575, "drop": 0.03749999999999998, "rotation": {"subspace_principal_angle": 1.307138927137268, "mean_class_cosine": 0.5860658652565125}}, "length": {"acc": 0.6164383561643836, "drop": -0.02143835616438361, "rotation": {"subspace_principal_angle": 0.9884986790030585, "mean_class_cosine": 0.9027214333318725}}}, "iid_acc": 0.595, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2061930031896504, "mean_class_cosine": 0.9278347514353975}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.844106463878327, "drop": 0.009643536121673013}, "domain": {"acc": 0.66875, "drop": 0.18500000000000005}, "length": {"acc": 0.8493150684931506, "drop": 0.004434931506849371}}, "iid_acc": 0.85375, "selectivity": 0.35375} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8642745709828393, "drop": 0.005725429017160666, "rotation": {"subspace_principal_angle": 1.3957840280918499, "mean_class_cosine": 0.27977067989217674}}}, "iid_acc": 0.87, "selectivity": 0.5974999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3533239187236508, "mean_class_cosine": 0.32998309013733307}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5460218408736349, "drop": -0.0035218408736349405, "rotation": {"subspace_principal_angle": 1.5216614620859656, "mean_class_cosine": 0.8734402521973395}}}, "iid_acc": 0.5425, "selectivity": 0.26999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.0674047358789585, "mean_class_cosine": 0.9542985146974496}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.890795631825273, "drop": -0.008295631825273042}}, "iid_acc": 0.8825, "selectivity": 0.6099999999999999} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9376498800959233, "drop": 0.023600119904076733, "rotation": {"subspace_principal_angle": 1.1479338257378555, "mean_class_cosine": 0.6070660513833379}}}, "iid_acc": 0.96125, "selectivity": 0.89375, "iid_split_rotation": {"subspace_principal_angle": 1.1171811389695125, "mean_class_cosine": 0.6559427626785893}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6139088729016786, "drop": 0.008591127098321438, "rotation": {"subspace_principal_angle": 1.2623715388420953, "mean_class_cosine": 0.8819688751478222}}}, "iid_acc": 0.6225, "selectivity": 0.555, "iid_split_rotation": {"subspace_principal_angle": 1.2562343632402415, "mean_class_cosine": 0.8445383158129988}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.95375, "drop": 0.0}, "paraphrase": {"acc": 0.9328537170263789, "drop": 0.020896282973621116}}, "iid_acc": 0.95375, "selectivity": 0.88625} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": 0.011323529411764732, "rotation": {"subspace_principal_angle": 1.4587999072142643, "mean_class_cosine": 0.11176243416412668}}}, "iid_acc": 0.5525, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4400549078317155, "mean_class_cosine": 0.13036926971343207}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.538562091503268, "drop": -0.02231209150326796, "rotation": {"subspace_principal_angle": 0.9958229238018077, "mean_class_cosine": 0.9168202372116476}}}, "iid_acc": 0.51625, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.4327159822922633, "mean_class_cosine": -0.5177790256388579}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.49673202614379086, "drop": 0.054517973856209156}}, "iid_acc": 0.55125, "selectivity": 0.026249999999999996} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5619946091644205, "drop": 0.039255390835579496, "rotation": {"subspace_principal_angle": 1.5685155167026745, "mean_class_cosine": 0.2634805112669454}}}, "iid_acc": 0.60125, "selectivity": 0.3212499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.498476982651726, "mean_class_cosine": 0.251518325093681}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.26125, "drop": 0.0}, "paraphrase": {"acc": 0.261455525606469, "drop": -0.00020552560646902585, "rotation": {"subspace_principal_angle": 1.3916019392156598, "mean_class_cosine": 0.6531664347702235}}}, "iid_acc": 0.26125, "selectivity": -0.018750000000000044, "iid_split_rotation": {"subspace_principal_angle": 1.3668566895557186, "mean_class_cosine": 0.6689391992705035}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5700808625336927, "drop": 0.016169137466307326}}, "iid_acc": 0.58625, "selectivity": 0.30625} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7195325542570952, "drop": -0.009532554257095205, "rotation": {"subspace_principal_angle": 1.4001021400462081, "mean_class_cosine": 0.16986648813286045}}}, "iid_acc": 0.71, "selectivity": 0.21249999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.397132197602315, "mean_class_cosine": 0.17279251519584937}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.5859766277128547, "drop": -0.009726627712854707, "rotation": {"subspace_principal_angle": 1.328491446497459, "mean_class_cosine": 0.9377244571178811}}}, "iid_acc": 0.57625, "selectivity": 0.07875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4036636361795571, "mean_class_cosine": 0.938290713035285}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7479131886477463, "drop": -0.00416318864774623}}, "iid_acc": 0.74375, "selectivity": 0.24625000000000002} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.6375, "drop": 0.0}, "paraphrase": {"acc": 0.6105610561056105, "drop": 0.026938943894389444, "rotation": {"subspace_principal_angle": 1.490662860294879, "mean_class_cosine": 0.08004773289195696}}}, "iid_acc": 0.6375, "selectivity": 0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.516170603318211, "mean_class_cosine": 0.054598560612693414}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5412541254125413, "drop": 0.0049958745874587596, "rotation": {"subspace_principal_angle": 1.018522615405077, "mean_class_cosine": 0.998295366669647}}}, "iid_acc": 0.54625, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.799728898586944, "mean_class_cosine": 0.9985173926775679}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.68875, "drop": 0.0}, "paraphrase": {"acc": 0.6831683168316832, "drop": 0.005581683168316753}}, "iid_acc": 0.68875, "selectivity": 0.19874999999999998} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.68125, "drop": 0.0}, "paraphrase": {"acc": 0.677115987460815, "drop": 0.004134012539184995, "rotation": {"subspace_principal_angle": 1.513447961089972, "mean_class_cosine": 0.05731693598787185}}}, "iid_acc": 0.68125, "selectivity": 0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5579856772962786, "mean_class_cosine": 0.012810299103026977}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5250783699059561, "drop": -0.02382836990595616, "rotation": {"subspace_principal_angle": 1.5103655352700989, "mean_class_cosine": 0.7751093816111925}}}, "iid_acc": 0.50125, "selectivity": -0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9404374749778482, "mean_class_cosine": 0.48400831659040194}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7570532915360502, "drop": -0.027053291536050184}}, "iid_acc": 0.73, "selectivity": 0.15749999999999997} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.91625, "drop": 0.0}, "paraphrase": {"acc": 0.9135285913528591, "drop": 0.002721408647140877, "rotation": {"subspace_principal_angle": 1.2733184253579535, "mean_class_cosine": 0.29310981653055773}}}, "iid_acc": 0.91625, "selectivity": 0.37250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1920176049103017, "mean_class_cosine": 0.36978603528128035}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.83, "drop": 0.0}, "paraphrase": {"acc": 0.8145048814504882, "drop": 0.0154951185495118, "rotation": {"subspace_principal_angle": 1.4437573872568787, "mean_class_cosine": 0.9676721148007581}}}, "iid_acc": 0.83, "selectivity": 0.28625, "iid_split_rotation": {"subspace_principal_angle": 1.4314363898953524, "mean_class_cosine": 0.9613334451017563}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.91625, "drop": 0.0}, "paraphrase": {"acc": 0.9135285913528591, "drop": 0.002721408647140877}}, "iid_acc": 0.91625, "selectivity": 0.37250000000000005} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.0015466448445171643, "rotation": {"subspace_principal_angle": 1.0288570623244413, "mean_class_cosine": 0.5157983478150161}}}, "iid_acc": 0.995, "selectivity": 0.22999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.069772811988795, "mean_class_cosine": 0.4803235060741538}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.734860883797054, "drop": -0.03611088379705407, "rotation": {"subspace_principal_angle": 1.4159558052810821, "mean_class_cosine": 0.998661624282861}}}, "iid_acc": 0.69875, "selectivity": -0.06625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.1246427235862093, "mean_class_cosine": 0.9995471474354687}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9918166939443536, "drop": -0.0018166939443535757}}, "iid_acc": 0.99, "selectivity": 0.22499999999999998} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6845070422535211, "drop": 0.009242957746478875, "rotation": {"subspace_principal_angle": 1.5353921160260495, "mean_class_cosine": 0.035396814949682484}}}, "iid_acc": 0.69375, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4561316678570049, "mean_class_cosine": 0.11441355588600202}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5295774647887324, "drop": -0.012077464788732395, "rotation": {"subspace_principal_angle": 1.1952282463550592, "mean_class_cosine": 0.34100773822159186}}}, "iid_acc": 0.5175, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.5260408088060127, "mean_class_cosine": 0.9004767260950717}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7535211267605634, "drop": -0.017271126760563416}}, "iid_acc": 0.73625, "selectivity": 0.09124999999999994} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.6127167630057804, "drop": -0.050216763005780374, "rotation": {"subspace_principal_angle": 1.4402506816259675, "mean_class_cosine": 0.1541271484023787}}}, "iid_acc": 0.5625, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.373294097014231, "mean_class_cosine": 0.19890553489034377}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.49710982658959535, "drop": -0.035571365051133785, "rotation": {"subspace_principal_angle": 1.1768674031248585, "mean_class_cosine": 0.7104337840281887}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.07692307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.1855767421726293, "mean_class_cosine": 0.8092177583847361}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5336538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5549132947976878, "drop": -0.02125944864384166}}, "iid_acc": 0.5336538461538461, "selectivity": 0.1490384615384615} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.9001386962552012, "drop": -0.01013869625520114, "rotation": {"subspace_principal_angle": 1.3883765772077856, "mean_class_cosine": 0.18140970234637815}}}, "iid_acc": 0.89, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3442390049803465, "mean_class_cosine": 0.22462415910132738}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5436893203883495, "drop": 0.018810679611650505, "rotation": {"subspace_principal_angle": 0.9241498843813925, "mean_class_cosine": -0.17909041195198278}}}, "iid_acc": 0.5625, "selectivity": -0.135, "iid_split_rotation": {"subspace_principal_angle": 0.7140290968434547, "mean_class_cosine": 0.9686729997957932}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8890429958391124, "drop": -0.010292995839112362}}, "iid_acc": 0.87875, "selectivity": 0.18125000000000002} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8325, "drop": 0.0}, "paraphrase": {"acc": 0.7746478873239436, "drop": 0.05785211267605639, "rotation": {"subspace_principal_angle": 1.2990940486645208, "mean_class_cosine": 0.26837165555553494}}, "domain": {"acc": 0.7975, "drop": 0.03500000000000003, "rotation": {"subspace_principal_angle": 1.3921776610825947, "mean_class_cosine": 0.17767038602968485}}}, "iid_acc": 0.8325, "selectivity": 0.32625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2834752018104276, "mean_class_cosine": 0.28338418654918335}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5309859154929577, "drop": -0.002235915492957652, "rotation": {"subspace_principal_angle": 1.0586485465355164, "mean_class_cosine": 0.9956582503304628}}, "domain": {"acc": 0.4875, "drop": 0.041250000000000064, "rotation": {"subspace_principal_angle": 1.361095756118237, "mean_class_cosine": 0.3175504588818687}}}, "iid_acc": 0.52875, "selectivity": 0.022500000000000075, "iid_split_rotation": {"subspace_principal_angle": 0.8302811510741681, "mean_class_cosine": 0.9988764885217147}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.819718309859155, "drop": 0.04653169014084502}, "domain": {"acc": 0.8425, "drop": 0.023749999999999938}}, "iid_acc": 0.86625, "selectivity": 0.36} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8829953198127926, "drop": -0.0029953198127925518, "rotation": {"subspace_principal_angle": 1.3716186408201203, "mean_class_cosine": 0.26184135314471124}}}, "iid_acc": 0.88, "selectivity": 0.59875, "iid_split_rotation": {"subspace_principal_angle": 1.3986787504321423, "mean_class_cosine": 0.27403222616028045}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.80625, "drop": 0.0}, "paraphrase": {"acc": 0.7862714508580343, "drop": 0.01997854914196573, "rotation": {"subspace_principal_angle": 1.1213682836223715, "mean_class_cosine": 0.9645304775406084}}}, "iid_acc": 0.80625, "selectivity": 0.525, "iid_split_rotation": {"subspace_principal_angle": 1.190218420873915, "mean_class_cosine": 0.9648000749883475}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8923556942277691, "drop": 0.00014430577223090157}}, "iid_acc": 0.8925, "selectivity": 0.61125} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5437908496732026, "drop": -0.002540849673202561, "rotation": {"subspace_principal_angle": 1.4224273033339936, "mean_class_cosine": 0.14782527199893394}}}, "iid_acc": 0.54125, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.4416538946189268, "mean_class_cosine": 0.1287837634804017}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.538562091503268, "drop": -0.021062091503267988, "rotation": {"subspace_principal_angle": 1.2749369531499732, "mean_class_cosine": 0.8687066652493326}}}, "iid_acc": 0.5175, "selectivity": 0.024999999999999967, "iid_split_rotation": {"subspace_principal_angle": 1.539011739741241, "mean_class_cosine": -0.5381342265082592}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": 0.016838235294117654}}, "iid_acc": 0.54625, "selectivity": 0.05375000000000002} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.5916442048517521, "drop": 0.023355795148247926, "rotation": {"subspace_principal_angle": 1.3790979385283946, "mean_class_cosine": 0.3195081442711161}}}, "iid_acc": 0.615, "selectivity": 0.36375, "iid_split_rotation": {"subspace_principal_angle": 1.4199621331697765, "mean_class_cosine": 0.29428965873721996}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.2875, "drop": 0.0}, "paraphrase": {"acc": 0.2803234501347709, "drop": 0.007176549865229087, "rotation": {"subspace_principal_angle": 1.1732397784339854, "mean_class_cosine": 0.6776335920045801}}}, "iid_acc": 0.2875, "selectivity": 0.036250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3510532914197495, "mean_class_cosine": 0.6935310313683204}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.63, "drop": 0.0}, "paraphrase": {"acc": 0.6226415094339622, "drop": 0.0073584905660377675}}, "iid_acc": 0.63, "selectivity": 0.37875000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.70375, "drop": 0.0}, "paraphrase": {"acc": 0.6828046744574291, "drop": 0.020945325542570892, "rotation": {"subspace_principal_angle": 1.4506727333643843, "mean_class_cosine": 0.1198349109994324}}}, "iid_acc": 0.70375, "selectivity": 0.20375, "iid_split_rotation": {"subspace_principal_angle": 1.4560638773991705, "mean_class_cosine": 0.1144809009150342}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.5442404006677797, "drop": -0.021740400667779713, "rotation": {"subspace_principal_angle": 1.3673823057780032, "mean_class_cosine": 0.9981711848367865}}}, "iid_acc": 0.5225, "selectivity": 0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 1.1930110372511171, "mean_class_cosine": 0.9981828476111254}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7362270450751253, "drop": 0.013772954924874736}}, "iid_acc": 0.75, "selectivity": 0.25} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.7084639498432602, "drop": -0.02096394984326022, "rotation": {"subspace_principal_angle": 1.5051962181158443, "mean_class_cosine": 0.06555306849859195}}}, "iid_acc": 0.6875, "selectivity": 0.14125, "iid_split_rotation": {"subspace_principal_angle": 1.5172834822024313, "mean_class_cosine": 0.053487308133283416}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5658307210031348, "drop": -0.010830721003134758, "rotation": {"subspace_principal_angle": 1.1030041242538011, "mean_class_cosine": 0.6806746541237996}}}, "iid_acc": 0.555, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.0411245457901148, "mean_class_cosine": 0.7191236806689514}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.7492163009404389, "drop": -0.0029663009404389262}}, "iid_acc": 0.74625, "selectivity": 0.19999999999999996} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9288702928870293, "drop": 0.012379707112970761, "rotation": {"subspace_principal_angle": 1.2301430406581242, "mean_class_cosine": 0.3341029094871022}}}, "iid_acc": 0.94125, "selectivity": 0.41500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.153997330636452, "mean_class_cosine": 0.4048355678035365}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7642956764295676, "drop": -0.005545676429567603, "rotation": {"subspace_principal_angle": 1.5114298942376507, "mean_class_cosine": 0.967424461326506}}}, "iid_acc": 0.75875, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9983998428171448, "mean_class_cosine": 0.9658650662106624}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.9428172942817294, "drop": 0.005932705718270537}}, "iid_acc": 0.94875, "selectivity": 0.4225} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.70375, "drop": 0.0}, "paraphrase": {"acc": 0.6732394366197183, "drop": 0.030510563380281686, "rotation": {"subspace_principal_angle": 1.5542729001951894, "mean_class_cosine": 0.01652267472899558}}}, "iid_acc": 0.70375, "selectivity": 0.13124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4874758177302438, "mean_class_cosine": 0.08322413642773308}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5464788732394367, "drop": 0.02602112676056334, "rotation": {"subspace_principal_angle": 1.5278946651673369, "mean_class_cosine": -0.5372610043917487}}}, "iid_acc": 0.5725, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.4388173212776232, "mean_class_cosine": 0.7758844286751576}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7183098591549296, "drop": 0.03419014084507033}}, "iid_acc": 0.7525, "selectivity": 0.17999999999999994} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5606936416184971, "drop": 0.0018063583815028927, "rotation": {"subspace_principal_angle": 1.4713914583801808, "mean_class_cosine": 0.1086073697472787}}}, "iid_acc": 0.5625, "selectivity": 0.20192307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.4025179262636047, "mean_class_cosine": 0.18370954160961964}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.47398843930635837, "drop": -0.012449977767896803, "rotation": {"subspace_principal_angle": 1.1101897330656618, "mean_class_cosine": 0.7275819529837535}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.10096153846153849, "iid_split_rotation": {"subspace_principal_angle": 1.5497284875073214, "mean_class_cosine": 0.7930942268672401}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.6009615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": 0.022926856380613536}}, "iid_acc": 0.6009615384615384, "selectivity": 0.24038461538461536} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8987517337031901, "drop": -1.733703190009983e-06, "rotation": {"subspace_principal_angle": 1.403522863323226, "mean_class_cosine": 0.16649449067720934}}}, "iid_acc": 0.89875, "selectivity": 0.21750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.420138412202417, "mean_class_cosine": 0.15008862700519784}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.57125, "drop": 0.0}, "paraphrase": {"acc": 0.5617198335644937, "drop": 0.009530166435506304, "rotation": {"subspace_principal_angle": 0.94423285674327, "mean_class_cosine": 0.24684962591703655}}}, "iid_acc": 0.57125, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5221499477378648, "mean_class_cosine": 0.9517198734637033}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9098474341192788, "drop": 0.012652565880721167}}, "iid_acc": 0.9225, "selectivity": 0.24124999999999996} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.85625, "drop": 0.0}, "paraphrase": {"acc": 0.8323943661971831, "drop": 0.023855633802816856, "rotation": {"subspace_principal_angle": 1.2594449210343017, "mean_class_cosine": 0.3063453465601831}}, "domain": {"acc": 0.8675, "drop": -0.011250000000000093, "rotation": {"subspace_principal_angle": 1.2223752440939522, "mean_class_cosine": 0.3414141888265658}}}, "iid_acc": 0.85625, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 1.1803322078545402, "mean_class_cosine": 0.38061764197303105}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5366197183098591, "drop": -0.004119718309859133, "rotation": {"subspace_principal_angle": 1.0843410916453302, "mean_class_cosine": 0.9995443885101234}}, "domain": {"acc": 0.4875, "drop": 0.044999999999999984, "rotation": {"subspace_principal_angle": 1.5626024602699073, "mean_class_cosine": 0.4830036010251207}}}, "iid_acc": 0.5325, "selectivity": 0.030000000000000027, "iid_split_rotation": {"subspace_principal_angle": 1.1107445344198659, "mean_class_cosine": 0.9998835461376956}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8028169014084507, "drop": 0.03468309859154928}, "domain": {"acc": 0.8225, "drop": 0.015000000000000013}}, "iid_acc": 0.8375, "selectivity": 0.3350000000000001} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.9125, "drop": 0.0}, "paraphrase": {"acc": 0.8954372623574145, "drop": 0.01706273764258548, "rotation": {"subspace_principal_angle": 1.0306140943276123, "mean_class_cosine": 0.5142922854845775}}, "domain": {"acc": 0.77875, "drop": 0.13374999999999992, "rotation": {"subspace_principal_angle": 1.2652458098581651, "mean_class_cosine": 0.30081823858041606}}, "length": {"acc": 0.9193302891933028, "drop": -0.006830289193302863, "rotation": {"subspace_principal_angle": 0.994782867288978, "mean_class_cosine": 0.5446849986678566}}}, "iid_acc": 0.9125, "selectivity": 0.4125, "iid_split_rotation": {"subspace_principal_angle": 1.0394992159651213, "mean_class_cosine": 0.5066520720072085}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.58125, "drop": 0.0}, "paraphrase": {"acc": 0.6254752851711026, "drop": -0.04422528517110258, "rotation": {"subspace_principal_angle": 0.9837366616608404, "mean_class_cosine": 0.9776968730440565}}, "domain": {"acc": 0.5575, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.3537717360413388, "mean_class_cosine": 0.5378898485059446}}, "length": {"acc": 0.6012176560121766, "drop": -0.019967656012176538, "rotation": {"subspace_principal_angle": 1.0664869055781216, "mean_class_cosine": 0.8881910220135161}}}, "iid_acc": 0.58125, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5132821504454912, "mean_class_cosine": 0.9660599604667153}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.91, "drop": 0.0}, "paraphrase": {"acc": 0.9049429657794676, "drop": 0.005057034220532386}, "domain": {"acc": 0.755, "drop": 0.15500000000000003}, "length": {"acc": 0.9178082191780822, "drop": -0.007808219178082165}}, "iid_acc": 0.91, "selectivity": 0.41000000000000003} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8783151326053042, "drop": 0.005434867394695786, "rotation": {"subspace_principal_angle": 1.4018526090844314, "mean_class_cosine": 0.21584227023789393}}}, "iid_acc": 0.88375, "selectivity": 0.6412500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3568370650579906, "mean_class_cosine": 0.277933277471287}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8798751950078003, "drop": 0.001374804992199663, "rotation": {"subspace_principal_angle": 1.5437428796709174, "mean_class_cosine": 0.9838278021789155}}}, "iid_acc": 0.88125, "selectivity": 0.6387499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5098184352195878, "mean_class_cosine": 0.9768614686576095}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8923556942277691, "drop": 0.007644305772230964}}, "iid_acc": 0.9, "selectivity": 0.6575} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.965, "drop": 0.0}, "paraphrase": {"acc": 0.9496402877697842, "drop": 0.015359712230215816, "rotation": {"subspace_principal_angle": 0.9402737529008706, "mean_class_cosine": 0.7722383597549551}}}, "iid_acc": 0.965, "selectivity": 0.9075, "iid_split_rotation": {"subspace_principal_angle": 0.8971810536010522, "mean_class_cosine": 0.8007298180881067}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.13875, "drop": 0.0}, "paraphrase": {"acc": 0.14628297362110312, "drop": -0.007532973621103112, "rotation": {"subspace_principal_angle": 1.1315401402990097, "mean_class_cosine": 0.6941035532694599}}}, "iid_acc": 0.13875, "selectivity": 0.08125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.385883009922417, "mean_class_cosine": 0.6014609571780535}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9136690647482014, "drop": 0.025080935251798553}}, "iid_acc": 0.93875, "selectivity": 0.88125} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.6300653594771242, "drop": 0.008684640522875875, "rotation": {"subspace_principal_angle": 1.2787080353422497, "mean_class_cosine": 0.28795269263212087}}}, "iid_acc": 0.63875, "selectivity": 0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.306568966622189, "mean_class_cosine": 0.2611635211442275}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": -0.028676470588235303, "rotation": {"subspace_principal_angle": 1.4646247494771925, "mean_class_cosine": 0.7483798246130362}}}, "iid_acc": 0.5125, "selectivity": -0.02750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5194053435641155, "mean_class_cosine": -0.9436278317396236}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5333333333333333, "drop": 0.0004166666666666208}}, "iid_acc": 0.53375, "selectivity": -0.006250000000000089} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.62, "drop": 0.0}, "paraphrase": {"acc": 0.5876010781671159, "drop": 0.032398921832884064, "rotation": {"subspace_principal_angle": 1.3946988954670643, "mean_class_cosine": 0.4588645470889185}}}, "iid_acc": 0.62, "selectivity": 0.31625, "iid_split_rotation": {"subspace_principal_angle": 1.4814328069291833, "mean_class_cosine": 0.3908719413079447}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.1825, "drop": 0.0}, "paraphrase": {"acc": 0.1765498652291105, "drop": 0.005950134770889487, "rotation": {"subspace_principal_angle": 1.428737416987744, "mean_class_cosine": 0.6603620114391797}}}, "iid_acc": 0.1825, "selectivity": -0.12125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3464001687915095, "mean_class_cosine": 0.917084029092293}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5175202156334232, "drop": 0.012479784366576818}}, "iid_acc": 0.53, "selectivity": 0.22625} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7612687813021702, "drop": -0.0037687813021702876, "rotation": {"subspace_principal_angle": 1.0841355564151438, "mean_class_cosine": 0.46767695778639146}}}, "iid_acc": 0.7575, "selectivity": 0.24624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1201130263537538, "mean_class_cosine": 0.4355807084230253}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6828046744574291, "drop": -0.012804674457429055, "rotation": {"subspace_principal_angle": 1.559246280818732, "mean_class_cosine": 0.8724262003165955}}}, "iid_acc": 0.67, "selectivity": 0.15875000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.208049879221657, "mean_class_cosine": 0.8740158426804394}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7395659432387313, "drop": 0.01418405676126877}}, "iid_acc": 0.75375, "selectivity": 0.24250000000000005} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6287128712871287, "drop": 0.05503712871287125, "rotation": {"subspace_principal_angle": 1.4304093704815346, "mean_class_cosine": 0.1399262745187435}}}, "iid_acc": 0.68375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4264870806603527, "mean_class_cosine": 0.14380889026388544}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5610561056105611, "drop": -0.008556105610561082, "rotation": {"subspace_principal_angle": 1.4505633244788272, "mean_class_cosine": 0.9993707117635033}}}, "iid_acc": 0.5525, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.562613516437802, "mean_class_cosine": 0.9992208351646927}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6171617161716172, "drop": 0.04283828382838284}}, "iid_acc": 0.66, "selectivity": 0.15625} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7429467084639498, "drop": -0.011696708463949879, "rotation": {"subspace_principal_angle": 1.4343600588327907, "mean_class_cosine": 0.1360133715328124}}}, "iid_acc": 0.73125, "selectivity": 0.12874999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3984660781436913, "mean_class_cosine": 0.17147854522739675}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5266457680250783, "drop": -0.017895768025078307, "rotation": {"subspace_principal_angle": 1.5522402276263103, "mean_class_cosine": 0.9681835964660888}}}, "iid_acc": 0.50875, "selectivity": -0.09375, "iid_split_rotation": {"subspace_principal_angle": 1.278589503289651, "mean_class_cosine": 0.36941996233066643}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7429467084639498, "drop": -0.009196708463949821}}, "iid_acc": 0.73375, "selectivity": 0.13124999999999998} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9163179916317992, "drop": 0.021182008368200833, "rotation": {"subspace_principal_angle": 1.190401875207584, "mean_class_cosine": 0.371286753786929}}}, "iid_acc": 0.9375, "selectivity": 0.39625, "iid_split_rotation": {"subspace_principal_angle": 1.0454137579769693, "mean_class_cosine": 0.5015440139447017}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5425383542538355, "drop": 0.008711645746164565, "rotation": {"subspace_principal_angle": 0.70122334453455, "mean_class_cosine": 0.9845649431106505}}}, "iid_acc": 0.55125, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 0.5014314350811042, "mean_class_cosine": 0.9308041543855714}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.945, "drop": 0.0}, "paraphrase": {"acc": 0.9456066945606695, "drop": -0.0006066945606695295}}, "iid_acc": 0.945, "selectivity": 0.40374999999999994} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9950900163666121, "drop": -9.001636661210011e-05, "rotation": {"subspace_principal_angle": 0.9545129430981294, "mean_class_cosine": 0.578006280576806}}}, "iid_acc": 0.995, "selectivity": 0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9634855001163742, "mean_class_cosine": 0.5706612157863881}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.723404255319149, "drop": -0.024654255319149, "rotation": {"subspace_principal_angle": 0.9334539669903851, "mean_class_cosine": 0.9999542616831976}}}, "iid_acc": 0.69875, "selectivity": -0.12875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4894565210879196, "mean_class_cosine": 0.9999767868513006}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.000296644844517191}}, "iid_acc": 0.99375, "selectivity": 0.16625} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7183098591549296, "drop": 0.025440140845070403, "rotation": {"subspace_principal_angle": 1.4257021861565011, "mean_class_cosine": 0.14458558178778447}}}, "iid_acc": 0.74375, "selectivity": 0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4515121208658293, "mean_class_cosine": 0.11900153014654812}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5183098591549296, "drop": 0.04419014084507045, "rotation": {"subspace_principal_angle": 1.5128863516688837, "mean_class_cosine": -0.06917035187866413}}}, "iid_acc": 0.5625, "selectivity": -0.03374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3373195084183762, "mean_class_cosine": 0.802090774256028}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.752112676056338, "drop": -0.01711267605633804}}, "iid_acc": 0.735, "selectivity": 0.13875000000000004} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.630057803468208, "drop": -0.005057803468208055, "rotation": {"subspace_principal_angle": 1.3578016617138275, "mean_class_cosine": 0.22714259169758785}}}, "iid_acc": 0.625, "selectivity": 0.24038461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.34596465147534, "mean_class_cosine": 0.24687761314743237}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.4519230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.45664739884393063, "drop": -0.004724321920853702, "rotation": {"subspace_principal_angle": 1.4589762703207936, "mean_class_cosine": 0.9288307050828664}}}, "iid_acc": 0.4519230769230769, "selectivity": 0.06730769230769229, "iid_split_rotation": {"subspace_principal_angle": 1.2273866113889293, "mean_class_cosine": 0.9582100809299986}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.5865384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.5838150289017341, "drop": 0.002723432636727474}}, "iid_acc": 0.5865384615384616, "selectivity": 0.20192307692307693} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8904299583911235, "drop": 0.0045700416088765294, "rotation": {"subspace_principal_angle": 1.3187059663512182, "mean_class_cosine": 0.24942879345042712}}}, "iid_acc": 0.895, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2441873711885174, "mean_class_cosine": 0.3208330998449743}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5325936199722607, "drop": 0.01740638002773931, "rotation": {"subspace_principal_angle": 1.2286707925235416, "mean_class_cosine": -0.644580232070691}}}, "iid_acc": 0.55, "selectivity": -0.20875, "iid_split_rotation": {"subspace_principal_angle": 0.7593618744036684, "mean_class_cosine": 0.9814176976119293}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8945908460471568, "drop": 0.004159153952843275}}, "iid_acc": 0.89875, "selectivity": 0.14} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7098150782361309, "drop": 0.04768492176386907, "rotation": {"subspace_principal_angle": 1.2799135817505813, "mean_class_cosine": 0.28679799861480304}}, "domain": {"acc": 0.73375, "drop": 0.023749999999999938, "rotation": {"subspace_principal_angle": 1.206450803740472, "mean_class_cosine": 0.3563378556067195}}}, "iid_acc": 0.7575, "selectivity": 0.2587499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2152898580705969, "mean_class_cosine": 0.34806520925621826}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.5889046941678521, "drop": -0.03265469416785205, "rotation": {"subspace_principal_angle": 1.51956074817224, "mean_class_cosine": 0.8695877050620235}}, "domain": {"acc": 0.54125, "drop": 0.015000000000000013, "rotation": {"subspace_principal_angle": 1.5651562278626985, "mean_class_cosine": 0.5820640074758123}}}, "iid_acc": 0.55625, "selectivity": 0.057499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.548228069141704, "mean_class_cosine": 0.8592588697201281}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.7475, "drop": 0.0}, "paraphrase": {"acc": 0.7226173541963016, "drop": 0.024882645803698478}, "domain": {"acc": 0.7575, "drop": -0.009999999999999898}}, "iid_acc": 0.7475, "selectivity": 0.24875000000000003} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.807843137254902, "drop": -0.005343137254901986, "rotation": {"subspace_principal_angle": 0.9349235883028191, "mean_class_cosine": 0.5938799054215346}}, "domain": {"acc": 0.66, "drop": 0.14249999999999996, "rotation": {"subspace_principal_angle": 1.2644579436182715, "mean_class_cosine": 0.30156951858956016}}, "length": {"acc": 0.802962962962963, "drop": -0.0004629629629629983, "rotation": {"subspace_principal_angle": 0.8439392303466156, "mean_class_cosine": 0.6645243339696073}}}, "iid_acc": 0.8025, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 0.8932244951414783, "mean_class_cosine": 0.626903094729581}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.7137254901960784, "drop": -0.017475490196078414, "rotation": {"subspace_principal_angle": 1.1712305596628225, "mean_class_cosine": 0.9467024508176543}}, "domain": {"acc": 0.57875, "drop": 0.11750000000000005, "rotation": {"subspace_principal_angle": 1.3012077657528451, "mean_class_cosine": 0.356404654810238}}, "length": {"acc": 0.7051851851851851, "drop": -0.008935185185185102, "rotation": {"subspace_principal_angle": 1.1652413786828095, "mean_class_cosine": 0.9420615872580638}}}, "iid_acc": 0.69625, "selectivity": 0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4630243777096839, "mean_class_cosine": 0.8983235883657518}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.78375, "drop": 0.0}, "paraphrase": {"acc": 0.7862745098039216, "drop": -0.002524509803921604}, "domain": {"acc": 0.61125, "drop": 0.1725}, "length": {"acc": 0.7881481481481482, "drop": -0.0043981481481482065}}, "iid_acc": 0.78375, "selectivity": 0.30999999999999994} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.8475, "drop": 0.0}, "paraphrase": {"acc": 0.8348484848484848, "drop": 0.012651515151515191, "rotation": {"subspace_principal_angle": 1.4031997490020873, "mean_class_cosine": 0.28396714753242047}}}, "iid_acc": 0.8475, "selectivity": 0.5975, "iid_split_rotation": {"subspace_principal_angle": 1.3806377410186017, "mean_class_cosine": 0.32568678636943665}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8181818181818182, "drop": 0.008068181818181808, "rotation": {"subspace_principal_angle": 1.4758819738130697, "mean_class_cosine": 0.9741084999680931}}}, "iid_acc": 0.82625, "selectivity": 0.57625, "iid_split_rotation": {"subspace_principal_angle": 1.1310624243298266, "mean_class_cosine": 0.9619284246782831}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.865, "drop": 0.0}, "paraphrase": {"acc": 0.8545454545454545, "drop": 0.010454545454545494}}, "iid_acc": 0.865, "selectivity": 0.615} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9343065693430657, "drop": 0.011943430656934373, "rotation": {"subspace_principal_angle": 1.1865093507054412, "mean_class_cosine": 0.7073522515216414}}}, "iid_acc": 0.94625, "selectivity": 0.865, "iid_split_rotation": {"subspace_principal_angle": 1.127510316714073, "mean_class_cosine": 0.7428054255764255}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5571776155717761, "drop": 0.00907238442822389, "rotation": {"subspace_principal_angle": 1.3920610393607629, "mean_class_cosine": 0.7826885130462339}}}, "iid_acc": 0.56625, "selectivity": 0.48500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3330395256421586, "mean_class_cosine": 0.8909496426331701}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8491484184914841, "drop": 0.022101581508515822}}, "iid_acc": 0.87125, "selectivity": 0.7899999999999999} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.4967490247074122, "drop": 0.027000975292587825, "rotation": {"subspace_principal_angle": 1.56369432512863, "mean_class_cosine": 0.0071019419641176074}}}, "iid_acc": 0.52375, "selectivity": -0.011249999999999982, "iid_split_rotation": {"subspace_principal_angle": 1.5297750902165248, "mean_class_cosine": 0.04100973285438006}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.48764629388816644, "drop": 0.014853706111833509, "rotation": {"subspace_principal_angle": 1.4431237989109693, "mean_class_cosine": -0.12706524234551198}}}, "iid_acc": 0.5025, "selectivity": -0.032500000000000084, "iid_split_rotation": {"subspace_principal_angle": 1.5637689056526396, "mean_class_cosine": -0.006306566848071626}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5006501950585176, "drop": 0.01684980494148236}}, "iid_acc": 0.5175, "selectivity": -0.01750000000000007} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5309973045822103, "drop": 0.0690026954177897, "rotation": {"subspace_principal_angle": 1.38978118980255, "mean_class_cosine": 0.3392687986531371}}}, "iid_acc": 0.6, "selectivity": 0.33875, "iid_split_rotation": {"subspace_principal_angle": 1.4386478796411275, "mean_class_cosine": 0.3036790886331675}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.1375, "drop": 0.0}, "paraphrase": {"acc": 0.1792452830188679, "drop": -0.0417452830188679, "rotation": {"subspace_principal_angle": 1.4906453725067592, "mean_class_cosine": 0.7942802395915041}}}, "iid_acc": 0.1375, "selectivity": -0.12374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4328823629930647, "mean_class_cosine": 0.6275373859348595}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.49326145552560646, "drop": 0.014238544474393489}}, "iid_acc": 0.5075, "selectivity": 0.24624999999999997} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7259136212624585, "drop": 0.0015863787375415273, "rotation": {"subspace_principal_angle": 1.3191750526863224, "mean_class_cosine": 0.24897450603487703}}}, "iid_acc": 0.7275, "selectivity": 0.21750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3632129148081515, "mean_class_cosine": 0.2060957956694738}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5647840531561462, "drop": -0.016034053156146255, "rotation": {"subspace_principal_angle": 1.2764535135475243, "mean_class_cosine": 0.9760866896549567}}}, "iid_acc": 0.54875, "selectivity": 0.03874999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4318977466650444, "mean_class_cosine": 0.980247865923659}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7325581395348837, "drop": 0.012441860465116306}}, "iid_acc": 0.745, "selectivity": 0.235} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.603648424543947, "drop": 0.047601575456053036, "rotation": {"subspace_principal_angle": 1.439031942391082, "mean_class_cosine": 0.13138343627378768}}}, "iid_acc": 0.65125, "selectivity": 0.12875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.281323973512102, "mean_class_cosine": 0.2854465710869707}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.615257048092869, "drop": -0.00025704809286897223, "rotation": {"subspace_principal_angle": 1.3313954355137612, "mean_class_cosine": 0.9237404825006307}}}, "iid_acc": 0.615, "selectivity": 0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.5707267568617127, "mean_class_cosine": 0.8603555241845086}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.6119402985074627, "drop": 0.05680970149253728}}, "iid_acc": 0.66875, "selectivity": 0.14625} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.7037643207855974, "drop": 0.013735679214402663, "rotation": {"subspace_principal_angle": 1.2968057041198295, "mean_class_cosine": 0.27057534861786225}}}, "iid_acc": 0.7175, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2964308527675759, "mean_class_cosine": 0.2709361985468884}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.485, "drop": 0.0}, "paraphrase": {"acc": 0.5073649754500819, "drop": -0.02236497545008187, "rotation": {"subspace_principal_angle": 1.0405608352231381, "mean_class_cosine": 0.7548288630584341}}}, "iid_acc": 0.485, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4829069517857016, "mean_class_cosine": 0.8524232187677423}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.6988543371522095, "drop": 0.03364566284779058}}, "iid_acc": 0.7325, "selectivity": 0.12} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8884353741496599, "drop": 0.017814625850340104, "rotation": {"subspace_principal_angle": 1.0443295050595705, "mean_class_cosine": 0.5024817409879795}}}, "iid_acc": 0.90625, "selectivity": 0.415, "iid_split_rotation": {"subspace_principal_angle": 1.0612122822978838, "mean_class_cosine": 0.4878141817807545}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.7877551020408163, "drop": 0.05724489795918364, "rotation": {"subspace_principal_angle": 1.3338632968935769, "mean_class_cosine": 0.8769564447689145}}}, "iid_acc": 0.845, "selectivity": 0.35374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3069371602368167, "mean_class_cosine": 0.9412077325237846}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.8789115646258503, "drop": 0.026088435374149732}}, "iid_acc": 0.905, "selectivity": 0.41375} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.0027821252059307966, "rotation": {"subspace_principal_angle": 0.6856933607020452, "mean_class_cosine": 0.7739801903730489}}}, "iid_acc": 0.99125, "selectivity": 0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6339194714608114, "mean_class_cosine": 0.8057121716043742}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8517298187808896, "drop": 0.007020181219110411, "rotation": {"subspace_principal_angle": 1.4167093648556626, "mean_class_cosine": 0.9633020957225444}}}, "iid_acc": 0.85875, "selectivity": -0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 1.5217003295605112, "mean_class_cosine": 0.9881251646589356}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.985, "drop": 0.0}, "paraphrase": {"acc": 0.9835255354200988, "drop": 0.0014744645799011913}}, "iid_acc": 0.985, "selectivity": 0.10375000000000001} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.65875, "drop": 0.0}, "paraphrase": {"acc": 0.6511299435028248, "drop": 0.007620056497175121, "rotation": {"subspace_principal_angle": 1.5394490513572545, "mean_class_cosine": 0.031342141781080114}}}, "iid_acc": 0.65875, "selectivity": 0.00374999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.529266372043682, "mean_class_cosine": 0.04151801773809659}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5268361581920904, "drop": -0.011836158192090385, "rotation": {"subspace_principal_angle": 1.5178073225930298, "mean_class_cosine": 0.21529478327501697}}}, "iid_acc": 0.515, "selectivity": -0.14, "iid_split_rotation": {"subspace_principal_angle": 1.0749998948016042, "mean_class_cosine": -0.653032824583841}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.7231638418079096, "drop": -0.0031638418079096287}}, "iid_acc": 0.72, "selectivity": 0.06499999999999995} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.6105769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": -0.004807692307692291, "rotation": {"subspace_principal_angle": 1.399178613094325, "mean_class_cosine": 0.18957379279214578}}}, "iid_acc": 0.6105769230769231, "selectivity": 0.13942307692307698, "iid_split_rotation": {"subspace_principal_angle": 1.4319729985001466, "mean_class_cosine": 0.17742658711691725}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.4951923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.5443786982248521, "drop": -0.049186390532544366, "rotation": {"subspace_principal_angle": 1.3438575364804408, "mean_class_cosine": 0.641655124351597}}}, "iid_acc": 0.4951923076923077, "selectivity": 0.024038461538461564, "iid_split_rotation": {"subspace_principal_angle": 1.2506869110680288, "mean_class_cosine": 0.4854077602261109}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": -0.033653846153846145}}, "iid_acc": 0.5817307692307693, "selectivity": 0.11057692307692313} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8706896551724138, "drop": 0.009310344827586192, "rotation": {"subspace_principal_angle": 1.2345926633445847, "mean_class_cosine": 0.3299056836099217}}}, "iid_acc": 0.88, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.219582897698587, "mean_class_cosine": 0.3440374169146968}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.735632183908046, "drop": 0.010617816091954002, "rotation": {"subspace_principal_angle": 1.5495300864536512, "mean_class_cosine": 0.7777397530079471}}}, "iid_acc": 0.74625, "selectivity": -0.028750000000000053, "iid_split_rotation": {"subspace_principal_angle": 1.4860573861271493, "mean_class_cosine": 0.7695588747682932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.85, "drop": 0.0}, "paraphrase": {"acc": 0.8520114942528736, "drop": -0.0020114942528736135}}, "iid_acc": 0.85, "selectivity": 0.07499999999999996} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.78, "drop": 0.0}, "paraphrase": {"acc": 0.732574679943101, "drop": 0.047425320056899034, "rotation": {"subspace_principal_angle": 1.2608414774243961, "mean_class_cosine": 0.30501563767234113}}, "domain": {"acc": 0.73, "drop": 0.050000000000000044, "rotation": {"subspace_principal_angle": 1.2628486007462063, "mean_class_cosine": 0.30310354600989187}}}, "iid_acc": 0.78, "selectivity": 0.275, "iid_split_rotation": {"subspace_principal_angle": 1.2752896290640992, "mean_class_cosine": 0.29122462225458356}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5419630156472262, "drop": -0.028213015647226114, "rotation": {"subspace_principal_angle": 1.2025920048727983, "mean_class_cosine": 0.9965846798687883}}, "domain": {"acc": 0.54125, "drop": -0.02749999999999997, "rotation": {"subspace_principal_angle": 1.4516540873608994, "mean_class_cosine": 0.4073452915520843}}}, "iid_acc": 0.51375, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 0.9220355133014713, "mean_class_cosine": 0.9960516328052383}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.7695590327169275, "drop": 0.04044096728307256}, "domain": {"acc": 0.81, "drop": 0.0}}, "iid_acc": 0.81, "selectivity": 0.30500000000000005} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8475, "drop": 0.0}, "paraphrase": {"acc": 0.8176470588235294, "drop": 0.029852941176470638, "rotation": {"subspace_principal_angle": 1.2030769237259111, "mean_class_cosine": 0.35948823052238327}}, "domain": {"acc": 0.52875, "drop": 0.31875, "rotation": {"subspace_principal_angle": 1.3485990887438482, "mean_class_cosine": 0.2203733735747998}}, "length": {"acc": 0.845925925925926, "drop": 0.001574074074074061, "rotation": {"subspace_principal_angle": 1.1802807331888672, "mean_class_cosine": 0.3806652417654258}}}, "iid_acc": 0.8475, "selectivity": 0.32125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1501816602046397, "mean_class_cosine": 0.40832162126076316}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.6941176470588235, "drop": 0.017132352941176543, "rotation": {"subspace_principal_angle": 1.2416590458376622, "mean_class_cosine": 0.9484259293470975}}, "domain": {"acc": 0.555, "drop": 0.15625, "rotation": {"subspace_principal_angle": 1.321419755161037, "mean_class_cosine": 0.2893218657278887}}, "length": {"acc": 0.72, "drop": -0.008749999999999925, "rotation": {"subspace_principal_angle": 1.35683655045995, "mean_class_cosine": 0.9567589393114461}}}, "iid_acc": 0.71125, "selectivity": 0.18500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5633547700573491, "mean_class_cosine": 0.9200623797757711}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8117647058823529, "drop": 0.014485294117647096}, "domain": {"acc": 0.58125, "drop": 0.245}, "length": {"acc": 0.8266666666666667, "drop": -0.0004166666666666208}}, "iid_acc": 0.82625, "selectivity": 0.30000000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.8575, "drop": 0.0}, "paraphrase": {"acc": 0.8515151515151516, "drop": 0.005984848484848482, "rotation": {"subspace_principal_angle": 1.4452099258783577, "mean_class_cosine": 0.24580852369730838}}}, "iid_acc": 0.8575, "selectivity": 0.6312500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3943169835208806, "mean_class_cosine": 0.29648597603430116}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.83625, "drop": 0.0}, "paraphrase": {"acc": 0.8181818181818182, "drop": 0.018068181818181817, "rotation": {"subspace_principal_angle": 1.4625849696271982, "mean_class_cosine": 0.976912859464512}}}, "iid_acc": 0.83625, "selectivity": 0.6100000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3260532260143207, "mean_class_cosine": 0.9681319283557925}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.8727272727272727, "drop": 0.01727272727272733}}, "iid_acc": 0.89, "selectivity": 0.6637500000000001} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9343065693430657, "drop": 0.020693430656934297, "rotation": {"subspace_principal_angle": 1.197650762223889, "mean_class_cosine": 0.684597725874047}}}, "iid_acc": 0.955, "selectivity": 0.8875, "iid_split_rotation": {"subspace_principal_angle": 1.55650137949423, "mean_class_cosine": 0.7066198619660665}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5328467153284672, "drop": -0.00909671532846712, "rotation": {"subspace_principal_angle": 1.3016766284172454, "mean_class_cosine": 0.7910526489249412}}}, "iid_acc": 0.52375, "selectivity": 0.45625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2176842386779647, "mean_class_cosine": 0.8907647870321528}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.8953771289537713, "drop": 0.024622871046228734}}, "iid_acc": 0.92, "selectivity": 0.8525} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5110533159947984, "drop": 0.025196684005201586, "rotation": {"subspace_principal_angle": 1.5229524083126806, "mean_class_cosine": 0.04782566779303716}}}, "iid_acc": 0.53625, "selectivity": 0.07124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.447992810822192, "mean_class_cosine": 0.12249508806050173}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.48875, "drop": 0.0}, "paraphrase": {"acc": 0.4967490247074122, "drop": -0.007999024707412206, "rotation": {"subspace_principal_angle": 1.0259677749536884, "mean_class_cosine": -0.6292438791105933}}}, "iid_acc": 0.48875, "selectivity": 0.023749999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.512376005309043, "mean_class_cosine": 0.4076926470016709}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.4759427828348505, "drop": 0.021557217165149523}}, "iid_acc": 0.4975, "selectivity": 0.03249999999999997} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.64, "drop": 0.0}, "paraphrase": {"acc": 0.6078167115902965, "drop": 0.032183288409703525, "rotation": {"subspace_principal_angle": 1.5433122888881867, "mean_class_cosine": 0.42874421872110796}}}, "iid_acc": 0.64, "selectivity": 0.34750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.5081587697047814, "mean_class_cosine": 0.4198126394888509}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.3575, "drop": 0.0}, "paraphrase": {"acc": 0.31266846361185985, "drop": 0.044831536388140136, "rotation": {"subspace_principal_angle": 1.530582416292941, "mean_class_cosine": 0.6444087690823997}}}, "iid_acc": 0.3575, "selectivity": 0.065, "iid_split_rotation": {"subspace_principal_angle": 1.459558734848993, "mean_class_cosine": 0.588656959531061}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.5053908355795148, "drop": -0.0003908355795148033}}, "iid_acc": 0.505, "selectivity": 0.21250000000000002} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.7408637873754153, "drop": -0.023363787375415224, "rotation": {"subspace_principal_angle": 1.3639308415465492, "mean_class_cosine": 0.20539322845895566}}}, "iid_acc": 0.7175, "selectivity": 0.20500000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.3710931416929857, "mean_class_cosine": 0.19837842368661718}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.6212624584717608, "drop": -0.006262458471760768, "rotation": {"subspace_principal_angle": 1.5406708151005613, "mean_class_cosine": 0.8985506716369838}}}, "iid_acc": 0.615, "selectivity": 0.10250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5633757536581958, "mean_class_cosine": 0.9281917050542527}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7558139534883721, "drop": -0.0020639534883720723}}, "iid_acc": 0.75375, "selectivity": 0.24125000000000008} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6086235489220564, "drop": 0.02637645107794362, "rotation": {"subspace_principal_angle": 1.490593398738083, "mean_class_cosine": 0.08011697135563739}}}, "iid_acc": 0.635, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 1.4638090458782234, "mean_class_cosine": 0.10678329666310886}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.593698175787728, "drop": -0.0036981757877280197, "rotation": {"subspace_principal_angle": 1.5381806288996867, "mean_class_cosine": 0.9646917936431184}}}, "iid_acc": 0.59, "selectivity": 0.07499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.47230491449444, "mean_class_cosine": 0.9335513939805022}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6650082918739635, "drop": 0.014991708126036563}}, "iid_acc": 0.68, "selectivity": 0.16500000000000004} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.6939443535188216, "drop": 0.02605564648117842, "rotation": {"subspace_principal_angle": 1.3786060400308688, "mean_class_cosine": 0.19100931112364286}}}, "iid_acc": 0.72, "selectivity": 0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.264300653737959, "mean_class_cosine": 0.3017194819812067}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.4828150572831424, "drop": 0.01718494271685761, "rotation": {"subspace_principal_angle": 1.5623310465881193, "mean_class_cosine": 0.5999071653626515}}}, "iid_acc": 0.5, "selectivity": -0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4335616759166658, "mean_class_cosine": 0.882837690979637}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7217675941080196, "drop": 0.014482405891980354}}, "iid_acc": 0.73625, "selectivity": 0.09375} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.92875, "drop": 0.0}, "paraphrase": {"acc": 0.9224489795918367, "drop": 0.0063010204081632715, "rotation": {"subspace_principal_angle": 1.0136784498875773, "mean_class_cosine": 0.5287421015976792}}}, "iid_acc": 0.92875, "selectivity": 0.46124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.0680431153367775, "mean_class_cosine": 0.4818398888533543}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.5931972789115646, "drop": 0.06680272108843543, "rotation": {"subspace_principal_angle": 1.4285742922049103, "mean_class_cosine": 0.8425489186964704}}}, "iid_acc": 0.66, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.0889790851620134, "mean_class_cosine": 0.7904713230276075}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.9414965986394558, "drop": -0.02024659863945577}}, "iid_acc": 0.92125, "selectivity": 0.45375} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9835255354200988, "drop": 0.008974464579901253, "rotation": {"subspace_principal_angle": 1.0095289992802334, "mean_class_cosine": 0.5322595207735084}}}, "iid_acc": 0.9925, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9575577798603117, "mean_class_cosine": 0.5755189198683394}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.68875, "drop": 0.0}, "paraphrase": {"acc": 0.5535420098846787, "drop": 0.13520799011532125, "rotation": {"subspace_principal_angle": 0.9140922175459604, "mean_class_cosine": 0.9973997191189731}}}, "iid_acc": 0.68875, "selectivity": -0.17000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2996635368722853, "mean_class_cosine": 0.9996246655982532}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.001929571663920937}}, "iid_acc": 0.98875, "selectivity": 0.13} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6511299435028248, "drop": 0.022620056497175134, "rotation": {"subspace_principal_angle": 1.4877740078390536, "mean_class_cosine": 0.08292697708829477}}}, "iid_acc": 0.67375, "selectivity": 0.039999999999999925, "iid_split_rotation": {"subspace_principal_angle": 1.4794201635002593, "mean_class_cosine": 0.09124905725263827}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.4971751412429379, "drop": 0.03032485875706209, "rotation": {"subspace_principal_angle": 1.4466641276425085, "mean_class_cosine": 0.1695136595431389}}}, "iid_acc": 0.5275, "selectivity": -0.10625000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.9893876475048883, "mean_class_cosine": -0.7941063384626064}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.730225988700565, "drop": -0.006475988700564983}}, "iid_acc": 0.72375, "selectivity": 0.08999999999999997} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.6346153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.6449704142011834, "drop": -0.010355029585798814, "rotation": {"subspace_principal_angle": 1.401130641487477, "mean_class_cosine": 0.20190956693568948}}}, "iid_acc": 0.6346153846153846, "selectivity": 0.14423076923076922, "iid_split_rotation": {"subspace_principal_angle": 1.5640320536219627, "mean_class_cosine": 0.22919831492919082}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.5480769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.5739644970414202, "drop": -0.025887573964497035, "rotation": {"subspace_principal_angle": 1.3084230282064335, "mean_class_cosine": 0.6381937638179063}}}, "iid_acc": 0.5480769230769231, "selectivity": 0.057692307692307765, "iid_split_rotation": {"subspace_principal_angle": 1.1829697589881627, "mean_class_cosine": 0.46077624553615687}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.5913461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.621301775147929, "drop": -0.029955621301775093}}, "iid_acc": 0.5913461538461539, "selectivity": 0.10096153846153849} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.896551724137931, "drop": 0.005948275862068919, "rotation": {"subspace_principal_angle": 1.1447057554079878, "mean_class_cosine": 0.41331408912500195}}}, "iid_acc": 0.9025, "selectivity": 0.11624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.074604861072015, "mean_class_cosine": 0.4760797664026806}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7775, "drop": 0.0}, "paraphrase": {"acc": 0.7873563218390804, "drop": -0.009856321839080473, "rotation": {"subspace_principal_angle": 1.4653362403173142, "mean_class_cosine": 0.5927810736166856}}}, "iid_acc": 0.7775, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.5560355491414277, "mean_class_cosine": 0.4882192393355668}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.81875, "drop": 0.0}, "paraphrase": {"acc": 0.8117816091954023, "drop": 0.006968390804597657}}, "iid_acc": 0.81875, "selectivity": 0.03249999999999997} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7937411095305832, "drop": 0.010008890469416776, "rotation": {"subspace_principal_angle": 1.3185889844917067, "mean_class_cosine": 0.2495420761739433}}, "domain": {"acc": 0.8825, "drop": -0.07874999999999999, "rotation": {"subspace_principal_angle": 1.3325721583114072, "mean_class_cosine": 0.23597731993808926}}}, "iid_acc": 0.80375, "selectivity": 0.29874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.269745108593276, "mean_class_cosine": 0.2965243102986965}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5391180654338549, "drop": -0.02661806543385492, "rotation": {"subspace_principal_angle": 0.6582429807151361, "mean_class_cosine": 0.99926299379126}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.5312218618656819, "mean_class_cosine": 0.3356468163658598}}}, "iid_acc": 0.5125, "selectivity": 0.007499999999999951, "iid_split_rotation": {"subspace_principal_angle": 0.8629499226938586, "mean_class_cosine": 0.9991531829071234}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.802275960170697, "drop": 0.008974039829303004}, "domain": {"acc": 0.84625, "drop": -0.03499999999999992}}, "iid_acc": 0.81125, "selectivity": 0.30625} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8666666666666667, "drop": 0.0008333333333333526, "rotation": {"subspace_principal_angle": 1.1792412886091128, "mean_class_cosine": 0.38162622377493116}}, "domain": {"acc": 0.54125, "drop": 0.32625000000000004, "rotation": {"subspace_principal_angle": 1.3149401585776963, "mean_class_cosine": 0.25307379864091606}}, "length": {"acc": 0.8725925925925926, "drop": -0.0050925925925925375, "rotation": {"subspace_principal_angle": 1.129874880545573, "mean_class_cosine": 0.42677296414992083}}}, "iid_acc": 0.8675, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.1059112988008373, "mean_class_cosine": 0.44832003402375775}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7941176470588235, "drop": 0.00963235294117648, "rotation": {"subspace_principal_angle": 1.2208647478151893, "mean_class_cosine": 0.9486786112928252}}, "domain": {"acc": 0.555, "drop": 0.24874999999999992, "rotation": {"subspace_principal_angle": 1.4628719512711352, "mean_class_cosine": 0.16922760359562494}}, "length": {"acc": 0.8103703703703704, "drop": -0.006620370370370443, "rotation": {"subspace_principal_angle": 1.4518778845432394, "mean_class_cosine": 0.9386719672471594}}}, "iid_acc": 0.80375, "selectivity": 0.31249999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.1012982204245179, "mean_class_cosine": 0.8734916313973369}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8666666666666667, "drop": 0.01083333333333325}, "domain": {"acc": 0.525, "drop": 0.3524999999999999}, "length": {"acc": 0.8740740740740741, "drop": 0.0034259259259258323}}, "iid_acc": 0.8775, "selectivity": 0.3862499999999999} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.8681818181818182, "drop": -0.0019318181818182012, "rotation": {"subspace_principal_angle": 1.348541899340855, "mean_class_cosine": 0.33742946528630835}}}, "iid_acc": 0.86625, "selectivity": 0.6087499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3243203014914424, "mean_class_cosine": 0.3890416789316658}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5772727272727273, "drop": 0.011477272727272725, "rotation": {"subspace_principal_angle": 1.1652446863372317, "mean_class_cosine": 0.943376790057927}}}, "iid_acc": 0.58875, "selectivity": 0.33125, "iid_split_rotation": {"subspace_principal_angle": 1.0212787219956532, "mean_class_cosine": 0.9231286978282006}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8727272727272727, "drop": 0.002272727272727315}}, "iid_acc": 0.875, "selectivity": 0.6174999999999999} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.97, "drop": 0.0}, "paraphrase": {"acc": 0.9586374695863747, "drop": 0.011362530413625227, "rotation": {"subspace_principal_angle": 1.0788515585794918, "mean_class_cosine": 0.680525963213341}}}, "iid_acc": 0.97, "selectivity": 0.91625, "iid_split_rotation": {"subspace_principal_angle": 1.0595584522895658, "mean_class_cosine": 0.7030153073160818}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.732360097323601, "drop": 0.0026399026763990197, "rotation": {"subspace_principal_angle": 1.5116341861027407, "mean_class_cosine": 0.873531191022212}}}, "iid_acc": 0.735, "selectivity": 0.68125, "iid_split_rotation": {"subspace_principal_angle": 1.4944238666561842, "mean_class_cosine": 0.9319808959810626}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9513381995133819, "drop": 0.004911800486618101}}, "iid_acc": 0.95625, "selectivity": 0.9025000000000001} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.56125, "drop": 0.0}, "paraphrase": {"acc": 0.5565669700910273, "drop": 0.004683029908972736, "rotation": {"subspace_principal_angle": 1.4553250675447744, "mean_class_cosine": 0.11521482211569181}}}, "iid_acc": 0.56125, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.4999753911814653, "mean_class_cosine": 0.0707617488193961}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5097529258777633, "drop": -0.01975292587776334, "rotation": {"subspace_principal_angle": 1.5663024506899672, "mean_class_cosine": 0.00763914984376122}}}, "iid_acc": 0.49, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.2427849178014092, "mean_class_cosine": 0.3406471787964678}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.48, "drop": 0.0}, "paraphrase": {"acc": 0.49154746423927176, "drop": -0.011547464239271776}}, "iid_acc": 0.48, "selectivity": -0.018750000000000044} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.610512129380054, "drop": 0.028237870619946093, "rotation": {"subspace_principal_angle": 1.430640605755257, "mean_class_cosine": 0.2631769683858712}}}, "iid_acc": 0.63875, "selectivity": 0.39625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.378148191691318, "mean_class_cosine": 0.297004025063675}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.12, "drop": 0.0}, "paraphrase": {"acc": 0.1280323450134771, "drop": -0.008032345013477105, "rotation": {"subspace_principal_angle": 1.543620776429251, "mean_class_cosine": 0.9288205176328912}}}, "iid_acc": 0.12, "selectivity": -0.1225, "iid_split_rotation": {"subspace_principal_angle": 1.43244607202274, "mean_class_cosine": 0.6582331054084928}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.6185983827493261, "drop": -0.0198483827493261}}, "iid_acc": 0.59875, "selectivity": 0.35625} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7657807308970099, "drop": -0.024530730897009967, "rotation": {"subspace_principal_angle": 1.3660163460568273, "mean_class_cosine": 0.20335174270412884}}}, "iid_acc": 0.74125, "selectivity": 0.18999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.413083036491105, "mean_class_cosine": 0.1570602898398632}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5475, "drop": 0.0}, "paraphrase": {"acc": 0.5647840531561462, "drop": -0.01728405315614623, "rotation": {"subspace_principal_angle": 0.8355530661327796, "mean_class_cosine": 0.9953186629963564}}}, "iid_acc": 0.5475, "selectivity": -0.003750000000000031, "iid_split_rotation": {"subspace_principal_angle": 1.4453486412673606, "mean_class_cosine": 0.996980243254297}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7774086378737541, "drop": -0.004908637873754174}}, "iid_acc": 0.7725, "selectivity": 0.22124999999999995} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6252072968490879, "drop": 0.044792703150912105, "rotation": {"subspace_principal_angle": 1.4695161231652571, "mean_class_cosine": 0.10110714243323128}}}, "iid_acc": 0.67, "selectivity": 0.17250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4419616097046257, "mean_class_cosine": 0.12847860473843392}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5174129353233831, "drop": 0.010087064676616908, "rotation": {"subspace_principal_angle": 0.9292634922449805, "mean_class_cosine": 0.9948800762828822}}}, "iid_acc": 0.5275, "selectivity": 0.02999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.044667492526286, "mean_class_cosine": 0.9982719754571383}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6882255389718076, "drop": 0.005524461028192373}}, "iid_acc": 0.69375, "selectivity": 0.19624999999999998} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.6890343698854338, "drop": 0.022215630114566287, "rotation": {"subspace_principal_angle": 1.44866296261917, "mean_class_cosine": 0.12182995563984489}}}, "iid_acc": 0.71125, "selectivity": 0.17875000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4353636321031356, "mean_class_cosine": 0.13501905614072562}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.45625, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.018381751227495913, "rotation": {"subspace_principal_angle": 1.4711476398340722, "mean_class_cosine": -0.12720149911565562}}}, "iid_acc": 0.45625, "selectivity": -0.07624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3619831712096353, "mean_class_cosine": 0.9877340756601919}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7495908346972177, "drop": 0.007909165302782295}}, "iid_acc": 0.7575, "selectivity": 0.22499999999999998} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9333333333333333, "drop": 0.009166666666666656, "rotation": {"subspace_principal_angle": 1.1452162894508524, "mean_class_cosine": 0.41284914883949264}}}, "iid_acc": 0.9425, "selectivity": 0.45375, "iid_split_rotation": {"subspace_principal_angle": 1.1544129671241614, "mean_class_cosine": 0.4044554792046704}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.82125, "drop": 0.0}, "paraphrase": {"acc": 0.8108843537414966, "drop": 0.010365646258503447, "rotation": {"subspace_principal_angle": 1.4778841101146312, "mean_class_cosine": 0.9112587174612317}}}, "iid_acc": 0.82125, "selectivity": 0.3325, "iid_split_rotation": {"subspace_principal_angle": 1.2685804950112414, "mean_class_cosine": 0.7130834799426322}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9374149659863945, "drop": 0.005085034013605494}}, "iid_acc": 0.9425, "selectivity": 0.45375} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9818780889621087, "drop": 0.011871911037891314, "rotation": {"subspace_principal_angle": 1.1022275614770123, "mean_class_cosine": 0.4516097785057144}}}, "iid_acc": 0.99375, "selectivity": 0.23875000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.0082315927465428, "mean_class_cosine": 0.5333574325360171}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9654036243822076, "drop": -0.00915362438220757, "rotation": {"subspace_principal_angle": 1.43644764939401, "mean_class_cosine": 0.9461248783860318}}}, "iid_acc": 0.95625, "selectivity": 0.20125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5120745206596544, "mean_class_cosine": 0.9885336177553897}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.008577018121911029}}, "iid_acc": 0.99375, "selectivity": 0.23875000000000002} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.693502824858757, "drop": 0.01774717514124302, "rotation": {"subspace_principal_angle": 1.4474830685331326, "mean_class_cosine": 0.12300097560505913}}}, "iid_acc": 0.71125, "selectivity": 0.09625000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3987832690834892, "mean_class_cosine": 0.1711660439449748}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5155367231638418, "drop": -0.015536723163841804, "rotation": {"subspace_principal_angle": 1.3401905016501925, "mean_class_cosine": 0.29339747507161273}}}, "iid_acc": 0.5, "selectivity": -0.11499999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6179518034185482, "mean_class_cosine": -0.9377606995696546}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.7259887005649718, "drop": -0.009738700564971725}}, "iid_acc": 0.71625, "selectivity": 0.10125000000000006} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.6586538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.6568047337278107, "drop": 0.0018491124260354708, "rotation": {"subspace_principal_angle": 1.4172417633253025, "mean_class_cosine": 0.17412301577924724}}}, "iid_acc": 0.6586538461538461, "selectivity": 0.21634615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.3955447682800144, "mean_class_cosine": 0.20505793596312397}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.49038461538461536, "drop": 0.0}, "paraphrase": {"acc": 0.4319526627218935, "drop": 0.05843195266272189, "rotation": {"subspace_principal_angle": 1.0501848723626588, "mean_class_cosine": 0.8199704533724792}}}, "iid_acc": 0.49038461538461536, "selectivity": 0.04807692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.1738154155015759, "mean_class_cosine": 0.567012811542814}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.6778846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.0625}}, "iid_acc": 0.6778846153846154, "selectivity": 0.23557692307692313} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.8951149425287356, "drop": 0.016135057471264425, "rotation": {"subspace_principal_angle": 1.3118705655633425, "mean_class_cosine": 0.2560422701778298}}}, "iid_acc": 0.91125, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3182833995213499, "mean_class_cosine": 0.24983798198888385}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5804597701149425, "drop": 0.02079022988505741, "rotation": {"subspace_principal_angle": 1.0152273026792804, "mean_class_cosine": 0.8731250404929112}}}, "iid_acc": 0.60125, "selectivity": -0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9386885075423576, "mean_class_cosine": 0.5900924520237005}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.9037356321839081, "drop": 0.00501436781609188}}, "iid_acc": 0.90875, "selectivity": 0.15999999999999992} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.8421052631578947, "drop": 0.026644736842105332, "rotation": {"subspace_principal_angle": 1.3051008608954664, "mean_class_cosine": 0.262580393680383}}, "domain": {"acc": 0.865, "drop": 0.003750000000000031, "rotation": {"subspace_principal_angle": 1.4287631781026644, "mean_class_cosine": 0.1415560812007921}}}, "iid_acc": 0.86875, "selectivity": 0.35750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2792444334922417, "mean_class_cosine": 0.28743897242092714}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.51125, "drop": 0.0}, "paraphrase": {"acc": 0.5391180654338549, "drop": -0.027868065433854894, "rotation": {"subspace_principal_angle": 1.5601359902088487, "mean_class_cosine": 0.9979053013441271}}, "domain": {"acc": 0.54125, "drop": -0.030000000000000027, "rotation": {"subspace_principal_angle": 1.4395920974393557, "mean_class_cosine": 0.23107881587582263}}}, "iid_acc": 0.51125, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.1884475439272368, "mean_class_cosine": 0.9977411853469901}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8492176386913229, "drop": 0.020782361308677055}, "domain": {"acc": 0.83375, "drop": 0.036250000000000004}}, "iid_acc": 0.87, "selectivity": 0.35875} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8647058823529412, "drop": 0.02904411764705883, "rotation": {"subspace_principal_angle": 1.3891107177208437, "mean_class_cosine": 0.1806876941601903}}, "domain": {"acc": 0.49, "drop": 0.40375000000000005, "rotation": {"subspace_principal_angle": 1.470272439761965, "mean_class_cosine": 0.10035467270286502}}, "length": {"acc": 0.8933333333333333, "drop": 0.0004166666666667318, "rotation": {"subspace_principal_angle": 1.336894896267935, "mean_class_cosine": 0.23177447065599882}}}, "iid_acc": 0.89375, "selectivity": 0.40125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.274478152252649, "mean_class_cosine": 0.2920008293919314}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8117647058823529, "drop": 0.0194852941176471, "rotation": {"subspace_principal_angle": 1.3046911642334313, "mean_class_cosine": 0.9572532608121695}}, "domain": {"acc": 0.48125, "drop": 0.35000000000000003, "rotation": {"subspace_principal_angle": 1.2118756513387043, "mean_class_cosine": -0.011293529250558751}}, "length": {"acc": 0.8355555555555556, "drop": -0.0043055555555555625, "rotation": {"subspace_principal_angle": 1.2439536531709152, "mean_class_cosine": 0.9646290213720963}}}, "iid_acc": 0.83125, "selectivity": 0.33875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5181831450402061, "mean_class_cosine": 0.9548748861745605}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8862745098039215, "drop": 0.007475490196078516}, "domain": {"acc": 0.7925, "drop": 0.10125000000000006}, "length": {"acc": 0.8962962962962963, "drop": -0.0025462962962962132}}, "iid_acc": 0.89375, "selectivity": 0.40125000000000005} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8818181818181818, "drop": 0.009431818181818152, "rotation": {"subspace_principal_angle": 1.4369187993598345, "mean_class_cosine": 0.22041085502998395}}}, "iid_acc": 0.89125, "selectivity": 0.675, "iid_split_rotation": {"subspace_principal_angle": 1.4078654977043972, "mean_class_cosine": 0.2586407992914274}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.7727272727272727, "drop": 0.037272727272727346, "rotation": {"subspace_principal_angle": 1.0175502832205285, "mean_class_cosine": 0.972353587476467}}}, "iid_acc": 0.81, "selectivity": 0.59375, "iid_split_rotation": {"subspace_principal_angle": 0.7420502510291406, "mean_class_cosine": 0.9634510604904396}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8833333333333333, "drop": 0.004166666666666652}}, "iid_acc": 0.8875, "selectivity": 0.6712499999999999} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9683698296836983, "drop": 0.007880170316301616, "rotation": {"subspace_principal_angle": 1.0992412190119936, "mean_class_cosine": 0.6867400981100495}}}, "iid_acc": 0.97625, "selectivity": 0.91875, "iid_split_rotation": {"subspace_principal_angle": 1.0034069622469282, "mean_class_cosine": 0.7160948548382569}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.80125, "drop": 0.0}, "paraphrase": {"acc": 0.8175182481751825, "drop": -0.016268248175182465, "rotation": {"subspace_principal_angle": 1.379334954656771, "mean_class_cosine": 0.8678086437576017}}}, "iid_acc": 0.80125, "selectivity": 0.74375, "iid_split_rotation": {"subspace_principal_angle": 0.8811167846778503, "mean_class_cosine": 0.9382525612852962}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9635036496350365, "drop": 0.01274635036496341}}, "iid_acc": 0.97625, "selectivity": 0.91875} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.5925, "drop": 0.0}, "paraphrase": {"acc": 0.599479843953186, "drop": -0.00697984395318596, "rotation": {"subspace_principal_angle": 1.4685167248622502, "mean_class_cosine": 0.10210136870418976}}}, "iid_acc": 0.5925, "selectivity": 0.09625, "iid_split_rotation": {"subspace_principal_angle": 1.492714330722181, "mean_class_cosine": 0.07800267856136348}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5097529258777633, "drop": -0.01975292587776334, "rotation": {"subspace_principal_angle": 1.3905538624434135, "mean_class_cosine": 0.43937650291700403}}}, "iid_acc": 0.49, "selectivity": -0.006250000000000033, "iid_split_rotation": {"subspace_principal_angle": 1.4581021828043095, "mean_class_cosine": -0.1678392152718658}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.577373211963589, "drop": -0.017373211963588986}}, "iid_acc": 0.56, "selectivity": 0.06375000000000003} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.65625, "drop": 0.0}, "paraphrase": {"acc": 0.6320754716981132, "drop": 0.024174528301886822, "rotation": {"subspace_principal_angle": 1.4488303719308924, "mean_class_cosine": 0.2608030814489897}}}, "iid_acc": 0.65625, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.4365260084834874, "mean_class_cosine": 0.27287586977099626}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.12, "drop": 0.0}, "paraphrase": {"acc": 0.12533692722371967, "drop": -0.005336927223719673, "rotation": {"subspace_principal_angle": 1.4252705910030379, "mean_class_cosine": 0.9453360722802503}}}, "iid_acc": 0.12, "selectivity": -0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.444135686833852, "mean_class_cosine": 0.6596132535425966}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6792452830188679, "drop": 0.0007547169811321641}}, "iid_acc": 0.68, "selectivity": 0.41375000000000006} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.760797342192691, "drop": -0.007047342192690986, "rotation": {"subspace_principal_angle": 1.4082028248717875, "mean_class_cosine": 0.16187804379501777}}}, "iid_acc": 0.75375, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.4629230727487192, "mean_class_cosine": 0.10766416205491924}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5813953488372093, "drop": -0.021395348837209283, "rotation": {"subspace_principal_angle": 0.7137539003362412, "mean_class_cosine": 0.9509360173967503}}}, "iid_acc": 0.56, "selectivity": 0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.0125517193691644, "mean_class_cosine": 0.9691372563411869}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.782392026578073, "drop": -0.008642026578073003}}, "iid_acc": 0.77375, "selectivity": 0.23875000000000002} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6451077943615257, "drop": 0.032392205638474336, "rotation": {"subspace_principal_angle": 1.4944467452146757, "mean_class_cosine": 0.0762754262877726}}}, "iid_acc": 0.6775, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.510090306300586, "mean_class_cosine": 0.06066874151454521}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5190713101160862, "drop": 0.005928689883913818, "rotation": {"subspace_principal_angle": 0.9851631097477104, "mean_class_cosine": 0.9870998388876457}}}, "iid_acc": 0.525, "selectivity": 0.01375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.8525820765093011, "mean_class_cosine": 0.9943272408460271}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.681592039800995, "drop": 0.028407960199004934}}, "iid_acc": 0.71, "selectivity": 0.19874999999999998} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7184942716857611, "drop": 0.01650572831423891, "rotation": {"subspace_principal_angle": 1.4867843877013787, "mean_class_cosine": 0.08391314783631819}}}, "iid_acc": 0.735, "selectivity": 0.18499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4727261303836074, "mean_class_cosine": 0.09791306933287178}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.4575, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.017131751227495884, "rotation": {"subspace_principal_angle": 1.4184867997055783, "mean_class_cosine": -0.12546411952087166}}}, "iid_acc": 0.4575, "selectivity": -0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0826072069356356, "mean_class_cosine": 0.9756432630642227}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7446808510638298, "drop": -0.004680851063829761}}, "iid_acc": 0.74, "selectivity": 0.18999999999999995} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9401360544217687, "drop": 0.01861394557823126, "rotation": {"subspace_principal_angle": 1.1843775125331726, "mean_class_cosine": 0.3768737131337715}}}, "iid_acc": 0.95875, "selectivity": 0.5037499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1630113814353649, "mean_class_cosine": 0.3965768790298291}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.7755102040816326, "drop": 0.026989795918367365, "rotation": {"subspace_principal_angle": 0.9611960505376331, "mean_class_cosine": 0.9267288005728522}}}, "iid_acc": 0.8025, "selectivity": 0.3475, "iid_split_rotation": {"subspace_principal_angle": 0.6069137112776986, "mean_class_cosine": 0.8431794446206526}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9523809523809523, "drop": 0.0038690476190477163}}, "iid_acc": 0.95625, "selectivity": 0.50125} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.005282125205930854, "rotation": {"subspace_principal_angle": 1.0067410745718517, "mean_class_cosine": 0.5346176539849523}}}, "iid_acc": 0.99375, "selectivity": 0.16125, "iid_split_rotation": {"subspace_principal_angle": 0.9448974604279329, "mean_class_cosine": 0.5858259839474748}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.66625, "drop": 0.0}, "paraphrase": {"acc": 0.40362438220757824, "drop": 0.2626256177924218, "rotation": {"subspace_principal_angle": 0.5567698167152252, "mean_class_cosine": 0.9809168363033169}}}, "iid_acc": 0.66625, "selectivity": -0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.4341855078247439, "mean_class_cosine": 0.9999348952126837}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.006077018121910971}}, "iid_acc": 0.99125, "selectivity": 0.15874999999999995} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.6878531073446328, "drop": 0.030896892655367214, "rotation": {"subspace_principal_angle": 1.4853872573002862, "mean_class_cosine": 0.08530526830833424}}}, "iid_acc": 0.71875, "selectivity": 0.11499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3987163989876108, "mean_class_cosine": 0.1712319268023629}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.5240112994350282, "drop": -0.019011299435028217, "rotation": {"subspace_principal_angle": 1.4845790668606034, "mean_class_cosine": 0.3310855041248773}}}, "iid_acc": 0.505, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.4077945456675773, "mean_class_cosine": -0.8399837311477907}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7372881355932204, "drop": 0.026461864406779667}}, "iid_acc": 0.76375, "selectivity": 0.16000000000000003} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6686390532544378, "drop": 0.018860946745562157, "rotation": {"subspace_principal_angle": 1.4431099310758673, "mean_class_cosine": 0.13799385145730425}}}, "iid_acc": 0.6875, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.4632019943344636, "mean_class_cosine": 0.12184824535087391}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.5913461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.5502958579881657, "drop": 0.04105029585798814, "rotation": {"subspace_principal_angle": 1.0140814812234873, "mean_class_cosine": 0.7438713263454885}}}, "iid_acc": 0.5913461538461539, "selectivity": 0.1634615384615385, "iid_split_rotation": {"subspace_principal_angle": 0.9988210743938787, "mean_class_cosine": 0.6629369773883759}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6745562130177515, "drop": 0.012943786982248517}}, "iid_acc": 0.6875, "selectivity": 0.25961538461538464} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8936781609195402, "drop": 0.013821839080459752, "rotation": {"subspace_principal_angle": 1.3954839068042866, "mean_class_cosine": 0.17441577734760833}}}, "iid_acc": 0.9075, "selectivity": 0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3326472336535684, "mean_class_cosine": 0.23590436416653315}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6135057471264368, "drop": 0.021494252873563213, "rotation": {"subspace_principal_angle": 0.9373099242351036, "mean_class_cosine": 0.8137785112536557}}}, "iid_acc": 0.635, "selectivity": -0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3179859570982269, "mean_class_cosine": 0.6358483449489087}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9051724137931034, "drop": 0.00982758620689661}}, "iid_acc": 0.915, "selectivity": 0.17500000000000004} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.795, "drop": 0.0}, "paraphrase": {"acc": 0.786628733997155, "drop": 0.00837126600284499, "rotation": {"subspace_principal_angle": 1.310108733461078, "mean_class_cosine": 0.2577449743747703}}, "domain": {"acc": 0.79, "drop": 0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.3864147968628893, "mean_class_cosine": 0.18333858173678275}}}, "iid_acc": 0.795, "selectivity": 0.28500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3577658636568546, "mean_class_cosine": 0.2114228247278906}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5362731152204836, "drop": -0.023773115220483643, "rotation": {"subspace_principal_angle": 1.4911474327942673, "mean_class_cosine": 0.9991258530703471}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.4889114257860925, "mean_class_cosine": 0.43982299953686566}}}, "iid_acc": 0.5125, "selectivity": 0.0024999999999999467, "iid_split_rotation": {"subspace_principal_angle": 1.5599854028146705, "mean_class_cosine": 0.9990812404452749}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.82, "drop": 0.0}, "paraphrase": {"acc": 0.7923186344238976, "drop": 0.027681365576102346}, "domain": {"acc": 0.83125, "drop": -0.011250000000000093}}, "iid_acc": 0.82, "selectivity": 0.30999999999999994} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8058823529411765, "drop": 0.03161764705882353, "rotation": {"subspace_principal_angle": 1.3708466120257015, "mean_class_cosine": 0.19862004766983304}}, "domain": {"acc": 0.68125, "drop": 0.15625, "rotation": {"subspace_principal_angle": 1.384027843480055, "mean_class_cosine": 0.1856845513313584}}, "length": {"acc": 0.8488888888888889, "drop": -0.011388888888888893, "rotation": {"subspace_principal_angle": 1.3282357766524422, "mean_class_cosine": 0.24018900406085658}}}, "iid_acc": 0.8375, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 1.3949678570258965, "mean_class_cosine": 0.17492389391787147}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.77625, "drop": 0.0}, "paraphrase": {"acc": 0.7823529411764706, "drop": -0.006102941176470589, "rotation": {"subspace_principal_angle": 0.8158694259835028, "mean_class_cosine": 0.9437326197394826}}, "domain": {"acc": 0.555, "drop": 0.22124999999999995, "rotation": {"subspace_principal_angle": 1.4680655032048706, "mean_class_cosine": 0.22680383044754415}}, "length": {"acc": 0.7866666666666666, "drop": -0.01041666666666663, "rotation": {"subspace_principal_angle": 1.075088915720483, "mean_class_cosine": 0.9383123548837031}}}, "iid_acc": 0.77625, "selectivity": 0.26749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3496789801104079, "mean_class_cosine": 0.92006570726007}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.8431372549019608, "drop": 0.008112745098039165}, "domain": {"acc": 0.55, "drop": 0.3012499999999999}, "length": {"acc": 0.8548148148148148, "drop": -0.003564814814814854}}, "iid_acc": 0.85125, "selectivity": 0.3424999999999999} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8363636363636363, "drop": 0.0023863636363636864, "rotation": {"subspace_principal_angle": 1.3907909651765882, "mean_class_cosine": 0.27160223854429794}}}, "iid_acc": 0.83875, "selectivity": 0.58375, "iid_split_rotation": {"subspace_principal_angle": 1.374984073519207, "mean_class_cosine": 0.3084279719861868}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5393939393939394, "drop": -0.008143939393939426, "rotation": {"subspace_principal_angle": 1.3095900045768096, "mean_class_cosine": 0.9315569392551215}}}, "iid_acc": 0.53125, "selectivity": 0.27625, "iid_split_rotation": {"subspace_principal_angle": 1.3512999965296404, "mean_class_cosine": 0.9302928191576891}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8712121212121212, "drop": 0.002537878787878811}}, "iid_acc": 0.87375, "selectivity": 0.61875} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9416058394160584, "drop": 0.017144160583941637, "rotation": {"subspace_principal_angle": 1.169967417171358, "mean_class_cosine": 0.6021107227776437}}}, "iid_acc": 0.95875, "selectivity": 0.88, "iid_split_rotation": {"subspace_principal_angle": 1.0879671022275426, "mean_class_cosine": 0.6463882048277961}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5717761557177615, "drop": -0.005526155717761494, "rotation": {"subspace_principal_angle": 1.5297708862576174, "mean_class_cosine": 0.8076540626245426}}}, "iid_acc": 0.56625, "selectivity": 0.48750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.31250485165038, "mean_class_cosine": 0.8967378658864777}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9245742092457421, "drop": 0.03417579075425792}}, "iid_acc": 0.95875, "selectivity": 0.88} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5370611183355006, "drop": 0.030438881664499373, "rotation": {"subspace_principal_angle": 1.4333280858176263, "mean_class_cosine": 0.13703568181783926}}}, "iid_acc": 0.5675, "selectivity": 0.08000000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5372359367014987, "mean_class_cosine": 0.03355409060481619}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5019505851755527, "drop": -0.0019505851755526882, "rotation": {"subspace_principal_angle": 1.5166639134271087, "mean_class_cosine": 0.7284040860032104}}}, "iid_acc": 0.5, "selectivity": 0.012500000000000011, "iid_split_rotation": {"subspace_principal_angle": 1.5345005286615787, "mean_class_cosine": -0.9244877025726085}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5136540962288687, "drop": -0.0011540962288687506}}, "iid_acc": 0.5125, "selectivity": 0.024999999999999967} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.6475, "drop": 0.0}, "paraphrase": {"acc": 0.578167115902965, "drop": 0.06933288409703497, "rotation": {"subspace_principal_angle": 1.4396284306360443, "mean_class_cosine": 0.2858703332393371}}}, "iid_acc": 0.6475, "selectivity": 0.39249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4241221420081531, "mean_class_cosine": 0.3414796507506154}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.2, "drop": 0.0}, "paraphrase": {"acc": 0.2169811320754717, "drop": -0.016981132075471694, "rotation": {"subspace_principal_angle": 1.478985421617415, "mean_class_cosine": 0.7260803181295756}}}, "iid_acc": 0.2, "selectivity": -0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3657973968851722, "mean_class_cosine": 0.6865447650543287}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.62375, "drop": 0.0}, "paraphrase": {"acc": 0.5956873315363881, "drop": 0.02806266846361194}}, "iid_acc": 0.62375, "selectivity": 0.36875} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7541528239202658, "drop": -0.004152823920265836, "rotation": {"subspace_principal_angle": 1.3055947110598365, "mean_class_cosine": 0.26210384065900993}}}, "iid_acc": 0.75, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4135255349018365, "mean_class_cosine": 0.15662326790731992}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6295681063122923, "drop": -0.007068106312292266, "rotation": {"subspace_principal_angle": 1.412398245668166, "mean_class_cosine": 0.9872471083520573}}}, "iid_acc": 0.6225, "selectivity": 0.06875000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.0308446538703597, "mean_class_cosine": 0.9830930319085469}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.7458471760797342, "drop": 0.0004028239202658046}}, "iid_acc": 0.74625, "selectivity": 0.1925} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.64625, "drop": 0.0}, "paraphrase": {"acc": 0.593698175787728, "drop": 0.052551824212272, "rotation": {"subspace_principal_angle": 1.4774953080617963, "mean_class_cosine": 0.09316571216572649}}}, "iid_acc": 0.64625, "selectivity": 0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4132962154086606, "mean_class_cosine": 0.15684975311512517}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5174129353233831, "drop": 0.010087064676616908, "rotation": {"subspace_principal_angle": 1.4727791381339745, "mean_class_cosine": 0.9962811296518983}}}, "iid_acc": 0.5275, "selectivity": 0.023749999999999938, "iid_split_rotation": {"subspace_principal_angle": 1.324949369047214, "mean_class_cosine": 0.9986827286127624}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.6633499170812603, "drop": 0.0379000829187397}}, "iid_acc": 0.70125, "selectivity": 0.1975} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6972176759410802, "drop": 0.007782324058919765, "rotation": {"subspace_principal_angle": 1.5034674219806008, "mean_class_cosine": 0.06727804731954892}}}, "iid_acc": 0.705, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4178710295321864, "mean_class_cosine": 0.1523299382782468}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.46125, "drop": 0.0}, "paraphrase": {"acc": 0.469721767594108, "drop": -0.008471767594108004, "rotation": {"subspace_principal_angle": 1.2701697435580983, "mean_class_cosine": -0.4177592757825601}}}, "iid_acc": 0.46125, "selectivity": -0.10749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4728680034154755, "mean_class_cosine": 0.9927211138662637}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.723404255319149, "drop": 0.0015957446808509967}}, "iid_acc": 0.725, "selectivity": 0.15625} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8993197278911564, "drop": 0.00693027210884356, "rotation": {"subspace_principal_angle": 1.2974026238207301, "mean_class_cosine": 0.2700006465670901}}}, "iid_acc": 0.90625, "selectivity": 0.44125, "iid_split_rotation": {"subspace_principal_angle": 1.2752165891310814, "mean_class_cosine": 0.29129449547307984}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8761904761904762, "drop": -0.0024404761904761596, "rotation": {"subspace_principal_angle": 1.0942486888882677, "mean_class_cosine": 0.942913260405956}}}, "iid_acc": 0.87375, "selectivity": 0.40875, "iid_split_rotation": {"subspace_principal_angle": 0.9061762984312886, "mean_class_cosine": 0.8552558967615458}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9306122448979591, "drop": -0.008112244897959142}}, "iid_acc": 0.9225, "selectivity": 0.45749999999999996} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.007327018121911055, "rotation": {"subspace_principal_angle": 1.072930156314306, "mean_class_cosine": 0.4775518369824139}}}, "iid_acc": 0.9925, "selectivity": 0.22375, "iid_split_rotation": {"subspace_principal_angle": 1.0267208542859751, "mean_class_cosine": 0.5176272803692485}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.7100494233937397, "drop": -0.022549423393739665, "rotation": {"subspace_principal_angle": 1.2433728178446086, "mean_class_cosine": 0.997937410347562}}}, "iid_acc": 0.6875, "selectivity": -0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4234682182286518, "mean_class_cosine": 0.9995278492579525}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.005679571663920968}}, "iid_acc": 0.9925, "selectivity": 0.22375} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.6675, "drop": 0.0}, "paraphrase": {"acc": 0.6652542372881356, "drop": 0.0022457627118643853, "rotation": {"subspace_principal_angle": 1.5101964181742233, "mean_class_cosine": 0.0605628247623603}}}, "iid_acc": 0.6675, "selectivity": 0.06499999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4116411659990913, "mean_class_cosine": 0.1584841015440742}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.495, "drop": 0.0}, "paraphrase": {"acc": 0.5112994350282486, "drop": -0.0162994350282486, "rotation": {"subspace_principal_angle": 1.2440118776947484, "mean_class_cosine": 0.4397585570740069}}}, "iid_acc": 0.495, "selectivity": -0.10750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.0189797005862058, "mean_class_cosine": -0.9017093672978498}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7090395480225988, "drop": 0.01721045197740112}}, "iid_acc": 0.72625, "selectivity": 0.12374999999999992} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.5976331360946746, "drop": 0.04178994082840226, "rotation": {"subspace_principal_angle": 1.3670769632091069, "mean_class_cosine": 0.221961109804074}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.13942307692307687, "iid_split_rotation": {"subspace_principal_angle": 1.3869516790493233, "mean_class_cosine": 0.2430729393469231}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.3269230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.35502958579881655, "drop": -0.028106508875739622, "rotation": {"subspace_principal_angle": 1.4355728624955333, "mean_class_cosine": 0.7267142033085312}}}, "iid_acc": 0.3269230769230769, "selectivity": -0.17307692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.1849716820736076, "mean_class_cosine": 0.7548318595334315}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.033653846153846145}}, "iid_acc": 0.6490384615384616, "selectivity": 0.14903846153846156} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8936781609195402, "drop": 0.007571839080459775, "rotation": {"subspace_principal_angle": 1.404231999409768, "mean_class_cosine": 0.1657952106350371}}}, "iid_acc": 0.90125, "selectivity": 0.19125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.415744792030871, "mean_class_cosine": 0.1544310160272101}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5919540229885057, "drop": 0.011795977011494263, "rotation": {"subspace_principal_angle": 0.829164366422975, "mean_class_cosine": 0.8319128806446898}}}, "iid_acc": 0.60375, "selectivity": -0.10624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1149004124569097, "mean_class_cosine": 0.34312396741171547}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9123563218390804, "drop": 0.00514367816091954}}, "iid_acc": 0.9175, "selectivity": 0.20750000000000002} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.7837837837837838, "drop": 0.04246621621621627, "rotation": {"subspace_principal_angle": 1.3737563564873747, "mean_class_cosine": 0.1957674385129709}}, "domain": {"acc": 0.80125, "drop": 0.025000000000000022, "rotation": {"subspace_principal_angle": 1.376738696045229, "mean_class_cosine": 0.19284193990423557}}}, "iid_acc": 0.82625, "selectivity": 0.32000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3152150432578507, "mean_class_cosine": 0.2528078527302202}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5362731152204836, "drop": -0.023773115220483643, "rotation": {"subspace_principal_angle": 0.8284726401777219, "mean_class_cosine": 0.9987588658556924}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.559015217957762, "mean_class_cosine": 0.37022421141777156}}}, "iid_acc": 0.5125, "selectivity": 0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.4590601880015512, "mean_class_cosine": 0.9985735288523132}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.84125, "drop": 0.0}, "paraphrase": {"acc": 0.7809388335704125, "drop": 0.06031116642958756}, "domain": {"acc": 0.84125, "drop": 0.0}}, "iid_acc": 0.84125, "selectivity": 0.3350000000000001} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8606060606060606, "drop": 0.019393939393939408, "rotation": {"subspace_principal_angle": 1.3448125803660131, "mean_class_cosine": 0.38575539047723395}}}, "iid_acc": 0.88, "selectivity": 0.67125, "iid_split_rotation": {"subspace_principal_angle": 1.2768731344200692, "mean_class_cosine": 0.4190057988176132}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.355, "drop": 0.0}, "paraphrase": {"acc": 0.3106060606060606, "drop": 0.044393939393939374, "rotation": {"subspace_principal_angle": 1.1355626344445542, "mean_class_cosine": 0.905045264153325}}}, "iid_acc": 0.355, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.5495618026124416, "mean_class_cosine": 0.9314215323578507}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8833333333333333, "drop": 0.0016666666666667052}}, "iid_acc": 0.885, "selectivity": 0.67625} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.56375, "drop": 0.0}, "paraphrase": {"acc": 0.540962288686606, "drop": 0.022787711313393966, "rotation": {"subspace_principal_angle": 1.4370366006567117, "mean_class_cosine": 0.13336121878297752}}}, "iid_acc": 0.56375, "selectivity": 0.039999999999999925, "iid_split_rotation": {"subspace_principal_angle": 1.49957415048444, "mean_class_cosine": 0.07116197799766466}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5019505851755527, "drop": -0.0019505851755526882, "rotation": {"subspace_principal_angle": 1.5354758515901374, "mean_class_cosine": 0.6588408805283696}}}, "iid_acc": 0.5, "selectivity": -0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1055065787936926, "mean_class_cosine": -0.9020263448542752}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.4837451235370611, "drop": 0.04625487646293891}}, "iid_acc": 0.53, "selectivity": 0.006249999999999978} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6226415094339622, "drop": 0.054858490566037754, "rotation": {"subspace_principal_angle": 1.4132260830128582, "mean_class_cosine": 0.315857679610795}}}, "iid_acc": 0.6775, "selectivity": 0.4175, "iid_split_rotation": {"subspace_principal_angle": 1.41912790604836, "mean_class_cosine": 0.32533634812236173}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.245, "drop": 0.0}, "paraphrase": {"acc": 0.2628032345013477, "drop": -0.017803234501347687, "rotation": {"subspace_principal_angle": 1.2435931646346163, "mean_class_cosine": 0.704084265739651}}}, "iid_acc": 0.245, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.2112395431952154, "mean_class_cosine": 0.6851956806446887}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.6185983827493261, "drop": 0.020151617250673937}}, "iid_acc": 0.63875, "selectivity": 0.37875000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7475083056478405, "drop": -0.01750830564784056, "rotation": {"subspace_principal_angle": 1.4592105747301425, "mean_class_cosine": 0.11135433008645393}}}, "iid_acc": 0.73, "selectivity": 0.23249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.5311242177574582, "mean_class_cosine": 0.03966170335806373}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5830564784053156, "drop": -0.018056478405315657, "rotation": {"subspace_principal_angle": 1.3688794358413638, "mean_class_cosine": 0.9651761136342984}}}, "iid_acc": 0.565, "selectivity": 0.06749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3920800768445374, "mean_class_cosine": 0.9821284088039223}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.7625, "drop": 0.0}, "paraphrase": {"acc": 0.7790697674418605, "drop": -0.01656976744186056}}, "iid_acc": 0.7625, "selectivity": 0.26499999999999996} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6677577741407529, "drop": 0.009742225859247111, "rotation": {"subspace_principal_angle": 1.5176621403016026, "mean_class_cosine": 0.05310918824653445}}}, "iid_acc": 0.6775, "selectivity": 0.11250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.457199891551373, "mean_class_cosine": 0.11335228191779778}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.46125, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.013381751227495908, "rotation": {"subspace_principal_angle": 1.5461043074662548, "mean_class_cosine": -0.13586943898113613}}}, "iid_acc": 0.46125, "selectivity": -0.10374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5005876536533236, "mean_class_cosine": 0.9790471285771307}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7364975450081833, "drop": 0.008502454991816677}}, "iid_acc": 0.745, "selectivity": 0.18000000000000005} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.9333333333333333, "drop": 0.0016666666666667052, "rotation": {"subspace_principal_angle": 1.2444931970939492, "mean_class_cosine": 0.32054342619861337}}}, "iid_acc": 0.935, "selectivity": 0.43750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2405200440265627, "mean_class_cosine": 0.32430439122172083}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.85, "drop": 0.0}, "paraphrase": {"acc": 0.8448979591836735, "drop": 0.005102040816326481, "rotation": {"subspace_principal_angle": 0.8606756929177377, "mean_class_cosine": 0.946865026478741}}}, "iid_acc": 0.85, "selectivity": 0.3525, "iid_split_rotation": {"subspace_principal_angle": 1.1743193451678584, "mean_class_cosine": 0.8825257604902363}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9442176870748299, "drop": 0.0020323129251701433}}, "iid_acc": 0.94625, "selectivity": 0.44875000000000004} +{"model": "gpt2-medium", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": 0.0011346787479407094, "rotation": {"subspace_principal_angle": 1.075164913899876, "mean_class_cosine": 0.4755871801312971}}}, "iid_acc": 0.99125, "selectivity": 0.20875, "iid_split_rotation": {"subspace_principal_angle": 0.9579353285296144, "mean_class_cosine": 0.575210123910989}} +{"model": "gpt2-medium", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6725, "drop": 0.0}, "paraphrase": {"acc": 0.6935749588138386, "drop": -0.021074958813838585, "rotation": {"subspace_principal_angle": 1.2610093649983862, "mean_class_cosine": 0.9991933388093968}}}, "iid_acc": 0.6725, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2635441752944665, "mean_class_cosine": 0.9997669069494608}} +{"model": "gpt2-medium", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.004429571663920884}}, "iid_acc": 0.99125, "selectivity": 0.20875} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.655367231638418, "drop": 0.03088276836158199, "rotation": {"subspace_principal_angle": 1.5026236950761325, "mean_class_cosine": 0.06811983851611561}}}, "iid_acc": 0.68625, "selectivity": 0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4122166058344536, "mean_class_cosine": 0.15791590818006312}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5098870056497176, "drop": -0.011137005649717535, "rotation": {"subspace_principal_angle": 1.2304337374701209, "mean_class_cosine": 0.42959183939934587}}}, "iid_acc": 0.49875, "selectivity": -0.08999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.504970426427495, "mean_class_cosine": -0.8759318900080222}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.7189265536723164, "drop": 0.0010734463276835804}}, "iid_acc": 0.72, "selectivity": 0.13124999999999998} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6449704142011834, "drop": -0.0007396449704141217, "rotation": {"subspace_principal_angle": 1.4713950529119475, "mean_class_cosine": 0.14817274811366463}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2211538461538462, "iid_split_rotation": {"subspace_principal_angle": 1.5076060847864945, "mean_class_cosine": 0.10760657740009509}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5048076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.4319526627218935, "drop": 0.07285502958579881, "rotation": {"subspace_principal_angle": 1.3335837272128281, "mean_class_cosine": 0.760631306748815}}}, "iid_acc": 0.5048076923076923, "selectivity": 0.08173076923076922, "iid_split_rotation": {"subspace_principal_angle": 1.2437180651878326, "mean_class_cosine": 0.6436823456850219}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6568047337278107, "drop": 0.030695266272189325}}, "iid_acc": 0.6875, "selectivity": 0.2644230769230769} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8836206896551724, "drop": 0.010129310344827669, "rotation": {"subspace_principal_angle": 1.3913906147486017, "mean_class_cosine": 0.17844485539062008}}}, "iid_acc": 0.89375, "selectivity": 0.1725000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4139643022148158, "mean_class_cosine": 0.15618990061442609}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.6120689655172413, "drop": 0.014181034482758648, "rotation": {"subspace_principal_angle": 0.7525157984620534, "mean_class_cosine": 0.7829464834163522}}}, "iid_acc": 0.62625, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6467108614757874, "mean_class_cosine": 0.4684027921683622}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9051724137931034, "drop": 0.01732758620689656}}, "iid_acc": 0.9225, "selectivity": 0.20125000000000004} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8264580369843528, "drop": 0.026041963015647207, "rotation": {"subspace_principal_angle": 1.205280976722454, "mean_class_cosine": 0.3574306478032766}}, "domain": {"acc": 0.85, "drop": 0.0025000000000000577, "rotation": {"subspace_principal_angle": 1.1833422855356077, "mean_class_cosine": 0.37783240518966293}}}, "iid_acc": 0.8525, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 1.1308070932804455, "mean_class_cosine": 0.4259297240456055}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5405405405405406, "drop": -0.025540540540540557, "rotation": {"subspace_principal_angle": 1.3060835605730696, "mean_class_cosine": 0.9997239873200385}}, "domain": {"acc": 0.54125, "drop": -0.026249999999999996, "rotation": {"subspace_principal_angle": 1.4432212340779689, "mean_class_cosine": 0.4463453977855254}}}, "iid_acc": 0.515, "selectivity": 0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.4900771909518655, "mean_class_cosine": 0.9997738232242801}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8122332859174964, "drop": 0.026516714082503556}, "domain": {"acc": 0.835, "drop": 0.003750000000000031}}, "iid_acc": 0.83875, "selectivity": 0.33999999999999997} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8333333333333334, "drop": 0.03416666666666668, "rotation": {"subspace_principal_angle": 1.3661248872725396, "mean_class_cosine": 0.20324546817811073}}, "domain": {"acc": 0.56875, "drop": 0.29875000000000007, "rotation": {"subspace_principal_angle": 1.424143642480157, "mean_class_cosine": 0.14612757252528197}}, "length": {"acc": 0.8725925925925926, "drop": -0.0050925925925925375, "rotation": {"subspace_principal_angle": 1.304086875035765, "mean_class_cosine": 0.26355866374052267}}}, "iid_acc": 0.8675, "selectivity": 0.34750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2174020916768917, "mean_class_cosine": 0.34608427795190094}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.7941176470588235, "drop": 0.017132352941176543, "rotation": {"subspace_principal_angle": 1.0543437038158343, "mean_class_cosine": 0.9698412280291595}}, "domain": {"acc": 0.5875, "drop": 0.22375, "rotation": {"subspace_principal_angle": 1.5050071211756375, "mean_class_cosine": 0.07636329311882721}}, "length": {"acc": 0.8177777777777778, "drop": -0.006527777777777799, "rotation": {"subspace_principal_angle": 1.0067642663923582, "mean_class_cosine": 0.9690199508303958}}}, "iid_acc": 0.81125, "selectivity": 0.29125, "iid_split_rotation": {"subspace_principal_angle": 1.3667918849667973, "mean_class_cosine": 0.9283366058714894}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8725490196078431, "drop": 0.004950980392156801}, "domain": {"acc": 0.53125, "drop": 0.34624999999999995}, "length": {"acc": 0.8785185185185185, "drop": -0.0010185185185185297}}, "iid_acc": 0.8775, "selectivity": 0.35749999999999993} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8803030303030303, "drop": 0.01344696969696979, "rotation": {"subspace_principal_angle": 1.3932379682070741, "mean_class_cosine": 0.2875718296573594}}}, "iid_acc": 0.89375, "selectivity": 0.6375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3409039719806688, "mean_class_cosine": 0.32901256562331466}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.5975, "drop": 0.0}, "paraphrase": {"acc": 0.5787878787878787, "drop": 0.01871212121212129, "rotation": {"subspace_principal_angle": 0.6350926877279145, "mean_class_cosine": 0.9549271920995692}}}, "iid_acc": 0.5975, "selectivity": 0.34125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.562730533540774, "mean_class_cosine": 0.945283431787009}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8712121212121212, "drop": 0.017537878787878824}}, "iid_acc": 0.88875, "selectivity": 0.6325000000000001} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.96625, "drop": 0.0}, "paraphrase": {"acc": 0.9586374695863747, "drop": 0.007612530413625307, "rotation": {"subspace_principal_angle": 1.1642922551102963, "mean_class_cosine": 0.6176664792575828}}}, "iid_acc": 0.96625, "selectivity": 0.89625, "iid_split_rotation": {"subspace_principal_angle": 1.0828409682713176, "mean_class_cosine": 0.6471064246781972}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8321167883211679, "drop": 0.006633211678832129, "rotation": {"subspace_principal_angle": 1.5226979113961525, "mean_class_cosine": 0.9230064437182347}}}, "iid_acc": 0.83875, "selectivity": 0.76875, "iid_split_rotation": {"subspace_principal_angle": 1.319128362171639, "mean_class_cosine": 0.9669627199357128}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.9625, "drop": 0.0}, "paraphrase": {"acc": 0.9464720194647201, "drop": 0.016027980535279873}}, "iid_acc": 0.9625, "selectivity": 0.8925000000000001} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.635890767230169, "drop": -0.009640767230169045, "rotation": {"subspace_principal_angle": 1.2265731114344571, "mean_class_cosine": 0.3374655623320725}}}, "iid_acc": 0.62625, "selectivity": 0.14125, "iid_split_rotation": {"subspace_principal_angle": 1.3292992619939612, "mean_class_cosine": 0.239156515406049}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.48875, "drop": 0.0}, "paraphrase": {"acc": 0.5084525357607282, "drop": -0.019702535760728224, "rotation": {"subspace_principal_angle": 1.5180213673447036, "mean_class_cosine": 0.7337434200066699}}}, "iid_acc": 0.48875, "selectivity": 0.003750000000000031, "iid_split_rotation": {"subspace_principal_angle": 0.9576309376243967, "mean_class_cosine": -0.9826509562242909}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5149544863459038, "drop": -0.002454486345903839}}, "iid_acc": 0.5125, "selectivity": 0.02749999999999997} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.62125, "drop": 0.0}, "paraphrase": {"acc": 0.5700808625336927, "drop": 0.051169137466307246, "rotation": {"subspace_principal_angle": 1.3744117706906038, "mean_class_cosine": 0.4523473567422777}}}, "iid_acc": 0.62125, "selectivity": 0.33499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3212695884395511, "mean_class_cosine": 0.4413545946757409}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.11875, "drop": 0.0}, "paraphrase": {"acc": 0.11455525606469003, "drop": 0.004194743935309969, "rotation": {"subspace_principal_angle": 1.3057209282680173, "mean_class_cosine": 0.9958332765626513}}}, "iid_acc": 0.11875, "selectivity": -0.1675, "iid_split_rotation": {"subspace_principal_angle": 1.5288056782240633, "mean_class_cosine": 0.6645736917680091}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5619946091644205, "drop": -0.006994609164420407}}, "iid_acc": 0.555, "selectivity": 0.26875000000000004} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7275747508305648, "drop": 0.007425249169435211, "rotation": {"subspace_principal_angle": 1.4274592750130581, "mean_class_cosine": 0.14284673359650063}}}, "iid_acc": 0.735, "selectivity": 0.21250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4743386316433527, "mean_class_cosine": 0.09630818958862122}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5631229235880398, "drop": -0.003122923588039783, "rotation": {"subspace_principal_angle": 1.2083317440135473, "mean_class_cosine": 0.995078207599845}}}, "iid_acc": 0.56, "selectivity": 0.03750000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.2327106774088032, "mean_class_cosine": 0.9974410369073173}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7574750830564784, "drop": -0.0012250830564783932}}, "iid_acc": 0.75625, "selectivity": 0.23375} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.648424543946932, "drop": 0.04407545605306795, "rotation": {"subspace_principal_angle": 1.4115718181974544, "mean_class_cosine": 0.15855257251441335}}}, "iid_acc": 0.6925, "selectivity": 0.20625, "iid_split_rotation": {"subspace_principal_angle": 1.3640799568494573, "mean_class_cosine": 0.20524729008148926}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5223880597014925, "drop": 0.001361940298507558, "rotation": {"subspace_principal_angle": 1.3460720235249697, "mean_class_cosine": 0.9994305298973165}}}, "iid_acc": 0.52375, "selectivity": 0.03750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.238382906553322, "mean_class_cosine": 0.999840093834897}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6401326699834162, "drop": -0.005132669983416216}}, "iid_acc": 0.635, "selectivity": 0.14875} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.6988543371522095, "drop": 0.018645662847790567, "rotation": {"subspace_principal_angle": 1.4345047521354637, "mean_class_cosine": 0.13587002144149035}}}, "iid_acc": 0.7175, "selectivity": 0.14875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.448693729965042, "mean_class_cosine": 0.12179941742267139}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.4675, "drop": 0.0}, "paraphrase": {"acc": 0.4779050736497545, "drop": -0.01040507364975446, "rotation": {"subspace_principal_angle": 1.4886165492494845, "mean_class_cosine": -0.5750441624077374}}}, "iid_acc": 0.4675, "selectivity": -0.10124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4864385355642995, "mean_class_cosine": 0.995645299950886}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7217675941080196, "drop": 0.003232405891980372}}, "iid_acc": 0.725, "selectivity": 0.15625} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9414965986394558, "drop": 0.0010034013605442205, "rotation": {"subspace_principal_angle": 1.0554320916574222, "mean_class_cosine": 0.49285180754683094}}}, "iid_acc": 0.9425, "selectivity": 0.50875, "iid_split_rotation": {"subspace_principal_angle": 1.0718759182874942, "mean_class_cosine": 0.4784778287309053}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5360544217687074, "drop": -0.0023044217687074875, "rotation": {"subspace_principal_angle": 0.6230061876035811, "mean_class_cosine": 0.8605846805322723}}}, "iid_acc": 0.53375, "selectivity": 0.09999999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.7433840996009756, "mean_class_cosine": 0.3607772763702933}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.9401360544217687, "drop": -0.007636054421768734}}, "iid_acc": 0.9325, "selectivity": 0.49874999999999997} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": 0.003634678747940767, "rotation": {"subspace_principal_angle": 1.0513045046339702, "mean_class_cosine": 0.49643906722894926}}}, "iid_acc": 0.99375, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9685923811624354, "mean_class_cosine": 0.5664600954062968}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.685337726523888, "drop": -0.01033772652388798, "rotation": {"subspace_principal_angle": 1.3240904740257078, "mean_class_cosine": 0.999937048647147}}}, "iid_acc": 0.675, "selectivity": -0.17374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5089394180470712, "mean_class_cosine": 0.9999798272733277}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.0015321252059308232}}, "iid_acc": 0.99, "selectivity": 0.14125} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7259887005649718, "drop": 0.005261299435028177, "rotation": {"subspace_principal_angle": 1.48146441422371, "mean_class_cosine": 0.08921314568865987}}}, "iid_acc": 0.73125, "selectivity": 0.1087499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4204367412204741, "mean_class_cosine": 0.1497936706206605}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5141242937853108, "drop": 0.018375706214689203, "rotation": {"subspace_principal_angle": 1.4237156435330802, "mean_class_cosine": -0.529348589575781}}}, "iid_acc": 0.5325, "selectivity": -0.09000000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.0016401557930925, "mean_class_cosine": -0.9687547350705455}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7245762711864406, "drop": -0.005826271186440635}}, "iid_acc": 0.71875, "selectivity": 0.09624999999999995} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6627218934911243, "drop": 0.010355029585798814, "rotation": {"subspace_principal_angle": 1.3926151506687268, "mean_class_cosine": 0.16905691896718814}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.21153846153846156, "iid_split_rotation": {"subspace_principal_angle": 1.455424707460169, "mean_class_cosine": 0.13141358197406708}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.5096153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.4378698224852071, "drop": 0.07174556213017746, "rotation": {"subspace_principal_angle": 1.2579697526624707, "mean_class_cosine": 0.8469173188143175}}}, "iid_acc": 0.5096153846153846, "selectivity": 0.04807692307692302, "iid_split_rotation": {"subspace_principal_angle": 1.545165675478419, "mean_class_cosine": 0.5995767861930784}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.024038461538461453}}, "iid_acc": 0.6394230769230769, "selectivity": 0.1778846153846153} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8922413793103449, "drop": 0.009008620689655134, "rotation": {"subspace_principal_angle": 1.2161263841678638, "mean_class_cosine": 0.34728086917165485}}}, "iid_acc": 0.90125, "selectivity": 0.11250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.242802856120801, "mean_class_cosine": 0.3221441156385047}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5975, "drop": 0.0}, "paraphrase": {"acc": 0.5747126436781609, "drop": 0.022787356321839147, "rotation": {"subspace_principal_angle": 1.0414849280010454, "mean_class_cosine": 0.8882768003569734}}}, "iid_acc": 0.5975, "selectivity": -0.19124999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.075377185201994, "mean_class_cosine": 0.00826524737963924}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8836206896551724, "drop": 0.01762931034482762}}, "iid_acc": 0.90125, "selectivity": 0.11250000000000004} +{"model": "pythia-6.9b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8545706371191135, "drop": 0.027929362880886432, "rotation": {"subspace_principal_angle": 1.2843658315181954, "mean_class_cosine": 0.2825299546199073}}, "domain": {"acc": 0.71875, "drop": 0.16374999999999995, "rotation": {"subspace_principal_angle": 1.4274597505649878, "mean_class_cosine": 0.14284626292142508}}}, "iid_acc": 0.8825, "selectivity": 0.3862499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1854350100069218, "mean_class_cosine": 0.3758939800584198}} +{"model": "pythia-6.9b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5318559556786704, "drop": -0.006855955678670367, "rotation": {"subspace_principal_angle": 1.3870121437737208, "mean_class_cosine": 0.9917248314020076}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.3286463453722377, "mean_class_cosine": 0.30726222746980725}}}, "iid_acc": 0.525, "selectivity": 0.028749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.514850907560883, "mean_class_cosine": 0.9745405250879291}} +{"model": "pythia-6.9b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8781163434903048, "drop": 0.024383656509695206}, "domain": {"acc": 0.75, "drop": 0.15249999999999997}}, "iid_acc": 0.9025, "selectivity": 0.40624999999999994} +{"model": "pythia-6.9b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8907692307692308, "drop": 0.0017307692307692024, "rotation": {"subspace_principal_angle": 1.3822125335258322, "mean_class_cosine": 0.280251720145784}}}, "iid_acc": 0.8925, "selectivity": 0.62625, "iid_split_rotation": {"subspace_principal_angle": 1.3453619680485411, "mean_class_cosine": 0.3176517400261344}} +{"model": "pythia-6.9b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.7475, "drop": 0.0}, "paraphrase": {"acc": 0.7569230769230769, "drop": -0.009423076923076867, "rotation": {"subspace_principal_angle": 1.0540699509625246, "mean_class_cosine": 0.9658592014821556}}}, "iid_acc": 0.7475, "selectivity": 0.48125000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.877740625219998, "mean_class_cosine": 0.9587274591133952}} +{"model": "pythia-6.9b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.9, "drop": 0.0024999999999999467}}, "iid_acc": 0.9025, "selectivity": 0.63625} +{"model": "pythia-6.9b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.62125, "drop": 0.0}, "paraphrase": {"acc": 0.6091503267973856, "drop": 0.012099673202614403, "rotation": {"subspace_principal_angle": 1.4323172982195795, "mean_class_cosine": 0.13803686358413447}}}, "iid_acc": 0.62125, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.473370357591302, "mean_class_cosine": 0.0972719173813966}} +{"model": "pythia-6.9b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5359477124183006, "drop": -0.007197712418300584, "rotation": {"subspace_principal_angle": 1.5296100396232466, "mean_class_cosine": 0.10858536033822105}}}, "iid_acc": 0.52875, "selectivity": 0.03375000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.097751036412192, "mean_class_cosine": -0.7830878155447331}} +{"model": "pythia-6.9b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5934640522875817, "drop": 0.01778594771241826}}, "iid_acc": 0.61125, "selectivity": 0.11624999999999996} +{"model": "pythia-6.9b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.6378162450066578, "drop": 0.04843375499334224, "rotation": {"subspace_principal_angle": 1.4209401948667373, "mean_class_cosine": 0.3239491922633613}}}, "iid_acc": 0.68625, "selectivity": 0.42000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.406430166854471, "mean_class_cosine": 0.2896893461708787}} +{"model": "pythia-6.9b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.18125, "drop": 0.0}, "paraphrase": {"acc": 0.1904127829560586, "drop": -0.009162782956058602, "rotation": {"subspace_principal_angle": 1.364681690199744, "mean_class_cosine": 0.3276963196758584}}}, "iid_acc": 0.18125, "selectivity": -0.08499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4358451906499528, "mean_class_cosine": 0.667868165289501}} +{"model": "pythia-6.9b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.6418109187749668, "drop": 0.021939081225033186}}, "iid_acc": 0.66375, "selectivity": 0.39749999999999996} +{"model": "pythia-6.9b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.95, "drop": 0.0}, "paraphrase": {"acc": 0.9382716049382716, "drop": 0.011728395061728403, "rotation": {"subspace_principal_angle": 1.224317317977924, "mean_class_cosine": 0.3395881657352101}}}, "iid_acc": 0.95, "selectivity": 0.42499999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.1632830218961225, "mean_class_cosine": 0.39632749804734413}} +{"model": "pythia-6.9b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.61875, "drop": 0.0}, "paraphrase": {"acc": 0.607681755829904, "drop": 0.011068244170095998, "rotation": {"subspace_principal_angle": 1.0087746185974722, "mean_class_cosine": 0.5517542959114724}}}, "iid_acc": 0.61875, "selectivity": 0.09375, "iid_split_rotation": {"subspace_principal_angle": 1.1932237166442734, "mean_class_cosine": 0.9672191891667108}} +{"model": "pythia-6.9b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.9525, "drop": 0.0}, "paraphrase": {"acc": 0.9396433470507545, "drop": 0.01285665294924554}}, "iid_acc": 0.9525, "selectivity": 0.4275} +{"model": "pythia-6.9b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.97875, "drop": 0.0}, "paraphrase": {"acc": 0.972568578553616, "drop": 0.006181421446384006, "rotation": {"subspace_principal_angle": 1.082750414325476, "mean_class_cosine": 0.6771447699248909}}}, "iid_acc": 0.97875, "selectivity": 0.8775, "iid_split_rotation": {"subspace_principal_angle": 1.0007322776397722, "mean_class_cosine": 0.716957258121963}} +{"model": "pythia-6.9b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6084788029925187, "drop": 0.04152119700748136, "rotation": {"subspace_principal_angle": 1.2715107394128569, "mean_class_cosine": 0.8465481480080757}}}, "iid_acc": 0.65, "selectivity": 0.5487500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5039960268389134, "mean_class_cosine": 0.8899973559179901}} +{"model": "pythia-6.9b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.9725, "drop": 0.0}, "paraphrase": {"acc": 0.9625935162094763, "drop": 0.009906483790523724}}, "iid_acc": 0.9725, "selectivity": 0.8712500000000001} +{"model": "pythia-6.9b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.7625, "drop": 0.0}, "paraphrase": {"acc": 0.7436332767402377, "drop": 0.01886672325976224, "rotation": {"subspace_principal_angle": 1.4931756045834128, "mean_class_cosine": 0.07754280185051747}}}, "iid_acc": 0.7625, "selectivity": 0.22999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.430776009151585, "mean_class_cosine": 0.13956323347084526}} +{"model": "pythia-6.9b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5314091680814941, "drop": -0.017659168081494037, "rotation": {"subspace_principal_angle": 1.2124127415949815, "mean_class_cosine": 0.990505627651225}}}, "iid_acc": 0.51375, "selectivity": -0.018749999999999933, "iid_split_rotation": {"subspace_principal_angle": 1.3996281155670207, "mean_class_cosine": 0.9911062384008227}} +{"model": "pythia-6.9b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7775891341256367, "drop": -0.006339134125636692}}, "iid_acc": 0.77125, "selectivity": 0.23875000000000002} diff --git a/repro_bundle/results_final/results/predictors.jsonl b/repro_bundle/results_final/results/predictors.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..c8125947df82fb3cc8384b6c73be57f887f7f9e1 --- /dev/null +++ b/repro_bundle/results_final/results/predictors.jsonl @@ -0,0 +1,497 @@ +{"model": "pythia-70m", "dataset": "sst2", "seed": 0, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921994833, "raptor_stability": 0.8375464081764221, "fragility": 0.25, "augmentation_robustness": 0.5702075197703796, "xie_feature_dispersion": 0.8227753483650888, "pac": 0.7038769639734008, "whitened_cosine_id": 0.8294597268104553}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 0, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998517, "raptor_stability": 0.9239972829818726, "fragility": 1.0, "augmentation_robustness": 0.8889069939502979, "xie_feature_dispersion": 1.0993059236404399, "pac": 0.9064521384660853, "whitened_cosine_id": 0.8396446704864502}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 0, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6360429269373853, "raptor_stability": 0.8923418521881104, "fragility": 1.5, "augmentation_robustness": 0.8610546015390961, "xie_feature_dispersion": 3.4534575785421744, "pac": 0.8766982268636032, "whitened_cosine_id": 0.9029156565666199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 0, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14189668585488155, "raptor_stability": 0.9457845091819763, "fragility": 1.5, "augmentation_robustness": 0.8719582541089957, "xie_feature_dispersion": 5.2630010355712145, "pac": 0.9088713816454861, "whitened_cosine_id": 0.9317190051078796}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 0, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974287, "raptor_stability": 0.7641952633857727, "fragility": 1.0, "augmentation_robustness": 0.6648445683071406, "xie_feature_dispersion": 0.4779918901319739, "pac": 0.7145199158464566, "whitened_cosine_id": 0.6775354743003845}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 0, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5953587445214507, "raptor_stability": 0.8714625239372253, "fragility": 0.5, "augmentation_robustness": 0.7563053805185295, "xie_feature_dispersion": 1.5333196894740788, "pac": 0.8138839522278773, "whitened_cosine_id": 0.8480013012886047}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 0, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994247, "raptor_stability": 0.8230767250061035, "fragility": 0.5, "augmentation_robustness": 0.6879123320435471, "xie_feature_dispersion": 1.779944270255231, "pac": 0.7554945285248253, "whitened_cosine_id": 0.8223811388015747}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 0, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.711632992199055, "raptor_stability": 0.8190855383872986, "fragility": 0.25, "augmentation_robustness": 0.6171877949935642, "xie_feature_dispersion": 2.512801472259043, "pac": 0.7181366666904314, "whitened_cosine_id": 0.7729045748710632}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989692, "raptor_stability": 0.8473518490791321, "fragility": 1.0, "augmentation_robustness": 0.7986513028646275, "xie_feature_dispersion": 1.163949469847951, "pac": 0.8230015759718798, "whitened_cosine_id": 0.7645858526229858}} +{"model": "pythia-70m", "dataset": "subj", "seed": 0, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002171, "raptor_stability": 0.9112874269485474, "fragility": 1.5, "augmentation_robustness": 0.808805287561379, "xie_feature_dispersion": 2.3756762827181857, "pac": 0.8600463572549633, "whitened_cosine_id": 0.9034857153892517}} +{"model": "pythia-70m", "dataset": "spam", "seed": 0, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004674, "raptor_stability": 0.9211491942405701, "fragility": 1.5, "augmentation_robustness": 0.7947530573404484, "xie_feature_dispersion": 10.408354824969607, "pac": 0.8579511257905093, "whitened_cosine_id": 0.9505216479301453}} +{"model": "pythia-70m", "dataset": "cola", "seed": 0, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921980065, "raptor_stability": 0.7842410206794739, "fragility": 1.0, "augmentation_robustness": 0.7059047928449348, "xie_feature_dispersion": 0.5820240387291664, "pac": 0.7450729067622044, "whitened_cosine_id": 0.7026472687721252}} +{"model": "pythia-70m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.7516991834926783, "raptor_stability": 0.8351565003395081, "fragility": 1.0, "augmentation_robustness": 0.783567124817086, "xie_feature_dispersion": 1.9650544929284912, "pac": 0.8093618125782971, "whitened_cosine_id": 0.8273500800132751}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 0, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998, "raptor_stability": 0.8865668177604675, "fragility": 1.5, "augmentation_robustness": 0.7726527290870765, "xie_feature_dispersion": 1.822712974246315, "pac": 0.829609773423772, "whitened_cosine_id": 0.8655645847320557}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350087, "raptor_stability": 0.8526354432106018, "fragility": 0.25, "augmentation_robustness": 0.5730135750873497, "xie_feature_dispersion": 1.4101369914642357, "pac": 0.7128245091489758, "whitened_cosine_id": 0.8668140769004822}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 0, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350252, "raptor_stability": 0.8803290128707886, "fragility": 1.0, "augmentation_robustness": 0.7816552796335788, "xie_feature_dispersion": 1.3391749548314755, "pac": 0.8309921462521836, "whitened_cosine_id": 0.8713250756263733}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3953680487548481, "raptor_stability": 0.8455236554145813, "fragility": 0.25, "augmentation_robustness": 0.7613131386356026, "xie_feature_dispersion": 0.9318287616318928, "pac": 0.8034183970250919, "whitened_cosine_id": 0.8953917026519775}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 0, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10457984207175601, "raptor_stability": 0.9457212686538696, "fragility": 1.5, "augmentation_robustness": 0.8741609981641726, "xie_feature_dispersion": 5.199477497190765, "pac": 0.909941133409021, "whitened_cosine_id": 0.9181342720985413}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 0, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342227, "raptor_stability": 0.7698441743850708, "fragility": 0.25, "augmentation_robustness": 0.6134415015890253, "xie_feature_dispersion": 0.25779876141354974, "pac": 0.6916428379870481, "whitened_cosine_id": 0.7203242182731628}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.574379658862361, "raptor_stability": 0.8597204685211182, "fragility": 0.25, "augmentation_robustness": 0.7115808236459255, "xie_feature_dispersion": 1.6192479136149953, "pac": 0.7856506460835218, "whitened_cosine_id": 0.8496574759483337}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348304, "raptor_stability": 0.8552483320236206, "fragility": 1.0, "augmentation_robustness": 0.735566226119687, "xie_feature_dispersion": 2.2631913339622987, "pac": 0.7954072790716538, "whitened_cosine_id": 0.8250994682312012}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 0, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934627, "raptor_stability": 0.8244509100914001, "fragility": 0.5, "augmentation_robustness": 0.6308774870695307, "xie_feature_dispersion": 1.9587240977086187, "pac": 0.7276641985804655, "whitened_cosine_id": 0.7910383939743042}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934696, "raptor_stability": 0.8365420699119568, "fragility": 1.0, "augmentation_robustness": 0.750756907821624, "xie_feature_dispersion": 1.5342572755652866, "pac": 0.7936494888667904, "whitened_cosine_id": 0.8055051565170288}} +{"model": "pythia-160m", "dataset": "subj", "seed": 0, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351844, "raptor_stability": 0.8689209222793579, "fragility": 1.0, "augmentation_robustness": 0.6997853402308795, "xie_feature_dispersion": 1.77134285145721, "pac": 0.7843531312551186, "whitened_cosine_id": 0.9039409160614014}} +{"model": "pythia-160m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352976, "raptor_stability": 0.943769633769989, "fragility": 3.0, "augmentation_robustness": 0.847956265457393, "xie_feature_dispersion": 6.963799146343127, "pac": 0.895862949613691, "whitened_cosine_id": 0.9455703496932983}} +{"model": "pythia-160m", "dataset": "cola", "seed": 0, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344465, "raptor_stability": 0.7988646626472473, "fragility": 0.25, "augmentation_robustness": 0.5594474756256655, "xie_feature_dispersion": 0.7758972276639047, "pac": 0.6791560691364564, "whitened_cosine_id": 0.7851360440254211}} +{"model": "pythia-160m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.4494122497402506, "raptor_stability": 0.8332012295722961, "fragility": 1.5, "augmentation_robustness": 0.7827926863582254, "xie_feature_dispersion": 2.5953837186456754, "pac": 0.8079969579652608, "whitened_cosine_id": 0.8253653049468994}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 0, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935011, "raptor_stability": 0.866767942905426, "fragility": 0.5, "augmentation_robustness": 0.7083306947816466, "xie_feature_dispersion": 1.5213101554825748, "pac": 0.7875493188435363, "whitened_cosine_id": 0.8781046867370605}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 0, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881098, "raptor_stability": 0.8615295886993408, "fragility": 0.25, "augmentation_robustness": 0.5615804178671275, "xie_feature_dispersion": 1.6884023301086555, "pac": 0.7115550032832342, "whitened_cosine_id": 0.8637628555297852}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881813, "raptor_stability": 0.8620725274085999, "fragility": 1.0, "augmentation_robustness": 0.7514519723873301, "xie_feature_dispersion": 1.923545301649778, "pac": 0.806762249897965, "whitened_cosine_id": 0.8767724633216858}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.32280286192962354, "raptor_stability": 0.8491001725196838, "fragility": 1.0, "augmentation_robustness": 0.7680864251666882, "xie_feature_dispersion": 6.130394237589204, "pac": 0.8085932988431861, "whitened_cosine_id": 0.8844512701034546}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.1116614706117857, "raptor_stability": 0.945551872253418, "fragility": 2.0, "augmentation_robustness": 0.8767712545738563, "xie_feature_dispersion": 7.687914979274747, "pac": 0.9111615634136372, "whitened_cosine_id": 0.8968205451965332}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 0, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879029, "raptor_stability": 0.787240743637085, "fragility": 0.25, "augmentation_robustness": 0.6335049509446696, "xie_feature_dispersion": 0.12206039095754939, "pac": 0.7103728472908772, "whitened_cosine_id": 0.7804538607597351}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 0, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.4882081459032719, "raptor_stability": 0.8632135391235352, "fragility": 0.25, "augmentation_robustness": 0.7301808505878356, "xie_feature_dispersion": 2.173352441480384, "pac": 0.7966971948556854, "whitened_cosine_id": 0.8471425175666809}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 0, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880458, "raptor_stability": 0.8126577138900757, "fragility": 0.25, "augmentation_robustness": 0.6377721250064905, "xie_feature_dispersion": 4.485569514407045, "pac": 0.7252149194482831, "whitened_cosine_id": 0.827558696269989}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 0, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879928, "raptor_stability": 0.7987161874771118, "fragility": 0.25, "augmentation_robustness": 0.59189319563508, "xie_feature_dispersion": 2.5329861784126293, "pac": 0.6953046915560959, "whitened_cosine_id": 0.810127854347229}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 0, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880347, "raptor_stability": 0.814401388168335, "fragility": 0.25, "augmentation_robustness": 0.6407410712281117, "xie_feature_dispersion": 0.7524282237703519, "pac": 0.7275712296982233, "whitened_cosine_id": 0.8191361427307129}} +{"model": "pythia-410m", "dataset": "subj", "seed": 0, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882566, "raptor_stability": 0.8862898945808411, "fragility": 0.5, "augmentation_robustness": 0.7428187785144447, "xie_feature_dispersion": 1.428972872571398, "pac": 0.8145543365476429, "whitened_cosine_id": 0.8959206342697144}} +{"model": "pythia-410m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882899, "raptor_stability": 0.9429970979690552, "fragility": 3.0, "augmentation_robustness": 0.8404945670527885, "xie_feature_dispersion": 6.72893787243834, "pac": 0.8917458325109218, "whitened_cosine_id": 0.933451235294342}} +{"model": "pythia-410m", "dataset": "cola", "seed": 0, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879473, "raptor_stability": 0.7842450737953186, "fragility": 0.25, "augmentation_robustness": 0.4686022827832364, "xie_feature_dispersion": 1.1149557488538653, "pac": 0.6264236782892775, "whitened_cosine_id": 0.8007057905197144}} +{"model": "pythia-410m", "dataset": "stance", "seed": 0, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.5795818589574081, "raptor_stability": 0.8285464644432068, "fragility": 1.0, "augmentation_robustness": 0.7555137866104972, "xie_feature_dispersion": 1.8361604596405292, "pac": 0.792030125526852, "whitened_cosine_id": 0.8221842646598816}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 0, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881513, "raptor_stability": 0.8521958589553833, "fragility": 0.5, "augmentation_robustness": 0.6802066641488503, "xie_feature_dispersion": 1.5630966219328584, "pac": 0.7662012615521168, "whitened_cosine_id": 0.8685176968574524}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.8599012494087219, "fragility": 0.25, "augmentation_robustness": 0.5241726726161993, "xie_feature_dispersion": 2.122662038410848, "pac": 0.6920369610124606, "whitened_cosine_id": 0.8232980370521545}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.825421154499054, "fragility": 1.0, "augmentation_robustness": 0.6705289577483847, "xie_feature_dispersion": 2.576726583434504, "pac": 0.7479750561237193, "whitened_cosine_id": 0.8465724587440491}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 0, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22041139949027433, "raptor_stability": 0.8484893441200256, "fragility": 1.0, "augmentation_robustness": 0.7715857915506817, "xie_feature_dispersion": 5.917035974988243, "pac": 0.8100375678353537, "whitened_cosine_id": 0.8215498328208923}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06228035557873579, "raptor_stability": 0.9404226541519165, "fragility": 3.0, "augmentation_robustness": 0.8674278851786935, "xie_feature_dispersion": 9.493553367505593, "pac": 0.9039252696653051, "whitened_cosine_id": 0.8300829529762268}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 0, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.7848396301269531, "fragility": 0.25, "augmentation_robustness": 0.6320998218394226, "xie_feature_dispersion": 0.2653617366201949, "pac": 0.7084697259831878, "whitened_cosine_id": 0.817557156085968}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.019767617442798265, "raptor_stability": 0.8671453595161438, "fragility": 0.25, "augmentation_robustness": 0.6900615345666754, "xie_feature_dispersion": 2.8877683311134636, "pac": 0.7786034470414096, "whitened_cosine_id": 0.830573558807373}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 0, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010617, "raptor_stability": 0.8155287504196167, "fragility": 0.25, "augmentation_robustness": 0.5947305676343537, "xie_feature_dispersion": 5.342310320027777, "pac": 0.7051296590269852, "whitened_cosine_id": 0.8247697353363037}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 0, "layer": 22, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.7958077192306519, "fragility": 0.25, "augmentation_robustness": 0.5722304723886765, "xie_feature_dispersion": 2.217429056075072, "pac": 0.6840190958096641, "whitened_cosine_id": 0.8194364309310913}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 0, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8089179396629333, "fragility": 0.5, "augmentation_robustness": 0.6530652465185355, "xie_feature_dispersion": 1.5482508686542642, "pac": 0.7309915930907345, "whitened_cosine_id": 0.8221121430397034}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 0, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010872, "raptor_stability": 0.8672937750816345, "fragility": 1.0, "augmentation_robustness": 0.694108081388879, "xie_feature_dispersion": 2.5476390335776844, "pac": 0.7807009282352568, "whitened_cosine_id": 0.8334078788757324}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 0, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010434, "raptor_stability": 0.9216615557670593, "fragility": 1.0, "augmentation_robustness": 0.8105170199587214, "xie_feature_dispersion": 23.432722674772293, "pac": 0.8660892878628903, "whitened_cosine_id": 0.8588989973068237}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 0, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010818, "raptor_stability": 0.8071193099021912, "fragility": 0.25, "augmentation_robustness": 0.4910409126593769, "xie_feature_dispersion": 1.549092757263006, "pac": 0.649080111280784, "whitened_cosine_id": 0.8179035186767578}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 0, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.36672114356076757, "raptor_stability": 0.8304018974304199, "fragility": 1.0, "augmentation_robustness": 0.7571886661476842, "xie_feature_dispersion": 3.593233328751099, "pac": 0.7937952817890521, "whitened_cosine_id": 0.8171507716178894}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8476687073707581, "fragility": 0.5, "augmentation_robustness": 0.6670456540942392, "xie_feature_dispersion": 2.045523026748139, "pac": 0.7573571807324986, "whitened_cosine_id": 0.8353421092033386}} +{"model": "gpt2", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349679, "raptor_stability": 0.8233915567398071, "fragility": 0.25, "augmentation_robustness": 0.4566662273595645, "xie_feature_dispersion": 1.6815976407755642, "pac": 0.6400288920496858, "whitened_cosine_id": 0.857738196849823}} +{"model": "gpt2", "dataset": "imdb", "seed": 0, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350731, "raptor_stability": 0.7829911708831787, "fragility": 0.25, "augmentation_robustness": 0.530870321802199, "xie_feature_dispersion": 1.6579993440971876, "pac": 0.6569307463426888, "whitened_cosine_id": 0.8754125237464905}} +{"model": "gpt2", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.34635368740834577, "raptor_stability": 0.8845524787902832, "fragility": 0.25, "augmentation_robustness": 0.8737520489801199, "xie_feature_dispersion": 0.8678892394256513, "pac": 0.8791522638852016, "whitened_cosine_id": 0.910406231880188}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 0, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.12115672881499416, "raptor_stability": 0.9302389025688171, "fragility": 1.5, "augmentation_robustness": 0.8117314786216752, "xie_feature_dispersion": 6.121815304004312, "pac": 0.8709851905952462, "whitened_cosine_id": 0.9043624997138977}} +{"model": "gpt2", "dataset": "counterfact", "seed": 0, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859340033, "raptor_stability": 0.7690413594245911, "fragility": 0.25, "augmentation_robustness": 0.7125524136549481, "xie_feature_dispersion": 0.1095429604074235, "pac": 0.7407968865397696, "whitened_cosine_id": 0.7072374820709229}} +{"model": "gpt2", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.7366388390202255, "raptor_stability": 0.8725014328956604, "fragility": 0.25, "augmentation_robustness": 0.7768639767211045, "xie_feature_dispersion": 1.77009486868241, "pac": 0.8246827048083825, "whitened_cosine_id": 0.8444538116455078}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347128, "raptor_stability": 0.7693442106246948, "fragility": 0.25, "augmentation_robustness": 0.4900059843963644, "xie_feature_dispersion": 3.694561283935865, "pac": 0.6296750975105296, "whitened_cosine_id": 0.8187122344970703}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 0, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346073, "raptor_stability": 0.7867448925971985, "fragility": 0.25, "augmentation_robustness": 0.6153990422956858, "xie_feature_dispersion": 2.309585917543921, "pac": 0.7010719674464421, "whitened_cosine_id": 0.8086420297622681}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 0, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346306, "raptor_stability": 0.7805789113044739, "fragility": 0.25, "augmentation_robustness": 0.636728510469453, "xie_feature_dispersion": 0.25866709333672444, "pac": 0.7086537108869635, "whitened_cosine_id": 0.803403377532959}} +{"model": "gpt2", "dataset": "subj", "seed": 0, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352106, "raptor_stability": 0.8574187755584717, "fragility": 0.25, "augmentation_robustness": 0.6556398588385755, "xie_feature_dispersion": 0.951143558340801, "pac": 0.7565293171985236, "whitened_cosine_id": 0.9090107679367065}} +{"model": "gpt2", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593527, "raptor_stability": 0.9161890149116516, "fragility": 2.0, "augmentation_robustness": 0.8054413549470808, "xie_feature_dispersion": 6.497801646724056, "pac": 0.8608151849293662, "whitened_cosine_id": 0.9386647939682007}} +{"model": "gpt2", "dataset": "cola", "seed": 0, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859346528, "raptor_stability": 0.7938734889030457, "fragility": 0.25, "augmentation_robustness": 0.46439785386376703, "xie_feature_dispersion": 0.9649847502240455, "pac": 0.6291356713834063, "whitened_cosine_id": 0.807263970375061}} +{"model": "gpt2", "dataset": "stance", "seed": 0, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.490010700608385, "raptor_stability": 0.8106656670570374, "fragility": 0.25, "augmentation_robustness": 0.7245922829104868, "xie_feature_dispersion": 2.664520632491477, "pac": 0.7676289749837621, "whitened_cosine_id": 0.8153484463691711}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935004, "raptor_stability": 0.8306007981300354, "fragility": 0.25, "augmentation_robustness": 0.5872621905723052, "xie_feature_dispersion": 1.1399162300669312, "pac": 0.7089314943511703, "whitened_cosine_id": 0.8771829605102539}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 0, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688191, "raptor_stability": 0.8413395881652832, "fragility": 0.25, "augmentation_robustness": 0.5136767036968108, "xie_feature_dispersion": 1.7719582872902115, "pac": 0.677508145931047, "whitened_cosine_id": 0.8595471382141113}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 0, "layer": 23, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881988, "raptor_stability": 0.9016227722167969, "fragility": 0.25, "augmentation_robustness": 0.7456883958977025, "xie_feature_dispersion": 1.107946359548132, "pac": 0.8236555840572497, "whitened_cosine_id": 0.8640339970588684}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.26407762597782747, "raptor_stability": 0.8502908945083618, "fragility": 1.0, "augmentation_robustness": 0.7617860523249808, "xie_feature_dispersion": 4.268927113799556, "pac": 0.8060384734166712, "whitened_cosine_id": 0.8772361874580383}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.097419516062648, "raptor_stability": 0.928425133228302, "fragility": 1.5, "augmentation_robustness": 0.8139603155889766, "xie_feature_dispersion": 6.700705765199243, "pac": 0.8711927244086393, "whitened_cosine_id": 0.8741976618766785}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 0, "layer": 3, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878758, "raptor_stability": 0.7550429105758667, "fragility": 0.25, "augmentation_robustness": 0.7384688990847975, "xie_feature_dispersion": 0.2304978644612739, "pac": 0.746755904830332, "whitened_cosine_id": 0.7729927897453308}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 0, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6820984754002586, "raptor_stability": 0.8745980262756348, "fragility": 0.25, "augmentation_robustness": 0.713824744543151, "xie_feature_dispersion": 0.8621445356293869, "pac": 0.7942113854093928, "whitened_cosine_id": 0.8492467999458313}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880176, "raptor_stability": 0.8115680813789368, "fragility": 0.25, "augmentation_robustness": 0.5889832766438082, "xie_feature_dispersion": 1.137989115163849, "pac": 0.7002756790113724, "whitened_cosine_id": 0.8243570923805237}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 0, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878862, "raptor_stability": 0.777171790599823, "fragility": 0.25, "augmentation_robustness": 0.5409047648889594, "xie_feature_dispersion": 2.1506698277748155, "pac": 0.6590382777443913, "whitened_cosine_id": 0.8085216283798218}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 0, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880017, "raptor_stability": 0.7901369333267212, "fragility": 0.25, "augmentation_robustness": 0.5844201102203882, "xie_feature_dispersion": 0.462184243542827, "pac": 0.6872785217735546, "whitened_cosine_id": 0.8134031891822815}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 0, "layer": 24, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688252, "raptor_stability": 0.8832855224609375, "fragility": 0.25, "augmentation_robustness": 0.7291403967687851, "xie_feature_dispersion": 0.4906864607581503, "pac": 0.8062129596148613, "whitened_cosine_id": 0.8952683210372925}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882066, "raptor_stability": 0.914933979511261, "fragility": 1.0, "augmentation_robustness": 0.8072328905323266, "xie_feature_dispersion": 4.701874448406257, "pac": 0.8610834350217937, "whitened_cosine_id": 0.915591835975647}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 0, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880159, "raptor_stability": 0.7916848063468933, "fragility": 0.25, "augmentation_robustness": 0.4476010668333125, "xie_feature_dispersion": 0.79552886696272, "pac": 0.6196429365901029, "whitened_cosine_id": 0.8178503513336182}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 0, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.42249695846338925, "raptor_stability": 0.8128862977027893, "fragility": 0.5, "augmentation_robustness": 0.7398084898037611, "xie_feature_dispersion": 2.2610696467251814, "pac": 0.7763473937532752, "whitened_cosine_id": 0.7861542701721191}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 0, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881549, "raptor_stability": 0.8164692521095276, "fragility": 0.25, "augmentation_robustness": 0.5951602317050955, "xie_feature_dispersion": 1.1005465582305016, "pac": 0.7058147419073115, "whitened_cosine_id": 0.8596013784408569}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648414, "raptor_stability": 0.8818655014038086, "fragility": 0.25, "augmentation_robustness": 0.6289366997740752, "xie_feature_dispersion": 1.8204152059660144, "pac": 0.7554011005889418, "whitened_cosine_id": 0.879472017288208}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 0, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.29387292376502, "raptor_stability": 0.8775492906570435, "fragility": 1.0, "augmentation_robustness": 0.7836625681170646, "xie_feature_dispersion": 2.218615090585379, "pac": 0.830605929387054, "whitened_cosine_id": 0.8882995247840881}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.33450754332538074, "raptor_stability": 0.8566421270370483, "fragility": 1.0, "augmentation_robustness": 0.7926578157046991, "xie_feature_dispersion": 3.7561628051141023, "pac": 0.8246499713708737, "whitened_cosine_id": 0.9043422937393188}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07399960414320991, "raptor_stability": 0.9483939409255981, "fragility": 1.5, "augmentation_robustness": 0.881505837440225, "xie_feature_dispersion": 7.327642739764713, "pac": 0.9149498891829115, "whitened_cosine_id": 0.9209960699081421}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 0, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237645932, "raptor_stability": 0.8066194653511047, "fragility": 0.25, "augmentation_robustness": 0.7512795380738843, "xie_feature_dispersion": 0.06081582657176685, "pac": 0.7789495017124946, "whitened_cosine_id": 0.7630186080932617}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6131208141710327, "raptor_stability": 0.8868826031684875, "fragility": 0.25, "augmentation_robustness": 0.848475526665586, "xie_feature_dispersion": 1.992273238599221, "pac": 0.8676790649170367, "whitened_cosine_id": 0.8511517643928528}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645996, "raptor_stability": 0.8185003995895386, "fragility": 0.25, "augmentation_robustness": 0.6978679311233282, "xie_feature_dispersion": 4.466765724760196, "pac": 0.7581841653564334, "whitened_cosine_id": 0.8152397871017456}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 0, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647149, "raptor_stability": 0.811665415763855, "fragility": 0.25, "augmentation_robustness": 0.6512162425704373, "xie_feature_dispersion": 2.8843177774025395, "pac": 0.7314408291671461, "whitened_cosine_id": 0.8048377633094788}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 0, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647015, "raptor_stability": 0.8171485662460327, "fragility": 0.25, "augmentation_robustness": 0.7362926278030426, "xie_feature_dispersion": 0.15381271196675575, "pac": 0.7767205970245377, "whitened_cosine_id": 0.8043486475944519}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 0, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651123, "raptor_stability": 0.8921712040901184, "fragility": 0.5, "augmentation_robustness": 0.7485023298249472, "xie_feature_dispersion": 0.9955751494789675, "pac": 0.8203367669575328, "whitened_cosine_id": 0.9154570698738098}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 0, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923765166, "raptor_stability": 0.9529117941856384, "fragility": 3.0, "augmentation_robustness": 0.87503706844258, "xie_feature_dispersion": 9.854132791030557, "pac": 0.9139744313141092, "whitened_cosine_id": 0.9523190855979919}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 0, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764728, "raptor_stability": 0.8035885691642761, "fragility": 0.25, "augmentation_robustness": 0.5005855255082153, "xie_feature_dispersion": 1.0666210514085848, "pac": 0.6520870473362457, "whitened_cosine_id": 0.818631649017334}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 0, "layer": 20, "concept": "stance", "predictors": {"sip_eigengap": 0.4157006657150525, "raptor_stability": 0.8240557312965393, "fragility": 0.5, "augmentation_robustness": 0.7331066076432546, "xie_feature_dispersion": 3.234317390572246, "pac": 0.778581169469897, "whitened_cosine_id": 0.8179960250854492}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 0, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649436, "raptor_stability": 0.8530629873275757, "fragility": 0.25, "augmentation_robustness": 0.7025335558592776, "xie_feature_dispersion": 1.1742203784808831, "pac": 0.7777982715934266, "whitened_cosine_id": 0.8717560768127441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 1, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921994895, "raptor_stability": 0.8314626216888428, "fragility": 0.25, "augmentation_robustness": 0.5808293731010495, "xie_feature_dispersion": 1.6689369031429147, "pac": 0.7061459973949462, "whitened_cosine_id": 0.824091911315918}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998223, "raptor_stability": 0.8874242901802063, "fragility": 1.0, "augmentation_robustness": 0.8228104146559615, "xie_feature_dispersion": 1.271550857614664, "pac": 0.855117352418084, "whitened_cosine_id": 0.8205482363700867}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6806422452118264, "raptor_stability": 0.8989484906196594, "fragility": 1.5, "augmentation_robustness": 0.8787513256973658, "xie_feature_dispersion": 3.4157347611498774, "pac": 0.8888499081585126, "whitened_cosine_id": 0.9129272699356079}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 1, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.15525918027531763, "raptor_stability": 0.9445532560348511, "fragility": 1.5, "augmentation_robustness": 0.8628763882540946, "xie_feature_dispersion": 5.228730991204411, "pac": 0.9037148221444729, "whitened_cosine_id": 0.9321261644363403}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 1, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974072, "raptor_stability": 0.7612232565879822, "fragility": 0.25, "augmentation_robustness": 0.6405111879943068, "xie_feature_dispersion": 0.7015570516775577, "pac": 0.7008672222911445, "whitened_cosine_id": 0.6881313323974609}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 1, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5841926037551581, "raptor_stability": 0.8477641940116882, "fragility": 0.5, "augmentation_robustness": 0.6474137169579742, "xie_feature_dispersion": 1.484633799742414, "pac": 0.7475889554848312, "whitened_cosine_id": 0.8390121459960938}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 1, "layer": 3, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994224, "raptor_stability": 0.8421536087989807, "fragility": 0.5, "augmentation_robustness": 0.7239927602573658, "xie_feature_dispersion": 2.981639036084935, "pac": 0.7830731845281733, "whitened_cosine_id": 0.8215951323509216}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 1, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987687, "raptor_stability": 0.819472074508667, "fragility": 1.0, "augmentation_robustness": 0.6380691012431728, "xie_feature_dispersion": 2.073055712645641, "pac": 0.7287705878759199, "whitened_cosine_id": 0.7341217398643494}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 1, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989095, "raptor_stability": 0.843525767326355, "fragility": 1.0, "augmentation_robustness": 0.8007193546782, "xie_feature_dispersion": 1.5447695946615063, "pac": 0.8221225610022775, "whitened_cosine_id": 0.7792574167251587}} +{"model": "pythia-70m", "dataset": "subj", "seed": 1, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002709, "raptor_stability": 0.8913278579711914, "fragility": 1.0, "augmentation_robustness": 0.7332091297611866, "xie_feature_dispersion": 1.980053369512756, "pac": 0.812268493866189, "whitened_cosine_id": 0.9182124137878418}} +{"model": "pythia-70m", "dataset": "spam", "seed": 1, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922003908, "raptor_stability": 0.9203832745552063, "fragility": 1.5, "augmentation_robustness": 0.7922849095394744, "xie_feature_dispersion": 10.611895676124647, "pac": 0.8563340920473403, "whitened_cosine_id": 0.950851321220398}} +{"model": "pythia-70m", "dataset": "cola", "seed": 1, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.7986350655555725, "fragility": 1.0, "augmentation_robustness": 0.7043344717903556, "xie_feature_dispersion": 0.7104026437376014, "pac": 0.751484768672964, "whitened_cosine_id": 0.7400056719779968}} +{"model": "pythia-70m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6754527167162487, "raptor_stability": 0.8382856249809265, "fragility": 1.0, "augmentation_robustness": 0.8019897778297428, "xie_feature_dispersion": 1.8257872743628552, "pac": 0.8201377014053346, "whitened_cosine_id": 0.8341732621192932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 1, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997764, "raptor_stability": 0.8679760098457336, "fragility": 0.5, "augmentation_robustness": 0.7205293551358967, "xie_feature_dispersion": 0.9711058139858004, "pac": 0.7942526824908152, "whitened_cosine_id": 0.8618689775466919}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 1, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349757, "raptor_stability": 0.8496716618537903, "fragility": 0.25, "augmentation_robustness": 0.5842367312683088, "xie_feature_dispersion": 1.8689219377096642, "pac": 0.7169541965610495, "whitened_cosine_id": 0.8571608662605286}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.39754248593502, "raptor_stability": 0.8464771509170532, "fragility": 0.5, "augmentation_robustness": 0.6850682519540424, "xie_feature_dispersion": 1.7078889299200177, "pac": 0.7657727014355478, "whitened_cosine_id": 0.8704335689544678}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 1, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.3976585069526139, "raptor_stability": 0.8461608290672302, "fragility": 1.0, "augmentation_robustness": 0.7765248733340492, "xie_feature_dispersion": 2.8951389779509134, "pac": 0.8113428512006398, "whitened_cosine_id": 0.9028950929641724}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.12810819210433996, "raptor_stability": 0.9377708435058594, "fragility": 3.0, "augmentation_robustness": 0.833317612225739, "xie_feature_dispersion": 10.50628418134277, "pac": 0.8855442278657992, "whitened_cosine_id": 0.9117757678031921}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 1, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.397542485934175, "raptor_stability": 0.7833874821662903, "fragility": 0.25, "augmentation_robustness": 0.6108469106562627, "xie_feature_dispersion": 0.3790863394835801, "pac": 0.6971171964112766, "whitened_cosine_id": 0.7308042049407959}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 1, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.5499358318667639, "raptor_stability": 0.8611683249473572, "fragility": 0.25, "augmentation_robustness": 0.7033012402006117, "xie_feature_dispersion": 1.9883915877863603, "pac": 0.7822347825739844, "whitened_cosine_id": 0.8481037616729736}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 1, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347079, "raptor_stability": 0.8096680641174316, "fragility": 0.25, "augmentation_robustness": 0.6360886444392633, "xie_feature_dispersion": 2.3558677799388796, "pac": 0.7228783542783475, "whitened_cosine_id": 0.8148756623268127}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 1, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345162, "raptor_stability": 0.7992299795150757, "fragility": 0.25, "augmentation_robustness": 0.5915203833323763, "xie_feature_dispersion": 3.2725825993057094, "pac": 0.695375181423726, "whitened_cosine_id": 0.7864978909492493}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 1, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345495, "raptor_stability": 0.8322464823722839, "fragility": 1.0, "augmentation_robustness": 0.768225719172241, "xie_feature_dispersion": 1.8790072833719011, "pac": 0.8002361007722625, "whitened_cosine_id": 0.7750239968299866}} +{"model": "pythia-160m", "dataset": "subj", "seed": 1, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352137, "raptor_stability": 0.8875827193260193, "fragility": 1.0, "augmentation_robustness": 0.7169304038534543, "xie_feature_dispersion": 1.9523142841304229, "pac": 0.8022565615897368, "whitened_cosine_id": 0.9101552963256836}} +{"model": "pythia-160m", "dataset": "spam", "seed": 1, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859347847, "raptor_stability": 0.913774847984314, "fragility": 1.0, "augmentation_robustness": 0.8208256677071334, "xie_feature_dispersion": 14.489404204949587, "pac": 0.8673002578457236, "whitened_cosine_id": 0.93967604637146}} +{"model": "pythia-160m", "dataset": "cola", "seed": 1, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859342416, "raptor_stability": 0.7926198244094849, "fragility": 1.0, "augmentation_robustness": 0.6447861391769918, "xie_feature_dispersion": 0.9165363040019758, "pac": 0.7187029817932383, "whitened_cosine_id": 0.7564545273780823}} +{"model": "pythia-160m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6971690127369661, "raptor_stability": 0.8396480679512024, "fragility": 1.0, "augmentation_robustness": 0.7875725989177652, "xie_feature_dispersion": 2.4375720598203166, "pac": 0.8136103334344837, "whitened_cosine_id": 0.8327767848968506}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 1, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349908, "raptor_stability": 0.8594450950622559, "fragility": 0.5, "augmentation_robustness": 0.6977040298979517, "xie_feature_dispersion": 0.7745815425512376, "pac": 0.7785745624801037, "whitened_cosine_id": 0.8690269589424133}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 1, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688131, "raptor_stability": 0.8608813881874084, "fragility": 0.25, "augmentation_robustness": 0.5666047397490934, "xie_feature_dispersion": 2.4202065461787297, "pac": 0.7137430639682509, "whitened_cosine_id": 0.8642251491546631}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 1, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881893, "raptor_stability": 0.8481273055076599, "fragility": 1.0, "augmentation_robustness": 0.7130798265252153, "xie_feature_dispersion": 2.320769849966657, "pac": 0.7806035660164377, "whitened_cosine_id": 0.8769779801368713}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 1, "layer": 13, "concept": "topic", "predictors": {"sip_eigengap": 0.30704864737844834, "raptor_stability": 0.8611399531364441, "fragility": 1.0, "augmentation_robustness": 0.7903941411680551, "xie_feature_dispersion": 3.4011674827208287, "pac": 0.8257670471522496, "whitened_cosine_id": 0.8893525004386902}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 1, "layer": 19, "concept": "topic", "predictors": {"sip_eigengap": 0.08244206897104008, "raptor_stability": 0.9409602880477905, "fragility": 2.0, "augmentation_robustness": 0.8566799506256193, "xie_feature_dispersion": 8.808736850172968, "pac": 0.8988201193367049, "whitened_cosine_id": 0.891150951385498}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 1, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878758, "raptor_stability": 0.7924683690071106, "fragility": 0.25, "augmentation_robustness": 0.6334062400656503, "xie_feature_dispersion": 0.6634691181873394, "pac": 0.7129373045363805, "whitened_cosine_id": 0.7721397280693054}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 1, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.519816512768835, "raptor_stability": 0.8667904734611511, "fragility": 0.25, "augmentation_robustness": 0.7478177367365705, "xie_feature_dispersion": 2.8557298314061703, "pac": 0.8073041050988607, "whitened_cosine_id": 0.8457608222961426}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 1, "layer": 10, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956878145, "raptor_stability": 0.8067510724067688, "fragility": 0.25, "augmentation_robustness": 0.6562650597345253, "xie_feature_dispersion": 5.470678455409298, "pac": 0.731508066070647, "whitened_cosine_id": 0.8200975656509399}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 1, "layer": 21, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879326, "raptor_stability": 0.7820147275924683, "fragility": 0.25, "augmentation_robustness": 0.587986767048995, "xie_feature_dispersion": 2.5969424874908773, "pac": 0.6850007473207316, "whitened_cosine_id": 0.7967674136161804}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 1, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879308, "raptor_stability": 0.7950331568717957, "fragility": 0.25, "augmentation_robustness": 0.6646175791835158, "xie_feature_dispersion": 2.059673087772778, "pac": 0.7298253680276557, "whitened_cosine_id": 0.8044969439506531}} +{"model": "pythia-410m", "dataset": "subj", "seed": 1, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882595, "raptor_stability": 0.8856076598167419, "fragility": 0.5, "augmentation_robustness": 0.7139391967127188, "xie_feature_dispersion": 1.5476544469738573, "pac": 0.7997734282647304, "whitened_cosine_id": 0.8976815342903137}} +{"model": "pythia-410m", "dataset": "spam", "seed": 1, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956874568, "raptor_stability": 0.9077131152153015, "fragility": 0.5, "augmentation_robustness": 0.8020763670174436, "xie_feature_dispersion": 17.967343252945344, "pac": 0.8548947411163725, "whitened_cosine_id": 0.9222725629806519}} +{"model": "pythia-410m", "dataset": "cola", "seed": 1, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879513, "raptor_stability": 0.7959128022193909, "fragility": 0.25, "augmentation_robustness": 0.45896954560148306, "xie_feature_dispersion": 0.6838657969146573, "pac": 0.627441173910437, "whitened_cosine_id": 0.8098257780075073}} +{"model": "pythia-410m", "dataset": "stance", "seed": 1, "layer": 19, "concept": "stance", "predictors": {"sip_eigengap": 0.520575334062942, "raptor_stability": 0.8262770771980286, "fragility": 0.5, "augmentation_robustness": 0.7565572182499646, "xie_feature_dispersion": 2.363460584912377, "pac": 0.7914171477239966, "whitened_cosine_id": 0.8227141499519348}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 1, "layer": 15, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881407, "raptor_stability": 0.8368763327598572, "fragility": 0.5, "augmentation_robustness": 0.6726761343337394, "xie_feature_dispersion": 0.8861988690029199, "pac": 0.7547762335467982, "whitened_cosine_id": 0.8615521788597107}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010796, "raptor_stability": 0.8518760800361633, "fragility": 0.25, "augmentation_robustness": 0.5469758605057219, "xie_feature_dispersion": 3.1855570820407504, "pac": 0.6994259702709427, "whitened_cosine_id": 0.8287058472633362}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8483022451400757, "fragility": 1.0, "augmentation_robustness": 0.6553477126801958, "xie_feature_dispersion": 2.6155304376742743, "pac": 0.7518249789101357, "whitened_cosine_id": 0.8468784689903259}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22290771506820922, "raptor_stability": 0.8515928983688354, "fragility": 1.5, "augmentation_robustness": 0.7746267660688861, "xie_feature_dispersion": 5.927804802925069, "pac": 0.8131098322188608, "whitened_cosine_id": 0.8236908912658691}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.058822548518337014, "raptor_stability": 0.9422677159309387, "fragility": 2.0, "augmentation_robustness": 0.8595274869237916, "xie_feature_dispersion": 8.534984669917208, "pac": 0.9008976014273651, "whitened_cosine_id": 0.8359106779098511}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.7920423746109009, "fragility": 0.25, "augmentation_robustness": 0.6413142308147114, "xie_feature_dispersion": 0.860340632017618, "pac": 0.7166783027128061, "whitened_cosine_id": 0.8168047666549683}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 1, "layer": 9, "concept": "emotion", "predictors": {"sip_eigengap": 0.4313770796442383, "raptor_stability": 0.8590365052223206, "fragility": 0.25, "augmentation_robustness": 0.694955833506802, "xie_feature_dispersion": 3.6526929613427925, "pac": 0.7769961693645613, "whitened_cosine_id": 0.8269678950309753}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 1, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 0.855816496101069, "raptor_stability": 0.810732364654541, "fragility": 0.25, "augmentation_robustness": 0.6270267466287328, "xie_feature_dispersion": 6.292509564537482, "pac": 0.718879555641637, "whitened_cosine_id": 0.8283994793891907}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 1, "layer": 13, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010526, "raptor_stability": 0.8113015294075012, "fragility": 0.25, "augmentation_robustness": 0.6088162548597921, "xie_feature_dispersion": 3.2324734107222866, "pac": 0.7100588921336466, "whitened_cosine_id": 0.8244034051895142}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 1, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010755, "raptor_stability": 0.791317880153656, "fragility": 0.25, "augmentation_robustness": 0.6080442076224152, "xie_feature_dispersion": 2.2409906655814193, "pac": 0.6996810438880356, "whitened_cosine_id": 0.818917989730835}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 1, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010871, "raptor_stability": 0.8893104195594788, "fragility": 1.0, "augmentation_robustness": 0.696504331799587, "xie_feature_dispersion": 2.547582289724496, "pac": 0.7929073756795328, "whitened_cosine_id": 0.832260251045227}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 1, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010578, "raptor_stability": 0.9184004664421082, "fragility": 1.0, "augmentation_robustness": 0.7929347080961948, "xie_feature_dispersion": 23.49076580939176, "pac": 0.8556675872691515, "whitened_cosine_id": 0.8580988049507141}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 1, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961009699, "raptor_stability": 0.8054239749908447, "fragility": 0.25, "augmentation_robustness": 0.46022627880181477, "xie_feature_dispersion": 0.7167139402414644, "pac": 0.6328251268963297, "whitened_cosine_id": 0.8232492804527283}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 1, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.4012148911144739, "raptor_stability": 0.8381373882293701, "fragility": 1.0, "augmentation_robustness": 0.7598258349622733, "xie_feature_dispersion": 2.6386343651247124, "pac": 0.7989816115958217, "whitened_cosine_id": 0.8214442133903503}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 1, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010833, "raptor_stability": 0.8418669700622559, "fragility": 1.0, "augmentation_robustness": 0.6524348508596854, "xie_feature_dispersion": 1.6761850609669111, "pac": 0.7471509104609706, "whitened_cosine_id": 0.8340412378311157}} +{"model": "gpt2", "dataset": "sst2", "seed": 1, "layer": 8, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485935016, "raptor_stability": 0.8106276392936707, "fragility": 0.25, "augmentation_robustness": 0.49014044266582973, "xie_feature_dispersion": 1.6638828788327318, "pac": 0.6503840409797502, "whitened_cosine_id": 0.8658217191696167}} +{"model": "gpt2", "dataset": "imdb", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350674, "raptor_stability": 0.8155258893966675, "fragility": 0.25, "augmentation_robustness": 0.582233966857331, "xie_feature_dispersion": 1.679121568516715, "pac": 0.6988799281269993, "whitened_cosine_id": 0.8795225024223328}} +{"model": "gpt2", "dataset": "ag_news", "seed": 1, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.31952355566470164, "raptor_stability": 0.8801834583282471, "fragility": 0.25, "augmentation_robustness": 0.8715217181573922, "xie_feature_dispersion": 0.9325824128470682, "pac": 0.8758525882428196, "whitened_cosine_id": 0.9088584184646606}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.13768748508218773, "raptor_stability": 0.9254063367843628, "fragility": 1.5, "augmentation_robustness": 0.7871717482295453, "xie_feature_dispersion": 5.392898240224899, "pac": 0.856289042506954, "whitened_cosine_id": 0.9057890176773071}} +{"model": "gpt2", "dataset": "counterfact", "seed": 1, "layer": 7, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343133, "raptor_stability": 0.7852461338043213, "fragility": 0.25, "augmentation_robustness": 0.6622392322460525, "xie_feature_dispersion": 0.5647271084995943, "pac": 0.723742683025187, "whitened_cosine_id": 0.7448322176933289}} +{"model": "gpt2", "dataset": "emotion", "seed": 1, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5436635722625862, "raptor_stability": 0.8717889785766602, "fragility": 0.25, "augmentation_robustness": 0.7592733911032505, "xie_feature_dispersion": 2.524501993095808, "pac": 0.8155311848399553, "whitened_cosine_id": 0.8443379998207092}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348176, "raptor_stability": 0.8326132297515869, "fragility": 0.25, "augmentation_robustness": 0.7215291300706207, "xie_feature_dispersion": 0.8694719793739331, "pac": 0.7770711799111039, "whitened_cosine_id": 0.8273347020149231}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 1, "layer": 10, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346122, "raptor_stability": 0.7642949819564819, "fragility": 0.25, "augmentation_robustness": 0.5635955618554486, "xie_feature_dispersion": 2.167418535028266, "pac": 0.6639452719059653, "whitened_cosine_id": 0.7877233624458313}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 1, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346217, "raptor_stability": 0.7664420008659363, "fragility": 0.25, "augmentation_robustness": 0.4934067076281954, "xie_feature_dispersion": 1.3905365467341602, "pac": 0.6299243542470658, "whitened_cosine_id": 0.7959058284759521}} +{"model": "gpt2", "dataset": "subj", "seed": 1, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352203, "raptor_stability": 0.853984534740448, "fragility": 0.5, "augmentation_robustness": 0.6319009208667878, "xie_feature_dispersion": 1.4255917684674573, "pac": 0.7429427278036179, "whitened_cosine_id": 0.9104035496711731}} +{"model": "gpt2", "dataset": "spam", "seed": 1, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353107, "raptor_stability": 0.9109200239181519, "fragility": 1.0, "augmentation_robustness": 0.7966288421897234, "xie_feature_dispersion": 14.38720114784055, "pac": 0.8537744330539376, "whitened_cosine_id": 0.9343566298484802}} +{"model": "gpt2", "dataset": "cola", "seed": 1, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859345384, "raptor_stability": 0.7919601202011108, "fragility": 0.25, "augmentation_robustness": 0.5130744044422636, "xie_feature_dispersion": 0.34394405556746344, "pac": 0.6525172623216873, "whitened_cosine_id": 0.7920956015586853}} +{"model": "gpt2", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.5721762371971069, "raptor_stability": 0.8337757587432861, "fragility": 0.5, "augmentation_robustness": 0.7456673050417738, "xie_feature_dispersion": 1.1954515513989168, "pac": 0.7897215318925299, "whitened_cosine_id": 0.8244547843933105}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 1, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349934, "raptor_stability": 0.8276537656784058, "fragility": 0.25, "augmentation_robustness": 0.6092711286506372, "xie_feature_dispersion": 0.655742490317081, "pac": 0.7184624471645215, "whitened_cosine_id": 0.8719987273216248}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.21030729568816, "raptor_stability": 0.8340166807174683, "fragility": 0.25, "augmentation_robustness": 0.5218252189937355, "xie_feature_dispersion": 1.7679503238709116, "pac": 0.6779209498556018, "whitened_cosine_id": 0.8651541471481323}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 1, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882157, "raptor_stability": 0.8258817195892334, "fragility": 0.25, "augmentation_robustness": 0.5752874121442174, "xie_feature_dispersion": 1.3475744664628955, "pac": 0.7005845658667254, "whitened_cosine_id": 0.8705449104309082}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 1, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.26782873777331245, "raptor_stability": 0.841077983379364, "fragility": 1.0, "augmentation_robustness": 0.7470633913167877, "xie_feature_dispersion": 4.180838494364298, "pac": 0.7940706873480758, "whitened_cosine_id": 0.8797165155410767}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.10399119928589491, "raptor_stability": 0.9248352646827698, "fragility": 1.5, "augmentation_robustness": 0.7945295897496373, "xie_feature_dispersion": 5.61182173262905, "pac": 0.8596824272162036, "whitened_cosine_id": 0.8758760094642639}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 1, "layer": 14, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878629, "raptor_stability": 0.7786945700645447, "fragility": 0.25, "augmentation_robustness": 0.6144804383339777, "xie_feature_dispersion": 0.5748324075475697, "pac": 0.6965875041992612, "whitened_cosine_id": 0.7664059400558472}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5592522550238864, "raptor_stability": 0.8738617300987244, "fragility": 0.25, "augmentation_robustness": 0.720919459890958, "xie_feature_dispersion": 0.9112899816215909, "pac": 0.7973905949948412, "whitened_cosine_id": 0.8513531684875488}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 1, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880974, "raptor_stability": 0.8126772046089172, "fragility": 0.25, "augmentation_robustness": 0.7050632297874023, "xie_feature_dispersion": 1.322075514358211, "pac": 0.7588702171981598, "whitened_cosine_id": 0.8305314779281616}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 1, "layer": 13, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879664, "raptor_stability": 0.7759368419647217, "fragility": 0.25, "augmentation_robustness": 0.5622012822384077, "xie_feature_dispersion": 2.3188187063045684, "pac": 0.6690690621015647, "whitened_cosine_id": 0.8039388656616211}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 1, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880128, "raptor_stability": 0.7699597477912903, "fragility": 0.25, "augmentation_robustness": 0.6476643493242761, "xie_feature_dispersion": 0.7754531474850158, "pac": 0.7088120485577831, "whitened_cosine_id": 0.8113746047019958}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 1, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882637, "raptor_stability": 0.8628992438316345, "fragility": 1.0, "augmentation_robustness": 0.6397751392051434, "xie_feature_dispersion": 2.0316047842765594, "pac": 0.751337191518389, "whitened_cosine_id": 0.8833309412002563}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 1, "layer": 16, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882794, "raptor_stability": 0.9118919968605042, "fragility": 1.5, "augmentation_robustness": 0.7830820986182032, "xie_feature_dispersion": 13.506990469815198, "pac": 0.8474870477393537, "whitened_cosine_id": 0.908318042755127}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 1, "layer": 16, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880065, "raptor_stability": 0.8000022768974304, "fragility": 0.25, "augmentation_robustness": 0.4811244638624909, "xie_feature_dispersion": 0.37533178694301084, "pac": 0.6405633703799607, "whitened_cosine_id": 0.8167566061019897}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 1, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.4372306810871382, "raptor_stability": 0.8121774792671204, "fragility": 0.5, "augmentation_robustness": 0.7197757150389671, "xie_feature_dispersion": 1.6812559297796246, "pac": 0.7659765971530437, "whitened_cosine_id": 0.7655757069587708}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 1, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881318, "raptor_stability": 0.8076351881027222, "fragility": 0.25, "augmentation_robustness": 0.6171652138408624, "xie_feature_dispersion": 0.9086262406903829, "pac": 0.7124002009717922, "whitened_cosine_id": 0.8581160306930542}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648992, "raptor_stability": 0.8649455904960632, "fragility": 0.25, "augmentation_robustness": 0.6131424180860278, "xie_feature_dispersion": 2.15799302448076, "pac": 0.7390440042910456, "whitened_cosine_id": 0.8817294239997864}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 1, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650475, "raptor_stability": 0.8865694999694824, "fragility": 1.0, "augmentation_robustness": 0.7989302272108586, "xie_feature_dispersion": 2.4686831296081597, "pac": 0.8427498635901705, "whitened_cosine_id": 0.8952975869178772}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 1, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.32396935001382504, "raptor_stability": 0.8518233299255371, "fragility": 1.0, "augmentation_robustness": 0.7712947063545464, "xie_feature_dispersion": 4.51435777726234, "pac": 0.8115590181400418, "whitened_cosine_id": 0.902996301651001}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.0634212938889773, "raptor_stability": 0.9348612427711487, "fragility": 3.0, "augmentation_robustness": 0.8289800859783713, "xie_feature_dispersion": 10.742096589376187, "pac": 0.8819206643747599, "whitened_cosine_id": 0.8965702056884766}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646287, "raptor_stability": 0.8297770023345947, "fragility": 0.25, "augmentation_robustness": 0.7628796605466598, "xie_feature_dispersion": 0.5968347455995571, "pac": 0.7963283314406273, "whitened_cosine_id": 0.7946618795394897}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 1, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.5646054321397148, "raptor_stability": 0.877227246761322, "fragility": 0.25, "augmentation_robustness": 0.8188598337200217, "xie_feature_dispersion": 2.8243621674803006, "pac": 0.8480435402406719, "whitened_cosine_id": 0.849128246307373}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 1, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237648017, "raptor_stability": 0.7998038530349731, "fragility": 0.25, "augmentation_robustness": 0.6147531583223901, "xie_feature_dispersion": 5.29862303981088, "pac": 0.7072785056786817, "whitened_cosine_id": 0.8292998671531677}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 1, "layer": 10, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237642246, "raptor_stability": 0.8075653314590454, "fragility": 0.25, "augmentation_robustness": 0.6484486881494425, "xie_feature_dispersion": 3.06991914353266, "pac": 0.7280070098042439, "whitened_cosine_id": 0.801740288734436}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 1, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237646085, "raptor_stability": 0.8232854008674622, "fragility": 0.25, "augmentation_robustness": 0.7361198321668313, "xie_feature_dispersion": 2.0361420955712815, "pac": 0.7797026165171468, "whitened_cosine_id": 0.8059210181236267}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 1, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237650999, "raptor_stability": 0.9056962132453918, "fragility": 0.5, "augmentation_robustness": 0.7644724917188301, "xie_feature_dispersion": 1.206422317161318, "pac": 0.835084352482111, "whitened_cosine_id": 0.9149098992347717}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 1, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237633957, "raptor_stability": 0.9107893109321594, "fragility": 0.5, "augmentation_robustness": 0.8112238006915025, "xie_feature_dispersion": 18.322389911250944, "pac": 0.8610065558118309, "whitened_cosine_id": 0.9440927505493164}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 1, "layer": 18, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764712, "raptor_stability": 0.8045408129692078, "fragility": 0.25, "augmentation_robustness": 0.48280717514059407, "xie_feature_dispersion": 0.5417228937870184, "pac": 0.6436739940549009, "whitened_cosine_id": 0.8147454857826233}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 1, "layer": 20, "concept": "stance", "predictors": {"sip_eigengap": 0.4812620447442306, "raptor_stability": 0.8222463130950928, "fragility": 0.5, "augmentation_robustness": 0.7353172970670624, "xie_feature_dispersion": 2.449239368783959, "pac": 0.7787818050810775, "whitened_cosine_id": 0.8223068118095398}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 1, "layer": 17, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764944, "raptor_stability": 0.8373395800590515, "fragility": 0.25, "augmentation_robustness": 0.6654539817611203, "xie_feature_dispersion": 0.6952088262981461, "pac": 0.7513967809100859, "whitened_cosine_id": 0.8643338680267334}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921995475, "raptor_stability": 0.826153576374054, "fragility": 0.25, "augmentation_robustness": 0.5550600140566192, "xie_feature_dispersion": 1.2685202535875624, "pac": 0.6906067952153365, "whitened_cosine_id": 0.8190510272979736}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998612, "raptor_stability": 0.8727290630340576, "fragility": 1.0, "augmentation_robustness": 0.8023565468625505, "xie_feature_dispersion": 1.1304006682636438, "pac": 0.837542804948304, "whitened_cosine_id": 0.8432037830352783}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 2, "layer": 1, "concept": "topic", "predictors": {"sip_eigengap": 0.6563267159870434, "raptor_stability": 0.9060066938400269, "fragility": 1.5, "augmentation_robustness": 0.8994135427836584, "xie_feature_dispersion": 3.2940888567741142, "pac": 0.9027101183118427, "whitened_cosine_id": 0.9018473625183105}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 2, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.1409559124523405, "raptor_stability": 0.9160086512565613, "fragility": 0.5, "augmentation_robustness": 0.8052056873288177, "xie_feature_dispersion": 1.129241758791748, "pac": 0.8606071692926895, "whitened_cosine_id": 0.9244332909584045}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 2, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921978088, "raptor_stability": 0.7689433693885803, "fragility": 0.5, "augmentation_robustness": 0.664331639648433, "xie_feature_dispersion": 0.4906246722259355, "pac": 0.7166375045185067, "whitened_cosine_id": 0.6844558715820312}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6049173382330815, "raptor_stability": 0.8735766410827637, "fragility": 1.0, "augmentation_robustness": 0.7882025082836114, "xie_feature_dispersion": 1.5346222774203366, "pac": 0.8308895746831875, "whitened_cosine_id": 0.8501214385032654}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 2, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921991698, "raptor_stability": 0.8122056126594543, "fragility": 0.5, "augmentation_robustness": 0.6586846606652739, "xie_feature_dispersion": 1.9447361037026398, "pac": 0.7354451366623641, "whitened_cosine_id": 0.7998017072677612}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987974, "raptor_stability": 0.8192676901817322, "fragility": 1.0, "augmentation_robustness": 0.6847445561526596, "xie_feature_dispersion": 1.857559017759172, "pac": 0.752006123167196, "whitened_cosine_id": 0.7365460991859436}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 2, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989806, "raptor_stability": 0.7996385097503662, "fragility": 0.5, "augmentation_robustness": 0.673138203990154, "xie_feature_dispersion": 1.2956329589634799, "pac": 0.73638835687026, "whitened_cosine_id": 0.7707058787345886}} +{"model": "pythia-70m", "dataset": "subj", "seed": 2, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.711632992200211, "raptor_stability": 0.8931885957717896, "fragility": 1.5, "augmentation_robustness": 0.7723343164195375, "xie_feature_dispersion": 2.385960236411005, "pac": 0.8327614560956635, "whitened_cosine_id": 0.9023992419242859}} +{"model": "pythia-70m", "dataset": "spam", "seed": 2, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004394, "raptor_stability": 0.9427627921104431, "fragility": 1.5, "augmentation_robustness": 0.8262282460478666, "xie_feature_dispersion": 10.434188689895425, "pac": 0.8844955190791548, "whitened_cosine_id": 0.9575247764587402}} +{"model": "pythia-70m", "dataset": "cola", "seed": 2, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921978055, "raptor_stability": 0.7819207310676575, "fragility": 1.0, "augmentation_robustness": 0.6892368500077494, "xie_feature_dispersion": 0.5849615167254145, "pac": 0.7355787905377034, "whitened_cosine_id": 0.7173330187797546}} +{"model": "pythia-70m", "dataset": "stance", "seed": 2, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.467708916223557, "raptor_stability": 0.823031485080719, "fragility": 1.0, "augmentation_robustness": 0.7788880873984344, "xie_feature_dispersion": 1.6869566033029149, "pac": 0.8009597862395768, "whitened_cosine_id": 0.8191723823547363}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 2, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998568, "raptor_stability": 0.8905608654022217, "fragility": 1.5, "augmentation_robustness": 0.7918130449642642, "xie_feature_dispersion": 1.7648002764851518, "pac": 0.8411869551832429, "whitened_cosine_id": 0.8727987408638}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 2, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349053, "raptor_stability": 0.8420531153678894, "fragility": 0.25, "augmentation_robustness": 0.5578391048029369, "xie_feature_dispersion": 1.9830546413363874, "pac": 0.6999461100854132, "whitened_cosine_id": 0.8557217121124268}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 2, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350019, "raptor_stability": 0.8570092916488647, "fragility": 1.0, "augmentation_robustness": 0.7611505060754636, "xie_feature_dispersion": 1.332977638352146, "pac": 0.8090798988621641, "whitened_cosine_id": 0.8642057180404663}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 2, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.5427793527803491, "raptor_stability": 0.8923298716545105, "fragility": 1.0, "augmentation_robustness": 0.8591541728510862, "xie_feature_dispersion": 3.279140555736114, "pac": 0.8757420222527983, "whitened_cosine_id": 0.9014958739280701}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.10936523074513556, "raptor_stability": 0.938740074634552, "fragility": 2.0, "augmentation_robustness": 0.8369704413137592, "xie_feature_dispersion": 8.126249426404561, "pac": 0.8878552579741557, "whitened_cosine_id": 0.9102428555488586}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 2, "layer": 5, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342813, "raptor_stability": 0.7747091054916382, "fragility": 0.25, "augmentation_robustness": 0.6417974566818402, "xie_feature_dispersion": 0.2710043288534134, "pac": 0.7082532810867392, "whitened_cosine_id": 0.7429371476173401}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 2, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.536126639231569, "raptor_stability": 0.8551614880561829, "fragility": 0.25, "augmentation_robustness": 0.7039385026715385, "xie_feature_dispersion": 2.1343450512794986, "pac": 0.7795499953638607, "whitened_cosine_id": 0.8513872623443604}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 2, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859346717, "raptor_stability": 0.8405966758728027, "fragility": 1.0, "augmentation_robustness": 0.7310447997932741, "xie_feature_dispersion": 2.1154557972849912, "pac": 0.7858207378330384, "whitened_cosine_id": 0.7988827228546143}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934474, "raptor_stability": 0.809672474861145, "fragility": 0.25, "augmentation_robustness": 0.6381545155767033, "xie_feature_dispersion": 3.2921137785738233, "pac": 0.7239134952189241, "whitened_cosine_id": 0.7898976802825928}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 2, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345826, "raptor_stability": 0.8115760087966919, "fragility": 0.25, "augmentation_robustness": 0.6743335197367628, "xie_feature_dispersion": 2.0708787972712512, "pac": 0.7429547642667274, "whitened_cosine_id": 0.7857599854469299}} +{"model": "pythia-160m", "dataset": "subj", "seed": 2, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352095, "raptor_stability": 0.8895910382270813, "fragility": 1.0, "augmentation_robustness": 0.7135129309774878, "xie_feature_dispersion": 1.9799099583102993, "pac": 0.8015519846022845, "whitened_cosine_id": 0.9134976267814636}} +{"model": "pythia-160m", "dataset": "spam", "seed": 2, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353311, "raptor_stability": 0.923969566822052, "fragility": 3.0, "augmentation_robustness": 0.7712028495795479, "xie_feature_dispersion": 10.474610541637501, "pac": 0.8475862082007999, "whitened_cosine_id": 0.9371472001075745}} +{"model": "pythia-160m", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934449, "raptor_stability": 0.7818026542663574, "fragility": 0.25, "augmentation_robustness": 0.5589858962588444, "xie_feature_dispersion": 0.36909274612641374, "pac": 0.6703942752626009, "whitened_cosine_id": 0.7660055756568909}} +{"model": "pythia-160m", "dataset": "stance", "seed": 2, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.550133319968111, "raptor_stability": 0.8319461345672607, "fragility": 1.5, "augmentation_robustness": 0.7869835159029682, "xie_feature_dispersion": 2.567878497868356, "pac": 0.8094648252351144, "whitened_cosine_id": 0.8248243927955627}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350152, "raptor_stability": 0.8500860929489136, "fragility": 0.5, "augmentation_robustness": 0.7049551414527425, "xie_feature_dispersion": 1.015287990938809, "pac": 0.777520617200828, "whitened_cosine_id": 0.8699454069137573}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880925, "raptor_stability": 0.8507713079452515, "fragility": 0.25, "augmentation_robustness": 0.56023098369132, "xie_feature_dispersion": 2.5382673428676514, "pac": 0.7055011458182857, "whitened_cosine_id": 0.8639923334121704}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881815, "raptor_stability": 0.8507294654846191, "fragility": 1.0, "augmentation_robustness": 0.6945310589611032, "xie_feature_dispersion": 2.1999878871216194, "pac": 0.7726302622228611, "whitened_cosine_id": 0.876013994216919}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 2, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.26365195418554105, "raptor_stability": 0.8327406644821167, "fragility": 1.0, "augmentation_robustness": 0.7542952777812159, "xie_feature_dispersion": 6.585161526633954, "pac": 0.7935179711316662, "whitened_cosine_id": 0.8844253420829773}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 2, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.09434498750832387, "raptor_stability": 0.9433990716934204, "fragility": 3.0, "augmentation_robustness": 0.8439835278167953, "xie_feature_dispersion": 13.538073364650904, "pac": 0.8936912997551079, "whitened_cosine_id": 0.882512092590332}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 2, "layer": 21, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878156, "raptor_stability": 0.7554473280906677, "fragility": 0.25, "augmentation_robustness": 0.5795980822750073, "xie_feature_dispersion": 0.25302338379103195, "pac": 0.6675227051828375, "whitened_cosine_id": 0.7533318400382996}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 2, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.4286496963727474, "raptor_stability": 0.8515519499778748, "fragility": 0.25, "augmentation_robustness": 0.7217973601571539, "xie_feature_dispersion": 2.4117533415561345, "pac": 0.7866746550675143, "whitened_cosine_id": 0.8424076437950134}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 2, "layer": 10, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880234, "raptor_stability": 0.8253244757652283, "fragility": 0.25, "augmentation_robustness": 0.6525579225683095, "xie_feature_dispersion": 3.286387642880714, "pac": 0.7389411991667689, "whitened_cosine_id": 0.8348305821418762}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295687971, "raptor_stability": 0.7986416220664978, "fragility": 0.5, "augmentation_robustness": 0.6102543792217037, "xie_feature_dispersion": 1.8402433518812902, "pac": 0.7044480006441007, "whitened_cosine_id": 0.7969480156898499}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 2, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687956, "raptor_stability": 0.8041142225265503, "fragility": 0.25, "augmentation_robustness": 0.6348348918631781, "xie_feature_dispersion": 2.8843251781307306, "pac": 0.7194745571948642, "whitened_cosine_id": 0.8083853721618652}} +{"model": "pythia-410m", "dataset": "subj", "seed": 2, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882677, "raptor_stability": 0.8861598372459412, "fragility": 1.0, "augmentation_robustness": 0.7199515086420826, "xie_feature_dispersion": 1.5647630291131758, "pac": 0.8030556729440119, "whitened_cosine_id": 0.9003214240074158}} +{"model": "pythia-410m", "dataset": "spam", "seed": 2, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882948, "raptor_stability": 0.9486513733863831, "fragility": 3.0, "augmentation_robustness": 0.8147735901930376, "xie_feature_dispersion": 8.118654593687403, "pac": 0.8817124817897104, "whitened_cosine_id": 0.933245062828064}} +{"model": "pythia-410m", "dataset": "cola", "seed": 2, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879588, "raptor_stability": 0.801864504814148, "fragility": 0.25, "augmentation_robustness": 0.5568185846503796, "xie_feature_dispersion": 0.47412534702494263, "pac": 0.6793415447322637, "whitened_cosine_id": 0.8032734394073486}} +{"model": "pythia-410m", "dataset": "stance", "seed": 2, "layer": 10, "concept": "stance", "predictors": {"sip_eigengap": 0.47549258334344463, "raptor_stability": 0.8262801170349121, "fragility": 0.5, "augmentation_robustness": 0.7669381088312476, "xie_feature_dispersion": 3.2823490583418935, "pac": 0.7966091129330799, "whitened_cosine_id": 0.8195136189460754}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 2, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881449, "raptor_stability": 0.8327568769454956, "fragility": 0.5, "augmentation_robustness": 0.6987375826417322, "xie_feature_dispersion": 1.0417000979829214, "pac": 0.7657472297936139, "whitened_cosine_id": 0.864019513130188}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 0.855816496101061, "raptor_stability": 0.8567962050437927, "fragility": 0.25, "augmentation_robustness": 0.5671640624141537, "xie_feature_dispersion": 2.9934403270968586, "pac": 0.7119801337289733, "whitened_cosine_id": 0.8268300294876099}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010875, "raptor_stability": 0.8453766107559204, "fragility": 1.0, "augmentation_robustness": 0.6545753307454034, "xie_feature_dispersion": 2.1868653519527914, "pac": 0.7499759707506619, "whitened_cosine_id": 0.8455048203468323}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 2, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.25959190862405174, "raptor_stability": 0.8485456705093384, "fragility": 1.5, "augmentation_robustness": 0.7696896788348585, "xie_feature_dispersion": 6.11360094217724, "pac": 0.8091176746720985, "whitened_cosine_id": 0.82154780626297}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 2, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.054741897344356236, "raptor_stability": 0.9400308728218079, "fragility": 3.0, "augmentation_robustness": 0.8422214207799361, "xie_feature_dispersion": 25.830217797796784, "pac": 0.891126146800872, "whitened_cosine_id": 0.8212482333183289}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.7967193126678467, "fragility": 0.25, "augmentation_robustness": 0.6376161223358435, "xie_feature_dispersion": 0.32762786809964206, "pac": 0.7171677175018452, "whitened_cosine_id": 0.8186400532722473}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 2, "layer": 14, "concept": "emotion", "predictors": {"sip_eigengap": 0.3939989620104987, "raptor_stability": 0.8490006923675537, "fragility": 0.5, "augmentation_robustness": 0.6994232263413324, "xie_feature_dispersion": 2.8039065456029375, "pac": 0.774211959354443, "whitened_cosine_id": 0.8225671648979187}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 2, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010765, "raptor_stability": 0.8104034066200256, "fragility": 0.25, "augmentation_robustness": 0.6280345727884112, "xie_feature_dispersion": 3.5385168087713166, "pac": 0.7192189897042185, "whitened_cosine_id": 0.8256248831748962}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010356, "raptor_stability": 0.8042046427726746, "fragility": 0.25, "augmentation_robustness": 0.5798721253653247, "xie_feature_dispersion": 4.885251661924462, "pac": 0.6920383840689996, "whitened_cosine_id": 0.8246341943740845}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 2, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010591, "raptor_stability": 0.7971631288528442, "fragility": 0.25, "augmentation_robustness": 0.6109212963398408, "xie_feature_dispersion": 3.0958681822178455, "pac": 0.7040422125963426, "whitened_cosine_id": 0.825272798538208}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 2, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010877, "raptor_stability": 0.8769533038139343, "fragility": 1.0, "augmentation_robustness": 0.686316534527354, "xie_feature_dispersion": 2.781727808686068, "pac": 0.7816349191706442, "whitened_cosine_id": 0.8311170339584351}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 2, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010558, "raptor_stability": 0.9453728795051575, "fragility": 1.0, "augmentation_robustness": 0.820275837399458, "xie_feature_dispersion": 23.518955158553926, "pac": 0.8828243584523077, "whitened_cosine_id": 0.8724843859672546}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 2, "layer": 13, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010875, "raptor_stability": 0.8096633553504944, "fragility": 0.25, "augmentation_robustness": 0.4820399124790271, "xie_feature_dispersion": 0.6272978464384222, "pac": 0.6458516339147607, "whitened_cosine_id": 0.8209949731826782}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 2, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.33865901445727387, "raptor_stability": 0.8298296332359314, "fragility": 1.0, "augmentation_robustness": 0.7626508775947324, "xie_feature_dispersion": 3.230859359898638, "pac": 0.7962402554153318, "whitened_cosine_id": 0.8170763850212097}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 2, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010808, "raptor_stability": 0.8269712924957275, "fragility": 1.0, "augmentation_robustness": 0.6705085135623837, "xie_feature_dispersion": 1.7905194345731703, "pac": 0.7487399030290556, "whitened_cosine_id": 0.8267220854759216}} +{"model": "gpt2", "dataset": "sst2", "seed": 2, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349719, "raptor_stability": 0.8054147958755493, "fragility": 0.25, "augmentation_robustness": 0.452677434814117, "xie_feature_dispersion": 1.9950046355062208, "pac": 0.6290461153448331, "whitened_cosine_id": 0.8579275012016296}} +{"model": "gpt2", "dataset": "imdb", "seed": 2, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350525, "raptor_stability": 0.8147286772727966, "fragility": 0.25, "augmentation_robustness": 0.5420364534350757, "xie_feature_dispersion": 1.5795679944440126, "pac": 0.6783825653539362, "whitened_cosine_id": 0.8795053362846375}} +{"model": "gpt2", "dataset": "ag_news", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.37606835826916657, "raptor_stability": 0.831476092338562, "fragility": 1.0, "augmentation_robustness": 0.7160835130077355, "xie_feature_dispersion": 3.9948278981935426, "pac": 0.7737798026731488, "whitened_cosine_id": 0.8991539478302002}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 2, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.1066838897374052, "raptor_stability": 0.926025390625, "fragility": 1.5, "augmentation_robustness": 0.7998744401134587, "xie_feature_dispersion": 5.30333228141077, "pac": 0.8629499153692293, "whitened_cosine_id": 0.9026139974594116}} +{"model": "gpt2", "dataset": "counterfact", "seed": 2, "layer": 5, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342753, "raptor_stability": 0.7606139779090881, "fragility": 0.25, "augmentation_robustness": 0.6853486887621179, "xie_feature_dispersion": 0.33002840995944466, "pac": 0.722981333335603, "whitened_cosine_id": 0.73723304271698}} +{"model": "gpt2", "dataset": "emotion", "seed": 2, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5291604831105917, "raptor_stability": 0.8539204597473145, "fragility": 0.25, "augmentation_robustness": 0.6936184867435239, "xie_feature_dispersion": 2.010880210817857, "pac": 0.7737694732454192, "whitened_cosine_id": 0.852493941783905}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 2, "layer": 3, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859346874, "raptor_stability": 0.7885839939117432, "fragility": 0.25, "augmentation_robustness": 0.6401645091390288, "xie_feature_dispersion": 2.7009497954700477, "pac": 0.7143742515253859, "whitened_cosine_id": 0.8214506506919861}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 2, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859344792, "raptor_stability": 0.7899907827377319, "fragility": 0.25, "augmentation_robustness": 0.6988157056142623, "xie_feature_dispersion": 1.1126097391994054, "pac": 0.7444032441759971, "whitened_cosine_id": 0.7636452317237854}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 2, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346346, "raptor_stability": 0.8108178973197937, "fragility": 0.25, "augmentation_robustness": 0.7288817265620731, "xie_feature_dispersion": 0.9421790599719136, "pac": 0.7698498119409334, "whitened_cosine_id": 0.7841159701347351}} +{"model": "gpt2", "dataset": "subj", "seed": 2, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935236, "raptor_stability": 0.8722711205482483, "fragility": 0.5, "augmentation_robustness": 0.671483073623656, "xie_feature_dispersion": 1.095645923787708, "pac": 0.7718770970859521, "whitened_cosine_id": 0.9157454967498779}} +{"model": "gpt2", "dataset": "spam", "seed": 2, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935328, "raptor_stability": 0.9371232390403748, "fragility": 1.0, "augmentation_robustness": 0.796134201470319, "xie_feature_dispersion": 13.868407438115556, "pac": 0.8666287202553469, "whitened_cosine_id": 0.9425339698791504}} +{"model": "gpt2", "dataset": "cola", "seed": 2, "layer": 8, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344947, "raptor_stability": 0.7808783650398254, "fragility": 0.25, "augmentation_robustness": 0.4942423131561597, "xie_feature_dispersion": 0.38598825222094824, "pac": 0.6375603390979926, "whitened_cosine_id": 0.7848895788192749}} +{"model": "gpt2", "dataset": "stance", "seed": 2, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.5282860197072011, "raptor_stability": 0.8207774758338928, "fragility": 0.25, "augmentation_robustness": 0.7408748851031105, "xie_feature_dispersion": 2.2792793583875905, "pac": 0.7808261804685017, "whitened_cosine_id": 0.8131434321403503}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 2, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350183, "raptor_stability": 0.8160192370414734, "fragility": 0.25, "augmentation_robustness": 0.5989172462486304, "xie_feature_dispersion": 0.7330282513183846, "pac": 0.707468241645052, "whitened_cosine_id": 0.8756778240203857}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881522, "raptor_stability": 0.8280349969863892, "fragility": 0.25, "augmentation_robustness": 0.5057969177918578, "xie_feature_dispersion": 2.142674805730929, "pac": 0.6669159573891235, "whitened_cosine_id": 0.8551852107048035}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 2, "layer": 17, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882153, "raptor_stability": 0.8116974234580994, "fragility": 0.25, "augmentation_robustness": 0.5686588039011631, "xie_feature_dispersion": 1.607819045435618, "pac": 0.6901781136796312, "whitened_cosine_id": 0.8731050491333008}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 2, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.31591694292321826, "raptor_stability": 0.8379892706871033, "fragility": 1.0, "augmentation_robustness": 0.742913021052252, "xie_feature_dispersion": 4.204689835168313, "pac": 0.7904511458696777, "whitened_cosine_id": 0.8770087361335754}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 2, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.09406374121141926, "raptor_stability": 0.9260669350624084, "fragility": 1.5, "augmentation_robustness": 0.7954546313237745, "xie_feature_dispersion": 6.374717972535422, "pac": 0.8607607831930915, "whitened_cosine_id": 0.8749038577079773}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 2, "layer": 15, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878642, "raptor_stability": 0.7624801993370056, "fragility": 0.25, "augmentation_robustness": 0.6618500550845569, "xie_feature_dispersion": 0.3458533220193969, "pac": 0.7121651272107812, "whitened_cosine_id": 0.7618997693061829}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5517110732147434, "raptor_stability": 0.8649501204490662, "fragility": 0.25, "augmentation_robustness": 0.7080522584774469, "xie_feature_dispersion": 0.9242099717833849, "pac": 0.7865011894632565, "whitened_cosine_id": 0.8501837849617004}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 2, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879795, "raptor_stability": 0.8065077662467957, "fragility": 0.25, "augmentation_robustness": 0.6075362981901802, "xie_feature_dispersion": 2.637665840049803, "pac": 0.7070220322184879, "whitened_cosine_id": 0.8130406141281128}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 2, "layer": 24, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879593, "raptor_stability": 0.7929912805557251, "fragility": 0.25, "augmentation_robustness": 0.687746633761177, "xie_feature_dispersion": 1.7178207547218787, "pac": 0.7403689571584511, "whitened_cosine_id": 0.8002586364746094}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 2, "layer": 10, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879999, "raptor_stability": 0.7896746397018433, "fragility": 0.25, "augmentation_robustness": 0.6055377963859948, "xie_feature_dispersion": 2.184985172419946, "pac": 0.6976062180439191, "whitened_cosine_id": 0.8094580769538879}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 2, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882808, "raptor_stability": 0.8819090127944946, "fragility": 0.5, "augmentation_robustness": 0.671921924564452, "xie_feature_dispersion": 1.4175250970378874, "pac": 0.7769154686794733, "whitened_cosine_id": 0.8918017148971558}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 2, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956883163, "raptor_stability": 0.9394904971122742, "fragility": 1.0, "augmentation_robustness": 0.789939038367017, "xie_feature_dispersion": 13.48331423128814, "pac": 0.8647147677396456, "whitened_cosine_id": 0.917801022529602}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 2, "layer": 20, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880179, "raptor_stability": 0.7839298248291016, "fragility": 0.25, "augmentation_robustness": 0.5005989571125445, "xie_feature_dispersion": 0.4452074705477342, "pac": 0.642264390970823, "whitened_cosine_id": 0.8129088282585144}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 2, "layer": 19, "concept": "stance", "predictors": {"sip_eigengap": 0.42059041720080603, "raptor_stability": 0.8081520199775696, "fragility": 0.5, "augmentation_robustness": 0.7271794999740994, "xie_feature_dispersion": 2.1040215143183647, "pac": 0.7676657599758345, "whitened_cosine_id": 0.7815761566162109}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 2, "layer": 11, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881302, "raptor_stability": 0.800855815410614, "fragility": 0.25, "augmentation_robustness": 0.6448667243445377, "xie_feature_dispersion": 0.9884816369621833, "pac": 0.7228612698775758, "whitened_cosine_id": 0.8472304344177246}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237649451, "raptor_stability": 0.848286509513855, "fragility": 0.25, "augmentation_robustness": 0.5634090093483565, "xie_feature_dispersion": 2.429164692187356, "pac": 0.7058477594311057, "whitened_cosine_id": 0.8725934624671936}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923765047, "raptor_stability": 0.8775961995124817, "fragility": 1.0, "augmentation_robustness": 0.774986382310033, "xie_feature_dispersion": 2.073190764464318, "pac": 0.8262912909112574, "whitened_cosine_id": 0.8976234793663025}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 2, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.39588586732534015, "raptor_stability": 0.8559584021568298, "fragility": 1.0, "augmentation_robustness": 0.7820646549848398, "xie_feature_dispersion": 3.6437178512980712, "pac": 0.8190115285708348, "whitened_cosine_id": 0.9048267006874084}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 2, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.058756522874283214, "raptor_stability": 0.9501736760139465, "fragility": 1.5, "augmentation_robustness": 0.8683291628685149, "xie_feature_dispersion": 6.332238643735282, "pac": 0.9092514194412307, "whitened_cosine_id": 0.9210717082023621}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 2, "layer": 14, "concept": "truth", "predictors": {"sip_eigengap": 1.29387292376465, "raptor_stability": 0.8335102796554565, "fragility": 0.25, "augmentation_robustness": 0.7324644634343902, "xie_feature_dispersion": 0.2367801772803424, "pac": 0.7829873715449234, "whitened_cosine_id": 0.7776349186897278}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 2, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.5021817224915929, "raptor_stability": 0.8779242038726807, "fragility": 0.25, "augmentation_robustness": 0.8391776121680871, "xie_feature_dispersion": 2.253764943264769, "pac": 0.8585509080203839, "whitened_cosine_id": 0.848435640335083}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 2, "layer": 7, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237646702, "raptor_stability": 0.8149078488349915, "fragility": 0.25, "augmentation_robustness": 0.6656494198846684, "xie_feature_dispersion": 3.241751193008559, "pac": 0.74027863435983, "whitened_cosine_id": 0.8149601221084595}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 2, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237645712, "raptor_stability": 0.8159747123718262, "fragility": 0.25, "augmentation_robustness": 0.6734582597657116, "xie_feature_dispersion": 3.9245245325964895, "pac": 0.744716486068769, "whitened_cosine_id": 0.7945672273635864}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647373, "raptor_stability": 0.8471121788024902, "fragility": 1.5, "augmentation_robustness": 0.7818095671541662, "xie_feature_dispersion": 2.234140033744398, "pac": 0.8144608729783283, "whitened_cosine_id": 0.8059022426605225}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 2, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651356, "raptor_stability": 0.8970006108283997, "fragility": 0.5, "augmentation_robustness": 0.7567696692336531, "xie_feature_dispersion": 1.2621909015930055, "pac": 0.8268851400310264, "whitened_cosine_id": 0.9196778535842896}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 2, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237647655, "raptor_stability": 0.9248191714286804, "fragility": 1.0, "augmentation_robustness": 0.82350992677322, "xie_feature_dispersion": 18.125591348203603, "pac": 0.8741645491009502, "whitened_cosine_id": 0.9478391408920288}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 2, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647155, "raptor_stability": 0.7994158864021301, "fragility": 0.25, "augmentation_robustness": 0.5431886973103349, "xie_feature_dispersion": 0.28432650457876973, "pac": 0.6713022918562326, "whitened_cosine_id": 0.8063071370124817}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 2, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.47687488644054804, "raptor_stability": 0.8371041417121887, "fragility": 0.25, "augmentation_robustness": 0.797300656029293, "xie_feature_dispersion": 2.6985372907022853, "pac": 0.8172023988707409, "whitened_cosine_id": 0.8233371376991272}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 2, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764932, "raptor_stability": 0.8735290765762329, "fragility": 0.25, "augmentation_robustness": 0.7483144226433699, "xie_feature_dispersion": 0.37743185068024515, "pac": 0.8109217496098013, "whitened_cosine_id": 0.8706146478652954}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996696, "raptor_stability": 0.8488397598266602, "fragility": 0.25, "augmentation_robustness": 0.5179823378479093, "xie_feature_dispersion": 2.9434633106744124, "pac": 0.6834110488372847, "whitened_cosine_id": 0.8510316610336304}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 3, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921997957, "raptor_stability": 0.9214267730712891, "fragility": 1.0, "augmentation_robustness": 0.8748364094526995, "xie_feature_dispersion": 1.0370053025918007, "pac": 0.8981315912619943, "whitened_cosine_id": 0.8452136516571045}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 3, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5774925151839428, "raptor_stability": 0.8593819737434387, "fragility": 1.5, "augmentation_robustness": 0.7534819327566528, "xie_feature_dispersion": 4.6426103470172615, "pac": 0.8064319532500457, "whitened_cosine_id": 0.914084255695343}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 3, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.13231466369730585, "raptor_stability": 0.9453434348106384, "fragility": 2.0, "augmentation_robustness": 0.8580053912222981, "xie_feature_dispersion": 7.46273693031987, "pac": 0.9016744130164682, "whitened_cosine_id": 0.930860161781311}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 3, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921978637, "raptor_stability": 0.7810012698173523, "fragility": 1.0, "augmentation_robustness": 0.7041587287479656, "xie_feature_dispersion": 0.5352386761634802, "pac": 0.7425799992826589, "whitened_cosine_id": 0.6924300193786621}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.7155300362438014, "raptor_stability": 0.8798012733459473, "fragility": 1.0, "augmentation_robustness": 0.7953262771468999, "xie_feature_dispersion": 1.4738735759860853, "pac": 0.8375637752464236, "whitened_cosine_id": 0.8543421626091003}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.711632992199115, "raptor_stability": 0.8559250831604004, "fragility": 1.0, "augmentation_robustness": 0.7877299458146467, "xie_feature_dispersion": 1.5138918067092333, "pac": 0.8218275144875236, "whitened_cosine_id": 0.7800295948982239}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 3, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921988689, "raptor_stability": 0.8289263248443604, "fragility": 1.0, "augmentation_robustness": 0.6909281594330826, "xie_feature_dispersion": 1.8985337702191092, "pac": 0.7599272421387215, "whitened_cosine_id": 0.7646318078041077}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 3, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.711632992198979, "raptor_stability": 0.8319928050041199, "fragility": 1.0, "augmentation_robustness": 0.7511852191768399, "xie_feature_dispersion": 1.1045209668465537, "pac": 0.7915890120904798, "whitened_cosine_id": 0.7830063700675964}} +{"model": "pythia-70m", "dataset": "subj", "seed": 3, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002202, "raptor_stability": 0.9258168935775757, "fragility": 1.5, "augmentation_robustness": 0.8378077411249061, "xie_feature_dispersion": 2.402632179496979, "pac": 0.8818123173512409, "whitened_cosine_id": 0.9089578986167908}} +{"model": "pythia-70m", "dataset": "spam", "seed": 3, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004663, "raptor_stability": 0.9411624073982239, "fragility": 3.0, "augmentation_robustness": 0.8491174045572158, "xie_feature_dispersion": 8.333983109527576, "pac": 0.8951399059777199, "whitened_cosine_id": 0.9507425427436829}} +{"model": "pythia-70m", "dataset": "cola", "seed": 3, "layer": 4, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921976148, "raptor_stability": 0.7598171830177307, "fragility": 0.25, "augmentation_robustness": 0.5844171391849297, "xie_feature_dispersion": 0.1566522503868137, "pac": 0.6721171611013301, "whitened_cosine_id": 0.6859370470046997}} +{"model": "pythia-70m", "dataset": "stance", "seed": 3, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.764248302629601, "raptor_stability": 0.8229038119316101, "fragility": 0.5, "augmentation_robustness": 0.7459009481197789, "xie_feature_dispersion": 2.080304032697393, "pac": 0.7844023800256945, "whitened_cosine_id": 0.8263651728630066}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 3, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.711632992199843, "raptor_stability": 0.8560130596160889, "fragility": 1.0, "augmentation_robustness": 0.6862800186116383, "xie_feature_dispersion": 1.5912468339045849, "pac": 0.7711465391138637, "whitened_cosine_id": 0.8750572800636292}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 3, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349443, "raptor_stability": 0.8484313488006592, "fragility": 0.25, "augmentation_robustness": 0.5622190281630771, "xie_feature_dispersion": 4.044131456746566, "pac": 0.7053251884818681, "whitened_cosine_id": 0.8518450260162354}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350076, "raptor_stability": 0.871891438961029, "fragility": 1.0, "augmentation_robustness": 0.7693997298355439, "xie_feature_dispersion": 1.4704300817182188, "pac": 0.8206455843982865, "whitened_cosine_id": 0.8666113018989563}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 3, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.36025022356685565, "raptor_stability": 0.8447644710540771, "fragility": 1.5, "augmentation_robustness": 0.7221689699035224, "xie_feature_dispersion": 6.085221682949123, "pac": 0.7834667204787997, "whitened_cosine_id": 0.9049122333526611}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 3, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10505729790354523, "raptor_stability": 0.9525120854377747, "fragility": 1.5, "augmentation_robustness": 0.8703692378009392, "xie_feature_dispersion": 5.186371589560047, "pac": 0.9114406616193569, "whitened_cosine_id": 0.9208746552467346}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342074, "raptor_stability": 0.78183913230896, "fragility": 0.5, "augmentation_robustness": 0.6746312909390834, "xie_feature_dispersion": 0.5886846655394216, "pac": 0.7282352116240216, "whitened_cosine_id": 0.7248976230621338}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 3, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.6992640027033808, "raptor_stability": 0.8696766495704651, "fragility": 0.5, "augmentation_robustness": 0.7326524720518076, "xie_feature_dispersion": 1.5546891203751845, "pac": 0.8011645608111364, "whitened_cosine_id": 0.8558102250099182}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 3, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347303, "raptor_stability": 0.8296602964401245, "fragility": 0.25, "augmentation_robustness": 0.6793747577236712, "xie_feature_dispersion": 2.1630123007277384, "pac": 0.7545175270818978, "whitened_cosine_id": 0.8202237486839294}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 3, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346095, "raptor_stability": 0.8123674988746643, "fragility": 0.5, "augmentation_robustness": 0.6624179901863742, "xie_feature_dispersion": 1.81846361791661, "pac": 0.7373927445305193, "whitened_cosine_id": 0.7944690585136414}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 3, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346586, "raptor_stability": 0.8003823757171631, "fragility": 0.25, "augmentation_robustness": 0.6676009627104235, "xie_feature_dispersion": 0.7182211892213498, "pac": 0.7339916692137933, "whitened_cosine_id": 0.7915064692497253}} +{"model": "pythia-160m", "dataset": "subj", "seed": 3, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352172, "raptor_stability": 0.8996919989585876, "fragility": 1.0, "augmentation_robustness": 0.7410086684825309, "xie_feature_dispersion": 1.9890978207832326, "pac": 0.8203503337205593, "whitened_cosine_id": 0.9151029586791992}} +{"model": "pythia-160m", "dataset": "spam", "seed": 3, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859350205, "raptor_stability": 0.9280425310134888, "fragility": 1.0, "augmentation_robustness": 0.8124903597056025, "xie_feature_dispersion": 13.64857545430944, "pac": 0.8702664453595457, "whitened_cosine_id": 0.9369125366210938}} +{"model": "pythia-160m", "dataset": "cola", "seed": 3, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343062, "raptor_stability": 0.7884514331817627, "fragility": 0.25, "augmentation_robustness": 0.5611738456088626, "xie_feature_dispersion": 0.17857729363649755, "pac": 0.6748126393953127, "whitened_cosine_id": 0.7649105191230774}} +{"model": "pythia-160m", "dataset": "stance", "seed": 3, "layer": 4, "concept": "stance", "predictors": {"sip_eigengap": 0.6506681129814839, "raptor_stability": 0.8327218890190125, "fragility": 0.5, "augmentation_robustness": 0.7643668163368509, "xie_feature_dispersion": 1.7873782154455453, "pac": 0.7985443526779317, "whitened_cosine_id": 0.8258171081542969}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 3, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350412, "raptor_stability": 0.8567770719528198, "fragility": 0.5, "augmentation_robustness": 0.690159632122668, "xie_feature_dispersion": 1.305808488628754, "pac": 0.7734683520377439, "whitened_cosine_id": 0.8785260319709778}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 3, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881342, "raptor_stability": 0.8563058376312256, "fragility": 0.25, "augmentation_robustness": 0.5454463736012634, "xie_feature_dispersion": 4.914029083398948, "pac": 0.7008761056162445, "whitened_cosine_id": 0.8632848858833313}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688189, "raptor_stability": 0.8555104732513428, "fragility": 1.0, "augmentation_robustness": 0.7200850357383629, "xie_feature_dispersion": 2.288214053878471, "pac": 0.7877977544948529, "whitened_cosine_id": 0.8782402873039246}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 3, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.3288136919930267, "raptor_stability": 0.8414013981819153, "fragility": 2.0, "augmentation_robustness": 0.7266437337043424, "xie_feature_dispersion": 12.434412392508191, "pac": 0.7840225659431288, "whitened_cosine_id": 0.8793582320213318}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 3, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.0776609124998989, "raptor_stability": 0.9545353055000305, "fragility": 2.0, "augmentation_robustness": 0.8723893129234274, "xie_feature_dispersion": 7.487139994618903, "pac": 0.9134623092117289, "whitened_cosine_id": 0.8982532620429993}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 3, "layer": 14, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879506, "raptor_stability": 0.8133618831634521, "fragility": 0.25, "augmentation_robustness": 0.6928116161487408, "xie_feature_dispersion": 0.25094480473639263, "pac": 0.7530867496560965, "whitened_cosine_id": 0.7940764427185059}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 3, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.647965606663094, "raptor_stability": 0.8681758046150208, "fragility": 0.5, "augmentation_robustness": 0.7102638120672203, "xie_feature_dispersion": 1.6253563675189826, "pac": 0.7892198083411206, "whitened_cosine_id": 0.8482173085212708}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 3, "layer": 2, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295687982, "raptor_stability": 0.8303574919700623, "fragility": 1.0, "augmentation_robustness": 0.6867166191679835, "xie_feature_dispersion": 1.7750280665756264, "pac": 0.758537055569023, "whitened_cosine_id": 0.8150797486305237}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 3, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879206, "raptor_stability": 0.8025309443473816, "fragility": 0.25, "augmentation_robustness": 0.6179124255783944, "xie_feature_dispersion": 4.2516203791362885, "pac": 0.710221684962888, "whitened_cosine_id": 0.8091989755630493}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 3, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879957, "raptor_stability": 0.8084143996238708, "fragility": 0.25, "augmentation_robustness": 0.6400424522618103, "xie_feature_dispersion": 0.98880123212057, "pac": 0.7242284259428406, "whitened_cosine_id": 0.8083922863006592}} +{"model": "pythia-410m", "dataset": "subj", "seed": 3, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882546, "raptor_stability": 0.9100797772407532, "fragility": 1.0, "augmentation_robustness": 0.7426358153217486, "xie_feature_dispersion": 1.4905914800590063, "pac": 0.826357796281251, "whitened_cosine_id": 0.8978877663612366}} +{"model": "pythia-410m", "dataset": "spam", "seed": 3, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882908, "raptor_stability": 0.9304795265197754, "fragility": 3.0, "augmentation_robustness": 0.8261601242245535, "xie_feature_dispersion": 8.165725328081457, "pac": 0.8783198253721645, "whitened_cosine_id": 0.9207106232643127}} +{"model": "pythia-410m", "dataset": "cola", "seed": 3, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.210307295687961, "raptor_stability": 0.7938510775566101, "fragility": 0.25, "augmentation_robustness": 0.5212292454913044, "xie_feature_dispersion": 0.90277616899984, "pac": 0.6575401615239573, "whitened_cosine_id": 0.8084408640861511}} +{"model": "pythia-410m", "dataset": "stance", "seed": 3, "layer": 17, "concept": "stance", "predictors": {"sip_eigengap": 0.4936540695216237, "raptor_stability": 0.8370916247367859, "fragility": 0.5, "augmentation_robustness": 0.7826824738040318, "xie_feature_dispersion": 2.485399050185102, "pac": 0.8098870492704089, "whitened_cosine_id": 0.8208258748054504}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 3, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881564, "raptor_stability": 0.8418619632720947, "fragility": 0.5, "augmentation_robustness": 0.6628138222662973, "xie_feature_dispersion": 1.7478308880725917, "pac": 0.7523378927691959, "whitened_cosine_id": 0.8692632913589478}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 3, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010637, "raptor_stability": 0.8572840690612793, "fragility": 0.25, "augmentation_robustness": 0.5195046452531485, "xie_feature_dispersion": 5.520447159983482, "pac": 0.688394357157214, "whitened_cosine_id": 0.8275399804115295}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 3, "layer": 20, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010867, "raptor_stability": 0.8348312973976135, "fragility": 0.5, "augmentation_robustness": 0.6270832143777221, "xie_feature_dispersion": 1.9157864290958875, "pac": 0.7309572558876678, "whitened_cosine_id": 0.8293268084526062}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 3, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.20484503848029767, "raptor_stability": 0.85955810546875, "fragility": 1.5, "augmentation_robustness": 0.7594005762230035, "xie_feature_dispersion": 5.718643787688656, "pac": 0.8094793408458767, "whitened_cosine_id": 0.828823447227478}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 3, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05273817789404076, "raptor_stability": 0.9474886655807495, "fragility": 3.0, "augmentation_robustness": 0.8532070565094912, "xie_feature_dispersion": 9.58314422780073, "pac": 0.9003478610451203, "whitened_cosine_id": 0.832189679145813}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 3, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010883, "raptor_stability": 0.8169371485710144, "fragility": 0.25, "augmentation_robustness": 0.6690915350986811, "xie_feature_dispersion": 0.42029681121277923, "pac": 0.7430143418348478, "whitened_cosine_id": 0.8196472525596619}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 3, "layer": 9, "concept": "emotion", "predictors": {"sip_eigengap": 0.473623796458844, "raptor_stability": 0.8568776249885559, "fragility": 0.25, "augmentation_robustness": 0.6824483342515307, "xie_feature_dispersion": 2.6608823301213778, "pac": 0.7696629796200434, "whitened_cosine_id": 0.83025723695755}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 3, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010806, "raptor_stability": 0.8051812052726746, "fragility": 0.25, "augmentation_robustness": 0.6006977265762596, "xie_feature_dispersion": 4.0725358408621375, "pac": 0.7029394659244671, "whitened_cosine_id": 0.8263804912567139}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 3, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010875, "raptor_stability": 0.8060323596000671, "fragility": 0.5, "augmentation_robustness": 0.6189902560303079, "xie_feature_dispersion": 2.2388020889990288, "pac": 0.7125113078151875, "whitened_cosine_id": 0.8258398771286011}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 3, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010296, "raptor_stability": 0.7851132154464722, "fragility": 0.25, "augmentation_robustness": 0.5954573958337406, "xie_feature_dispersion": 1.0772966237421127, "pac": 0.6902853056401064, "whitened_cosine_id": 0.8228062987327576}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 3, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.8939151167869568, "fragility": 1.0, "augmentation_robustness": 0.7047598539681612, "xie_feature_dispersion": 2.9162655683918848, "pac": 0.799337485377559, "whitened_cosine_id": 0.8292585611343384}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 3, "layer": 3, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010838, "raptor_stability": 0.925938069820404, "fragility": 3.0, "augmentation_robustness": 0.8210687839915044, "xie_feature_dispersion": 10.71642677493898, "pac": 0.8735034269059543, "whitened_cosine_id": 0.8716909289360046}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 3, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010886, "raptor_stability": 0.7957292199134827, "fragility": 0.25, "augmentation_robustness": 0.45052250604191046, "xie_feature_dispersion": 0.7106391233544392, "pac": 0.6231258629776966, "whitened_cosine_id": 0.8221802115440369}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 3, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.37531177284512046, "raptor_stability": 0.8395811915397644, "fragility": 1.0, "augmentation_robustness": 0.7703567449624081, "xie_feature_dispersion": 2.787120846289539, "pac": 0.8049689682510863, "whitened_cosine_id": 0.8196287155151367}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 3, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.855816496101079, "raptor_stability": 0.8419731855392456, "fragility": 1.0, "augmentation_robustness": 0.6515139260319379, "xie_feature_dispersion": 2.265789759921114, "pac": 0.7467435557855917, "whitened_cosine_id": 0.8332340121269226}} +{"model": "gpt2", "dataset": "sst2", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934796, "raptor_stability": 0.8351662158966064, "fragility": 0.25, "augmentation_robustness": 0.46886485771385517, "xie_feature_dispersion": 4.2622215814473865, "pac": 0.6520155368052308, "whitened_cosine_id": 0.8634011149406433}} +{"model": "gpt2", "dataset": "imdb", "seed": 3, "layer": 8, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350514, "raptor_stability": 0.8268616795539856, "fragility": 0.5, "augmentation_robustness": 0.6116287845851899, "xie_feature_dispersion": 1.47086614284479, "pac": 0.7192452320695877, "whitened_cosine_id": 0.876645565032959}} +{"model": "gpt2", "dataset": "ag_news", "seed": 3, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.37970761633998423, "raptor_stability": 0.8910131454467773, "fragility": 0.25, "augmentation_robustness": 0.8581536528187313, "xie_feature_dispersion": 0.8261918893870008, "pac": 0.8745833991327543, "whitened_cosine_id": 0.9110751152038574}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 3, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.10991110313414279, "raptor_stability": 0.9330034255981445, "fragility": 1.5, "augmentation_robustness": 0.8025370623665369, "xie_feature_dispersion": 5.468890231915903, "pac": 0.8677702439823407, "whitened_cosine_id": 0.9063879251480103}} +{"model": "gpt2", "dataset": "counterfact", "seed": 3, "layer": 7, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343042, "raptor_stability": 0.7817078232765198, "fragility": 0.25, "augmentation_robustness": 0.6861783480825473, "xie_feature_dispersion": 0.1934064316642644, "pac": 0.7339430856795335, "whitened_cosine_id": 0.7454338073730469}} +{"model": "gpt2", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6182210878004821, "raptor_stability": 0.8550314903259277, "fragility": 0.25, "augmentation_robustness": 0.6785777254477671, "xie_feature_dispersion": 1.1654170019303518, "pac": 0.7668046078868473, "whitened_cosine_id": 0.8550549149513245}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 3, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347265, "raptor_stability": 0.8183044791221619, "fragility": 0.25, "augmentation_robustness": 0.7349596865260767, "xie_feature_dispersion": 0.7005527274806919, "pac": 0.7766320828241193, "whitened_cosine_id": 0.8098105788230896}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 3, "layer": 7, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346741, "raptor_stability": 0.7654048204421997, "fragility": 0.25, "augmentation_robustness": 0.5347742087458891, "xie_feature_dispersion": 3.241040228671077, "pac": 0.6500895145940444, "whitened_cosine_id": 0.8011150360107422}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 3, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346404, "raptor_stability": 0.7987147569656372, "fragility": 0.25, "augmentation_robustness": 0.7470243212681252, "xie_feature_dispersion": 0.9553360233273932, "pac": 0.7728695391168812, "whitened_cosine_id": 0.7885229587554932}} +{"model": "gpt2", "dataset": "subj", "seed": 3, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352408, "raptor_stability": 0.8722814321517944, "fragility": 0.5, "augmentation_robustness": 0.645032578503623, "xie_feature_dispersion": 1.9236159050902961, "pac": 0.7586570053277087, "whitened_cosine_id": 0.9171640276908875}} +{"model": "gpt2", "dataset": "spam", "seed": 3, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593528, "raptor_stability": 0.9187144041061401, "fragility": 1.5, "augmentation_robustness": 0.812376554161553, "xie_feature_dispersion": 12.08932579677981, "pac": 0.8655454791338466, "whitened_cosine_id": 0.9336515665054321}} +{"model": "gpt2", "dataset": "cola", "seed": 3, "layer": 8, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344077, "raptor_stability": 0.7706639766693115, "fragility": 0.25, "augmentation_robustness": 0.5024593786840387, "xie_feature_dispersion": 0.689280134170416, "pac": 0.6365616776766752, "whitened_cosine_id": 0.782214343547821}} +{"model": "gpt2", "dataset": "stance", "seed": 3, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.6228451679003366, "raptor_stability": 0.8234939575195312, "fragility": 0.5, "augmentation_robustness": 0.7234382249579733, "xie_feature_dispersion": 1.7228146499923553, "pac": 0.7734660912387523, "whitened_cosine_id": 0.8058009147644043}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 3, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935006, "raptor_stability": 0.8277425765991211, "fragility": 0.25, "augmentation_robustness": 0.6110880945528435, "xie_feature_dispersion": 1.2604141020751956, "pac": 0.7194153355759823, "whitened_cosine_id": 0.8723833560943604}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 3, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881538, "raptor_stability": 0.8475278615951538, "fragility": 0.25, "augmentation_robustness": 0.5298442019954345, "xie_feature_dispersion": 4.500703580339019, "pac": 0.6886860317952941, "whitened_cosine_id": 0.8626024723052979}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 3, "layer": 24, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881826, "raptor_stability": 0.896838903427124, "fragility": 0.25, "augmentation_robustness": 0.6259364774466486, "xie_feature_dispersion": 0.14721586133397682, "pac": 0.7613876904368864, "whitened_cosine_id": 0.8744872808456421}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 3, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.27417661774965685, "raptor_stability": 0.8498548269271851, "fragility": 1.0, "augmentation_robustness": 0.7432784283933686, "xie_feature_dispersion": 2.8627715211803557, "pac": 0.7965666276602767, "whitened_cosine_id": 0.8807106018066406}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 3, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.07699019521608971, "raptor_stability": 0.9369117021560669, "fragility": 1.5, "augmentation_robustness": 0.8223812079136419, "xie_feature_dispersion": 6.380122368170902, "pac": 0.8796464550348544, "whitened_cosine_id": 0.877031683921814}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 3, "layer": 20, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878205, "raptor_stability": 0.765117883682251, "fragility": 0.25, "augmentation_robustness": 0.6346673610552594, "xie_feature_dispersion": 0.2291144061643789, "pac": 0.6998926223687552, "whitened_cosine_id": 0.7576854825019836}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6234881751702148, "raptor_stability": 0.8669646382331848, "fragility": 0.25, "augmentation_robustness": 0.7006085301347155, "xie_feature_dispersion": 0.8652954065662566, "pac": 0.7837865841839502, "whitened_cosine_id": 0.8504673838615417}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 3, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880274, "raptor_stability": 0.813629686832428, "fragility": 0.25, "augmentation_robustness": 0.7004445795324097, "xie_feature_dispersion": 1.0577105967646572, "pac": 0.7570371331824188, "whitened_cosine_id": 0.8194169998168945}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 3, "layer": 16, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956880063, "raptor_stability": 0.7668350338935852, "fragility": 0.25, "augmentation_robustness": 0.5387038041591353, "xie_feature_dispersion": 3.0633639663829166, "pac": 0.6527694190263602, "whitened_cosine_id": 0.8101463317871094}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 3, "layer": 24, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687966, "raptor_stability": 0.7879273891448975, "fragility": 0.25, "augmentation_robustness": 0.7205035666208062, "xie_feature_dispersion": 0.933971243574555, "pac": 0.7542154778828518, "whitened_cosine_id": 0.7968817353248596}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 3, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882648, "raptor_stability": 0.8908157348632812, "fragility": 0.5, "augmentation_robustness": 0.6905186912348897, "xie_feature_dispersion": 1.3649873458587798, "pac": 0.7906672130490855, "whitened_cosine_id": 0.8880553841590881}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 3, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688045, "raptor_stability": 0.9237306714057922, "fragility": 1.0, "augmentation_robustness": 0.8008774808792517, "xie_feature_dispersion": 14.196302533817459, "pac": 0.862304076142522, "whitened_cosine_id": 0.9053286910057068}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 3, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879595, "raptor_stability": 0.786989688873291, "fragility": 0.25, "augmentation_robustness": 0.4592027295956805, "xie_feature_dispersion": 0.6920920489375771, "pac": 0.6230962092344857, "whitened_cosine_id": 0.8074944615364075}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 3, "layer": 16, "concept": "stance", "predictors": {"sip_eigengap": 0.4985627032597986, "raptor_stability": 0.8289973139762878, "fragility": 0.5, "augmentation_robustness": 0.7594746024395039, "xie_feature_dispersion": 1.849522827224973, "pac": 0.7942359582078959, "whitened_cosine_id": 0.7846319079399109}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 3, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881336, "raptor_stability": 0.8215765357017517, "fragility": 0.25, "augmentation_robustness": 0.5954299844747202, "xie_feature_dispersion": 1.2573193221997883, "pac": 0.7085032600882359, "whitened_cosine_id": 0.8571715950965881}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 3, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237647006, "raptor_stability": 0.8808207511901855, "fragility": 0.25, "augmentation_robustness": 0.6288068642959735, "xie_feature_dispersion": 5.001977030703906, "pac": 0.7548138077430795, "whitened_cosine_id": 0.8865893483161926}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650275, "raptor_stability": 0.8786996006965637, "fragility": 1.0, "augmentation_robustness": 0.7573141076406408, "xie_feature_dispersion": 2.4507641389974713, "pac": 0.8180068541686023, "whitened_cosine_id": 0.8870015740394592}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 3, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.27907133932759765, "raptor_stability": 0.8452746272087097, "fragility": 1.5, "augmentation_robustness": 0.7357258977124443, "xie_feature_dispersion": 8.01989836662422, "pac": 0.790500262460577, "whitened_cosine_id": 0.8956002593040466}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 3, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.07227979648704244, "raptor_stability": 0.9585252404212952, "fragility": 1.0, "augmentation_robustness": 0.8819010246269907, "xie_feature_dispersion": 6.547993580545896, "pac": 0.9202131325241429, "whitened_cosine_id": 0.9235095977783203}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 3, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646929, "raptor_stability": 0.8464949131011963, "fragility": 0.25, "augmentation_robustness": 0.7631963084432197, "xie_feature_dispersion": 0.07258280090844692, "pac": 0.804845610772208, "whitened_cosine_id": 0.8042153120040894}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 3, "layer": 11, "concept": "emotion", "predictors": {"sip_eigengap": 0.5364682673815797, "raptor_stability": 0.8717088103294373, "fragility": 0.25, "augmentation_robustness": 0.8095188795391789, "xie_feature_dispersion": 1.9155672487876931, "pac": 0.8406138449343081, "whitened_cosine_id": 0.8489202857017517}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 3, "layer": 6, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237646567, "raptor_stability": 0.8316985368728638, "fragility": 0.25, "augmentation_robustness": 0.6779121169500498, "xie_feature_dispersion": 4.322967748214793, "pac": 0.7548053269114567, "whitened_cosine_id": 0.8318672776222229}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 3, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237646929, "raptor_stability": 0.8263404369354248, "fragility": 1.0, "augmentation_robustness": 0.7174827260397666, "xie_feature_dispersion": 2.0942342832154073, "pac": 0.7719115814875956, "whitened_cosine_id": 0.791140615940094}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 3, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.293872923764742, "raptor_stability": 0.8533427119255066, "fragility": 2.0, "augmentation_robustness": 0.8330055291842721, "xie_feature_dispersion": 2.823288787076928, "pac": 0.8431741205548893, "whitened_cosine_id": 0.7997245192527771}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 3, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651214, "raptor_stability": 0.900323212146759, "fragility": 0.5, "augmentation_robustness": 0.7438287558717409, "xie_feature_dispersion": 1.6449955999096242, "pac": 0.82207598400925, "whitened_cosine_id": 0.9154244661331177}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 3, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237634952, "raptor_stability": 0.9340345859527588, "fragility": 0.5, "augmentation_robustness": 0.8432807256360871, "xie_feature_dispersion": 18.65723826433361, "pac": 0.888657655794423, "whitened_cosine_id": 0.9443875551223755}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 3, "layer": 14, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646847, "raptor_stability": 0.8058444857597351, "fragility": 0.25, "augmentation_robustness": 0.5270067886058506, "xie_feature_dispersion": 0.7691357194942805, "pac": 0.6664256371827928, "whitened_cosine_id": 0.8194195032119751}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 3, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.5814899203429595, "raptor_stability": 0.8490622639656067, "fragility": 0.25, "augmentation_robustness": 0.8046175979185222, "xie_feature_dispersion": 2.1435570572895988, "pac": 0.8268399309420644, "whitened_cosine_id": 0.8264880776405334}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 3, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764957, "raptor_stability": 0.8473092317581177, "fragility": 0.25, "augmentation_robustness": 0.6972751540596228, "xie_feature_dispersion": 1.2429462131243278, "pac": 0.7722921929088702, "whitened_cosine_id": 0.8770169615745544}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 4, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.711632992199527, "raptor_stability": 0.8324896097183228, "fragility": 0.25, "augmentation_robustness": 0.5885938862832164, "xie_feature_dispersion": 1.3807377703846688, "pac": 0.7105417480007696, "whitened_cosine_id": 0.8337145447731018}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 4, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.711632992199886, "raptor_stability": 0.9050891995429993, "fragility": 1.0, "augmentation_robustness": 0.8489347430216386, "xie_feature_dispersion": 1.2126650382704005, "pac": 0.877011971282319, "whitened_cosine_id": 0.850266695022583}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 4, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6455833206335958, "raptor_stability": 0.9030647277832031, "fragility": 1.5, "augmentation_robustness": 0.8698773067049781, "xie_feature_dispersion": 3.394769818942791, "pac": 0.8864710172440906, "whitened_cosine_id": 0.9047799110412598}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.1519243074719573, "raptor_stability": 0.9499274492263794, "fragility": 1.5, "augmentation_robustness": 0.874702657535151, "xie_feature_dispersion": 5.282104233636098, "pac": 0.9123150533807651, "whitened_cosine_id": 0.9344291687011719}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 4, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.711632992197464, "raptor_stability": 0.7690514922142029, "fragility": 0.5, "augmentation_robustness": 0.646014274798961, "xie_feature_dispersion": 0.46179379874179, "pac": 0.7075328835065819, "whitened_cosine_id": 0.6736164689064026}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6855318380406782, "raptor_stability": 0.8837265372276306, "fragility": 1.0, "augmentation_robustness": 0.7879341856049312, "xie_feature_dispersion": 1.5036821374561455, "pac": 0.835830361416281, "whitened_cosine_id": 0.8548471331596375}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 4, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921993338, "raptor_stability": 0.858695387840271, "fragility": 1.0, "augmentation_robustness": 0.7810593679700619, "xie_feature_dispersion": 1.841863426224599, "pac": 0.8198773779051665, "whitened_cosine_id": 0.7971394658088684}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 4, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921988385, "raptor_stability": 0.81805419921875, "fragility": 0.5, "augmentation_robustness": 0.6270972563452893, "xie_feature_dispersion": 2.0918980749205067, "pac": 0.7225757277820197, "whitened_cosine_id": 0.7637650966644287}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 4, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990021, "raptor_stability": 0.8543492555618286, "fragility": 1.0, "augmentation_robustness": 0.8033994268781987, "xie_feature_dispersion": 2.0592689851956822, "pac": 0.8288743412200137, "whitened_cosine_id": 0.7868671417236328}} +{"model": "pythia-70m", "dataset": "subj", "seed": 4, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002564, "raptor_stability": 0.8824465274810791, "fragility": 1.0, "augmentation_robustness": 0.7248629327605404, "xie_feature_dispersion": 1.925210310106005, "pac": 0.8036547301208097, "whitened_cosine_id": 0.9232692122459412}} +{"model": "pythia-70m", "dataset": "spam", "seed": 4, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004778, "raptor_stability": 0.9337431788444519, "fragility": 3.0, "augmentation_robustness": 0.8271455702691285, "xie_feature_dispersion": 8.1620184111796, "pac": 0.8804443745567903, "whitened_cosine_id": 0.9525339603424072}} +{"model": "pythia-70m", "dataset": "cola", "seed": 4, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921978768, "raptor_stability": 0.7940040826797485, "fragility": 1.0, "augmentation_robustness": 0.7165809555085785, "xie_feature_dispersion": 0.5974938164403546, "pac": 0.7552925190941635, "whitened_cosine_id": 0.7271562218666077}} +{"model": "pythia-70m", "dataset": "stance", "seed": 4, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.8148210225469336, "raptor_stability": 0.8456482291221619, "fragility": 1.0, "augmentation_robustness": 0.8025558898811386, "xie_feature_dispersion": 1.966541164919816, "pac": 0.8241020595016502, "whitened_cosine_id": 0.8281933665275574}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 4, "layer": 3, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997497, "raptor_stability": 0.8669647574424744, "fragility": 0.5, "augmentation_robustness": 0.7238996722036884, "xie_feature_dispersion": 1.0213916251568427, "pac": 0.7954322148230815, "whitened_cosine_id": 0.8732953667640686}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 4, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349261, "raptor_stability": 0.842431366443634, "fragility": 0.25, "augmentation_robustness": 0.6006899879923536, "xie_feature_dispersion": 2.8701281427080465, "pac": 0.7215606772179939, "whitened_cosine_id": 0.8559513688087463}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 4, "layer": 8, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485935049, "raptor_stability": 0.8715534806251526, "fragility": 0.5, "augmentation_robustness": 0.7578237900499349, "xie_feature_dispersion": 1.468960362896107, "pac": 0.8146886353375438, "whitened_cosine_id": 0.8794116377830505}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 4, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3108708237634092, "raptor_stability": 0.8495153784751892, "fragility": 0.25, "augmentation_robustness": 0.7581191296084264, "xie_feature_dispersion": 0.9310101861217775, "pac": 0.8038172540418078, "whitened_cosine_id": 0.8954182863235474}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 4, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.11711906515590792, "raptor_stability": 0.9491503834724426, "fragility": 1.5, "augmentation_robustness": 0.872645584642886, "xie_feature_dispersion": 5.1845819525161145, "pac": 0.9108979840576643, "whitened_cosine_id": 0.9199619293212891}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 4, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.397542485934307, "raptor_stability": 0.7986763715744019, "fragility": 0.25, "augmentation_robustness": 0.6679982833044269, "xie_feature_dispersion": 0.5584164688455334, "pac": 0.7333373274394144, "whitened_cosine_id": 0.7404189705848694}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.7058740799382045, "raptor_stability": 0.8810455799102783, "fragility": 1.0, "augmentation_robustness": 0.7601115407446962, "xie_feature_dispersion": 2.0356355439733544, "pac": 0.8205785603274873, "whitened_cosine_id": 0.8614587187767029}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 4, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.397542485934676, "raptor_stability": 0.8066017627716064, "fragility": 0.5, "augmentation_robustness": 0.6529396166694911, "xie_feature_dispersion": 2.9800122485144014, "pac": 0.7297706897205487, "whitened_cosine_id": 0.8008232712745667}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 4, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934582, "raptor_stability": 0.8202213644981384, "fragility": 0.25, "augmentation_robustness": 0.6057950002763687, "xie_feature_dispersion": 3.616328383472754, "pac": 0.7130081823872536, "whitened_cosine_id": 0.806983470916748}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 4, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346277, "raptor_stability": 0.8139558434486389, "fragility": 0.25, "augmentation_robustness": 0.6972874057878297, "xie_feature_dispersion": 1.5561310456535857, "pac": 0.7556216246182343, "whitened_cosine_id": 0.8032292723655701}} +{"model": "pythia-160m", "dataset": "subj", "seed": 4, "layer": 5, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351835, "raptor_stability": 0.8953569531440735, "fragility": 1.0, "augmentation_robustness": 0.7336658853694705, "xie_feature_dispersion": 1.8870415060935894, "pac": 0.814511419256772, "whitened_cosine_id": 0.9113974571228027}} +{"model": "pythia-160m", "dataset": "spam", "seed": 4, "layer": 12, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353225, "raptor_stability": 0.9130350351333618, "fragility": 0.5, "augmentation_robustness": 0.7635738024186676, "xie_feature_dispersion": 1.0575657286947788, "pac": 0.8383044187760147, "whitened_cosine_id": 0.9334843754768372}} +{"model": "pythia-160m", "dataset": "cola", "seed": 4, "layer": 2, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343628, "raptor_stability": 0.7938650250434875, "fragility": 0.5, "augmentation_robustness": 0.6480257075060164, "xie_feature_dispersion": 0.4811558212116989, "pac": 0.720945366274752, "whitened_cosine_id": 0.7700868844985962}} +{"model": "pythia-160m", "dataset": "stance", "seed": 4, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.591622106050008, "raptor_stability": 0.8417823314666748, "fragility": 1.0, "augmentation_robustness": 0.7769617698504535, "xie_feature_dispersion": 1.8092532859980797, "pac": 0.8093720506585642, "whitened_cosine_id": 0.8301091194152832}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 4, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935008, "raptor_stability": 0.8665009140968323, "fragility": 0.5, "augmentation_robustness": 0.7086095119691946, "xie_feature_dispersion": 0.8634855022918472, "pac": 0.7875552130330135, "whitened_cosine_id": 0.8738407492637634}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 4, "layer": 20, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688061, "raptor_stability": 0.8289727568626404, "fragility": 0.25, "augmentation_robustness": 0.5141338849042595, "xie_feature_dispersion": 3.3676014696218965, "pac": 0.6715533208834499, "whitened_cosine_id": 0.8497597575187683}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 4, "layer": 18, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882124, "raptor_stability": 0.8689995408058167, "fragility": 0.5, "augmentation_robustness": 0.7288625926699196, "xie_feature_dispersion": 1.9498286768695636, "pac": 0.7989310667378682, "whitened_cosine_id": 0.8895055055618286}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 4, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.2618240585223393, "raptor_stability": 0.8512345552444458, "fragility": 1.0, "augmentation_robustness": 0.7582979472125323, "xie_feature_dispersion": 5.093989432901596, "pac": 0.804766251228489, "whitened_cosine_id": 0.8913780450820923}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 4, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.10955555398923562, "raptor_stability": 0.9442237615585327, "fragility": 3.0, "augmentation_robustness": 0.8451578901315185, "xie_feature_dispersion": 10.914315920866802, "pac": 0.8946908258450257, "whitened_cosine_id": 0.887626588344574}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 4, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 1.210307295687852, "raptor_stability": 0.7869765162467957, "fragility": 0.25, "augmentation_robustness": 0.6309538054038855, "xie_feature_dispersion": 0.4977947174675206, "pac": 0.7089651608253406, "whitened_cosine_id": 0.7760900855064392}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 4, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.5396430100267917, "raptor_stability": 0.8690226674079895, "fragility": 0.25, "augmentation_robustness": 0.7275986700802664, "xie_feature_dispersion": 1.771013879173235, "pac": 0.7983106687441279, "whitened_cosine_id": 0.8532538414001465}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 4, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879974, "raptor_stability": 0.8093463182449341, "fragility": 0.25, "augmentation_robustness": 0.6522689919211225, "xie_feature_dispersion": 4.342720640505695, "pac": 0.7308076550830283, "whitened_cosine_id": 0.8254486918449402}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879697, "raptor_stability": 0.801374614238739, "fragility": 0.5, "augmentation_robustness": 0.5927650536655581, "xie_feature_dispersion": 1.97527441416868, "pac": 0.6970698339521486, "whitened_cosine_id": 0.8055334091186523}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 4, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687872, "raptor_stability": 0.8144592642784119, "fragility": 0.25, "augmentation_robustness": 0.6563598700902292, "xie_feature_dispersion": 3.7665662682483, "pac": 0.7354095671843206, "whitened_cosine_id": 0.8185945153236389}} +{"model": "pythia-410m", "dataset": "subj", "seed": 4, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882532, "raptor_stability": 0.8692127466201782, "fragility": 0.5, "augmentation_robustness": 0.7034189811361117, "xie_feature_dispersion": 1.4726101086077301, "pac": 0.7863158638781449, "whitened_cosine_id": 0.8998460173606873}} +{"model": "pythia-410m", "dataset": "spam", "seed": 4, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882872, "raptor_stability": 0.9380085468292236, "fragility": 3.0, "augmentation_robustness": 0.8306074892035639, "xie_feature_dispersion": 6.631280295102294, "pac": 0.8843080180163938, "whitened_cosine_id": 0.9329432845115662}} +{"model": "pythia-410m", "dataset": "cola", "seed": 4, "layer": 13, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879954, "raptor_stability": 0.8124500513076782, "fragility": 0.25, "augmentation_robustness": 0.6007360344180946, "xie_feature_dispersion": 0.12968683683090573, "pac": 0.7065930428628864, "whitened_cosine_id": 0.8110272288322449}} +{"model": "pythia-410m", "dataset": "stance", "seed": 4, "layer": 16, "concept": "stance", "predictors": {"sip_eigengap": 0.5699871301215584, "raptor_stability": 0.8456130027770996, "fragility": 0.5, "augmentation_robustness": 0.7784342554427507, "xie_feature_dispersion": 2.7721624801456297, "pac": 0.8120236291099252, "whitened_cosine_id": 0.8285333514213562}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 4, "layer": 17, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881238, "raptor_stability": 0.8406742215156555, "fragility": 0.5, "augmentation_robustness": 0.6639522820533309, "xie_feature_dispersion": 1.2414049521527235, "pac": 0.7523132517844933, "whitened_cosine_id": 0.8544057011604309}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 4, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010629, "raptor_stability": 0.8421841859817505, "fragility": 0.25, "augmentation_robustness": 0.5255393994669172, "xie_feature_dispersion": 3.4302386066080803, "pac": 0.6838617927243338, "whitened_cosine_id": 0.8283126950263977}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 4, "layer": 17, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8507398962974548, "fragility": 0.5, "augmentation_robustness": 0.6571983080736206, "xie_feature_dispersion": 2.186069124316738, "pac": 0.7539691021855377, "whitened_cosine_id": 0.8371143341064453}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 4, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.2265558176551069, "raptor_stability": 0.8569103479385376, "fragility": 1.5, "augmentation_robustness": 0.775218514152037, "xie_feature_dispersion": 5.3593709934723845, "pac": 0.8160644310452874, "whitened_cosine_id": 0.8336088061332703}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 4, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.07870194711620748, "raptor_stability": 0.9465585350990295, "fragility": 3.0, "augmentation_robustness": 0.8529859375455171, "xie_feature_dispersion": 9.52666970556469, "pac": 0.8997722363222733, "whitened_cosine_id": 0.8307517170906067}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010886, "raptor_stability": 0.8017974495887756, "fragility": 0.25, "augmentation_robustness": 0.6581476958880402, "xie_feature_dispersion": 0.5566355201054835, "pac": 0.7299725727384079, "whitened_cosine_id": 0.8239882588386536}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.02238370320048682, "raptor_stability": 0.861964762210846, "fragility": 0.25, "augmentation_robustness": 0.683838498205867, "xie_feature_dispersion": 2.404190193181982, "pac": 0.7729016302083564, "whitened_cosine_id": 0.8354349136352539}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 4, "layer": 13, "concept": "hate", "predictors": {"sip_eigengap": 0.85581649610108, "raptor_stability": 0.815185010433197, "fragility": 0.25, "augmentation_robustness": 0.6213317320340237, "xie_feature_dispersion": 4.537467582180153, "pac": 0.7182583712336104, "whitened_cosine_id": 0.8315821886062622}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 4, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010853, "raptor_stability": 0.8050772547721863, "fragility": 0.5, "augmentation_robustness": 0.6219013344721133, "xie_feature_dispersion": 2.2302886171133345, "pac": 0.7134892946221498, "whitened_cosine_id": 0.825539231300354}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 4, "layer": 21, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010789, "raptor_stability": 0.8045533895492554, "fragility": 0.25, "augmentation_robustness": 0.6038154742221061, "xie_feature_dispersion": 2.038988255260799, "pac": 0.7041844318856807, "whitened_cosine_id": 0.8270390629768372}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 4, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.8694700002670288, "fragility": 1.0, "augmentation_robustness": 0.7011428092065101, "xie_feature_dispersion": 2.50098235434014, "pac": 0.7853064047367695, "whitened_cosine_id": 0.8374378681182861}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 4, "layer": 8, "concept": "spam", "predictors": {"sip_eigengap": 0.85581649610102, "raptor_stability": 0.9178888201713562, "fragility": 1.0, "augmentation_robustness": 0.7928219231647947, "xie_feature_dispersion": 23.000550231187145, "pac": 0.8553553716680755, "whitened_cosine_id": 0.8662302494049072}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 4, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010873, "raptor_stability": 0.8157230615615845, "fragility": 0.25, "augmentation_robustness": 0.4993857440040359, "xie_feature_dispersion": 0.31676967447098947, "pac": 0.6575544027828102, "whitened_cosine_id": 0.8299345970153809}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 4, "layer": 17, "concept": "stance", "predictors": {"sip_eigengap": 0.33898643987040145, "raptor_stability": 0.828758716583252, "fragility": 0.5, "augmentation_robustness": 0.7561307052369252, "xie_feature_dispersion": 2.4747830566647955, "pac": 0.7924447109100885, "whitened_cosine_id": 0.825951099395752}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 4, "layer": 17, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010853, "raptor_stability": 0.8331989645957947, "fragility": 1.0, "augmentation_robustness": 0.627322928762955, "xie_feature_dispersion": 1.9407650108278485, "pac": 0.7302609466793748, "whitened_cosine_id": 0.8166413307189941}} +{"model": "gpt2", "dataset": "sst2", "seed": 4, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349996, "raptor_stability": 0.8296358585357666, "fragility": 0.25, "augmentation_robustness": 0.538035138872437, "xie_feature_dispersion": 3.0119285802495273, "pac": 0.6838354987041018, "whitened_cosine_id": 0.8716096878051758}} +{"model": "gpt2", "dataset": "imdb", "seed": 4, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350842, "raptor_stability": 0.8212791681289673, "fragility": 0.25, "augmentation_robustness": 0.5719860488360142, "xie_feature_dispersion": 1.6921644874690207, "pac": 0.6966326084824908, "whitened_cosine_id": 0.8852264881134033}} +{"model": "gpt2", "dataset": "ag_news", "seed": 4, "layer": 1, "concept": "topic", "predictors": {"sip_eigengap": 0.5333226182809293, "raptor_stability": 0.8636436462402344, "fragility": 0.5, "augmentation_robustness": 0.7962607413710602, "xie_feature_dispersion": 2.079954695672695, "pac": 0.8299521938056473, "whitened_cosine_id": 0.9045447707176208}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.1380779079157012, "raptor_stability": 0.932190477848053, "fragility": 1.5, "augmentation_robustness": 0.8059798202543121, "xie_feature_dispersion": 5.433633428902341, "pac": 0.8690851490511826, "whitened_cosine_id": 0.9079684615135193}} +{"model": "gpt2", "dataset": "counterfact", "seed": 4, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341972, "raptor_stability": 0.778740644454956, "fragility": 0.25, "augmentation_robustness": 0.689096563147253, "xie_feature_dispersion": 0.14506327750165868, "pac": 0.7339186038011045, "whitened_cosine_id": 0.7351542115211487}} +{"model": "gpt2", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5149693378247766, "raptor_stability": 0.8657316565513611, "fragility": 0.25, "augmentation_robustness": 0.6810631291311643, "xie_feature_dispersion": 1.1678519242893954, "pac": 0.7733973928412627, "whitened_cosine_id": 0.8591601848602295}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 4, "layer": 7, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347578, "raptor_stability": 0.7834518551826477, "fragility": 0.25, "augmentation_robustness": 0.5369999970113795, "xie_feature_dispersion": 3.433541094108048, "pac": 0.6602259260970136, "whitened_cosine_id": 0.8224130868911743}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 4, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345638, "raptor_stability": 0.7776834964752197, "fragility": 0.25, "augmentation_robustness": 0.5766816651217024, "xie_feature_dispersion": 2.7506262893948623, "pac": 0.677182580798461, "whitened_cosine_id": 0.790766179561615}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 4, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859344674, "raptor_stability": 0.779086172580719, "fragility": 0.25, "augmentation_robustness": 0.636466393301742, "xie_feature_dispersion": 2.815574398399473, "pac": 0.7077762829412305, "whitened_cosine_id": 0.8022896647453308}} +{"model": "gpt2", "dataset": "subj", "seed": 4, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352041, "raptor_stability": 0.8775368332862854, "fragility": 0.25, "augmentation_robustness": 0.7423765328860231, "xie_feature_dispersion": 0.37989537180206523, "pac": 0.8099566830861542, "whitened_cosine_id": 0.9062554836273193}} +{"model": "gpt2", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935321, "raptor_stability": 0.9084468483924866, "fragility": 1.5, "augmentation_robustness": 0.7197045600505599, "xie_feature_dispersion": 10.910132668764545, "pac": 0.8140757042215232, "whitened_cosine_id": 0.9353939294815063}} +{"model": "gpt2", "dataset": "cola", "seed": 4, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934498, "raptor_stability": 0.8069943189620972, "fragility": 0.25, "augmentation_robustness": 0.5145383818654423, "xie_feature_dispersion": 0.1311472028168222, "pac": 0.6607663504137697, "whitened_cosine_id": 0.8018531203269958}} +{"model": "gpt2", "dataset": "stance", "seed": 4, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5528580887739023, "raptor_stability": 0.8414139747619629, "fragility": 0.25, "augmentation_robustness": 0.7880240633832077, "xie_feature_dispersion": 0.9522759018719159, "pac": 0.8147190190725853, "whitened_cosine_id": 0.8301456570625305}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 4, "layer": 11, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934983, "raptor_stability": 0.8080868124961853, "fragility": 0.5, "augmentation_robustness": 0.5740397734709618, "xie_feature_dispersion": 1.3792581222344824, "pac": 0.6910632929835736, "whitened_cosine_id": 0.8652515411376953}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 4, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881142, "raptor_stability": 0.8341349363327026, "fragility": 0.25, "augmentation_robustness": 0.522748227401526, "xie_feature_dispersion": 3.1215463045771212, "pac": 0.6784415818671143, "whitened_cosine_id": 0.8599640727043152}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 4, "layer": 23, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882204, "raptor_stability": 0.9387363791465759, "fragility": 0.25, "augmentation_robustness": 0.8625580835416802, "xie_feature_dispersion": 2.3252826460499807, "pac": 0.900647231344128, "whitened_cosine_id": 0.8744732737541199}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 4, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.2490734387361061, "raptor_stability": 0.8433712124824524, "fragility": 1.0, "augmentation_robustness": 0.728063364939073, "xie_feature_dispersion": 3.704335352266776, "pac": 0.7857172887107626, "whitened_cosine_id": 0.881250262260437}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 4, "layer": 19, "concept": "topic", "predictors": {"sip_eigengap": 0.1068009553418093, "raptor_stability": 0.93490070104599, "fragility": 1.5, "augmentation_robustness": 0.8151994842437004, "xie_feature_dispersion": 6.637992417041358, "pac": 0.8750500926448452, "whitened_cosine_id": 0.8799629211425781}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 4, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878789, "raptor_stability": 0.7866268754005432, "fragility": 0.25, "augmentation_robustness": 0.6947319575923, "xie_feature_dispersion": 0.1466747323393826, "pac": 0.7406794164964217, "whitened_cosine_id": 0.7769638895988464}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5850775560898813, "raptor_stability": 0.8753568530082703, "fragility": 0.25, "augmentation_robustness": 0.7021984842516064, "xie_feature_dispersion": 0.8499569074700448, "pac": 0.7887776686299384, "whitened_cosine_id": 0.8553555011749268}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 4, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880465, "raptor_stability": 0.8168267607688904, "fragility": 0.25, "augmentation_robustness": 0.6873079329545261, "xie_feature_dispersion": 1.292586904966985, "pac": 0.7520673468617083, "whitened_cosine_id": 0.8249880075454712}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956880108, "raptor_stability": 0.7787988185882568, "fragility": 0.25, "augmentation_robustness": 0.535601105761814, "xie_feature_dispersion": 3.144268329198213, "pac": 0.6571999621750354, "whitened_cosine_id": 0.8181383609771729}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 4, "layer": 19, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880503, "raptor_stability": 0.7947708964347839, "fragility": 0.25, "augmentation_robustness": 0.5712740238571347, "xie_feature_dispersion": 2.3591332178783015, "pac": 0.6830224601459594, "whitened_cosine_id": 0.8295685052871704}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 4, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882628, "raptor_stability": 0.8655866980552673, "fragility": 0.5, "augmentation_robustness": 0.658993852125021, "xie_feature_dispersion": 1.294561616399908, "pac": 0.7622902750901441, "whitened_cosine_id": 0.8870449066162109}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 4, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882712, "raptor_stability": 0.9078684449195862, "fragility": 1.5, "augmentation_robustness": 0.7717328160030563, "xie_feature_dispersion": 4.942103977943902, "pac": 0.8398006304613213, "whitened_cosine_id": 0.9151633977890015}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 4, "layer": 16, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.210307295688009, "raptor_stability": 0.8008159399032593, "fragility": 0.25, "augmentation_robustness": 0.4862514101376035, "xie_feature_dispersion": 0.25648875129106014, "pac": 0.6435336750204314, "whitened_cosine_id": 0.8185151219367981}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 4, "layer": 15, "concept": "stance", "predictors": {"sip_eigengap": 0.4591650118237964, "raptor_stability": 0.818568766117096, "fragility": 0.5, "augmentation_robustness": 0.7404297489156003, "xie_feature_dispersion": 1.9692841257546414, "pac": 0.7794992575163482, "whitened_cosine_id": 0.7844472527503967}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 4, "layer": 23, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881433, "raptor_stability": 0.821359395980835, "fragility": 0.5, "augmentation_robustness": 0.5781830082719558, "xie_feature_dispersion": 1.435474963980124, "pac": 0.6997712021263953, "whitened_cosine_id": 0.8453944325447083}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 4, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237647728, "raptor_stability": 0.8522672653198242, "fragility": 0.25, "augmentation_robustness": 0.5718218717950756, "xie_feature_dispersion": 3.34994169851392, "pac": 0.7120445685574499, "whitened_cosine_id": 0.8843919038772583}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 4, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650468, "raptor_stability": 0.8942583203315735, "fragility": 1.0, "augmentation_robustness": 0.7825299636425157, "xie_feature_dispersion": 2.138162107858256, "pac": 0.8383941419870446, "whitened_cosine_id": 0.904426097869873}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 4, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.34415459384006464, "raptor_stability": 0.8814667463302612, "fragility": 0.5, "augmentation_robustness": 0.8472979677532054, "xie_feature_dispersion": 2.489404569197885, "pac": 0.8643823570417333, "whitened_cosine_id": 0.9101595878601074}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 4, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07232027374703373, "raptor_stability": 0.9524329900741577, "fragility": 1.5, "augmentation_robustness": 0.8775071051796265, "xie_feature_dispersion": 7.195039088518054, "pac": 0.914970047626892, "whitened_cosine_id": 0.9238596558570862}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 4, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646105, "raptor_stability": 0.8380523920059204, "fragility": 0.25, "augmentation_robustness": 0.7531953570003166, "xie_feature_dispersion": 0.1475593716528689, "pac": 0.7956238745031186, "whitened_cosine_id": 0.7818605303764343}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 4, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6524644536078231, "raptor_stability": 0.8897152543067932, "fragility": 0.25, "augmentation_robustness": 0.8390363852029633, "xie_feature_dispersion": 1.6422070698599567, "pac": 0.8643758197548783, "whitened_cosine_id": 0.854346513748169}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 4, "layer": 6, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647435, "raptor_stability": 0.8176067471504211, "fragility": 0.25, "augmentation_robustness": 0.7060153262170692, "xie_feature_dispersion": 4.1435032919274235, "pac": 0.7618110366837452, "whitened_cosine_id": 0.8246268033981323}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 4, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237636255, "raptor_stability": 0.8205522894859314, "fragility": 0.25, "augmentation_robustness": 0.6705621714892224, "xie_feature_dispersion": 3.9671135412282537, "pac": 0.7455572304875768, "whitened_cosine_id": 0.8058186173439026}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 4, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237646216, "raptor_stability": 0.8083575367927551, "fragility": 0.25, "augmentation_robustness": 0.685570372616926, "xie_feature_dispersion": 3.445721700576395, "pac": 0.7469639547048406, "whitened_cosine_id": 0.810322642326355}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 4, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651185, "raptor_stability": 0.886184573173523, "fragility": 0.5, "augmentation_robustness": 0.7504325815914692, "xie_feature_dispersion": 0.9492959086809997, "pac": 0.8183085773824961, "whitened_cosine_id": 0.919140100479126}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 4, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237647513, "raptor_stability": 0.9303396940231323, "fragility": 0.5, "augmentation_robustness": 0.8326478848343983, "xie_feature_dispersion": 18.28477413612892, "pac": 0.8814937894287653, "whitened_cosine_id": 0.9515261054039001}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 4, "layer": 14, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646718, "raptor_stability": 0.8116881251335144, "fragility": 0.25, "augmentation_robustness": 0.5360347240146708, "xie_feature_dispersion": 0.15135759875003865, "pac": 0.6738614245740926, "whitened_cosine_id": 0.8114071488380432}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.5271400318991657, "raptor_stability": 0.8213255405426025, "fragility": 0.5, "augmentation_robustness": 0.7172001724128508, "xie_feature_dispersion": 2.6328433172283368, "pac": 0.7692628564777266, "whitened_cosine_id": 0.826519787311554}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 4, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649207, "raptor_stability": 0.853298008441925, "fragility": 0.25, "augmentation_robustness": 0.6940637989225106, "xie_feature_dispersion": 0.7459643011965803, "pac": 0.7736809036822179, "whitened_cosine_id": 0.8680458068847656}} +{"model": "pythia-6.9b", "dataset": "sst2", "seed": 0, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.6051536478445365, "raptor_stability": 0.866299569606781, "fragility": 0.5, "augmentation_robustness": 0.637465302331378, "xie_feature_dispersion": 3.003399315085323, "pac": 0.7518824359690794, "whitened_cosine_id": 0.8191360235214233}} +{"model": "pythia-6.9b", "dataset": "ag_news", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.14898247898485747, "raptor_stability": 0.8647598624229431, "fragility": 1.0, "augmentation_robustness": 0.8786821950624223, "xie_feature_dispersion": 4.831002014369458, "pac": 0.8717210287426826, "whitened_cosine_id": 0.8155225515365601}} +{"model": "pythia-6.9b", "dataset": "counterfact", "seed": 0, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 0.6051536478445407, "raptor_stability": 0.8013729453086853, "fragility": 0.25, "augmentation_robustness": 0.7941799292141412, "xie_feature_dispersion": 0.2684332577388649, "pac": 0.7977764372614132, "whitened_cosine_id": 0.8149519562721252}} +{"model": "pythia-6.9b", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.0015580962176310153, "raptor_stability": 0.8656815886497498, "fragility": 0.25, "augmentation_robustness": 0.768174003000928, "xie_feature_dispersion": 3.785119247452656, "pac": 0.8169277958253389, "whitened_cosine_id": 0.8165521621704102}} +{"model": "pythia-6.9b", "dataset": "subj", "seed": 0, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.6051536478445388, "raptor_stability": 0.8781737685203552, "fragility": 1.0, "augmentation_robustness": 0.7621547288102961, "xie_feature_dispersion": 2.594084569527733, "pac": 0.8201642486653257, "whitened_cosine_id": 0.8201795816421509}} +{"model": "pythia-6.9b", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.045861524455193044, "raptor_stability": 0.9449687600135803, "fragility": 2.0, "augmentation_robustness": 0.8846214928355313, "xie_feature_dispersion": 7.980539713743011, "pac": 0.9147951264245557, "whitened_cosine_id": 0.8197609186172485}} +{"model": "pythia-6.9b", "dataset": "tweet_hate", "seed": 0, "layer": 16, "concept": "hate", "predictors": {"sip_eigengap": 0.6051536478445219, "raptor_stability": 0.8099347949028015, "fragility": 0.5, "augmentation_robustness": 0.6843764099541061, "xie_feature_dispersion": 5.482455155869018, "pac": 0.7471556024284538, "whitened_cosine_id": 0.8192850947380066}} diff --git a/repro_bundle/results_final/results/predictors_1aug.jsonl b/repro_bundle/results_final/results/predictors_1aug.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..bc5227abf3302ef9d7c185585f63b50ee47870b3 --- /dev/null +++ b/repro_bundle/results_final/results/predictors_1aug.jsonl @@ -0,0 +1,472 @@ +{"model": "pythia-70m", "dataset": "sst2", "seed": 0, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921994833, "raptor_stability": 0.8375464081764221, "fragility": 0.25, "augmentation_robustness": 0.6545744286726296, "xie_feature_dispersion": 0.8227753483650888, "pac": 0.7460604184245259, "whitened_cosine_id": 0.8294597268104553}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 0, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998517, "raptor_stability": 0.9239972829818726, "fragility": 1.0, "augmentation_robustness": 0.9364739773993105, "xie_feature_dispersion": 1.0993059236404399, "pac": 0.9302356301905915, "whitened_cosine_id": 0.8396446704864502}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 0, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6360429269373853, "raptor_stability": 0.8923418521881104, "fragility": 1.5, "augmentation_robustness": 0.9079627357769793, "xie_feature_dispersion": 3.4534575785421744, "pac": 0.9001522939825448, "whitened_cosine_id": 0.9029156565666199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 0, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14189668585488155, "raptor_stability": 0.9457845091819763, "fragility": 1.5, "augmentation_robustness": 0.8839691673684479, "xie_feature_dispersion": 5.2630010355712145, "pac": 0.9148768382752122, "whitened_cosine_id": 0.9317190051078796}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 0, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974287, "raptor_stability": 0.7641952633857727, "fragility": 1.0, "augmentation_robustness": 0.7320095613414648, "xie_feature_dispersion": 0.4779918901319739, "pac": 0.7481024123636187, "whitened_cosine_id": 0.6775354743003845}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 0, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5953587445214507, "raptor_stability": 0.8714625239372253, "fragility": 0.5, "augmentation_robustness": 0.8065598522498184, "xie_feature_dispersion": 1.5333196894740788, "pac": 0.8390111880935218, "whitened_cosine_id": 0.8480013012886047}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 0, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994247, "raptor_stability": 0.8230767250061035, "fragility": 0.5, "augmentation_robustness": 0.7221715232065848, "xie_feature_dispersion": 1.779944270255231, "pac": 0.7726241241063442, "whitened_cosine_id": 0.8223811388015747}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 0, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.711632992199055, "raptor_stability": 0.8190855383872986, "fragility": 0.25, "augmentation_robustness": 0.676819096553469, "xie_feature_dispersion": 2.512801472259043, "pac": 0.7479523174703837, "whitened_cosine_id": 0.7729045748710632}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989692, "raptor_stability": 0.8473518490791321, "fragility": 1.0, "augmentation_robustness": 0.8442697431706384, "xie_feature_dispersion": 1.163949469847951, "pac": 0.8458107961248853, "whitened_cosine_id": 0.7645858526229858}} +{"model": "pythia-70m", "dataset": "subj", "seed": 0, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002171, "raptor_stability": 0.9112874269485474, "fragility": 1.5, "augmentation_robustness": 0.8646995541446441, "xie_feature_dispersion": 2.3756762827181857, "pac": 0.8879934905465957, "whitened_cosine_id": 0.9034857153892517}} +{"model": "pythia-70m", "dataset": "spam", "seed": 0, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004674, "raptor_stability": 0.9211491942405701, "fragility": 1.5, "augmentation_robustness": 0.884000762861815, "xie_feature_dispersion": 10.408354824969607, "pac": 0.9025749785511925, "whitened_cosine_id": 0.9505216479301453}} +{"model": "pythia-70m", "dataset": "cola", "seed": 0, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921980065, "raptor_stability": 0.7842410206794739, "fragility": 1.0, "augmentation_robustness": 0.751228525306698, "xie_feature_dispersion": 0.5820240387291664, "pac": 0.7677347729930859, "whitened_cosine_id": 0.7026472687721252}} +{"model": "pythia-70m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.7516991834926783, "raptor_stability": 0.8351565003395081, "fragility": 1.0, "augmentation_robustness": 0.8460384036839802, "xie_feature_dispersion": 1.9650544929284912, "pac": 0.8405974520117441, "whitened_cosine_id": 0.8273500800132751}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 0, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998, "raptor_stability": 0.8865668177604675, "fragility": 1.5, "augmentation_robustness": 0.8695133740159424, "xie_feature_dispersion": 1.822712974246315, "pac": 0.878040095888205, "whitened_cosine_id": 0.8655645847320557}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350087, "raptor_stability": 0.8526354432106018, "fragility": 0.25, "augmentation_robustness": 0.6188354080698831, "xie_feature_dispersion": 1.4101369914642357, "pac": 0.7357354256402424, "whitened_cosine_id": 0.8668140769004822}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 0, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350252, "raptor_stability": 0.8803290128707886, "fragility": 1.0, "augmentation_robustness": 0.8724963955930902, "xie_feature_dispersion": 1.3391749548314755, "pac": 0.8764127042319394, "whitened_cosine_id": 0.8713250756263733}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3953680487548481, "raptor_stability": 0.8455236554145813, "fragility": 0.25, "augmentation_robustness": 0.8483093395593638, "xie_feature_dispersion": 0.9318287616318928, "pac": 0.8469164974869725, "whitened_cosine_id": 0.8953917026519775}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 0, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10457984207175601, "raptor_stability": 0.9457212686538696, "fragility": 1.5, "augmentation_robustness": 0.8857764499429269, "xie_feature_dispersion": 5.199477497190765, "pac": 0.9157488592983982, "whitened_cosine_id": 0.9181342720985413}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 0, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342227, "raptor_stability": 0.7698441743850708, "fragility": 0.25, "augmentation_robustness": 0.6816650768598687, "xie_feature_dispersion": 0.25779876141354974, "pac": 0.7257546256224697, "whitened_cosine_id": 0.7203242182731628}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.574379658862361, "raptor_stability": 0.8597204685211182, "fragility": 0.25, "augmentation_robustness": 0.7682780634518579, "xie_feature_dispersion": 1.6192479136149953, "pac": 0.8139992659864881, "whitened_cosine_id": 0.8496574759483337}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348304, "raptor_stability": 0.8552483320236206, "fragility": 1.0, "augmentation_robustness": 0.7851229050246351, "xie_feature_dispersion": 2.2631913339622987, "pac": 0.8201856185241279, "whitened_cosine_id": 0.8250994682312012}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 0, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934627, "raptor_stability": 0.8244509100914001, "fragility": 0.5, "augmentation_robustness": 0.7136331067294028, "xie_feature_dispersion": 1.9587240977086187, "pac": 0.7690420084104015, "whitened_cosine_id": 0.7910383939743042}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934696, "raptor_stability": 0.8365420699119568, "fragility": 1.0, "augmentation_robustness": 0.7890852919390294, "xie_feature_dispersion": 1.5342572755652866, "pac": 0.8128136809254931, "whitened_cosine_id": 0.8055051565170288}} +{"model": "pythia-160m", "dataset": "subj", "seed": 0, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351844, "raptor_stability": 0.8689209222793579, "fragility": 1.0, "augmentation_robustness": 0.7746642653183333, "xie_feature_dispersion": 1.77134285145721, "pac": 0.8217925937988456, "whitened_cosine_id": 0.9039409160614014}} +{"model": "pythia-160m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352976, "raptor_stability": 0.943769633769989, "fragility": 3.0, "augmentation_robustness": 0.9172945300081299, "xie_feature_dispersion": 6.963799146343127, "pac": 0.9305320818890594, "whitened_cosine_id": 0.9455703496932983}} +{"model": "pythia-160m", "dataset": "cola", "seed": 0, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344465, "raptor_stability": 0.7988646626472473, "fragility": 0.25, "augmentation_robustness": 0.5972243911972416, "xie_feature_dispersion": 0.7758972276639047, "pac": 0.6980445269222444, "whitened_cosine_id": 0.7851360440254211}} +{"model": "pythia-160m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.4494122497402506, "raptor_stability": 0.8332012295722961, "fragility": 1.5, "augmentation_robustness": 0.8344763697080788, "xie_feature_dispersion": 2.5953837186456754, "pac": 0.8338387996401875, "whitened_cosine_id": 0.8253653049468994}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 0, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935011, "raptor_stability": 0.866767942905426, "fragility": 0.5, "augmentation_robustness": 0.7943799335400991, "xie_feature_dispersion": 1.5213101554825748, "pac": 0.8305739382227626, "whitened_cosine_id": 0.8781046867370605}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 0, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881098, "raptor_stability": 0.8615295886993408, "fragility": 0.25, "augmentation_robustness": 0.64759763388926, "xie_feature_dispersion": 1.6884023301086555, "pac": 0.7545636112943004, "whitened_cosine_id": 0.8637628555297852}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881813, "raptor_stability": 0.8620725274085999, "fragility": 1.0, "augmentation_robustness": 0.8437283807712266, "xie_feature_dispersion": 1.923545301649778, "pac": 0.8529004540899132, "whitened_cosine_id": 0.8767724633216858}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.32280286192962354, "raptor_stability": 0.8491001725196838, "fragility": 1.0, "augmentation_robustness": 0.8717545371887874, "xie_feature_dispersion": 6.130394237589204, "pac": 0.8604273548542356, "whitened_cosine_id": 0.8844512701034546}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.1116614706117857, "raptor_stability": 0.945551872253418, "fragility": 2.0, "augmentation_robustness": 0.8856969264519378, "xie_feature_dispersion": 7.687914979274747, "pac": 0.9156243993526778, "whitened_cosine_id": 0.8968205451965332}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 0, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879029, "raptor_stability": 0.787240743637085, "fragility": 0.25, "augmentation_robustness": 0.7337481325439938, "xie_feature_dispersion": 0.12206039095754939, "pac": 0.7604944380905394, "whitened_cosine_id": 0.7804538607597351}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 0, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.4882081459032719, "raptor_stability": 0.8632135391235352, "fragility": 0.25, "augmentation_robustness": 0.7884647975902315, "xie_feature_dispersion": 2.173352441480384, "pac": 0.8258391683568833, "whitened_cosine_id": 0.8471425175666809}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 0, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880458, "raptor_stability": 0.8126577138900757, "fragility": 0.25, "augmentation_robustness": 0.7034424602081344, "xie_feature_dispersion": 4.485569514407045, "pac": 0.758050087049105, "whitened_cosine_id": 0.827558696269989}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 0, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879928, "raptor_stability": 0.7987161874771118, "fragility": 0.25, "augmentation_robustness": 0.6753132678040145, "xie_feature_dispersion": 2.5329861784126293, "pac": 0.7370147276405632, "whitened_cosine_id": 0.810127854347229}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 0, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880347, "raptor_stability": 0.814401388168335, "fragility": 0.25, "augmentation_robustness": 0.6923253341445008, "xie_feature_dispersion": 0.7524282237703519, "pac": 0.7533633611564179, "whitened_cosine_id": 0.8191361427307129}} +{"model": "pythia-410m", "dataset": "subj", "seed": 0, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882566, "raptor_stability": 0.8862898945808411, "fragility": 0.5, "augmentation_robustness": 0.7988711466010014, "xie_feature_dispersion": 1.428972872571398, "pac": 0.8425805205909213, "whitened_cosine_id": 0.8959206342697144}} +{"model": "pythia-410m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882899, "raptor_stability": 0.9429970979690552, "fragility": 3.0, "augmentation_robustness": 0.9052157104985206, "xie_feature_dispersion": 6.72893787243834, "pac": 0.9241064042337879, "whitened_cosine_id": 0.933451235294342}} +{"model": "pythia-410m", "dataset": "cola", "seed": 0, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879473, "raptor_stability": 0.7842450737953186, "fragility": 0.25, "augmentation_robustness": 0.5039294079390225, "xie_feature_dispersion": 1.1149557488538653, "pac": 0.6440872408671705, "whitened_cosine_id": 0.8007057905197144}} +{"model": "pythia-410m", "dataset": "stance", "seed": 0, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.5795818589574081, "raptor_stability": 0.8285464644432068, "fragility": 1.0, "augmentation_robustness": 0.8175707678986336, "xie_feature_dispersion": 1.8361604596405292, "pac": 0.8230586161709201, "whitened_cosine_id": 0.8221842646598816}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 0, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881513, "raptor_stability": 0.8521958589553833, "fragility": 0.5, "augmentation_robustness": 0.7625019501536046, "xie_feature_dispersion": 1.5630966219328584, "pac": 0.8073489045544939, "whitened_cosine_id": 0.8685176968574524}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.8599012494087219, "fragility": 0.25, "augmentation_robustness": 0.5848209179858852, "xie_feature_dispersion": 2.122662038410848, "pac": 0.7223610836973036, "whitened_cosine_id": 0.8232980370521545}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.825421154499054, "fragility": 1.0, "augmentation_robustness": 0.7804717428621477, "xie_feature_dispersion": 2.576726583434504, "pac": 0.8029464486806008, "whitened_cosine_id": 0.8465724587440491}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 0, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22041139949027433, "raptor_stability": 0.8484893441200256, "fragility": 1.0, "augmentation_robustness": 0.8650683659984831, "xie_feature_dispersion": 5.917035974988243, "pac": 0.8567788550592543, "whitened_cosine_id": 0.8215498328208923}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06228035557873579, "raptor_stability": 0.9404226541519165, "fragility": 3.0, "augmentation_robustness": 0.8779983715906402, "xie_feature_dispersion": 9.493553367505593, "pac": 0.9092105128712784, "whitened_cosine_id": 0.8300829529762268}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 0, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.7848396301269531, "fragility": 0.25, "augmentation_robustness": 0.7408315387590054, "xie_feature_dispersion": 0.2653617366201949, "pac": 0.7628355844429793, "whitened_cosine_id": 0.817557156085968}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.019767617442798265, "raptor_stability": 0.8671453595161438, "fragility": 0.25, "augmentation_robustness": 0.755285317721517, "xie_feature_dispersion": 2.8877683311134636, "pac": 0.8112153386188303, "whitened_cosine_id": 0.830573558807373}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 0, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010617, "raptor_stability": 0.8155287504196167, "fragility": 0.25, "augmentation_robustness": 0.6718229230312025, "xie_feature_dispersion": 5.342310320027777, "pac": 0.7436758367254096, "whitened_cosine_id": 0.8247697353363037}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 0, "layer": 22, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.7958077192306519, "fragility": 0.25, "augmentation_robustness": 0.6470101589682584, "xie_feature_dispersion": 2.217429056075072, "pac": 0.7214089390994551, "whitened_cosine_id": 0.8194364309310913}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 0, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8089179396629333, "fragility": 0.5, "augmentation_robustness": 0.7119613985713426, "xie_feature_dispersion": 1.5482508686542642, "pac": 0.760439669117138, "whitened_cosine_id": 0.8221121430397034}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 0, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010872, "raptor_stability": 0.8672937750816345, "fragility": 1.0, "augmentation_robustness": 0.7720944652439713, "xie_feature_dispersion": 2.5476390335776844, "pac": 0.819694120162803, "whitened_cosine_id": 0.8334078788757324}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 0, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010434, "raptor_stability": 0.9216615557670593, "fragility": 1.0, "augmentation_robustness": 0.895838461689926, "xie_feature_dispersion": 23.432722674772293, "pac": 0.9087500087284927, "whitened_cosine_id": 0.8588989973068237}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 0, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010818, "raptor_stability": 0.8071193099021912, "fragility": 0.25, "augmentation_robustness": 0.5249275775049415, "xie_feature_dispersion": 1.549092757263006, "pac": 0.6660234437035664, "whitened_cosine_id": 0.8179035186767578}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 0, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.36672114356076757, "raptor_stability": 0.8304018974304199, "fragility": 1.0, "augmentation_robustness": 0.8219455087263939, "xie_feature_dispersion": 3.593233328751099, "pac": 0.8261737030784069, "whitened_cosine_id": 0.8171507716178894}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8476687073707581, "fragility": 0.5, "augmentation_robustness": 0.7513937676225657, "xie_feature_dispersion": 2.045523026748139, "pac": 0.7995312374966619, "whitened_cosine_id": 0.8353421092033386}} +{"model": "gpt2", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349679, "raptor_stability": 0.8233915567398071, "fragility": 0.25, "augmentation_robustness": 0.5123065291062753, "xie_feature_dispersion": 1.6815976407755642, "pac": 0.6678490429230413, "whitened_cosine_id": 0.857738196849823}} +{"model": "gpt2", "dataset": "imdb", "seed": 0, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350731, "raptor_stability": 0.7829911708831787, "fragility": 0.25, "augmentation_robustness": 0.6517344633023527, "xie_feature_dispersion": 1.6579993440971876, "pac": 0.7173628170927657, "whitened_cosine_id": 0.8754125237464905}} +{"model": "gpt2", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.34635368740834577, "raptor_stability": 0.8845524787902832, "fragility": 0.25, "augmentation_robustness": 0.9187097662639041, "xie_feature_dispersion": 0.8678892394256513, "pac": 0.9016311225270937, "whitened_cosine_id": 0.910406231880188}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 0, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.12115672881499416, "raptor_stability": 0.9302389025688171, "fragility": 1.5, "augmentation_robustness": 0.8321901130650252, "xie_feature_dispersion": 6.121815304004312, "pac": 0.8812145078169211, "whitened_cosine_id": 0.9043624997138977}} +{"model": "gpt2", "dataset": "counterfact", "seed": 0, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859340033, "raptor_stability": 0.7690413594245911, "fragility": 0.25, "augmentation_robustness": 0.809669912657523, "xie_feature_dispersion": 0.1095429604074235, "pac": 0.789355636041057, "whitened_cosine_id": 0.7072374820709229}} +{"model": "gpt2", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.7366388390202255, "raptor_stability": 0.8725014328956604, "fragility": 0.25, "augmentation_robustness": 0.8145765018284364, "xie_feature_dispersion": 1.77009486868241, "pac": 0.8435389673620484, "whitened_cosine_id": 0.8444538116455078}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347128, "raptor_stability": 0.7693442106246948, "fragility": 0.25, "augmentation_robustness": 0.551602767544549, "xie_feature_dispersion": 3.694561283935865, "pac": 0.6604734890846219, "whitened_cosine_id": 0.8187122344970703}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 0, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346073, "raptor_stability": 0.7867448925971985, "fragility": 0.25, "augmentation_robustness": 0.6861444856119838, "xie_feature_dispersion": 2.309585917543921, "pac": 0.7364446891045912, "whitened_cosine_id": 0.8086420297622681}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 0, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346306, "raptor_stability": 0.7805789113044739, "fragility": 0.25, "augmentation_robustness": 0.657242871403201, "xie_feature_dispersion": 0.25866709333672444, "pac": 0.7189108913538375, "whitened_cosine_id": 0.803403377532959}} +{"model": "gpt2", "dataset": "subj", "seed": 0, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352106, "raptor_stability": 0.8574187755584717, "fragility": 0.25, "augmentation_robustness": 0.7295931094481585, "xie_feature_dispersion": 0.951143558340801, "pac": 0.7935059425033151, "whitened_cosine_id": 0.9090107679367065}} +{"model": "gpt2", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593527, "raptor_stability": 0.9161890149116516, "fragility": 2.0, "augmentation_robustness": 0.9103539019844458, "xie_feature_dispersion": 6.497801646724056, "pac": 0.9132714584480487, "whitened_cosine_id": 0.9386647939682007}} +{"model": "gpt2", "dataset": "cola", "seed": 0, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859346528, "raptor_stability": 0.7938734889030457, "fragility": 0.25, "augmentation_robustness": 0.49778314433790527, "xie_feature_dispersion": 0.9649847502240455, "pac": 0.6458283166204755, "whitened_cosine_id": 0.807263970375061}} +{"model": "gpt2", "dataset": "stance", "seed": 0, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.490010700608385, "raptor_stability": 0.8106656670570374, "fragility": 0.25, "augmentation_robustness": 0.7954562938419368, "xie_feature_dispersion": 2.664520632491477, "pac": 0.803060980449487, "whitened_cosine_id": 0.8153484463691711}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935004, "raptor_stability": 0.8306007981300354, "fragility": 0.25, "augmentation_robustness": 0.6972854536891078, "xie_feature_dispersion": 1.1399162300669312, "pac": 0.7639431259095716, "whitened_cosine_id": 0.8771829605102539}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 0, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688191, "raptor_stability": 0.8413395881652832, "fragility": 0.25, "augmentation_robustness": 0.5985845188957462, "xie_feature_dispersion": 1.7719582872902115, "pac": 0.7199620535305147, "whitened_cosine_id": 0.8595471382141113}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 0, "layer": 3, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878758, "raptor_stability": 0.7550429105758667, "fragility": 0.25, "augmentation_robustness": 0.8450710203949588, "xie_feature_dispersion": 0.2304978644612739, "pac": 0.8000569654854128, "whitened_cosine_id": 0.7729927897453308}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 0, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6820984754002586, "raptor_stability": 0.8745980262756348, "fragility": 0.25, "augmentation_robustness": 0.7780721034137521, "xie_feature_dispersion": 0.8621445356293869, "pac": 0.8263350648446934, "whitened_cosine_id": 0.8492467999458313}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880176, "raptor_stability": 0.8115680813789368, "fragility": 0.25, "augmentation_robustness": 0.6729005392577236, "xie_feature_dispersion": 1.137989115163849, "pac": 0.7422343103183302, "whitened_cosine_id": 0.8243570923805237}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 0, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878862, "raptor_stability": 0.777171790599823, "fragility": 0.25, "augmentation_robustness": 0.6061778394389484, "xie_feature_dispersion": 2.1506698277748155, "pac": 0.6916748150193857, "whitened_cosine_id": 0.8085216283798218}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 0, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880017, "raptor_stability": 0.7901369333267212, "fragility": 0.25, "augmentation_robustness": 0.6555666038259415, "xie_feature_dispersion": 0.462184243542827, "pac": 0.7228517685763314, "whitened_cosine_id": 0.8134031891822815}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 0, "layer": 24, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688252, "raptor_stability": 0.8832855224609375, "fragility": 0.25, "augmentation_robustness": 0.7951101769713113, "xie_feature_dispersion": 0.4906864607581503, "pac": 0.8391978497161243, "whitened_cosine_id": 0.8952683210372925}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 0, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880159, "raptor_stability": 0.7916848063468933, "fragility": 0.25, "augmentation_robustness": 0.5224171661394365, "xie_feature_dispersion": 0.79552886696272, "pac": 0.6570509862431648, "whitened_cosine_id": 0.8178503513336182}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 0, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.42249695846338925, "raptor_stability": 0.8128862977027893, "fragility": 0.5, "augmentation_robustness": 0.8037561319272016, "xie_feature_dispersion": 2.2610696467251814, "pac": 0.8083212148149954, "whitened_cosine_id": 0.7861542701721191}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 0, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881549, "raptor_stability": 0.8164692521095276, "fragility": 0.25, "augmentation_robustness": 0.6904412835869388, "xie_feature_dispersion": 1.1005465582305016, "pac": 0.7534552678482331, "whitened_cosine_id": 0.8596013784408569}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648414, "raptor_stability": 0.8818655014038086, "fragility": 0.25, "augmentation_robustness": 0.67470015678154, "xie_feature_dispersion": 1.8204152059660144, "pac": 0.7782828290926743, "whitened_cosine_id": 0.879472017288208}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 0, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.29387292376502, "raptor_stability": 0.8775492906570435, "fragility": 1.0, "augmentation_robustness": 0.8720676344266011, "xie_feature_dispersion": 2.218615090585379, "pac": 0.8748084625418222, "whitened_cosine_id": 0.8882995247840881}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.33450754332538074, "raptor_stability": 0.8566421270370483, "fragility": 1.0, "augmentation_robustness": 0.8744251513008834, "xie_feature_dispersion": 3.7561628051141023, "pac": 0.8655336391689659, "whitened_cosine_id": 0.9043422937393188}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07399960414320991, "raptor_stability": 0.9483939409255981, "fragility": 1.5, "augmentation_robustness": 0.891918719098521, "xie_feature_dispersion": 7.327642739764713, "pac": 0.9201563300120597, "whitened_cosine_id": 0.9209960699081421}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 0, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237645932, "raptor_stability": 0.8066194653511047, "fragility": 0.25, "augmentation_robustness": 0.8373520059306747, "xie_feature_dispersion": 0.06081582657176685, "pac": 0.8219857356408897, "whitened_cosine_id": 0.7630186080932617}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6131208141710327, "raptor_stability": 0.8868826031684875, "fragility": 0.25, "augmentation_robustness": 0.8904501661922035, "xie_feature_dispersion": 1.992273238599221, "pac": 0.8886663846803455, "whitened_cosine_id": 0.8511517643928528}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645996, "raptor_stability": 0.8185003995895386, "fragility": 0.25, "augmentation_robustness": 0.7607385932812474, "xie_feature_dispersion": 4.466765724760196, "pac": 0.789619496435393, "whitened_cosine_id": 0.8152397871017456}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 0, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647149, "raptor_stability": 0.811665415763855, "fragility": 0.25, "augmentation_robustness": 0.7276199921424138, "xie_feature_dispersion": 2.8843177774025395, "pac": 0.7696427039531344, "whitened_cosine_id": 0.8048377633094788}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 0, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647015, "raptor_stability": 0.8171485662460327, "fragility": 0.25, "augmentation_robustness": 0.7939897357140582, "xie_feature_dispersion": 0.15381271196675575, "pac": 0.8055691509800454, "whitened_cosine_id": 0.8043486475944519}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 0, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651123, "raptor_stability": 0.8921712040901184, "fragility": 0.5, "augmentation_robustness": 0.8243175866171568, "xie_feature_dispersion": 0.9955751494789675, "pac": 0.8582443953536376, "whitened_cosine_id": 0.9154570698738098}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 0, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923765166, "raptor_stability": 0.9529117941856384, "fragility": 3.0, "augmentation_robustness": 0.94099314226979, "xie_feature_dispersion": 9.854132791030557, "pac": 0.9469524682277142, "whitened_cosine_id": 0.9523190855979919}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 0, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764728, "raptor_stability": 0.8035885691642761, "fragility": 0.25, "augmentation_robustness": 0.5589292088936479, "xie_feature_dispersion": 1.0666210514085848, "pac": 0.681258889028962, "whitened_cosine_id": 0.818631649017334}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 0, "layer": 20, "concept": "stance", "predictors": {"sip_eigengap": 0.4157006657150525, "raptor_stability": 0.8240557312965393, "fragility": 0.5, "augmentation_robustness": 0.8015463605814978, "xie_feature_dispersion": 3.234317390572246, "pac": 0.8128010459390185, "whitened_cosine_id": 0.8179960250854492}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 0, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649436, "raptor_stability": 0.8530629873275757, "fragility": 0.25, "augmentation_robustness": 0.7821972078583704, "xie_feature_dispersion": 1.1742203784808831, "pac": 0.8176300975929731, "whitened_cosine_id": 0.8717560768127441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921995341, "raptor_stability": 0.8336094617843628, "fragility": 0.25, "augmentation_robustness": 0.6377618570084167, "xie_feature_dispersion": 1.7265608037193576, "pac": 0.7356856593963897, "whitened_cosine_id": 0.8218536972999573}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998223, "raptor_stability": 0.8874242901802063, "fragility": 1.0, "augmentation_robustness": 0.8992101261321568, "xie_feature_dispersion": 1.271550857614664, "pac": 0.8933172081561815, "whitened_cosine_id": 0.8205482363700867}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6806422452118264, "raptor_stability": 0.8989484906196594, "fragility": 1.5, "augmentation_robustness": 0.9168698848955824, "xie_feature_dispersion": 3.4157347611498774, "pac": 0.9079091877576209, "whitened_cosine_id": 0.9129272699356079}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.18807531413389164, "raptor_stability": 0.9559001326560974, "fragility": 2.0, "augmentation_robustness": 0.9200740132537277, "xie_feature_dispersion": 5.534601025030786, "pac": 0.9379870729549126, "whitened_cosine_id": 0.933467447757721}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 1, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.777370810508728, "fragility": 0.5, "augmentation_robustness": 0.7673906542908651, "xie_feature_dispersion": 0.6104756443427384, "pac": 0.7723807323997965, "whitened_cosine_id": 0.6849023103713989}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 1, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.5332073250620768, "raptor_stability": 0.8536012172698975, "fragility": 0.25, "augmentation_robustness": 0.7542190503987464, "xie_feature_dispersion": 1.7320048864765314, "pac": 0.803910133834322, "whitened_cosine_id": 0.8371078372001648}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 1, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994555, "raptor_stability": 0.8290720582008362, "fragility": 0.5, "augmentation_robustness": 0.7840608903875346, "xie_feature_dispersion": 2.6910245205930696, "pac": 0.8065664742941854, "whitened_cosine_id": 0.8217921853065491}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 1, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987687, "raptor_stability": 0.819472074508667, "fragility": 1.0, "augmentation_robustness": 0.698958315186553, "xie_feature_dispersion": 2.073055712645641, "pac": 0.75921519484761, "whitened_cosine_id": 0.7341217398643494}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 1, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989486, "raptor_stability": 0.7951779961585999, "fragility": 0.5, "augmentation_robustness": 0.7589872979724266, "xie_feature_dispersion": 1.2637336763777984, "pac": 0.7770826470655132, "whitened_cosine_id": 0.7604337334632874}} +{"model": "pythia-70m", "dataset": "subj", "seed": 1, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922001947, "raptor_stability": 0.8965046405792236, "fragility": 1.5, "augmentation_robustness": 0.8297915129525963, "xie_feature_dispersion": 2.3654705725557443, "pac": 0.86314807676591, "whitened_cosine_id": 0.9005034565925598}} +{"model": "pythia-70m", "dataset": "spam", "seed": 1, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004856, "raptor_stability": 0.9503062963485718, "fragility": 3.0, "augmentation_robustness": 0.9222602495310279, "xie_feature_dispersion": 6.604643787479847, "pac": 0.9362832729397998, "whitened_cosine_id": 0.9495452046394348}} +{"model": "pythia-70m", "dataset": "cola", "seed": 1, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.7986350655555725, "fragility": 1.0, "augmentation_robustness": 0.720600733310639, "xie_feature_dispersion": 0.7104026437376014, "pac": 0.7596178994331058, "whitened_cosine_id": 0.7400056719779968}} +{"model": "pythia-70m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6754527167162487, "raptor_stability": 0.8382856249809265, "fragility": 1.0, "augmentation_robustness": 0.8615565989812328, "xie_feature_dispersion": 1.8257872743628552, "pac": 0.8499211119810797, "whitened_cosine_id": 0.8341732621192932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 1, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997502, "raptor_stability": 0.8919016122817993, "fragility": 1.5, "augmentation_robustness": 0.8621425329457455, "xie_feature_dispersion": 1.7499668565574926, "pac": 0.8770220726137724, "whitened_cosine_id": 0.8624804019927979}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 1, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349208, "raptor_stability": 0.8506006002426147, "fragility": 0.25, "augmentation_robustness": 0.637601559557224, "xie_feature_dispersion": 1.9524296354090338, "pac": 0.7441010798999194, "whitened_cosine_id": 0.8579458594322205}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 1, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350347, "raptor_stability": 0.8709588050842285, "fragility": 1.0, "augmentation_robustness": 0.8681998579081558, "xie_feature_dispersion": 1.4529956718728516, "pac": 0.8695793314961922, "whitened_cosine_id": 0.8694334030151367}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 1, "layer": 7, "concept": "topic", "predictors": {"sip_eigengap": 0.45824187669700944, "raptor_stability": 0.8648454546928406, "fragility": 1.0, "augmentation_robustness": 0.883777230516235, "xie_feature_dispersion": 2.411760223810931, "pac": 0.8743113426045378, "whitened_cosine_id": 0.9022970199584961}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.12810819210433996, "raptor_stability": 0.9377708435058594, "fragility": 3.0, "augmentation_robustness": 0.8310133850459603, "xie_feature_dispersion": 10.50628418134277, "pac": 0.8843921142759099, "whitened_cosine_id": 0.9117757678031921}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 1, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341767, "raptor_stability": 0.7650079131126404, "fragility": 0.5, "augmentation_robustness": 0.7779032363265277, "xie_feature_dispersion": 0.6428540179632124, "pac": 0.771455574719584, "whitened_cosine_id": 0.7203051447868347}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 1, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5407329041738655, "raptor_stability": 0.8731709122657776, "fragility": 0.5, "augmentation_robustness": 0.8006721595864167, "xie_feature_dispersion": 1.6646203889081308, "pac": 0.8369215359260971, "whitened_cosine_id": 0.854195773601532}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 1, "layer": 3, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348242, "raptor_stability": 0.8365353345870972, "fragility": 1.0, "augmentation_robustness": 0.7581229958494136, "xie_feature_dispersion": 1.9822842109390966, "pac": 0.7973291652182554, "whitened_cosine_id": 0.8330134749412537}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 1, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346446, "raptor_stability": 0.8064267039299011, "fragility": 0.25, "augmentation_robustness": 0.6783393674021946, "xie_feature_dispersion": 3.1602498907933048, "pac": 0.7423830356660479, "whitened_cosine_id": 0.7950520515441895}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 1, "layer": 8, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934659, "raptor_stability": 0.7950870990753174, "fragility": 0.25, "augmentation_robustness": 0.7522003888311254, "xie_feature_dispersion": 1.0110021271866068, "pac": 0.7736437439532213, "whitened_cosine_id": 0.7926974892616272}} +{"model": "pythia-160m", "dataset": "subj", "seed": 1, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351837, "raptor_stability": 0.8696022033691406, "fragility": 1.0, "augmentation_robustness": 0.7639011148909334, "xie_feature_dispersion": 2.2528502880398076, "pac": 0.816751659130037, "whitened_cosine_id": 0.9035648703575134}} +{"model": "pythia-160m", "dataset": "spam", "seed": 1, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352847, "raptor_stability": 0.9360153079032898, "fragility": 3.0, "augmentation_robustness": 0.9010665268870804, "xie_feature_dispersion": 7.136742674643747, "pac": 0.9185409173951851, "whitened_cosine_id": 0.9397127628326416}} +{"model": "pythia-160m", "dataset": "cola", "seed": 1, "layer": 8, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343877, "raptor_stability": 0.799544095993042, "fragility": 0.25, "augmentation_robustness": 0.6070554401031638, "xie_feature_dispersion": 0.34675513205319614, "pac": 0.7032997680481029, "whitened_cosine_id": 0.7765609622001648}} +{"model": "pythia-160m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6971690127369661, "raptor_stability": 0.8396480679512024, "fragility": 1.0, "augmentation_robustness": 0.8287048173260642, "xie_feature_dispersion": 2.4375720598203166, "pac": 0.8341764426386333, "whitened_cosine_id": 0.8327767848968506}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 1, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349912, "raptor_stability": 0.8510211110115051, "fragility": 0.5, "augmentation_robustness": 0.7777689965959116, "xie_feature_dispersion": 0.793449068148806, "pac": 0.8143950538037084, "whitened_cosine_id": 0.873329222202301}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 1, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688131, "raptor_stability": 0.8608813881874084, "fragility": 0.25, "augmentation_robustness": 0.6344172991669839, "xie_feature_dispersion": 2.4202065461787297, "pac": 0.7476493436771962, "whitened_cosine_id": 0.8642251491546631}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 1, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881893, "raptor_stability": 0.8481273055076599, "fragility": 1.0, "augmentation_robustness": 0.8003220169247562, "xie_feature_dispersion": 2.320769849966657, "pac": 0.8242246612162081, "whitened_cosine_id": 0.8769779801368713}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.31678079404561216, "raptor_stability": 0.8443418741226196, "fragility": 1.5, "augmentation_robustness": 0.8372749145627374, "xie_feature_dispersion": 6.763574998208828, "pac": 0.8408083943426785, "whitened_cosine_id": 0.8846130967140198}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.08895215281835345, "raptor_stability": 0.9383633732795715, "fragility": 3.0, "augmentation_robustness": 0.8350381950983168, "xie_feature_dispersion": 10.874814823071377, "pac": 0.8867007841889442, "whitened_cosine_id": 0.8853277564048767}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 1, "layer": 16, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878747, "raptor_stability": 0.7895128726959229, "fragility": 0.25, "augmentation_robustness": 0.7721623022098749, "xie_feature_dispersion": 0.6522465207916613, "pac": 0.7808375874528989, "whitened_cosine_id": 0.7736839652061462}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 1, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5847740085808558, "raptor_stability": 0.8644301295280457, "fragility": 1.0, "augmentation_robustness": 0.7506737064210864, "xie_feature_dispersion": 2.0251662058673543, "pac": 0.807551917974566, "whitened_cosine_id": 0.8443336486816406}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879302, "raptor_stability": 0.8073345422744751, "fragility": 0.25, "augmentation_robustness": 0.7273212412855212, "xie_feature_dispersion": 5.356927418014672, "pac": 0.7673278917799982, "whitened_cosine_id": 0.8217640519142151}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 1, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956880057, "raptor_stability": 0.8053412437438965, "fragility": 0.25, "augmentation_robustness": 0.6599905871716598, "xie_feature_dispersion": 2.9578595858885146, "pac": 0.7326659154577781, "whitened_cosine_id": 0.8191479444503784}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 1, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687935, "raptor_stability": 0.7920674681663513, "fragility": 0.25, "augmentation_robustness": 0.7530199776918786, "xie_feature_dispersion": 2.0304992840234406, "pac": 0.772543722929115, "whitened_cosine_id": 0.8043911457061768}} +{"model": "pythia-410m", "dataset": "subj", "seed": 1, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882515, "raptor_stability": 0.8890897631645203, "fragility": 0.5, "augmentation_robustness": 0.774157157496381, "xie_feature_dispersion": 1.4231569822537715, "pac": 0.8316234603304506, "whitened_cosine_id": 0.8959898948669434}} +{"model": "pythia-410m", "dataset": "spam", "seed": 1, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882854, "raptor_stability": 0.9181560277938843, "fragility": 3.0, "augmentation_robustness": 0.8613923232327275, "xie_feature_dispersion": 13.35429844074093, "pac": 0.889774175513306, "whitened_cosine_id": 0.9216764569282532}} +{"model": "pythia-410m", "dataset": "cola", "seed": 1, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879537, "raptor_stability": 0.8038613796234131, "fragility": 0.25, "augmentation_robustness": 0.5332025416049512, "xie_feature_dispersion": 0.5611200077386973, "pac": 0.6685319606141822, "whitened_cosine_id": 0.8066012859344482}} +{"model": "pythia-410m", "dataset": "stance", "seed": 1, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.5266347984206482, "raptor_stability": 0.8334753513336182, "fragility": 0.5, "augmentation_robustness": 0.8255680326919436, "xie_feature_dispersion": 2.548552169168803, "pac": 0.8295216920127808, "whitened_cosine_id": 0.82421875}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 1, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.210307295688135, "raptor_stability": 0.8675035238265991, "fragility": 1.0, "augmentation_robustness": 0.7796142351617439, "xie_feature_dispersion": 2.10629803849922, "pac": 0.8235588794941715, "whitened_cosine_id": 0.8668701648712158}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 1, "layer": 22, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8375368714332581, "fragility": 0.25, "augmentation_robustness": 0.5707958641951811, "xie_feature_dispersion": 2.2166913426405066, "pac": 0.7041663678142196, "whitened_cosine_id": 0.8229580521583557}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8483022451400757, "fragility": 1.0, "augmentation_robustness": 0.773674431811487, "xie_feature_dispersion": 2.6155304376742743, "pac": 0.8109883384757813, "whitened_cosine_id": 0.8468784689903259}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 1, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.34505361071401985, "raptor_stability": 0.8623528480529785, "fragility": 1.5, "augmentation_robustness": 0.866225226560754, "xie_feature_dispersion": 4.689887830271909, "pac": 0.8642890373068662, "whitened_cosine_id": 0.8534741997718811}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.058822548518337014, "raptor_stability": 0.9422677159309387, "fragility": 2.0, "augmentation_robustness": 0.8626061839556527, "xie_feature_dispersion": 8.534984669917208, "pac": 0.9024369499432957, "whitened_cosine_id": 0.8359106779098511}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 1, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010858, "raptor_stability": 0.7976512312889099, "fragility": 0.25, "augmentation_robustness": 0.743646934055148, "xie_feature_dispersion": 0.8601541662036642, "pac": 0.7706490826720289, "whitened_cosine_id": 0.8195480704307556}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 1, "layer": 19, "concept": "emotion", "predictors": {"sip_eigengap": 0.40445308965852644, "raptor_stability": 0.8435872197151184, "fragility": 0.5, "augmentation_robustness": 0.7387802241177431, "xie_feature_dispersion": 2.4410346204498894, "pac": 0.7911837219164308, "whitened_cosine_id": 0.8220205903053284}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 1, "layer": 23, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.7921804785728455, "fragility": 0.25, "augmentation_robustness": 0.6227259022979126, "xie_feature_dispersion": 3.0062653097311682, "pac": 0.707453190435379, "whitened_cosine_id": 0.8225616812705994}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 1, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010583, "raptor_stability": 0.8012144565582275, "fragility": 0.25, "augmentation_robustness": 0.6679021398605163, "xie_feature_dispersion": 3.9129339183725933, "pac": 0.7345582982093719, "whitened_cosine_id": 0.8250377774238586}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 1, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010749, "raptor_stability": 0.7956711649894714, "fragility": 0.25, "augmentation_robustness": 0.7021864744022904, "xie_feature_dispersion": 2.452541305670078, "pac": 0.748928819695881, "whitened_cosine_id": 0.8224895000457764}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 1, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010877, "raptor_stability": 0.8905167579650879, "fragility": 1.0, "augmentation_robustness": 0.7827455359108311, "xie_feature_dispersion": 2.6061836065846578, "pac": 0.8366311469379595, "whitened_cosine_id": 0.8333907723426819}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 1, "layer": 13, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010649, "raptor_stability": 0.9169431924819946, "fragility": 1.5, "augmentation_robustness": 0.8734181256563801, "xie_feature_dispersion": 22.457465915951406, "pac": 0.8951806590691873, "whitened_cosine_id": 0.8557054400444031}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 1, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.855816496100951, "raptor_stability": 0.8068552017211914, "fragility": 0.25, "augmentation_robustness": 0.5148135283708677, "xie_feature_dispersion": 0.6902828208408504, "pac": 0.6608343650460295, "whitened_cosine_id": 0.8203985691070557}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 1, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.4012383857529635, "raptor_stability": 0.8363339304924011, "fragility": 0.5, "augmentation_robustness": 0.8057210469674079, "xie_feature_dispersion": 3.057908787470783, "pac": 0.8210274887299045, "whitened_cosine_id": 0.8217589259147644}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 1, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010833, "raptor_stability": 0.8418669700622559, "fragility": 1.0, "augmentation_robustness": 0.7293225446950196, "xie_feature_dispersion": 1.6761850609669111, "pac": 0.7855947573786377, "whitened_cosine_id": 0.8340412378311157}} +{"model": "gpt2", "dataset": "sst2", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.39754248593502, "raptor_stability": 0.8069735169410706, "fragility": 0.25, "augmentation_robustness": 0.5092073410955438, "xie_feature_dispersion": 1.648414982559877, "pac": 0.6580904290183072, "whitened_cosine_id": 0.8618525862693787}} +{"model": "gpt2", "dataset": "imdb", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350674, "raptor_stability": 0.8155258893966675, "fragility": 0.25, "augmentation_robustness": 0.6918368797960317, "xie_feature_dispersion": 1.679121568516715, "pac": 0.7536813845963496, "whitened_cosine_id": 0.8795225024223328}} +{"model": "gpt2", "dataset": "ag_news", "seed": 1, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.3994197076471763, "raptor_stability": 0.8592859506607056, "fragility": 0.5, "augmentation_robustness": 0.8643756652798957, "xie_feature_dispersion": 1.95723297240118, "pac": 0.8618308079703006, "whitened_cosine_id": 0.9026321768760681}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.13768748508218773, "raptor_stability": 0.9254063367843628, "fragility": 1.5, "augmentation_robustness": 0.7987703771800989, "xie_feature_dispersion": 5.392898240224899, "pac": 0.8620883569822309, "whitened_cosine_id": 0.9057890176773071}} +{"model": "gpt2", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342838, "raptor_stability": 0.7716324925422668, "fragility": 0.25, "augmentation_robustness": 0.7788427795247373, "xie_feature_dispersion": 0.5555909547173986, "pac": 0.775237636033502, "whitened_cosine_id": 0.7441529631614685}} +{"model": "gpt2", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5908930964643848, "raptor_stability": 0.8605509400367737, "fragility": 0.25, "augmentation_robustness": 0.7363054353676359, "xie_feature_dispersion": 1.3002246100849617, "pac": 0.7984281877022048, "whitened_cosine_id": 0.859701931476593}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348176, "raptor_stability": 0.8326132297515869, "fragility": 0.25, "augmentation_robustness": 0.7692316541993737, "xie_feature_dispersion": 0.8694719793739331, "pac": 0.8009224419754803, "whitened_cosine_id": 0.8273347020149231}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 1, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934648, "raptor_stability": 0.775840163230896, "fragility": 0.25, "augmentation_robustness": 0.6870212631337806, "xie_feature_dispersion": 2.3737390065905095, "pac": 0.7314307131823383, "whitened_cosine_id": 0.7987164855003357}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 1, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346262, "raptor_stability": 0.8207367062568665, "fragility": 0.25, "augmentation_robustness": 0.7990409525750108, "xie_feature_dispersion": 1.0977706807422258, "pac": 0.8098888294159386, "whitened_cosine_id": 0.7928319573402405}} +{"model": "gpt2", "dataset": "subj", "seed": 1, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351993, "raptor_stability": 0.8593925833702087, "fragility": 0.5, "augmentation_robustness": 0.6926262488173032, "xie_feature_dispersion": 1.2545052058156476, "pac": 0.776009416093756, "whitened_cosine_id": 0.9041862487792969}} +{"model": "gpt2", "dataset": "spam", "seed": 1, "layer": 12, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353376, "raptor_stability": 0.9029030203819275, "fragility": 0.25, "augmentation_robustness": 0.835250541372843, "xie_feature_dispersion": 1.4550691568045606, "pac": 0.8690767808773853, "whitened_cosine_id": 0.9343510270118713}} +{"model": "gpt2", "dataset": "cola", "seed": 1, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344225, "raptor_stability": 0.804100751876831, "fragility": 0.25, "augmentation_robustness": 0.6480747813745288, "xie_feature_dispersion": 0.35759111190860965, "pac": 0.72608776662568, "whitened_cosine_id": 0.7744043469429016}} +{"model": "gpt2", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.5721762371971069, "raptor_stability": 0.8337757587432861, "fragility": 0.5, "augmentation_robustness": 0.7933240817840598, "xie_feature_dispersion": 1.1954515513989168, "pac": 0.8135499202636729, "whitened_cosine_id": 0.8244547843933105}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 1, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934976, "raptor_stability": 0.854945182800293, "fragility": 0.25, "augmentation_robustness": 0.8063951328046506, "xie_feature_dispersion": 0.46463961416788957, "pac": 0.8306701578024718, "whitened_cosine_id": 0.8690304160118103}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 1, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688203, "raptor_stability": 0.8271149396896362, "fragility": 0.25, "augmentation_robustness": 0.5702797141411624, "xie_feature_dispersion": 1.6809287391267713, "pac": 0.6986973269153993, "whitened_cosine_id": 0.8607844114303589}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 1, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 1.210307295687949, "raptor_stability": 0.791394829750061, "fragility": 0.25, "augmentation_robustness": 0.8160921898342837, "xie_feature_dispersion": 0.5899893741394086, "pac": 0.8037435097921724, "whitened_cosine_id": 0.795337975025177}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5592522550238864, "raptor_stability": 0.8738617300987244, "fragility": 0.25, "augmentation_robustness": 0.7757794731716269, "xie_feature_dispersion": 0.9112899816215909, "pac": 0.8248206016351756, "whitened_cosine_id": 0.8513531684875488}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 1, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880974, "raptor_stability": 0.8126772046089172, "fragility": 0.25, "augmentation_robustness": 0.7615894417916816, "xie_feature_dispersion": 1.322075514358211, "pac": 0.7871333232002995, "whitened_cosine_id": 0.8305314779281616}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 1, "layer": 9, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879932, "raptor_stability": 0.7768864035606384, "fragility": 0.25, "augmentation_robustness": 0.6407154851585959, "xie_feature_dispersion": 2.304316868570724, "pac": 0.7088009443596172, "whitened_cosine_id": 0.7996089458465576}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 1, "layer": 24, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687991, "raptor_stability": 0.808860182762146, "fragility": 0.25, "augmentation_robustness": 0.779057007645201, "xie_feature_dispersion": 1.0954582686409513, "pac": 0.7939585952036735, "whitened_cosine_id": 0.8093989491462708}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 1, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882626, "raptor_stability": 0.8778501749038696, "fragility": 0.5, "augmentation_robustness": 0.7528926011129462, "xie_feature_dispersion": 1.211853663340529, "pac": 0.8153713880084079, "whitened_cosine_id": 0.8855262398719788}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 1, "layer": 16, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880065, "raptor_stability": 0.8000022768974304, "fragility": 0.25, "augmentation_robustness": 0.5320479658836585, "xie_feature_dispersion": 0.37533178694301084, "pac": 0.6660251213905445, "whitened_cosine_id": 0.8167566061019897}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 1, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.4372306810871382, "raptor_stability": 0.8121774792671204, "fragility": 0.5, "augmentation_robustness": 0.7829438977858721, "xie_feature_dispersion": 1.6812559297796246, "pac": 0.7975606885264963, "whitened_cosine_id": 0.7655757069587708}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 1, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.210307295688115, "raptor_stability": 0.8233330845832825, "fragility": 0.25, "augmentation_robustness": 0.6852795330724812, "xie_feature_dispersion": 0.620429769989345, "pac": 0.7543063088278819, "whitened_cosine_id": 0.856203019618988}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 1, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237649584, "raptor_stability": 0.8650567531585693, "fragility": 0.25, "augmentation_robustness": 0.6688278069812248, "xie_feature_dispersion": 2.1604793022510873, "pac": 0.7669422800698971, "whitened_cosine_id": 0.8860111236572266}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 1, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923765019, "raptor_stability": 0.8548077940940857, "fragility": 1.0, "augmentation_robustness": 0.8296912251707328, "xie_feature_dispersion": 2.619265146348305, "pac": 0.8422495096324092, "whitened_cosine_id": 0.8861786127090454}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.3284735817230639, "raptor_stability": 0.865305483341217, "fragility": 1.0, "augmentation_robustness": 0.8707412223835058, "xie_feature_dispersion": 3.3146556682562625, "pac": 0.8680233528623614, "whitened_cosine_id": 0.9071072340011597}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 1, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.06700207827316693, "raptor_stability": 0.9375728368759155, "fragility": 3.0, "augmentation_robustness": 0.8370451297198389, "xie_feature_dispersion": 9.643792516378186, "pac": 0.8873089832978772, "whitened_cosine_id": 0.8979226350784302}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646287, "raptor_stability": 0.8297770023345947, "fragility": 0.25, "augmentation_robustness": 0.8462807577423243, "xie_feature_dispersion": 0.5968347455995571, "pac": 0.8380288800384594, "whitened_cosine_id": 0.7946618795394897}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 1, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.6161992460964921, "raptor_stability": 0.882070004940033, "fragility": 0.25, "augmentation_robustness": 0.8630147174643309, "xie_feature_dispersion": 2.820954214278998, "pac": 0.8725423612021819, "whitened_cosine_id": 0.8536849617958069}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 1, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645619, "raptor_stability": 0.8290610313415527, "fragility": 0.25, "augmentation_robustness": 0.7926104311252984, "xie_feature_dispersion": 5.467260144092137, "pac": 0.8108357312334256, "whitened_cosine_id": 0.8316267728805542}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 1, "layer": 20, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647093, "raptor_stability": 0.7731820940971375, "fragility": 0.25, "augmentation_robustness": 0.6355369799746609, "xie_feature_dispersion": 2.999448892433556, "pac": 0.7043595370358992, "whitened_cosine_id": 0.798532247543335}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 1, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237645867, "raptor_stability": 0.819866955280304, "fragility": 0.25, "augmentation_robustness": 0.8020143361928123, "xie_feature_dispersion": 2.0378126951408726, "pac": 0.8109406457365581, "whitened_cosine_id": 0.8144164681434631}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 1, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.293872923765124, "raptor_stability": 0.8871787190437317, "fragility": 0.5, "augmentation_robustness": 0.7845579254901294, "xie_feature_dispersion": 1.6425796792755814, "pac": 0.8358683222669305, "whitened_cosine_id": 0.9134312868118286}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 1, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237639786, "raptor_stability": 0.9124530553817749, "fragility": 0.5, "augmentation_robustness": 0.907099253659993, "xie_feature_dispersion": 18.266615233549604, "pac": 0.909776154520884, "whitened_cosine_id": 0.9418519139289856}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 1, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647346, "raptor_stability": 0.8199473023414612, "fragility": 0.25, "augmentation_robustness": 0.5916327937131608, "xie_feature_dispersion": 0.5437385315920449, "pac": 0.7057900480273109, "whitened_cosine_id": 0.8188185691833496}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 1, "layer": 22, "concept": "stance", "predictors": {"sip_eigengap": 0.5082992376979554, "raptor_stability": 0.8089979290962219, "fragility": 1.0, "augmentation_robustness": 0.7725685252374112, "xie_feature_dispersion": 2.217566060904384, "pac": 0.7907832271668165, "whitened_cosine_id": 0.8215915560722351}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 1, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649438, "raptor_stability": 0.8504278063774109, "fragility": 0.25, "augmentation_robustness": 0.7801897363682817, "xie_feature_dispersion": 0.4453040461938641, "pac": 0.8153087713728463, "whitened_cosine_id": 0.8711152076721191}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 2, "layer": 1, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996196, "raptor_stability": 0.8725253343582153, "fragility": 1.0, "augmentation_robustness": 0.7320160436555694, "xie_feature_dispersion": 1.8850577574441822, "pac": 0.8022706890068924, "whitened_cosine_id": 0.8443945646286011}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998612, "raptor_stability": 0.8727290630340576, "fragility": 1.0, "augmentation_robustness": 0.8893448732634346, "xie_feature_dispersion": 1.1304006682636438, "pac": 0.8810369681487461, "whitened_cosine_id": 0.8432037830352783}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5762362270014945, "raptor_stability": 0.8451294302940369, "fragility": 1.0, "augmentation_robustness": 0.8227084099515827, "xie_feature_dispersion": 4.598590726152205, "pac": 0.8339189201228098, "whitened_cosine_id": 0.9046899676322937}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.13862003115739582, "raptor_stability": 0.9396008253097534, "fragility": 2.0, "augmentation_robustness": 0.8714397420888463, "xie_feature_dispersion": 7.460918475973291, "pac": 0.9055202836992999, "whitened_cosine_id": 0.9276723265647888}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921976638, "raptor_stability": 0.7323077321052551, "fragility": 0.25, "augmentation_robustness": 0.6404787869477748, "xie_feature_dispersion": 0.06278074668041304, "pac": 0.686393259526515, "whitened_cosine_id": 0.7365422248840332}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 2, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.6201478550392475, "raptor_stability": 0.860109269618988, "fragility": 0.5, "augmentation_robustness": 0.7869247885334385, "xie_feature_dispersion": 1.544046125551409, "pac": 0.8235170290762133, "whitened_cosine_id": 0.841076672077179}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 2, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921991476, "raptor_stability": 0.8527593612670898, "fragility": 1.0, "augmentation_robustness": 0.794283550859188, "xie_feature_dispersion": 1.5619058370885555, "pac": 0.823521456063139, "whitened_cosine_id": 0.7776172161102295}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987974, "raptor_stability": 0.8192676901817322, "fragility": 1.0, "augmentation_robustness": 0.7427685123086125, "xie_feature_dispersion": 1.857559017759172, "pac": 0.7810181012451723, "whitened_cosine_id": 0.7365460991859436}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990956, "raptor_stability": 0.8345112800598145, "fragility": 1.0, "augmentation_robustness": 0.7886622062234387, "xie_feature_dispersion": 1.342889203260779, "pac": 0.8115867431416266, "whitened_cosine_id": 0.8005088567733765}} +{"model": "pythia-70m", "dataset": "subj", "seed": 2, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.711632992200211, "raptor_stability": 0.8931885957717896, "fragility": 1.5, "augmentation_robustness": 0.8300863608388453, "xie_feature_dispersion": 2.385960236411005, "pac": 0.8616374783053175, "whitened_cosine_id": 0.9023992419242859}} +{"model": "pythia-70m", "dataset": "spam", "seed": 2, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004394, "raptor_stability": 0.9427627921104431, "fragility": 1.5, "augmentation_robustness": 0.9205893937056647, "xie_feature_dispersion": 10.434188689895425, "pac": 0.9316760929080539, "whitened_cosine_id": 0.9575247764587402}} +{"model": "pythia-70m", "dataset": "cola", "seed": 2, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921978055, "raptor_stability": 0.7819207310676575, "fragility": 1.0, "augmentation_robustness": 0.7494339269084461, "xie_feature_dispersion": 0.5849615167254145, "pac": 0.7656773289880519, "whitened_cosine_id": 0.7173330187797546}} +{"model": "pythia-70m", "dataset": "stance", "seed": 2, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.467708916223557, "raptor_stability": 0.823031485080719, "fragility": 1.0, "augmentation_robustness": 0.8142662820172029, "xie_feature_dispersion": 1.6869566033029149, "pac": 0.8186488835489609, "whitened_cosine_id": 0.8191723823547363}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 2, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998414, "raptor_stability": 0.8615893721580505, "fragility": 0.5, "augmentation_robustness": 0.8216595763224332, "xie_feature_dispersion": 0.9880768598724781, "pac": 0.8416244742402419, "whitened_cosine_id": 0.8701478838920593}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859348855, "raptor_stability": 0.8407484292984009, "fragility": 0.25, "augmentation_robustness": 0.5918794762745934, "xie_feature_dispersion": 2.033621000673696, "pac": 0.7163139527864971, "whitened_cosine_id": 0.8444841504096985}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 2, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350019, "raptor_stability": 0.8570092916488647, "fragility": 1.0, "augmentation_robustness": 0.8472864724655493, "xie_feature_dispersion": 1.332977638352146, "pac": 0.852147882057207, "whitened_cosine_id": 0.8642057180404663}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.4756762658594781, "raptor_stability": 0.8755650520324707, "fragility": 1.0, "augmentation_robustness": 0.8736053793270782, "xie_feature_dispersion": 2.6831447841823626, "pac": 0.8745852156797744, "whitened_cosine_id": 0.9056297540664673}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.10936523074513556, "raptor_stability": 0.938740074634552, "fragility": 2.0, "augmentation_robustness": 0.849824439887086, "xie_feature_dispersion": 8.126249426404561, "pac": 0.894282257260819, "whitened_cosine_id": 0.9102428555488586}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343164, "raptor_stability": 0.7734526991844177, "fragility": 0.25, "augmentation_robustness": 0.7341807784351317, "xie_feature_dispersion": 0.24746600780733177, "pac": 0.7538167388097747, "whitened_cosine_id": 0.7335423827171326}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 2, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.5898968502609128, "raptor_stability": 0.8509833216667175, "fragility": 0.25, "augmentation_robustness": 0.7503952782199734, "xie_feature_dispersion": 1.9109009411804163, "pac": 0.8006892999433455, "whitened_cosine_id": 0.8450579643249512}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 2, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347338, "raptor_stability": 0.8163250684738159, "fragility": 0.25, "augmentation_robustness": 0.720428393459381, "xie_feature_dispersion": 1.672145195874313, "pac": 0.7683767309665985, "whitened_cosine_id": 0.8197868466377258}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345715, "raptor_stability": 0.8170239925384521, "fragility": 1.0, "augmentation_robustness": 0.7672939677957877, "xie_feature_dispersion": 2.1306331599217914, "pac": 0.7921589801671199, "whitened_cosine_id": 0.7781753540039062}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 2, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345553, "raptor_stability": 0.8093597888946533, "fragility": 0.25, "augmentation_robustness": 0.7487295667365286, "xie_feature_dispersion": 1.9731796308382967, "pac": 0.779044677815591, "whitened_cosine_id": 0.7859788537025452}} +{"model": "pythia-160m", "dataset": "subj", "seed": 2, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935202, "raptor_stability": 0.8897599577903748, "fragility": 1.0, "augmentation_robustness": 0.8051559896221852, "xie_feature_dispersion": 1.9497282630608124, "pac": 0.84745797370628, "whitened_cosine_id": 0.9099186658859253}} +{"model": "pythia-160m", "dataset": "spam", "seed": 2, "layer": 8, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935196, "raptor_stability": 0.9361570477485657, "fragility": 1.5, "augmentation_robustness": 0.9125687730559409, "xie_feature_dispersion": 12.123957991903673, "pac": 0.9243629104022533, "whitened_cosine_id": 0.9451765418052673}} +{"model": "pythia-160m", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934449, "raptor_stability": 0.7818026542663574, "fragility": 0.25, "augmentation_robustness": 0.6248388717067496, "xie_feature_dispersion": 0.36909274612641374, "pac": 0.7033207629865534, "whitened_cosine_id": 0.7660055756568909}} +{"model": "pythia-160m", "dataset": "stance", "seed": 2, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.550133319968111, "raptor_stability": 0.8319461345672607, "fragility": 1.5, "augmentation_robustness": 0.8296352226027173, "xie_feature_dispersion": 2.567878497868356, "pac": 0.830790678584989, "whitened_cosine_id": 0.8248243927955627}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350152, "raptor_stability": 0.8500860929489136, "fragility": 0.5, "augmentation_robustness": 0.7916745698533383, "xie_feature_dispersion": 1.015287990938809, "pac": 0.8208803314011259, "whitened_cosine_id": 0.8699454069137573}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880925, "raptor_stability": 0.8507713079452515, "fragility": 0.25, "augmentation_robustness": 0.6548658634701003, "xie_feature_dispersion": 2.5382673428676514, "pac": 0.7528185857076759, "whitened_cosine_id": 0.8639923334121704}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881815, "raptor_stability": 0.8507294654846191, "fragility": 1.0, "augmentation_robustness": 0.8077367041653364, "xie_feature_dispersion": 2.1999878871216194, "pac": 0.8292330848249778, "whitened_cosine_id": 0.876013994216919}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 2, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.48444830765922703, "raptor_stability": 0.8797903060913086, "fragility": 1.5, "augmentation_robustness": 0.8897077200880827, "xie_feature_dispersion": 4.318369085246188, "pac": 0.8847490130896957, "whitened_cosine_id": 0.8930330276489258}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06934869454749534, "raptor_stability": 0.9469916224479675, "fragility": 2.0, "augmentation_robustness": 0.8838004807942493, "xie_feature_dispersion": 7.256347007074097, "pac": 0.9153960516211084, "whitened_cosine_id": 0.8971635699272156}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878984, "raptor_stability": 0.7757969498634338, "fragility": 0.25, "augmentation_robustness": 0.7602377504416505, "xie_feature_dispersion": 0.20512071078637148, "pac": 0.7680173501525422, "whitened_cosine_id": 0.7685407996177673}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 2, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.5276419919506735, "raptor_stability": 0.8558531403541565, "fragility": 0.25, "augmentation_robustness": 0.7811194669553142, "xie_feature_dispersion": 2.490653998171255, "pac": 0.8184863036547354, "whitened_cosine_id": 0.8436367511749268}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 2, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880245, "raptor_stability": 0.8074635863304138, "fragility": 0.5, "augmentation_robustness": 0.7107950361323764, "xie_feature_dispersion": 1.9566993340581085, "pac": 0.759129311231395, "whitened_cosine_id": 0.8125779628753662}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295687971, "raptor_stability": 0.7986416220664978, "fragility": 0.5, "augmentation_robustness": 0.6903031510581926, "xie_feature_dispersion": 1.8402433518812902, "pac": 0.7444723865623453, "whitened_cosine_id": 0.7969480156898499}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 2, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879715, "raptor_stability": 0.8084327578544617, "fragility": 0.5, "augmentation_robustness": 0.7257825239499871, "xie_feature_dispersion": 1.7043242651025499, "pac": 0.7671076409022244, "whitened_cosine_id": 0.7994747161865234}} +{"model": "pythia-410m", "dataset": "subj", "seed": 2, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688262, "raptor_stability": 0.8808537721633911, "fragility": 0.5, "augmentation_robustness": 0.7730695710492665, "xie_feature_dispersion": 1.6627162013824885, "pac": 0.8269616716063288, "whitened_cosine_id": 0.8980955481529236}} +{"model": "pythia-410m", "dataset": "spam", "seed": 2, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956871068, "raptor_stability": 0.9453094601631165, "fragility": 1.0, "augmentation_robustness": 0.931993646437433, "xie_feature_dispersion": 17.657395386814123, "pac": 0.9386515533002747, "whitened_cosine_id": 0.9302859306335449}} +{"model": "pythia-410m", "dataset": "cola", "seed": 2, "layer": 17, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879637, "raptor_stability": 0.7948347926139832, "fragility": 0.25, "augmentation_robustness": 0.6269925244354974, "xie_feature_dispersion": 0.47979217522467377, "pac": 0.7109136585247402, "whitened_cosine_id": 0.8036185503005981}} +{"model": "pythia-410m", "dataset": "stance", "seed": 2, "layer": 10, "concept": "stance", "predictors": {"sip_eigengap": 0.47549258334344463, "raptor_stability": 0.8262801170349121, "fragility": 0.5, "augmentation_robustness": 0.8202575512360268, "xie_feature_dispersion": 3.2823490583418935, "pac": 0.8232688341354695, "whitened_cosine_id": 0.8195136189460754}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 2, "layer": 18, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881415, "raptor_stability": 0.8284711837768555, "fragility": 0.5, "augmentation_robustness": 0.7694913743994085, "xie_feature_dispersion": 1.2261896328299637, "pac": 0.798981279088132, "whitened_cosine_id": 0.8577603101730347}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010692, "raptor_stability": 0.8523985147476196, "fragility": 0.25, "augmentation_robustness": 0.6344970850996317, "xie_feature_dispersion": 3.125584115559324, "pac": 0.7434477999236256, "whitened_cosine_id": 0.8270771503448486}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8408520221710205, "fragility": 1.0, "augmentation_robustness": 0.7462513872391434, "xie_feature_dispersion": 2.261174552822513, "pac": 0.793551704705082, "whitened_cosine_id": 0.8419722318649292}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.3112674068238993, "raptor_stability": 0.8652853965759277, "fragility": 1.0, "augmentation_robustness": 0.8618662021334369, "xie_feature_dispersion": 4.2348607484097185, "pac": 0.8635757993546823, "whitened_cosine_id": 0.8362071514129639}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05012914539691602, "raptor_stability": 0.9452676773071289, "fragility": 3.0, "augmentation_robustness": 0.8590266216701495, "xie_feature_dispersion": 9.524616008618372, "pac": 0.9021471494886393, "whitened_cosine_id": 0.8305992484092712}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.7967193126678467, "fragility": 0.25, "augmentation_robustness": 0.755117458917892, "xie_feature_dispersion": 0.32762786809964206, "pac": 0.7759183857928693, "whitened_cosine_id": 0.8186400532722473}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 2, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.3604267252609762, "raptor_stability": 0.8496060371398926, "fragility": 0.5, "augmentation_robustness": 0.7497415128763687, "xie_feature_dispersion": 2.6993559873766775, "pac": 0.7996737750081306, "whitened_cosine_id": 0.8220704197883606}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 2, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8090983033180237, "fragility": 0.25, "augmentation_robustness": 0.6782138616148163, "xie_feature_dispersion": 3.2068116736663326, "pac": 0.74365608246642, "whitened_cosine_id": 0.8261517286300659}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 2, "layer": 7, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010822, "raptor_stability": 0.8073044419288635, "fragility": 0.25, "augmentation_robustness": 0.6867498455562944, "xie_feature_dispersion": 4.820388126528082, "pac": 0.747027143742579, "whitened_cosine_id": 0.8241050839424133}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 2, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010454, "raptor_stability": 0.7936219573020935, "fragility": 0.25, "augmentation_robustness": 0.6887170523418908, "xie_feature_dispersion": 2.728943510383841, "pac": 0.7411695048219922, "whitened_cosine_id": 0.8242641091346741}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 2, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010878, "raptor_stability": 0.8761361837387085, "fragility": 1.0, "augmentation_robustness": 0.7502694820898129, "xie_feature_dispersion": 2.7954401496665007, "pac": 0.8132028329142607, "whitened_cosine_id": 0.829967737197876}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 2, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010256, "raptor_stability": 0.9437923431396484, "fragility": 1.0, "augmentation_robustness": 0.9235808889611689, "xie_feature_dispersion": 23.76038075342399, "pac": 0.9336866160504087, "whitened_cosine_id": 0.8749403953552246}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 2, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.8045715689659119, "fragility": 0.25, "augmentation_robustness": 0.5269196391645194, "xie_feature_dispersion": 0.5939571817383262, "pac": 0.6657456040652157, "whitened_cosine_id": 0.8205759525299072}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 2, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.33865901445727387, "raptor_stability": 0.8298296332359314, "fragility": 1.0, "augmentation_robustness": 0.817334225300112, "xie_feature_dispersion": 3.230859359898638, "pac": 0.8235819292680218, "whitened_cosine_id": 0.8170763850212097}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 2, "layer": 14, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010837, "raptor_stability": 0.8219150900840759, "fragility": 0.5, "augmentation_robustness": 0.7537574160413394, "xie_feature_dispersion": 1.8431532727005182, "pac": 0.7878362530627077, "whitened_cosine_id": 0.8218834400177002}} +{"model": "gpt2", "dataset": "sst2", "seed": 2, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350012, "raptor_stability": 0.8311133980751038, "fragility": 0.25, "augmentation_robustness": 0.5968769748584443, "xie_feature_dispersion": 2.0407834350862544, "pac": 0.713995186466774, "whitened_cosine_id": 0.8748435378074646}} +{"model": "gpt2", "dataset": "imdb", "seed": 2, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350318, "raptor_stability": 0.8567911386489868, "fragility": 0.25, "augmentation_robustness": 0.8673792464607593, "xie_feature_dispersion": 0.26606041071725683, "pac": 0.862085192554873, "whitened_cosine_id": 0.8767779469490051}} +{"model": "gpt2", "dataset": "ag_news", "seed": 2, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.3821667228940741, "raptor_stability": 0.845832109451294, "fragility": 0.5, "augmentation_robustness": 0.8470052851683942, "xie_feature_dispersion": 1.979682604320564, "pac": 0.8464186973098441, "whitened_cosine_id": 0.903249979019165}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 2, "layer": 7, "concept": "topic", "predictors": {"sip_eigengap": 0.11558697704158609, "raptor_stability": 0.932966411113739, "fragility": 1.0, "augmentation_robustness": 0.8559305238325257, "xie_feature_dispersion": 4.299506274009608, "pac": 0.8944484674731323, "whitened_cosine_id": 0.9021454453468323}} +{"model": "gpt2", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343073, "raptor_stability": 0.7622057199478149, "fragility": 0.25, "augmentation_robustness": 0.7918529895588636, "xie_feature_dispersion": 0.33360574253280906, "pac": 0.7770293547533393, "whitened_cosine_id": 0.7459770441055298}} +{"model": "gpt2", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5200547285263808, "raptor_stability": 0.85597825050354, "fragility": 0.25, "augmentation_robustness": 0.7622544507506325, "xie_feature_dispersion": 1.2683994964221992, "pac": 0.8091163506270862, "whitened_cosine_id": 0.8535709977149963}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 2, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.397542485934766, "raptor_stability": 0.8122290968894958, "fragility": 0.25, "augmentation_robustness": 0.7701636575129145, "xie_feature_dispersion": 0.6036175754728635, "pac": 0.7911963772012052, "whitened_cosine_id": 0.8273720741271973}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 2, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859344792, "raptor_stability": 0.7899907827377319, "fragility": 0.25, "augmentation_robustness": 0.775070377256843, "xie_feature_dispersion": 1.1126097391994054, "pac": 0.7825305799972875, "whitened_cosine_id": 0.7636452317237854}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346548, "raptor_stability": 0.7969914674758911, "fragility": 0.25, "augmentation_robustness": 0.7083710583755256, "xie_feature_dispersion": 1.8684634814954304, "pac": 0.7526812629257084, "whitened_cosine_id": 0.8185012936592102}} +{"model": "gpt2", "dataset": "subj", "seed": 2, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352377, "raptor_stability": 0.8654428720474243, "fragility": 0.5, "augmentation_robustness": 0.7478951378849938, "xie_feature_dispersion": 1.4494284784328508, "pac": 0.8066690049662091, "whitened_cosine_id": 0.9153006076812744}} +{"model": "gpt2", "dataset": "spam", "seed": 2, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935319, "raptor_stability": 0.9391940832138062, "fragility": 1.0, "augmentation_robustness": 0.9242307247684519, "xie_feature_dispersion": 13.972658689577166, "pac": 0.9317124039911291, "whitened_cosine_id": 0.9425793290138245}} +{"model": "gpt2", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859345233, "raptor_stability": 0.7784033417701721, "fragility": 0.25, "augmentation_robustness": 0.5640885986826307, "xie_feature_dispersion": 0.38566559649240273, "pac": 0.6712459702264014, "whitened_cosine_id": 0.7816012501716614}} +{"model": "gpt2", "dataset": "stance", "seed": 2, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.5282860197072011, "raptor_stability": 0.8207774758338928, "fragility": 0.25, "augmentation_robustness": 0.7971149286838108, "xie_feature_dispersion": 2.2792793583875905, "pac": 0.8089462022588518, "whitened_cosine_id": 0.8131434321403503}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 2, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934994, "raptor_stability": 0.808012068271637, "fragility": 0.25, "augmentation_robustness": 0.7076387264686993, "xie_feature_dispersion": 0.8358972401809576, "pac": 0.7578253973701681, "whitened_cosine_id": 0.8693022727966309}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 2, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881729, "raptor_stability": 0.8265162706375122, "fragility": 0.25, "augmentation_robustness": 0.5469683098756675, "xie_feature_dispersion": 2.1253471965524975, "pac": 0.6867422902565898, "whitened_cosine_id": 0.8618488311767578}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 2, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.35097833794424904, "raptor_stability": 0.8510255813598633, "fragility": 0.5, "augmentation_robustness": 0.868126023664805, "xie_feature_dispersion": 1.511481734340918, "pac": 0.8595758025123341, "whitened_cosine_id": 0.8767551183700562}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 2, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878433, "raptor_stability": 0.7515677809715271, "fragility": 0.25, "augmentation_robustness": 0.8471055642765235, "xie_feature_dispersion": 0.3165629770195801, "pac": 0.7993366726240253, "whitened_cosine_id": 0.7615948915481567}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5517110732147434, "raptor_stability": 0.8649501204490662, "fragility": 0.25, "augmentation_robustness": 0.7727700403781212, "xie_feature_dispersion": 0.9242099717833849, "pac": 0.8188600804135937, "whitened_cosine_id": 0.8501837849617004}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 2, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880352, "raptor_stability": 0.7788109183311462, "fragility": 0.25, "augmentation_robustness": 0.602822870481301, "xie_feature_dispersion": 2.344924657926786, "pac": 0.6908168944062236, "whitened_cosine_id": 0.8237905502319336}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 2, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295688015, "raptor_stability": 0.7836035490036011, "fragility": 0.25, "augmentation_robustness": 0.6445047638465966, "xie_feature_dispersion": 2.829344259675601, "pac": 0.7140541564250988, "whitened_cosine_id": 0.8160997629165649}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880054, "raptor_stability": 0.7852885723114014, "fragility": 0.25, "augmentation_robustness": 0.6693173469866233, "xie_feature_dispersion": 0.831323350714088, "pac": 0.7273029596490124, "whitened_cosine_id": 0.8087374567985535}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 2, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882837, "raptor_stability": 0.8820376992225647, "fragility": 0.5, "augmentation_robustness": 0.7285834645747785, "xie_feature_dispersion": 1.6491215963892647, "pac": 0.8053105818986717, "whitened_cosine_id": 0.8916261792182922}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 2, "layer": 19, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880483, "raptor_stability": 0.7810653448104858, "fragility": 0.25, "augmentation_robustness": 0.5184999598231592, "xie_feature_dispersion": 0.4506855901276054, "pac": 0.6497826523168225, "whitened_cosine_id": 0.8220202326774597}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 2, "layer": 24, "concept": "stance", "predictors": {"sip_eigengap": 0.4048263633949508, "raptor_stability": 0.8287909626960754, "fragility": 0.25, "augmentation_robustness": 0.8148653549769995, "xie_feature_dispersion": 0.8817497017545193, "pac": 0.8218281588365375, "whitened_cosine_id": 0.8191941380500793}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881538, "raptor_stability": 0.8170264959335327, "fragility": 0.25, "augmentation_robustness": 0.6658562775951427, "xie_feature_dispersion": 0.6916798290341591, "pac": 0.7414413867643377, "whitened_cosine_id": 0.8560160398483276}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923764946, "raptor_stability": 0.8504462838172913, "fragility": 0.25, "augmentation_robustness": 0.6405801888029593, "xie_feature_dispersion": 2.428287550117154, "pac": 0.7455132363101253, "whitened_cosine_id": 0.8769184350967407}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650424, "raptor_stability": 0.8790234923362732, "fragility": 1.0, "augmentation_robustness": 0.873146163657816, "xie_feature_dispersion": 2.109825225717331, "pac": 0.8760848279970446, "whitened_cosine_id": 0.8947857618331909}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.4777053541327048, "raptor_stability": 0.8827468156814575, "fragility": 0.5, "augmentation_robustness": 0.9086818995974959, "xie_feature_dispersion": 1.9664542049888671, "pac": 0.8957143576394767, "whitened_cosine_id": 0.9089540839195251}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05912307025918351, "raptor_stability": 0.9463669061660767, "fragility": 1.5, "augmentation_robustness": 0.8757474533027418, "xie_feature_dispersion": 6.768079984244763, "pac": 0.9110571797344093, "whitened_cosine_id": 0.9140043258666992}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 2, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646398, "raptor_stability": 0.8356521129608154, "fragility": 0.25, "augmentation_robustness": 0.8160727049350692, "xie_feature_dispersion": 0.23710320099138982, "pac": 0.8258624089479423, "whitened_cosine_id": 0.7763546705245972}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 2, "layer": 16, "concept": "emotion", "predictors": {"sip_eigengap": 0.48285928508127296, "raptor_stability": 0.8633195757865906, "fragility": 0.25, "augmentation_robustness": 0.8474371685693614, "xie_feature_dispersion": 2.284945012773296, "pac": 0.855378372177976, "whitened_cosine_id": 0.8482233881950378}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 2, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645508, "raptor_stability": 0.8284585475921631, "fragility": 0.25, "augmentation_robustness": 0.7660624849853699, "xie_feature_dispersion": 3.252034579020222, "pac": 0.7972605162887665, "whitened_cosine_id": 0.8060616850852966}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237642506, "raptor_stability": 0.8207340240478516, "fragility": 0.25, "augmentation_robustness": 0.733566852009586, "xie_feature_dispersion": 3.71696516386688, "pac": 0.7771504380287189, "whitened_cosine_id": 0.8012250661849976}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 2, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237644577, "raptor_stability": 0.8056316375732422, "fragility": 0.25, "augmentation_robustness": 0.7408950314947991, "xie_feature_dispersion": 2.6206447748583725, "pac": 0.7732633345340206, "whitened_cosine_id": 0.8052534461021423}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 2, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.29387292376514, "raptor_stability": 0.8994405269622803, "fragility": 0.5, "augmentation_robustness": 0.8304811500465553, "xie_feature_dispersion": 1.2535193562102473, "pac": 0.8649608385044179, "whitened_cosine_id": 0.9212391972541809}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 2, "layer": 13, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237650766, "raptor_stability": 0.9231064319610596, "fragility": 1.0, "augmentation_robustness": 0.9155044674854933, "xie_feature_dispersion": 18.157099012256424, "pac": 0.9193054497232764, "whitened_cosine_id": 0.9482038021087646}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 2, "layer": 14, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647517, "raptor_stability": 0.8107694983482361, "fragility": 0.25, "augmentation_robustness": 0.6126361282502457, "xie_feature_dispersion": 0.29378764734852275, "pac": 0.7117028132992409, "whitened_cosine_id": 0.8085699677467346}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 2, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.47687488644054804, "raptor_stability": 0.8371041417121887, "fragility": 0.25, "augmentation_robustness": 0.8437004456467965, "xie_feature_dispersion": 2.6985372907022853, "pac": 0.8404022936794926, "whitened_cosine_id": 0.8233371376991272}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 2, "layer": 14, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764935, "raptor_stability": 0.8342365026473999, "fragility": 0.25, "augmentation_robustness": 0.7913619262318612, "xie_feature_dispersion": 0.5692187925507834, "pac": 0.8127992144396305, "whitened_cosine_id": 0.8644318580627441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996696, "raptor_stability": 0.8488397598266602, "fragility": 0.25, "augmentation_robustness": 0.5708345486394, "xie_feature_dispersion": 2.9434633106744124, "pac": 0.7098371542330301, "whitened_cosine_id": 0.8510316610336304}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998183, "raptor_stability": 0.8842445015907288, "fragility": 1.0, "augmentation_robustness": 0.8984364327435839, "xie_feature_dispersion": 1.2037141877282427, "pac": 0.8913404671671563, "whitened_cosine_id": 0.8669403195381165}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 3, "layer": 1, "concept": "topic", "predictors": {"sip_eigengap": 0.6283839672205405, "raptor_stability": 0.9187504649162292, "fragility": 1.5, "augmentation_robustness": 0.9445047803748948, "xie_feature_dispersion": 3.4172576932701224, "pac": 0.931627622645562, "whitened_cosine_id": 0.9079165458679199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 3, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14101513130976387, "raptor_stability": 0.9509273767471313, "fragility": 1.5, "augmentation_robustness": 0.8868008425779641, "xie_feature_dispersion": 5.186993220693557, "pac": 0.9188641096625477, "whitened_cosine_id": 0.9339398145675659}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921976405, "raptor_stability": 0.7710468173027039, "fragility": 0.5, "augmentation_robustness": 0.7989766303576111, "xie_feature_dispersion": 0.4743571278941463, "pac": 0.7850117238301575, "whitened_cosine_id": 0.6775692105293274}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 3, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.635597069567996, "raptor_stability": 0.8650970458984375, "fragility": 1.0, "augmentation_robustness": 0.7848846813705812, "xie_feature_dispersion": 1.4784056981569556, "pac": 0.8249908636345094, "whitened_cosine_id": 0.847564160823822}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 3, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.711632992199183, "raptor_stability": 0.8229442834854126, "fragility": 0.5, "augmentation_robustness": 0.7320233370031859, "xie_feature_dispersion": 1.57080462526856, "pac": 0.7774838102442992, "whitened_cosine_id": 0.8052536249160767}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 3, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921988689, "raptor_stability": 0.8289263248443604, "fragility": 1.0, "augmentation_robustness": 0.7538693401418205, "xie_feature_dispersion": 1.8985337702191092, "pac": 0.7913978324930904, "whitened_cosine_id": 0.7646318078041077}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 3, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989895, "raptor_stability": 0.8404061198234558, "fragility": 1.0, "augmentation_robustness": 0.8369988993389674, "xie_feature_dispersion": 1.1920844617447306, "pac": 0.8387025095812116, "whitened_cosine_id": 0.7623035311698914}} +{"model": "pythia-70m", "dataset": "subj", "seed": 3, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002775, "raptor_stability": 0.899202823638916, "fragility": 1.0, "augmentation_robustness": 0.7801516491279672, "xie_feature_dispersion": 2.0439990482699595, "pac": 0.8396772363834416, "whitened_cosine_id": 0.9309227466583252}} +{"model": "pythia-70m", "dataset": "spam", "seed": 3, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004678, "raptor_stability": 0.9567713141441345, "fragility": 3.0, "augmentation_robustness": 0.9371847229705604, "xie_feature_dispersion": 6.424426785927112, "pac": 0.9469780185573475, "whitened_cosine_id": 0.9486836791038513}} +{"model": "pythia-70m", "dataset": "cola", "seed": 3, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921973576, "raptor_stability": 0.7688994407653809, "fragility": 1.0, "augmentation_robustness": 0.7138657324958716, "xie_feature_dispersion": 0.7233065123875068, "pac": 0.7413825866306263, "whitened_cosine_id": 0.6779297590255737}} +{"model": "pythia-70m", "dataset": "stance", "seed": 3, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.764248302629601, "raptor_stability": 0.8229038119316101, "fragility": 0.5, "augmentation_robustness": 0.7822202756663522, "xie_feature_dispersion": 2.080304032697393, "pac": 0.8025620437989811, "whitened_cosine_id": 0.8263651728630066}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 3, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.711632992199799, "raptor_stability": 0.8503726124763489, "fragility": 0.5, "augmentation_robustness": 0.8033091683218101, "xie_feature_dispersion": 1.2407015450743701, "pac": 0.8268408903990795, "whitened_cosine_id": 0.866639256477356}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 3, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349443, "raptor_stability": 0.8484313488006592, "fragility": 0.25, "augmentation_robustness": 0.637225872199634, "xie_feature_dispersion": 4.044131456746566, "pac": 0.7428286105001466, "whitened_cosine_id": 0.8518450260162354}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 3, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350145, "raptor_stability": 0.8780309557914734, "fragility": 1.0, "augmentation_robustness": 0.874329700092187, "xie_feature_dispersion": 1.4259597904886765, "pac": 0.8761803279418302, "whitened_cosine_id": 0.8702338933944702}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 3, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.35191923415776655, "raptor_stability": 0.8508845567703247, "fragility": 0.25, "augmentation_robustness": 0.8265529420495982, "xie_feature_dispersion": 0.9266248016633525, "pac": 0.8387187494099615, "whitened_cosine_id": 0.8987531661987305}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 3, "layer": 9, "concept": "topic", "predictors": {"sip_eigengap": 0.11774124797544638, "raptor_stability": 0.9511698484420776, "fragility": 1.5, "augmentation_robustness": 0.8745928253851561, "xie_feature_dispersion": 6.276849376401518, "pac": 0.9128813369136168, "whitened_cosine_id": 0.917269766330719}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342074, "raptor_stability": 0.78183913230896, "fragility": 0.5, "augmentation_robustness": 0.7622129099393176, "xie_feature_dispersion": 0.5886846655394216, "pac": 0.7720260211241388, "whitened_cosine_id": 0.7248976230621338}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 3, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5119239787691542, "raptor_stability": 0.8616406917572021, "fragility": 0.25, "augmentation_robustness": 0.7504844246649606, "xie_feature_dispersion": 1.6907559164308787, "pac": 0.8060625582110814, "whitened_cosine_id": 0.852891206741333}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 3, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347243, "raptor_stability": 0.8252765536308289, "fragility": 0.25, "augmentation_robustness": 0.7347966668215313, "xie_feature_dispersion": 2.064566431309704, "pac": 0.7800366102261801, "whitened_cosine_id": 0.8165638446807861}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 3, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345138, "raptor_stability": 0.7875639796257019, "fragility": 0.25, "augmentation_robustness": 0.6763370074729932, "xie_feature_dispersion": 3.2860294351071695, "pac": 0.7319504935493475, "whitened_cosine_id": 0.7773585915565491}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 3, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934679, "raptor_stability": 0.7985202670097351, "fragility": 0.25, "augmentation_robustness": 0.7440947615507673, "xie_feature_dispersion": 0.7467058167702149, "pac": 0.7713075142802512, "whitened_cosine_id": 0.7971633076667786}} +{"model": "pythia-160m", "dataset": "subj", "seed": 3, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935211, "raptor_stability": 0.8905059099197388, "fragility": 1.0, "augmentation_robustness": 0.7853861456321554, "xie_feature_dispersion": 2.317051234221485, "pac": 0.8379460277759471, "whitened_cosine_id": 0.913381040096283}} +{"model": "pythia-160m", "dataset": "spam", "seed": 3, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859350205, "raptor_stability": 0.9280425310134888, "fragility": 1.0, "augmentation_robustness": 0.9003377239515604, "xie_feature_dispersion": 13.64857545430944, "pac": 0.9141901274825246, "whitened_cosine_id": 0.9369125366210938}} +{"model": "pythia-160m", "dataset": "cola", "seed": 3, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343062, "raptor_stability": 0.7884514331817627, "fragility": 0.25, "augmentation_robustness": 0.6450635280093344, "xie_feature_dispersion": 0.17857729363649755, "pac": 0.7167574805955486, "whitened_cosine_id": 0.7649105191230774}} +{"model": "pythia-160m", "dataset": "stance", "seed": 3, "layer": 4, "concept": "stance", "predictors": {"sip_eigengap": 0.6506681129814839, "raptor_stability": 0.8327218890190125, "fragility": 0.5, "augmentation_robustness": 0.8189402967776248, "xie_feature_dispersion": 1.7873782154455453, "pac": 0.8258310928983186, "whitened_cosine_id": 0.8258171081542969}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 3, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350412, "raptor_stability": 0.8567770719528198, "fragility": 0.5, "augmentation_robustness": 0.7799380638785056, "xie_feature_dispersion": 1.305808488628754, "pac": 0.8183575679156627, "whitened_cosine_id": 0.8785260319709778}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 3, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881342, "raptor_stability": 0.8563058376312256, "fragility": 0.25, "augmentation_robustness": 0.6228553515841018, "xie_feature_dispersion": 4.914029083398948, "pac": 0.7395805946076637, "whitened_cosine_id": 0.8632848858833313}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 3, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688201, "raptor_stability": 0.8571096062660217, "fragility": 1.0, "augmentation_robustness": 0.8391733545717237, "xie_feature_dispersion": 2.1649660520641487, "pac": 0.8481414804188727, "whitened_cosine_id": 0.8794662356376648}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 3, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.3940956743493873, "raptor_stability": 0.8732040524482727, "fragility": 1.0, "augmentation_robustness": 0.8558702619852715, "xie_feature_dispersion": 3.495736654891566, "pac": 0.8645371572167722, "whitened_cosine_id": 0.894050121307373}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 3, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.0777198899969492, "raptor_stability": 0.9526127576828003, "fragility": 2.0, "augmentation_robustness": 0.8767089502212883, "xie_feature_dispersion": 7.889853603758875, "pac": 0.9146608539520442, "whitened_cosine_id": 0.8949448466300964}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 3, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879801, "raptor_stability": 0.8159152269363403, "fragility": 0.25, "augmentation_robustness": 0.7880748749868189, "xie_feature_dispersion": 0.24343282857388523, "pac": 0.8019950509615796, "whitened_cosine_id": 0.800447404384613}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 3, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.48221468653288235, "raptor_stability": 0.8573822379112244, "fragility": 0.25, "augmentation_robustness": 0.7660417601070145, "xie_feature_dispersion": 1.9953148592463423, "pac": 0.8117119990091195, "whitened_cosine_id": 0.8465678095817566}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 3, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879983, "raptor_stability": 0.8090613484382629, "fragility": 0.25, "augmentation_robustness": 0.7199613185198703, "xie_feature_dispersion": 4.138422068582218, "pac": 0.7645113334790666, "whitened_cosine_id": 0.8162167072296143}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 3, "layer": 9, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956877805, "raptor_stability": 0.8041320443153381, "fragility": 0.25, "augmentation_robustness": 0.6772706695322654, "xie_feature_dispersion": 4.201391999844916, "pac": 0.7407013569238018, "whitened_cosine_id": 0.8077095150947571}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 3, "layer": 13, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880465, "raptor_stability": 0.8054513335227966, "fragility": 0.25, "augmentation_robustness": 0.7134721107475187, "xie_feature_dispersion": 0.942681786130636, "pac": 0.7594617221351576, "whitened_cosine_id": 0.8189021944999695}} +{"model": "pythia-410m", "dataset": "subj", "seed": 3, "layer": 15, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882677, "raptor_stability": 0.908913254737854, "fragility": 1.0, "augmentation_robustness": 0.791121231421306, "xie_feature_dispersion": 1.6016867212241452, "pac": 0.85001724307958, "whitened_cosine_id": 0.9041904807090759}} +{"model": "pythia-410m", "dataset": "spam", "seed": 3, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688292, "raptor_stability": 0.9435619115829468, "fragility": 3.0, "augmentation_robustness": 0.9066470040376822, "xie_feature_dispersion": 6.763845431690352, "pac": 0.9251044578103145, "whitened_cosine_id": 0.9281833171844482}} +{"model": "pythia-410m", "dataset": "cola", "seed": 3, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956878849, "raptor_stability": 0.7881181836128235, "fragility": 0.25, "augmentation_robustness": 0.600597854138349, "xie_feature_dispersion": 0.9458123063554009, "pac": 0.6943580188755862, "whitened_cosine_id": 0.7995412349700928}} +{"model": "pythia-410m", "dataset": "stance", "seed": 3, "layer": 17, "concept": "stance", "predictors": {"sip_eigengap": 0.4936540695216237, "raptor_stability": 0.8370916247367859, "fragility": 0.5, "augmentation_robustness": 0.8332617730055567, "xie_feature_dispersion": 2.485399050185102, "pac": 0.8351766988711713, "whitened_cosine_id": 0.8208258748054504}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 3, "layer": 13, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881398, "raptor_stability": 0.8489888310432434, "fragility": 0.5, "augmentation_robustness": 0.7844172937998297, "xie_feature_dispersion": 1.6180674325869053, "pac": 0.8167030624215366, "whitened_cosine_id": 0.865872859954834}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010287, "raptor_stability": 0.8647460341453552, "fragility": 0.25, "augmentation_robustness": 0.641727222279711, "xie_feature_dispersion": 5.808094646975757, "pac": 0.7532366282125331, "whitened_cosine_id": 0.8289788365364075}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8411266803741455, "fragility": 1.0, "augmentation_robustness": 0.7503016837377494, "xie_feature_dispersion": 2.282144440990373, "pac": 0.7957141820559475, "whitened_cosine_id": 0.8417641520500183}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 3, "layer": 13, "concept": "topic", "predictors": {"sip_eigengap": 0.20340792936838478, "raptor_stability": 0.8698489665985107, "fragility": 1.0, "augmentation_robustness": 0.8437237058947922, "xie_feature_dispersion": 4.342567776713613, "pac": 0.8567863362466515, "whitened_cosine_id": 0.8350374698638916}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 3, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.059803315773717704, "raptor_stability": 0.9504701495170593, "fragility": 2.0, "augmentation_robustness": 0.8680315611599807, "xie_feature_dispersion": 8.620883539713939, "pac": 0.90925085533852, "whitened_cosine_id": 0.8365957140922546}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 3, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010883, "raptor_stability": 0.8169371485710144, "fragility": 0.25, "augmentation_robustness": 0.7710432901435903, "xie_feature_dispersion": 0.42029681121277923, "pac": 0.7939902193573023, "whitened_cosine_id": 0.8196472525596619}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 3, "layer": 11, "concept": "emotion", "predictors": {"sip_eigengap": 0.4267118062158825, "raptor_stability": 0.8550241589546204, "fragility": 0.25, "augmentation_robustness": 0.7405822372807725, "xie_feature_dispersion": 2.564638903560827, "pac": 0.7978031981176965, "whitened_cosine_id": 0.8296079635620117}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 3, "layer": 15, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010767, "raptor_stability": 0.796916663646698, "fragility": 0.25, "augmentation_robustness": 0.6606039027652186, "xie_feature_dispersion": 3.912958891643261, "pac": 0.7287602832059583, "whitened_cosine_id": 0.8257460594177246}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 3, "layer": 18, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010796, "raptor_stability": 0.8012734055519104, "fragility": 0.25, "augmentation_robustness": 0.6623961560942748, "xie_feature_dispersion": 3.563898854267082, "pac": 0.7318347808230926, "whitened_cosine_id": 0.8225507736206055}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 3, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010296, "raptor_stability": 0.7851132154464722, "fragility": 0.25, "augmentation_robustness": 0.6617407773933088, "xie_feature_dispersion": 1.0772966237421127, "pac": 0.7234269964198905, "whitened_cosine_id": 0.8228062987327576}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 3, "layer": 17, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010871, "raptor_stability": 0.8912879228591919, "fragility": 1.0, "augmentation_robustness": 0.74955767566332, "xie_feature_dispersion": 3.0211175775214865, "pac": 0.8204227992612559, "whitened_cosine_id": 0.8278633952140808}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 3, "layer": 3, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010838, "raptor_stability": 0.925938069820404, "fragility": 3.0, "augmentation_robustness": 0.9098532007735074, "xie_feature_dispersion": 10.71642677493898, "pac": 0.9178956352969557, "whitened_cosine_id": 0.8716909289360046}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 3, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010827, "raptor_stability": 0.7985167503356934, "fragility": 0.25, "augmentation_robustness": 0.5412217711508899, "xie_feature_dispersion": 1.0588280189082442, "pac": 0.6698692607432917, "whitened_cosine_id": 0.823022723197937}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 3, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.37531177284512046, "raptor_stability": 0.8395811915397644, "fragility": 1.0, "augmentation_robustness": 0.8246880782030601, "xie_feature_dispersion": 2.787120846289539, "pac": 0.8321346348714123, "whitened_cosine_id": 0.8196287155151367}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 3, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010829, "raptor_stability": 0.8403837084770203, "fragility": 1.0, "augmentation_robustness": 0.7414207905077144, "xie_feature_dispersion": 2.400769158905071, "pac": 0.7909022494923673, "whitened_cosine_id": 0.8367124795913696}} +{"model": "gpt2", "dataset": "sst2", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934796, "raptor_stability": 0.8351662158966064, "fragility": 0.25, "augmentation_robustness": 0.5635262291495843, "xie_feature_dispersion": 4.2622215814473865, "pac": 0.6993462225230953, "whitened_cosine_id": 0.8634011149406433}} +{"model": "gpt2", "dataset": "imdb", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350287, "raptor_stability": 0.8339491486549377, "fragility": 0.25, "augmentation_robustness": 0.7662356686632448, "xie_feature_dispersion": 1.304014068501824, "pac": 0.8000924086590913, "whitened_cosine_id": 0.87335205078125}} +{"model": "gpt2", "dataset": "ag_news", "seed": 3, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.3778683465799438, "raptor_stability": 0.8665527105331421, "fragility": 0.5, "augmentation_robustness": 0.8531902817642327, "xie_feature_dispersion": 2.241047464066251, "pac": 0.8598714961486874, "whitened_cosine_id": 0.9046292304992676}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 3, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.10991110313414279, "raptor_stability": 0.9330034255981445, "fragility": 1.5, "augmentation_robustness": 0.8169811795296109, "xie_feature_dispersion": 5.468890231915903, "pac": 0.8749923025638777, "whitened_cosine_id": 0.9063879251480103}} +{"model": "gpt2", "dataset": "counterfact", "seed": 3, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343766, "raptor_stability": 0.7898539900779724, "fragility": 0.25, "augmentation_robustness": 0.8079306258304046, "xie_feature_dispersion": 0.19395964260888718, "pac": 0.7988923079541885, "whitened_cosine_id": 0.7610896825790405}} +{"model": "gpt2", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6182210878004821, "raptor_stability": 0.8550314903259277, "fragility": 0.25, "augmentation_robustness": 0.7431736186782528, "xie_feature_dispersion": 1.1654170019303518, "pac": 0.7991025545020902, "whitened_cosine_id": 0.8550549149513245}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347225, "raptor_stability": 0.8043380379676819, "fragility": 0.25, "augmentation_robustness": 0.677191562860757, "xie_feature_dispersion": 1.3127141118978531, "pac": 0.7407648004142194, "whitened_cosine_id": 0.8204189538955688}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 3, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934697, "raptor_stability": 0.780476987361908, "fragility": 0.25, "augmentation_robustness": 0.6631271115518341, "xie_feature_dispersion": 3.352185591368954, "pac": 0.7218020494568711, "whitened_cosine_id": 0.8065347075462341}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 3, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346575, "raptor_stability": 0.7473269104957581, "fragility": 0.25, "augmentation_robustness": 0.6123626549338375, "xie_feature_dispersion": 0.7096033674345994, "pac": 0.6798447827147978, "whitened_cosine_id": 0.7918455004692078}} +{"model": "gpt2", "dataset": "subj", "seed": 3, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935243, "raptor_stability": 0.8725587725639343, "fragility": 0.5, "augmentation_robustness": 0.7069137856416525, "xie_feature_dispersion": 1.7904837862778602, "pac": 0.7897362791027934, "whitened_cosine_id": 0.9169808626174927}} +{"model": "gpt2", "dataset": "spam", "seed": 3, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353023, "raptor_stability": 0.9019039273262024, "fragility": 1.5, "augmentation_robustness": 0.8805267705055504, "xie_feature_dispersion": 12.885975509137046, "pac": 0.8912153489158764, "whitened_cosine_id": 0.92914217710495}} +{"model": "gpt2", "dataset": "cola", "seed": 3, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344647, "raptor_stability": 0.796906054019928, "fragility": 0.25, "augmentation_robustness": 0.6909560170275559, "xie_feature_dispersion": 0.4966128694499476, "pac": 0.743931035523742, "whitened_cosine_id": 0.7800789475440979}} +{"model": "gpt2", "dataset": "stance", "seed": 3, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.6661294213646244, "raptor_stability": 0.8302767872810364, "fragility": 0.5, "augmentation_robustness": 0.8014307017054971, "xie_feature_dispersion": 1.9080832844053377, "pac": 0.8158537444932668, "whitened_cosine_id": 0.8170424103736877}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 3, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350134, "raptor_stability": 0.8213337063789368, "fragility": 0.25, "augmentation_robustness": 0.6974710476635542, "xie_feature_dispersion": 1.3078300052773906, "pac": 0.7594023770212455, "whitened_cosine_id": 0.8702872395515442}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295687936, "raptor_stability": 0.8499755263328552, "fragility": 0.25, "augmentation_robustness": 0.6261496459413354, "xie_feature_dispersion": 4.436216156416145, "pac": 0.7380625861370953, "whitened_cosine_id": 0.8625351190567017}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 3, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.2921886382497235, "raptor_stability": 0.8477186560630798, "fragility": 1.0, "augmentation_robustness": 0.8387418545839186, "xie_feature_dispersion": 4.131659598391053, "pac": 0.8432302553234992, "whitened_cosine_id": 0.88265460729599}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 3, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879386, "raptor_stability": 0.7917901277542114, "fragility": 0.25, "augmentation_robustness": 0.789771355720532, "xie_feature_dispersion": 0.17500181221361583, "pac": 0.7907807417373718, "whitened_cosine_id": 0.7891764640808105}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6234881751702148, "raptor_stability": 0.8669646382331848, "fragility": 0.25, "augmentation_robustness": 0.7744639183310552, "xie_feature_dispersion": 0.8652954065662566, "pac": 0.82071427828212, "whitened_cosine_id": 0.8504673838615417}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 3, "layer": 6, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295687904, "raptor_stability": 0.7907098531723022, "fragility": 0.25, "augmentation_robustness": 0.6575825596573526, "xie_feature_dispersion": 3.5458093952463834, "pac": 0.7241462064148274, "whitened_cosine_id": 0.8148408532142639}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 3, "layer": 22, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879737, "raptor_stability": 0.7582568526268005, "fragility": 0.25, "augmentation_robustness": 0.6081859135950609, "xie_feature_dispersion": 0.8063341929658796, "pac": 0.6832213831109307, "whitened_cosine_id": 0.7960034012794495}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 3, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688269, "raptor_stability": 0.8890963792800903, "fragility": 1.0, "augmentation_robustness": 0.7291926988187397, "xie_feature_dispersion": 2.080419773242272, "pac": 0.809144539049415, "whitened_cosine_id": 0.8867965340614319}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 3, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879426, "raptor_stability": 0.7859232425689697, "fragility": 0.25, "augmentation_robustness": 0.5254228710879436, "xie_feature_dispersion": 0.6803530491181083, "pac": 0.6556730568284567, "whitened_cosine_id": 0.8077235221862793}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 3, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5196901382587911, "raptor_stability": 0.8363353610038757, "fragility": 0.5, "augmentation_robustness": 0.8228205239645261, "xie_feature_dispersion": 1.7693304935455851, "pac": 0.8295779424842009, "whitened_cosine_id": 0.7967984080314636}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 3, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881522, "raptor_stability": 0.811364471912384, "fragility": 0.5, "augmentation_robustness": 0.6900060500051179, "xie_feature_dispersion": 1.3697277559493137, "pac": 0.750685260958751, "whitened_cosine_id": 0.8628206849098206}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648796, "raptor_stability": 0.8676397204399109, "fragility": 0.25, "augmentation_robustness": 0.6345770721983662, "xie_feature_dispersion": 5.001598619911871, "pac": 0.7511083963191385, "whitened_cosine_id": 0.8821530938148499}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650275, "raptor_stability": 0.8786996006965637, "fragility": 1.0, "augmentation_robustness": 0.8588046228179724, "xie_feature_dispersion": 2.4507641389974713, "pac": 0.8687521117572681, "whitened_cosine_id": 0.8870015740394592}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 3, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.2775334420918106, "raptor_stability": 0.8491312265396118, "fragility": 1.5, "augmentation_robustness": 0.8575692544379112, "xie_feature_dispersion": 8.067934468405575, "pac": 0.8533502404887615, "whitened_cosine_id": 0.8975652456283569}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 3, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.07983459673425802, "raptor_stability": 0.9590606093406677, "fragility": 1.0, "augmentation_robustness": 0.9129941790024725, "xie_feature_dispersion": 5.747537131789415, "pac": 0.9360273941715701, "whitened_cosine_id": 0.9247861504554749}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 3, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646678, "raptor_stability": 0.840959370136261, "fragility": 0.25, "augmentation_robustness": 0.8017592029882017, "xie_feature_dispersion": 0.07321360576071477, "pac": 0.8213592865622313, "whitened_cosine_id": 0.7885988354682922}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 3, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.6025878875187768, "raptor_stability": 0.8772271275520325, "fragility": 0.25, "augmentation_robustness": 0.865087290926423, "xie_feature_dispersion": 1.9052135627589408, "pac": 0.8711572092392277, "whitened_cosine_id": 0.8513849377632141}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647677, "raptor_stability": 0.8638144731521606, "fragility": 2.0, "augmentation_robustness": 0.8618714559295404, "xie_feature_dispersion": 3.073243341198275, "pac": 0.8628429645408505, "whitened_cosine_id": 0.8141002058982849}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 3, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237638515, "raptor_stability": 0.8214561939239502, "fragility": 0.25, "augmentation_robustness": 0.7518367724678281, "xie_feature_dispersion": 4.329624079200656, "pac": 0.7866464831958891, "whitened_cosine_id": 0.7974085807800293}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 3, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237646187, "raptor_stability": 0.804497480392456, "fragility": 0.25, "augmentation_robustness": 0.7534120265877415, "xie_feature_dispersion": 0.582349253613317, "pac": 0.7789547534900988, "whitened_cosine_id": 0.8017151355743408}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 3, "layer": 15, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651343, "raptor_stability": 0.9011402130126953, "fragility": 0.5, "augmentation_robustness": 0.8098720947850936, "xie_feature_dispersion": 1.5822189245994793, "pac": 0.8555061538988944, "whitened_cosine_id": 0.9193786978721619}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 3, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237650004, "raptor_stability": 0.9187747240066528, "fragility": 1.0, "augmentation_robustness": 0.9072265197438765, "xie_feature_dispersion": 18.77242405320582, "pac": 0.9130006218752647, "whitened_cosine_id": 0.9447034597396851}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 3, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646827, "raptor_stability": 0.8013314008712769, "fragility": 0.25, "augmentation_robustness": 0.5714415857266395, "xie_feature_dispersion": 0.7710445833759078, "pac": 0.6863864932989582, "whitened_cosine_id": 0.8173381686210632}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 3, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.5814899203429595, "raptor_stability": 0.8490622639656067, "fragility": 0.25, "augmentation_robustness": 0.8472747423617989, "xie_feature_dispersion": 2.1435570572895988, "pac": 0.8481685031637027, "whitened_cosine_id": 0.8264880776405334}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 3, "layer": 13, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649207, "raptor_stability": 0.8474602699279785, "fragility": 0.25, "augmentation_robustness": 0.765850633291224, "xie_feature_dispersion": 1.245317692001698, "pac": 0.8066554516096013, "whitened_cosine_id": 0.8748792409896851}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 4, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996045, "raptor_stability": 0.8566904067993164, "fragility": 1.0, "augmentation_robustness": 0.6395377546477263, "xie_feature_dispersion": 1.5854208982087719, "pac": 0.7481140807235214, "whitened_cosine_id": 0.8359927535057068}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 4, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.711632992199886, "raptor_stability": 0.9050891995429993, "fragility": 1.0, "augmentation_robustness": 0.9251332642220863, "xie_feature_dispersion": 1.2126650382704005, "pac": 0.9151112318825427, "whitened_cosine_id": 0.850266695022583}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 4, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5924204103995422, "raptor_stability": 0.8544853329658508, "fragility": 1.0, "augmentation_robustness": 0.8372965090354867, "xie_feature_dispersion": 4.656184075098706, "pac": 0.8458909210006688, "whitened_cosine_id": 0.9082167744636536}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.1519243074719573, "raptor_stability": 0.9499274492263794, "fragility": 1.5, "augmentation_robustness": 0.8830180866818871, "xie_feature_dispersion": 5.282104233636098, "pac": 0.9164727679541332, "whitened_cosine_id": 0.9344291687011719}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 4, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921975906, "raptor_stability": 0.7707024216651917, "fragility": 1.0, "augmentation_robustness": 0.7329255811603577, "xie_feature_dispersion": 0.5102383939416054, "pac": 0.7518140014127747, "whitened_cosine_id": 0.7052313089370728}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 4, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.6400948489658904, "raptor_stability": 0.8515534996986389, "fragility": 0.5, "augmentation_robustness": 0.7338919283811642, "xie_feature_dispersion": 1.307093856011509, "pac": 0.7927227140399016, "whitened_cosine_id": 0.8422322273254395}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 4, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.71163299219924, "raptor_stability": 0.8237500786781311, "fragility": 0.5, "augmentation_robustness": 0.7327459917351864, "xie_feature_dispersion": 2.5240514742338784, "pac": 0.7782480352066588, "whitened_cosine_id": 0.803077220916748}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 4, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987739, "raptor_stability": 0.8263923525810242, "fragility": 1.0, "augmentation_robustness": 0.7109290902303631, "xie_feature_dispersion": 2.0494666093180176, "pac": 0.7686607214056936, "whitened_cosine_id": 0.7474713325500488}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 4, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990436, "raptor_stability": 0.8242563605308533, "fragility": 0.5, "augmentation_robustness": 0.7821271817029374, "xie_feature_dispersion": 1.5098181981749643, "pac": 0.8031917711168953, "whitened_cosine_id": 0.7899174094200134}} +{"model": "pythia-70m", "dataset": "subj", "seed": 4, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002564, "raptor_stability": 0.8824465274810791, "fragility": 1.0, "augmentation_robustness": 0.8066508753882752, "xie_feature_dispersion": 1.925210310106005, "pac": 0.8445487014346771, "whitened_cosine_id": 0.9232692122459412}} +{"model": "pythia-70m", "dataset": "spam", "seed": 4, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004774, "raptor_stability": 0.9476951360702515, "fragility": 3.0, "augmentation_robustness": 0.9265937443188577, "xie_feature_dispersion": 6.387885559509479, "pac": 0.9371444401945546, "whitened_cosine_id": 0.9508467316627502}} +{"model": "pythia-70m", "dataset": "cola", "seed": 4, "layer": 5, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977023, "raptor_stability": 0.7777998447418213, "fragility": 0.25, "augmentation_robustness": 0.6579850681228125, "xie_feature_dispersion": 0.250601434273502, "pac": 0.7178924564323169, "whitened_cosine_id": 0.7004658579826355}} +{"model": "pythia-70m", "dataset": "stance", "seed": 4, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.7333740478467312, "raptor_stability": 0.8212578296661377, "fragility": 1.0, "augmentation_robustness": 0.7504520187465166, "xie_feature_dispersion": 1.7128280566290308, "pac": 0.7858549242063271, "whitened_cosine_id": 0.8275880217552185}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 4, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997524, "raptor_stability": 0.8540244102478027, "fragility": 1.0, "augmentation_robustness": 0.7819111270390232, "xie_feature_dispersion": 1.4981388024521016, "pac": 0.817967768643413, "whitened_cosine_id": 0.8604907989501953}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 4, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349261, "raptor_stability": 0.842431366443634, "fragility": 0.25, "augmentation_robustness": 0.670070029768046, "xie_feature_dispersion": 2.8701281427080465, "pac": 0.75625069810584, "whitened_cosine_id": 0.8559513688087463}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 4, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350782, "raptor_stability": 0.8713650703430176, "fragility": 0.5, "augmentation_robustness": 0.8521992774428392, "xie_feature_dispersion": 1.7964561061229958, "pac": 0.8617821738929283, "whitened_cosine_id": 0.8889275789260864}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.3656440339329693, "raptor_stability": 0.8521515130996704, "fragility": 1.5, "augmentation_robustness": 0.8214046671297884, "xie_feature_dispersion": 6.173757088063873, "pac": 0.8367780901147295, "whitened_cosine_id": 0.9043599367141724}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 4, "layer": 9, "concept": "topic", "predictors": {"sip_eigengap": 0.12756426041857358, "raptor_stability": 0.9445372819900513, "fragility": 1.5, "augmentation_robustness": 0.8622565473270213, "xie_feature_dispersion": 6.343488803218173, "pac": 0.9033969146585363, "whitened_cosine_id": 0.9183123707771301}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 4, "layer": 7, "concept": "truth", "predictors": {"sip_eigengap": 1.397542485934335, "raptor_stability": 0.7996639609336853, "fragility": 0.25, "augmentation_robustness": 0.7613788429632672, "xie_feature_dispersion": 0.5275399816201919, "pac": 0.7805214019484763, "whitened_cosine_id": 0.753498375415802}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.7058740799382045, "raptor_stability": 0.8810455799102783, "fragility": 1.0, "augmentation_robustness": 0.8152160353634006, "xie_feature_dispersion": 2.0356355439733544, "pac": 0.8481308076368395, "whitened_cosine_id": 0.8614587187767029}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 4, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348062, "raptor_stability": 0.8045304417610168, "fragility": 0.5, "augmentation_robustness": 0.7049071379157302, "xie_feature_dispersion": 2.0681692958689024, "pac": 0.7547187898383736, "whitened_cosine_id": 0.8201311230659485}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346577, "raptor_stability": 0.8003606200218201, "fragility": 0.25, "augmentation_robustness": 0.6612670699122942, "xie_feature_dispersion": 2.4099178377906756, "pac": 0.7308138449670571, "whitened_cosine_id": 0.8072301745414734}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 4, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345884, "raptor_stability": 0.8317489624023438, "fragility": 1.0, "augmentation_robustness": 0.772983120611079, "xie_feature_dispersion": 2.3238338563599275, "pac": 0.8023660415067113, "whitened_cosine_id": 0.8050597906112671}} +{"model": "pythia-160m", "dataset": "subj", "seed": 4, "layer": 5, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351835, "raptor_stability": 0.8953569531440735, "fragility": 1.0, "augmentation_robustness": 0.8059484746527689, "xie_feature_dispersion": 1.8870415060935894, "pac": 0.8506527138984212, "whitened_cosine_id": 0.9113974571228027}} +{"model": "pythia-160m", "dataset": "spam", "seed": 4, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352754, "raptor_stability": 0.9212907552719116, "fragility": 1.5, "augmentation_robustness": 0.8624898624867177, "xie_feature_dispersion": 12.624355801662276, "pac": 0.8918903088793146, "whitened_cosine_id": 0.9403319358825684}} +{"model": "pythia-160m", "dataset": "cola", "seed": 4, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934378, "raptor_stability": 0.7925094366073608, "fragility": 0.25, "augmentation_robustness": 0.6106269635092136, "xie_feature_dispersion": 0.1983095485933765, "pac": 0.7015682000582872, "whitened_cosine_id": 0.7796781063079834}} +{"model": "pythia-160m", "dataset": "stance", "seed": 4, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.591622106050008, "raptor_stability": 0.8417823314666748, "fragility": 1.0, "augmentation_robustness": 0.8225818656918379, "xie_feature_dispersion": 1.8092532859980797, "pac": 0.8321820985792563, "whitened_cosine_id": 0.8301091194152832}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 4, "layer": 3, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349856, "raptor_stability": 0.8758466839790344, "fragility": 1.0, "augmentation_robustness": 0.835043699708102, "xie_feature_dispersion": 1.5364396930791746, "pac": 0.8554451918435682, "whitened_cosine_id": 0.8725274801254272}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 4, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880225, "raptor_stability": 0.8483951687812805, "fragility": 0.25, "augmentation_robustness": 0.643421731590692, "xie_feature_dispersion": 3.4588103868670306, "pac": 0.7459084501859863, "whitened_cosine_id": 0.8561255931854248}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 4, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882035, "raptor_stability": 0.8837658166885376, "fragility": 1.0, "augmentation_robustness": 0.8596997806079171, "xie_feature_dispersion": 1.9038436246678319, "pac": 0.8717327986482273, "whitened_cosine_id": 0.894847571849823}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.3547804650921568, "raptor_stability": 0.8753689527511597, "fragility": 1.0, "augmentation_robustness": 0.8555244488391097, "xie_feature_dispersion": 3.4329927120747055, "pac": 0.8654467007951347, "whitened_cosine_id": 0.8951209783554077}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 4, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.10955555398923562, "raptor_stability": 0.9442237615585327, "fragility": 3.0, "augmentation_robustness": 0.8375015283715399, "xie_feature_dispersion": 10.914315920866802, "pac": 0.8908626449650363, "whitened_cosine_id": 0.887626588344574}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 4, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878207, "raptor_stability": 0.7893005013465881, "fragility": 0.25, "augmentation_robustness": 0.7388001024408721, "xie_feature_dispersion": 0.4902099764637905, "pac": 0.7640503018937301, "whitened_cosine_id": 0.7745464444160461}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.5474574770136091, "raptor_stability": 0.8616097569465637, "fragility": 0.25, "augmentation_robustness": 0.7404739041427045, "xie_feature_dispersion": 1.968628707347913, "pac": 0.8010418305446341, "whitened_cosine_id": 0.8528909087181091}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 4, "layer": 15, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295688051, "raptor_stability": 0.8146539330482483, "fragility": 0.25, "augmentation_robustness": 0.7026104422955255, "xie_feature_dispersion": 4.365462087324579, "pac": 0.7586321876718869, "whitened_cosine_id": 0.8268682956695557}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 4, "layer": 15, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878458, "raptor_stability": 0.8015760779380798, "fragility": 0.25, "augmentation_robustness": 0.6575946714155477, "xie_feature_dispersion": 3.8919081093582863, "pac": 0.7295853746768137, "whitened_cosine_id": 0.814882755279541}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 4, "layer": 16, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956878596, "raptor_stability": 0.7986729145050049, "fragility": 0.25, "augmentation_robustness": 0.700139464938641, "xie_feature_dispersion": 3.5880837975111466, "pac": 0.7494061897218229, "whitened_cosine_id": 0.8099539279937744}} +{"model": "pythia-410m", "dataset": "subj", "seed": 4, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882532, "raptor_stability": 0.8692127466201782, "fragility": 0.5, "augmentation_robustness": 0.7900862434288312, "xie_feature_dispersion": 1.4726101086077301, "pac": 0.8296494950245047, "whitened_cosine_id": 0.8998460173606873}} +{"model": "pythia-410m", "dataset": "spam", "seed": 4, "layer": 24, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688298, "raptor_stability": 0.9033204317092896, "fragility": 3.0, "augmentation_robustness": 0.8355633491173182, "xie_feature_dispersion": 12.679622904881825, "pac": 0.8694418904133039, "whitened_cosine_id": 0.9067055583000183}} +{"model": "pythia-410m", "dataset": "cola", "seed": 4, "layer": 17, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880077, "raptor_stability": 0.8151479363441467, "fragility": 0.25, "augmentation_robustness": 0.631437331569998, "xie_feature_dispersion": 0.15645623015703475, "pac": 0.7232926339570724, "whitened_cosine_id": 0.8250035643577576}} +{"model": "pythia-410m", "dataset": "stance", "seed": 4, "layer": 16, "concept": "stance", "predictors": {"sip_eigengap": 0.5699871301215584, "raptor_stability": 0.8456130027770996, "fragility": 0.5, "augmentation_robustness": 0.8244991343863916, "xie_feature_dispersion": 2.7721624801456297, "pac": 0.8350560685817456, "whitened_cosine_id": 0.8285333514213562}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 4, "layer": 11, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881264, "raptor_stability": 0.8462154269218445, "fragility": 0.5, "augmentation_robustness": 0.7743294062255062, "xie_feature_dispersion": 1.1628324109274197, "pac": 0.8102724165736753, "whitened_cosine_id": 0.8581516146659851}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 4, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010715, "raptor_stability": 0.8520708680152893, "fragility": 0.25, "augmentation_robustness": 0.6412965254676084, "xie_feature_dispersion": 4.357957129868121, "pac": 0.7466836967414489, "whitened_cosine_id": 0.8315754532814026}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 4, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010869, "raptor_stability": 0.8517562747001648, "fragility": 0.5, "augmentation_robustness": 0.771865951099836, "xie_feature_dispersion": 2.184545736254185, "pac": 0.8118111129000004, "whitened_cosine_id": 0.8343337774276733}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 4, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.24047302056829442, "raptor_stability": 0.8453668355941772, "fragility": 1.5, "augmentation_robustness": 0.8458315674904345, "xie_feature_dispersion": 5.873911972981273, "pac": 0.845599201542306, "whitened_cosine_id": 0.8292579054832458}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 4, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06567607847107426, "raptor_stability": 0.948825478553772, "fragility": 3.0, "augmentation_robustness": 0.853398753358697, "xie_feature_dispersion": 9.539501225119675, "pac": 0.9011121159562345, "whitened_cosine_id": 0.8352555632591248}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 4, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.7958840727806091, "fragility": 0.25, "augmentation_robustness": 0.7423103383870259, "xie_feature_dispersion": 0.5601797812591205, "pac": 0.7690972055838174, "whitened_cosine_id": 0.8237647414207458}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 4, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.01721014234208262, "raptor_stability": 0.8599122166633606, "fragility": 0.25, "augmentation_robustness": 0.7417171978776979, "xie_feature_dispersion": 2.3728522318398584, "pac": 0.8008147072705292, "whitened_cosine_id": 0.8342825770378113}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 4, "layer": 20, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.8056564331054688, "fragility": 0.25, "augmentation_robustness": 0.6716115591135512, "xie_feature_dispersion": 2.820489046498124, "pac": 0.73863399610951, "whitened_cosine_id": 0.8278409242630005}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010146, "raptor_stability": 0.7977588176727295, "fragility": 0.25, "augmentation_robustness": 0.64006899512257, "xie_feature_dispersion": 4.84100547742313, "pac": 0.7189139063976497, "whitened_cosine_id": 0.8262239098548889}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 4, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010499, "raptor_stability": 0.8048697710037231, "fragility": 0.25, "augmentation_robustness": 0.6973612069523063, "xie_feature_dispersion": 4.496528530562788, "pac": 0.7511154889780147, "whitened_cosine_id": 0.8300395607948303}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 4, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.8694700002670288, "fragility": 1.0, "augmentation_robustness": 0.7901078358260162, "xie_feature_dispersion": 2.50098235434014, "pac": 0.8297889180465226, "whitened_cosine_id": 0.8374378681182861}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 4, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010415, "raptor_stability": 0.9186072945594788, "fragility": 1.0, "augmentation_robustness": 0.8992452095528303, "xie_feature_dispersion": 23.667968510878133, "pac": 0.9089262520561545, "whitened_cosine_id": 0.869879961013794}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 4, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010873, "raptor_stability": 0.8157230615615845, "fragility": 0.25, "augmentation_robustness": 0.5471678545861385, "xie_feature_dispersion": 0.31676967447098947, "pac": 0.6814454580738615, "whitened_cosine_id": 0.8299345970153809}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.3335229215667947, "raptor_stability": 0.824205219745636, "fragility": 0.5, "augmentation_robustness": 0.7978037532307928, "xie_feature_dispersion": 2.0733724889181753, "pac": 0.8110044864882144, "whitened_cosine_id": 0.8242607712745667}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 4, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.855816496101083, "raptor_stability": 0.8407758474349976, "fragility": 1.0, "augmentation_robustness": 0.7768042275797109, "xie_feature_dispersion": 1.8641614757755678, "pac": 0.8087900375073542, "whitened_cosine_id": 0.8235855102539062}} +{"model": "gpt2", "dataset": "sst2", "seed": 4, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934977, "raptor_stability": 0.8240580558776855, "fragility": 0.25, "augmentation_robustness": 0.5955483532445387, "xie_feature_dispersion": 2.990446112484106, "pac": 0.7098032045611121, "whitened_cosine_id": 0.8673197031021118}} +{"model": "gpt2", "dataset": "imdb", "seed": 4, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859351042, "raptor_stability": 0.8240647912025452, "fragility": 0.25, "augmentation_robustness": 0.7103925323300748, "xie_feature_dispersion": 1.6507650137935161, "pac": 0.76722866176631, "whitened_cosine_id": 0.892749547958374}} +{"model": "gpt2", "dataset": "ag_news", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.37928405790056124, "raptor_stability": 0.8549799919128418, "fragility": 0.5, "augmentation_robustness": 0.8421595137240991, "xie_feature_dispersion": 2.2799844708419164, "pac": 0.8485697528184705, "whitened_cosine_id": 0.9051476716995239}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 4, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.15303585153401286, "raptor_stability": 0.9340819120407104, "fragility": 1.5, "augmentation_robustness": 0.8178668816980438, "xie_feature_dispersion": 6.187412713082707, "pac": 0.8759743968693772, "whitened_cosine_id": 0.907498836517334}} +{"model": "gpt2", "dataset": "counterfact", "seed": 4, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341972, "raptor_stability": 0.778740644454956, "fragility": 0.25, "augmentation_robustness": 0.7971651296687343, "xie_feature_dispersion": 0.14506327750165868, "pac": 0.7879528870618452, "whitened_cosine_id": 0.7351542115211487}} +{"model": "gpt2", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5149693378247766, "raptor_stability": 0.8657316565513611, "fragility": 0.25, "augmentation_robustness": 0.7431909261678031, "xie_feature_dispersion": 1.1678519242893954, "pac": 0.8044612913595821, "whitened_cosine_id": 0.8591601848602295}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 4, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347685, "raptor_stability": 0.8219725489616394, "fragility": 0.25, "augmentation_robustness": 0.7928502271629654, "xie_feature_dispersion": 0.8049332910819382, "pac": 0.8074113880623024, "whitened_cosine_id": 0.8097414970397949}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346366, "raptor_stability": 0.7918288111686707, "fragility": 0.25, "augmentation_robustness": 0.6646703017429912, "xie_feature_dispersion": 3.311280523732777, "pac": 0.7282495564558309, "whitened_cosine_id": 0.8104947209358215}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 4, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859344243, "raptor_stability": 0.8132399916648865, "fragility": 0.25, "augmentation_robustness": 0.7322499286526112, "xie_feature_dispersion": 2.8837150977712067, "pac": 0.7727449601587488, "whitened_cosine_id": 0.8200110197067261}} +{"model": "gpt2", "dataset": "subj", "seed": 4, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.39754248593522, "raptor_stability": 0.8405840992927551, "fragility": 0.5, "augmentation_robustness": 0.7275234344331813, "xie_feature_dispersion": 1.7854217365687841, "pac": 0.7840537668629682, "whitened_cosine_id": 0.9135460257530212}} +{"model": "gpt2", "dataset": "spam", "seed": 4, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353234, "raptor_stability": 0.9081653356552124, "fragility": 1.5, "augmentation_robustness": 0.8591723776230531, "xie_feature_dispersion": 12.538518032179876, "pac": 0.8836688566391327, "whitened_cosine_id": 0.9366497993469238}} +{"model": "gpt2", "dataset": "cola", "seed": 4, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344865, "raptor_stability": 0.7840872406959534, "fragility": 0.25, "augmentation_robustness": 0.5048661738238992, "xie_feature_dispersion": 0.16631055680345583, "pac": 0.6444767072599262, "whitened_cosine_id": 0.7860410809516907}} +{"model": "gpt2", "dataset": "stance", "seed": 4, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5528580887739023, "raptor_stability": 0.8414139747619629, "fragility": 0.25, "augmentation_robustness": 0.8322638497812136, "xie_feature_dispersion": 0.9522759018719159, "pac": 0.8368389122715882, "whitened_cosine_id": 0.8301456570625305}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 4, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349983, "raptor_stability": 0.8202458024024963, "fragility": 0.25, "augmentation_robustness": 0.7289728446571213, "xie_feature_dispersion": 0.931895464785709, "pac": 0.7746093235298088, "whitened_cosine_id": 0.8736944198608398}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 4, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881493, "raptor_stability": 0.8379729986190796, "fragility": 0.25, "augmentation_robustness": 0.6139063338760717, "xie_feature_dispersion": 3.178141409790626, "pac": 0.7259396662475757, "whitened_cosine_id": 0.8596592545509338}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 4, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.2485401486377617, "raptor_stability": 0.8904624581336975, "fragility": 0.25, "augmentation_robustness": 0.9007411868297208, "xie_feature_dispersion": 1.0727326936501331, "pac": 0.8956018224817092, "whitened_cosine_id": 0.892510175704956}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878993, "raptor_stability": 0.7903531193733215, "fragility": 0.25, "augmentation_robustness": 0.79632543598189, "xie_feature_dispersion": 0.15365171233992977, "pac": 0.7933392776776058, "whitened_cosine_id": 0.7768749594688416}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5850775560898813, "raptor_stability": 0.8753568530082703, "fragility": 0.25, "augmentation_robustness": 0.7625841583432605, "xie_feature_dispersion": 0.8499569074700448, "pac": 0.8189705056757655, "whitened_cosine_id": 0.8553555011749268}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 4, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880672, "raptor_stability": 0.7783800959587097, "fragility": 0.25, "augmentation_robustness": 0.6250450756800867, "xie_feature_dispersion": 2.9465184575878856, "pac": 0.7017125858193982, "whitened_cosine_id": 0.8258277177810669}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 4, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956877166, "raptor_stability": 0.7933056354522705, "fragility": 0.25, "augmentation_robustness": 0.6933130878602948, "xie_feature_dispersion": 2.7324461419877974, "pac": 0.7433093616562827, "whitened_cosine_id": 0.8186708092689514}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 4, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882608, "raptor_stability": 0.8612777590751648, "fragility": 0.5, "augmentation_robustness": 0.7114055951912336, "xie_feature_dispersion": 1.9179527386531523, "pac": 0.7863416771331992, "whitened_cosine_id": 0.8883224129676819}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956883079, "raptor_stability": 0.9192178249359131, "fragility": 1.0, "augmentation_robustness": 0.894155653371751, "xie_feature_dispersion": 13.303624175400875, "pac": 0.906686739153832, "whitened_cosine_id": 0.9148821830749512}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 4, "layer": 13, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.210307295688004, "raptor_stability": 0.8070831894874573, "fragility": 0.25, "augmentation_robustness": 0.5376584999703106, "xie_feature_dispersion": 0.1971499159632617, "pac": 0.6723708447288839, "whitened_cosine_id": 0.816491961479187}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 4, "layer": 15, "concept": "stance", "predictors": {"sip_eigengap": 0.4591650118237964, "raptor_stability": 0.818568766117096, "fragility": 0.5, "augmentation_robustness": 0.7938712936502795, "xie_feature_dispersion": 1.9692841257546414, "pac": 0.8062200298836877, "whitened_cosine_id": 0.7844472527503967}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 4, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881198, "raptor_stability": 0.812563955783844, "fragility": 0.25, "augmentation_robustness": 0.7084456019576522, "xie_feature_dispersion": 0.9841060606862964, "pac": 0.7605047788707481, "whitened_cosine_id": 0.8456441760063171}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 4, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237646625, "raptor_stability": 0.8596814870834351, "fragility": 0.25, "augmentation_robustness": 0.6898037113021047, "xie_feature_dispersion": 3.3502512176055483, "pac": 0.7747425991927699, "whitened_cosine_id": 0.8781615495681763}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 4, "layer": 20, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650564, "raptor_stability": 0.8608689308166504, "fragility": 0.5, "augmentation_robustness": 0.792066215002771, "xie_feature_dispersion": 2.5567349907892805, "pac": 0.8264675729097106, "whitened_cosine_id": 0.8956750631332397}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 4, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.35362597314605054, "raptor_stability": 0.8578630685806274, "fragility": 1.0, "augmentation_robustness": 0.8552613951921951, "xie_feature_dispersion": 3.846365008247924, "pac": 0.8565622318864112, "whitened_cosine_id": 0.9092753529548645}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 4, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.08672486871948791, "raptor_stability": 0.9325566291809082, "fragility": 3.0, "augmentation_robustness": 0.7997953113506687, "xie_feature_dispersion": 11.842901742955636, "pac": 0.8661759702657885, "whitened_cosine_id": 0.894465446472168}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.293872923764685, "raptor_stability": 0.8403885364532471, "fragility": 0.25, "augmentation_robustness": 0.8397827650909445, "xie_feature_dispersion": 0.14763852549031706, "pac": 0.8400856507720957, "whitened_cosine_id": 0.7999392747879028}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.6393559906046417, "raptor_stability": 0.8851574063301086, "fragility": 0.25, "augmentation_robustness": 0.8701497898578472, "xie_feature_dispersion": 1.6371745621378626, "pac": 0.8776535980939779, "whitened_cosine_id": 0.8543353080749512}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 4, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647368, "raptor_stability": 0.7992756366729736, "fragility": 0.25, "augmentation_robustness": 0.6877395999249556, "xie_feature_dispersion": 4.044282793013871, "pac": 0.7435076182989646, "whitened_cosine_id": 0.8199916481971741}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.293872923764363, "raptor_stability": 0.8176515698432922, "fragility": 0.25, "augmentation_robustness": 0.7400420269154885, "xie_feature_dispersion": 3.976714100479099, "pac": 0.7788467983793903, "whitened_cosine_id": 0.793383777141571}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 4, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647522, "raptor_stability": 0.8136802911758423, "fragility": 0.25, "augmentation_robustness": 0.7463283551572726, "xie_feature_dispersion": 3.432685813095943, "pac": 0.7800043231665574, "whitened_cosine_id": 0.8169722557067871}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 4, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651185, "raptor_stability": 0.886184573173523, "fragility": 0.5, "augmentation_robustness": 0.8108833430892144, "xie_feature_dispersion": 0.9492959086809997, "pac": 0.8485339581313687, "whitened_cosine_id": 0.919140100479126}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923764812, "raptor_stability": 0.9170556664466858, "fragility": 0.5, "augmentation_robustness": 0.9123323230914238, "xie_feature_dispersion": 18.31873452250279, "pac": 0.9146939947690548, "whitened_cosine_id": 0.9489960670471191}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 4, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646698, "raptor_stability": 0.8118268251419067, "fragility": 0.25, "augmentation_robustness": 0.6088240065593917, "xie_feature_dispersion": 0.15540364667380704, "pac": 0.7103254158506492, "whitened_cosine_id": 0.8048716187477112}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.5271400318991657, "raptor_stability": 0.8213255405426025, "fragility": 0.5, "augmentation_robustness": 0.7692249112172099, "xie_feature_dispersion": 2.6328433172283368, "pac": 0.7952752258799063, "whitened_cosine_id": 0.826519787311554}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 4, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649123, "raptor_stability": 0.8499941825866699, "fragility": 0.25, "augmentation_robustness": 0.8088275899418353, "xie_feature_dispersion": 0.7460673348935266, "pac": 0.8294108862642526, "whitened_cosine_id": 0.8643565773963928}} diff --git a/repro_bundle/results_final/results/stats_ranking.jsonl b/repro_bundle/results_final/results/stats_ranking.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..390b7571b54ed2a0e072ec18ec77883648d66c37 --- /dev/null +++ b/repro_bundle/results_final/results/stats_ranking.jsonl @@ -0,0 +1,21 @@ +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17628948336103584, "p": 0.00011810508174001726, "kendall": 0.11986019460492477, "ci95": [0.0864486183837455, 0.2656858828070762], "loco_mean": 0.06306440054904434, "n": 472}, {"predictor": "pac", "spearman": 0.11790738396920045, "p": 0.010354502513996207, "kendall": 0.07946596854652109, "ci95": [0.03199899524267362, 0.19739551469136], "loco_mean": 0.03027311003896434, "n": 472}, {"predictor": "fragility", "spearman": 0.10714381169126685, "p": 0.019896547808716226, "kendall": 0.0819652122104107, "ci95": [0.02184423176267955, 0.19486119226176976], "loco_mean": -0.08367198887077822, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.05468516346193012, "p": 0.23570006318676667, "kendall": 0.0401513209128409, "ci95": [-0.03209005896781303, 0.1425023272515047], "loco_mean": -0.012901634988281898, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.000509027245282521, "p": 0.9911998467763682, "kendall": 0.0014395350322827368, "ci95": [-0.08621837109724974, 0.09490063671198525], "loco_mean": 0.10342197319983981, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.062397712100437104, "p": 0.17594389124505538, "kendall": -0.04115892744258281, "ci95": [-0.14459614732709314, 0.02266644621275267], "loco_mean": -0.06425062206029349, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.07663916907355152, "p": 0.0963002843104417, "kendall": -0.05037133000022321, "ci95": [-0.16246440320558697, 0.010007628269806309], "loco_mean": -0.09396010191341637, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9449489763429444, "p": 3.6596555462199364e-230, "kendall": 0.7988412681276762, "ci95": [0.9320857918652855, 0.9546405882585741], "loco_mean": 0.714095456119887, "n": 472}, {"predictor": "pac", "spearman": 0.8386192401102935, "p": 4.389790032967171e-126, "kendall": 0.6566807009967972, "ci95": [0.8041701035011758, 0.8686758317302297], "loco_mean": 0.7336833473426215, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7690200093623065, "p": 2.168995457934309e-93, "kendall": 0.5761092518622476, "ci95": [0.7221527504365517, 0.8086111694029999], "loco_mean": 0.17490397542579705, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7108589060252217, "p": 7.570490911324105e-74, "kendall": 0.5244521213429774, "ci95": [0.6647373626280134, 0.7590285725368467], "loco_mean": 0.6466917724774698, "n": 472}, {"predictor": "fragility", "spearman": 0.6402369351573364, "p": 8.412937618018045e-56, "kendall": 0.5176148935000774, "ci95": [0.5762098860164254, 0.6965181301409472], "loco_mean": 0.34843798881555416, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4577071159234756, "p": 8.112397111719808e-26, "kendall": 0.30139623592068804, "ci95": [0.3719011054200997, 0.5362764648442968], "loco_mean": -0.15587895297039037, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.07233275965780582, "p": 0.11656426348042145, "kendall": -0.045983526648003925, "ci95": [-0.1676638264409657, 0.0276888325777201], "loco_mean": 0.27295969447731966, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9663956649462877, "p": 1.9769716421360586e-279, "kendall": 0.844200942819101, "ci95": [0.9582833668806262, 0.972249511018557], "loco_mean": 0.822013981169579, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8142104839116457, "p": 4.6349247787002345e-113, "kendall": 0.6190938860700277, "ci95": [0.7784315483263554, 0.8457827980307192], "loco_mean": 0.2507519638805218, "n": 472}, {"predictor": "pac", "spearman": 0.7989523437605205, "p": 7.370681520368353e-106, "kendall": 0.6131202993990427, "ci95": [0.7548202970707872, 0.8362192541519101], "loco_mean": 0.7092535567689743, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6528079273309736, "p": 1.1636128383807618e-58, "kendall": 0.47211126704811257, "ci95": [0.594467447577123, 0.7075880242965953], "loco_mean": 0.5938518661656708, "n": 472}, {"predictor": "fragility", "spearman": 0.6434621748707247, "p": 1.6002788335451959e-56, "kendall": 0.5162034040583596, "ci95": [0.5777199855374996, 0.6972436181916808], "loco_mean": 0.30639625747257265, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.46410427354367867, "p": 1.377354979546913e-26, "kendall": 0.31332541653172113, "ci95": [0.38047007067527516, 0.5387551616620547], "loco_mean": -0.07487413986761712, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.05687906812360867, "p": 0.21740724040384907, "kendall": -0.03140843113445152, "ci95": [-0.15064310621609026, 0.04683266118594591], "loco_mean": 0.24522656676053, "n": 472}]} +{"target": "excess_rotation", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.045730142358639335, "p": 0.3214915785134481, "kendall": 0.029813955162114502, "ci95": [-0.04291475943890026, 0.13280723679010745], "loco_mean": 0.11571837660547339, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.019548618802250367, "p": 0.6718436438563953, "kendall": 0.01587965653174074, "ci95": [-0.06592842183087877, 0.11021518205534078], "loco_mean": 0.11096384112403068, "n": 472}, {"predictor": "pac", "spearman": -0.028921926535337233, "p": 0.5307837018089083, "kendall": -0.022059088128396125, "ci95": [-0.11461587366461326, 0.05697485566708844], "loco_mean": 0.0566288049921826, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": -0.11740393878758819, "p": 0.010688017745394272, "kendall": -0.08521357371621864, "ci95": [-0.20901249520866028, -0.030594774861803075], "loco_mean": -0.17614889459054547, "n": 472}, {"predictor": "fragility", "spearman": -0.13634890441214695, "p": 0.0029947554168688454, "kendall": -0.0995317208632813, "ci95": [-0.2216850177242833, -0.05109443013239468], "loco_mean": 0.01926842737612711, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.19517738512287972, "p": 1.9527049508461923e-05, "kendall": -0.12812623700025189, "ci95": [-0.28507442645385095, -0.11649292054466626], "loco_mean": -0.11715629043368607, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.24497186857439832, "p": 7.030580997726293e-08, "kendall": -0.1661088920076289, "ci95": [-0.3367444423080759, -0.15504169609386367], "loco_mean": -0.11500811181491921, "n": 472}]} +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17628948336103584, "p": 0.00011810508174001726, "kendall": 0.11986019460492477, "ci95": [0.0864486183837455, 0.2656858828070762], "loco_mean": 0.06306440054904434, "n": 472}, {"predictor": "pac", "spearman": 0.11790738396920045, "p": 0.010354502513996207, "kendall": 0.07946596854652109, "ci95": [0.03199899524267362, 0.19739551469136], "loco_mean": 0.03027311003896434, "n": 472}, {"predictor": "fragility", "spearman": 0.10714381169126685, "p": 0.019896547808716226, "kendall": 0.0819652122104107, "ci95": [0.02184423176267955, 0.19486119226176976], "loco_mean": -0.08367198887077822, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.05468516346193012, "p": 0.23570006318676667, "kendall": 0.0401513209128409, "ci95": [-0.03209005896781303, 0.1425023272515047], "loco_mean": -0.012901634988281898, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.000509027245282521, "p": 0.9911998467763682, "kendall": 0.0014395350322827368, "ci95": [-0.08621837109724974, 0.09490063671198525], "loco_mean": 0.10342197319983981, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.062397712100437104, "p": 0.17594389124505538, "kendall": -0.04115892744258281, "ci95": [-0.14459614732709314, 0.02266644621275267], "loco_mean": -0.06425062206029349, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.07663916907355152, "p": 0.0963002843104417, "kendall": -0.05037133000022321, "ci95": [-0.16246440320558697, 0.010007628269806309], "loco_mean": -0.09396010191341637, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9449489763429444, "p": 3.6596555462199364e-230, "kendall": 0.7988412681276762, "ci95": [0.9320857918652855, 0.9546405882585741], "loco_mean": 0.714095456119887, "n": 472}, {"predictor": "pac", "spearman": 0.8386192401102935, "p": 4.389790032967171e-126, "kendall": 0.6566807009967972, "ci95": [0.8041701035011758, 0.8686758317302297], "loco_mean": 0.7336833473426215, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7690200093623065, "p": 2.168995457934309e-93, "kendall": 0.5761092518622476, "ci95": [0.7221527504365517, 0.8086111694029999], "loco_mean": 0.17490397542579705, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7108589060252217, "p": 7.570490911324105e-74, "kendall": 0.5244521213429774, "ci95": [0.6647373626280134, 0.7590285725368467], "loco_mean": 0.6466917724774698, "n": 472}, {"predictor": "fragility", "spearman": 0.6402369351573364, "p": 8.412937618018045e-56, "kendall": 0.5176148935000774, "ci95": [0.5762098860164254, 0.6965181301409472], "loco_mean": 0.34843798881555416, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4577071159234756, "p": 8.112397111719808e-26, "kendall": 0.30139623592068804, "ci95": [0.3719011054200997, 0.5362764648442968], "loco_mean": -0.15587895297039037, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.07233275965780582, "p": 0.11656426348042145, "kendall": -0.045983526648003925, "ci95": [-0.1676638264409657, 0.0276888325777201], "loco_mean": 0.27295969447731966, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9663956649462877, "p": 1.9769716421360586e-279, "kendall": 0.844200942819101, "ci95": [0.9582833668806262, 0.972249511018557], "loco_mean": 0.822013981169579, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8142104839116457, "p": 4.6349247787002345e-113, "kendall": 0.6190938860700277, "ci95": [0.7784315483263554, 0.8457827980307192], "loco_mean": 0.2507519638805218, "n": 472}, {"predictor": "pac", "spearman": 0.7989523437605205, "p": 7.370681520368353e-106, "kendall": 0.6131202993990427, "ci95": [0.7548202970707872, 0.8362192541519101], "loco_mean": 0.7092535567689743, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6528079273309736, "p": 1.1636128383807618e-58, "kendall": 0.47211126704811257, "ci95": [0.594467447577123, 0.7075880242965953], "loco_mean": 0.5938518661656708, "n": 472}, {"predictor": "fragility", "spearman": 0.6434621748707247, "p": 1.6002788335451959e-56, "kendall": 0.5162034040583596, "ci95": [0.5777199855374996, 0.6972436181916808], "loco_mean": 0.30639625747257265, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.46410427354367867, "p": 1.377354979546913e-26, "kendall": 0.31332541653172113, "ci95": [0.38047007067527516, 0.5387551616620547], "loco_mean": -0.07487413986761712, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.05687906812360867, "p": 0.21740724040384907, "kendall": -0.03140843113445152, "ci95": [-0.15064310621609026, 0.04683266118594591], "loco_mean": 0.24522656676053, "n": 472}]} +{"target": "excess_rotation", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.045730142358639335, "p": 0.3214915785134481, "kendall": 0.029813955162114502, "ci95": [-0.04291475943890026, 0.13280723679010745], "loco_mean": 0.11571837660547339, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.019548618802250367, "p": 0.6718436438563953, "kendall": 0.01587965653174074, "ci95": [-0.06592842183087877, 0.11021518205534078], "loco_mean": 0.11096384112403068, "n": 472}, {"predictor": "pac", "spearman": -0.028921926535337233, "p": 0.5307837018089083, "kendall": -0.022059088128396125, "ci95": [-0.11461587366461326, 0.05697485566708844], "loco_mean": 0.0566288049921826, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": -0.11740393878758819, "p": 0.010688017745394272, "kendall": -0.08521357371621864, "ci95": [-0.20901249520866028, -0.030594774861803075], "loco_mean": -0.17614889459054547, "n": 472}, {"predictor": "fragility", "spearman": -0.13634890441214695, "p": 0.0029947554168688454, "kendall": -0.0995317208632813, "ci95": [-0.2216850177242833, -0.05109443013239468], "loco_mean": 0.01926842737612711, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.19517738512287972, "p": 1.9527049508461923e-05, "kendall": -0.12812623700025189, "ci95": [-0.28507442645385095, -0.11649292054466626], "loco_mean": -0.11715629043368607, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.24497186857439832, "p": 7.030580997726293e-08, "kendall": -0.1661088920076289, "ci95": [-0.3367444423080759, -0.15504169609386367], "loco_mean": -0.11500811181491921, "n": 472}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.945277182014238, "p": 9.337459754910518e-231, "kendall": 0.7983554643923854, "ci95": [0.9333084997956226, 0.9543142070293422], "loco_mean": 0.7141063826933135, "n": 472}, {"predictor": "pac", "spearman": 0.829368199518008, "p": 6.627571105525153e-121, "kendall": 0.6507610925186225, "ci95": [0.7900683215434267, 0.8626805965121719], "loco_mean": 0.7336797051514793, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7722893988883461, "p": 1.1694016278709668e-94, "kendall": 0.5790420670049299, "ci95": [0.7261852611270078, 0.8115416236341599], "loco_mean": 0.1736109975703192, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7014051904425961, "p": 4.00772467868457e-71, "kendall": 0.5190722947929036, "ci95": [0.6523637437930375, 0.7519192922696618], "loco_mean": 0.6471397619879594, "n": 472}, {"predictor": "fragility", "spearman": 0.6363823982599154, "p": 5.9577092259848535e-55, "kendall": 0.5143793253952167, "ci95": [0.5725907009133878, 0.6937731505242939], "loco_mean": 0.34986796894079264, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4581480709700259, "p": 7.187635460885088e-26, "kendall": 0.30306955989780127, "ci95": [0.37208319007295093, 0.5348514178665623], "loco_mean": -0.15719014178157917, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.06573052678524917, "p": 0.1539294493094097, "kendall": -0.04105318569650595, "ci95": [-0.1621986615192486, 0.032661878174654775], "loco_mean": 0.26767276518643324, "n": 472}]} +{"target": "rot", "n_configs": 65, "table": [{"predictor": "raptor_stability", "spearman": 0.5911276223776224, "p": 2.1672603442221434e-07, "kendall": 0.4173076923076923, "ci95": [0.41440898366535817, 0.7344290509215793], "loco_mean": 0.5911276223776224, "n": 65}, {"predictor": "sip_eigengap", "spearman": 0.44897557844424857, "p": 0.0001758950578581869, "kendall": 0.2983639495498553, "ci95": [0.23603641574980944, 0.6343287327262498], "loco_mean": 0.44897557844424857, "n": 65}, {"predictor": "pac", "spearman": 0.30882867132867137, "p": 0.012312509028599924, "kendall": 0.22980769230769232, "ci95": [0.06200499586869199, 0.5260443637575132], "loco_mean": 0.30882867132867137, "n": 65}, {"predictor": "augmentation_robustness", "spearman": 0.2582167832167832, "p": 0.037821993052357034, "kendall": 0.20384615384615387, "ci95": [0.020557929438251295, 0.46731002998794663], "loco_mean": 0.2582167832167832, "n": 65}, {"predictor": "fragility", "spearman": 0.10679428041192313, "p": 0.39715011987161525, "kendall": 0.08337535720556229, "ci95": [-0.13908064801589717, 0.3307824275559058], "loco_mean": 0.10679428041192313, "n": 65}, {"predictor": "whitened_cosine_id", "spearman": 0.09256993006993007, "p": 0.4633032146744395, "kendall": 0.058653846153846154, "ci95": [-0.1762755938835871, 0.3483470158023297], "loco_mean": 0.09256993006993007, "n": 65}, {"predictor": "xie_feature_dispersion", "spearman": -0.2018793706293706, "p": 0.10681509556234108, "kendall": -0.12019230769230771, "ci95": [-0.41763718111997106, 0.05416957753817809], "loco_mean": -0.2018793706293706, "n": 65}]} +{"target": "rot", "n_configs": 30, "table": [{"predictor": "augmentation_robustness", "spearman": 0.9523915461624026, "p": 5.619668712783104e-16, "kendall": 0.8482758620689655, "ci95": [0.864686979660719, 0.9865953239369837], "loco_mean": 0.9523915461624026, "n": 30}, {"predictor": "pac", "spearman": 0.9492769744160177, "p": 1.3384988023003793e-15, "kendall": 0.8298850574712644, "ci95": [0.8589262081591627, 0.9825238459373974], "loco_mean": 0.9492769744160177, "n": 30}, {"predictor": "raptor_stability", "spearman": 0.9145717463848719, "p": 1.591848261552271e-12, "kendall": 0.7701149425287357, "ci95": [0.8019984353499967, 0.9703535542920806], "loco_mean": 0.9145717463848719, "n": 30}, {"predictor": "fragility", "spearman": 0.5781130008770354, "p": 0.0008200072678836242, "kendall": 0.48222918208303855, "ci95": [0.26840285452584595, 0.7832806431752672], "loco_mean": 0.5781130008770354, "n": 30}, {"predictor": "sip_eigengap", "spearman": 0.43680463920753176, "p": 0.015802027838103843, "kendall": 0.33410226938009097, "ci95": [-0.0031909130751061653, 0.7400731438760575], "loco_mean": 0.43680463920753176, "n": 30}, {"predictor": "whitened_cosine_id", "spearman": -0.09499443826473859, "p": 0.6175475919174652, "kendall": -0.06666666666666667, "ci95": [-0.5053692755201276, 0.2995932701894814], "loco_mean": -0.09499443826473859, "n": 30}, {"predictor": "xie_feature_dispersion", "spearman": -0.4558398220244716, "p": 0.011355265245928056, "kendall": -0.33333333333333337, "ci95": [-0.7491798041486974, -0.06494777422855233], "loco_mean": -0.4558398220244716, "n": 30}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "xie_feature_dispersion", "spearman": 0.4728519361053399, "p": 1.1464290456409484e-27, "kendall": 0.3253985389902479, "ci95": [0.3967563726075673, 0.542265881293586], "loco_mean": 0.016789323632302774, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.2862441882147689, "p": 2.3677522446575233e-10, "kendall": 0.2094893662960164, "ci95": [0.19278007884190945, 0.36980131480594247], "loco_mean": 0.03363734818312618, "n": 472}, {"predictor": "pac", "spearman": 0.2804438719231004, "p": 5.591892097984572e-10, "kendall": 0.19426751592356684, "ci95": [0.18701914871438588, 0.36463149935028044], "loco_mean": -0.007956839965853247, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.25572471258609414, "p": 1.7506347437926356e-08, "kendall": 0.1776242398071179, "ci95": [0.16331496782130528, 0.34704649902275786], "loco_mean": 0.026340683746993027, "n": 472}, {"predictor": "raptor_stability", "spearman": 0.25140485949807356, "p": 3.0840154954490405e-08, "kendall": 0.17542912663284033, "ci95": [0.15495279727696354, 0.34964140973153524], "loco_mean": -0.008636272013881119, "n": 472}, {"predictor": "fragility", "spearman": 0.20639010867268004, "p": 6.160185397919852e-06, "kendall": 0.16185440852066482, "ci95": [0.11576886871308321, 0.2946208683742071], "loco_mean": -0.1915255148436771, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.05727620177722808, "p": 0.2142077888687003, "kendall": 0.03383761372004358, "ci95": [-0.021117453461432404, 0.13897505421146655], "loco_mean": 0.07412953871153372, "n": 472}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.945277182014238, "p": 9.337459754910518e-231, "kendall": 0.7983554643923854, "ci95": [0.9333084997956226, 0.9543142070293422], "loco_mean": 0.7141063826933135, "n": 472}, {"predictor": "pac", "spearman": 0.829368199518008, "p": 6.627571105525153e-121, "kendall": 0.6507610925186225, "ci95": [0.7900683215434267, 0.8626805965121719], "loco_mean": 0.7336797051514793, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7722893988883461, "p": 1.1694016278709668e-94, "kendall": 0.5790420670049299, "ci95": [0.7261852611270078, 0.8115416236341599], "loco_mean": 0.1736109975703192, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7014051904425961, "p": 4.00772467868457e-71, "kendall": 0.5190722947929036, "ci95": [0.6523637437930375, 0.7519192922696618], "loco_mean": 0.6471397619879594, "n": 472}, {"predictor": "fragility", "spearman": 0.6363823982599154, "p": 5.9577092259848535e-55, "kendall": 0.5143793253952167, "ci95": [0.5725907009133878, 0.6937731505242939], "loco_mean": 0.34986796894079264, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4581480709700259, "p": 7.187635460885088e-26, "kendall": 0.30306955989780127, "ci95": [0.37208319007295093, 0.5348514178665623], "loco_mean": -0.15719014178157917, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.06573052678524917, "p": 0.1539294493094097, "kendall": -0.04105318569650595, "ci95": [-0.1621986615192486, 0.032661878174654775], "loco_mean": 0.26767276518643324, "n": 472}]} +{"target": "rot", "n_configs": 65, "table": [{"predictor": "raptor_stability", "spearman": 0.5911276223776224, "p": 2.1672603442221434e-07, "kendall": 0.4173076923076923, "ci95": [0.41440898366535817, 0.7344290509215793], "loco_mean": 0.5911276223776224, "n": 65}, {"predictor": "sip_eigengap", "spearman": 0.44897557844424857, "p": 0.0001758950578581869, "kendall": 0.2983639495498553, "ci95": [0.23603641574980944, 0.6343287327262498], "loco_mean": 0.44897557844424857, "n": 65}, {"predictor": "pac", "spearman": 0.30882867132867137, "p": 0.012312509028599924, "kendall": 0.22980769230769232, "ci95": [0.06200499586869199, 0.5260443637575132], "loco_mean": 0.30882867132867137, "n": 65}, {"predictor": "augmentation_robustness", "spearman": 0.2582167832167832, "p": 0.037821993052357034, "kendall": 0.20384615384615387, "ci95": [0.020557929438251295, 0.46731002998794663], "loco_mean": 0.2582167832167832, "n": 65}, {"predictor": "fragility", "spearman": 0.10679428041192313, "p": 0.39715011987161525, "kendall": 0.08337535720556229, "ci95": [-0.13908064801589717, 0.3307824275559058], "loco_mean": 0.10679428041192313, "n": 65}, {"predictor": "whitened_cosine_id", "spearman": 0.09256993006993007, "p": 0.4633032146744395, "kendall": 0.058653846153846154, "ci95": [-0.1762755938835871, 0.3483470158023297], "loco_mean": 0.09256993006993007, "n": 65}, {"predictor": "xie_feature_dispersion", "spearman": -0.2018793706293706, "p": 0.10681509556234108, "kendall": -0.12019230769230771, "ci95": [-0.41763718111997106, 0.05416957753817809], "loco_mean": -0.2018793706293706, "n": 65}]} +{"target": "rot", "n_configs": 30, "table": [{"predictor": "augmentation_robustness", "spearman": 0.9523915461624026, "p": 5.619668712783104e-16, "kendall": 0.8482758620689655, "ci95": [0.864686979660719, 0.9865953239369837], "loco_mean": 0.9523915461624026, "n": 30}, {"predictor": "pac", "spearman": 0.9492769744160177, "p": 1.3384988023003793e-15, "kendall": 0.8298850574712644, "ci95": [0.8589262081591627, 0.9825238459373974], "loco_mean": 0.9492769744160177, "n": 30}, {"predictor": "raptor_stability", "spearman": 0.9145717463848719, "p": 1.591848261552271e-12, "kendall": 0.7701149425287357, "ci95": [0.8019984353499967, 0.9703535542920806], "loco_mean": 0.9145717463848719, "n": 30}, {"predictor": "fragility", "spearman": 0.5781130008770354, "p": 0.0008200072678836242, "kendall": 0.48222918208303855, "ci95": [0.26840285452584595, 0.7832806431752672], "loco_mean": 0.5781130008770354, "n": 30}, {"predictor": "sip_eigengap", "spearman": 0.43680463920753176, "p": 0.015802027838103843, "kendall": 0.33410226938009097, "ci95": [-0.0031909130751061653, 0.7400731438760575], "loco_mean": 0.43680463920753176, "n": 30}, {"predictor": "whitened_cosine_id", "spearman": -0.09499443826473859, "p": 0.6175475919174652, "kendall": -0.06666666666666667, "ci95": [-0.5053692755201276, 0.2995932701894814], "loco_mean": -0.09499443826473859, "n": 30}, {"predictor": "xie_feature_dispersion", "spearman": -0.4558398220244716, "p": 0.011355265245928056, "kendall": -0.33333333333333337, "ci95": [-0.7491798041486974, -0.06494777422855233], "loco_mean": -0.4558398220244716, "n": 30}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "xie_feature_dispersion", "spearman": 0.4728519361053399, "p": 1.1464290456409484e-27, "kendall": 0.3253985389902479, "ci95": [0.3967563726075673, 0.542265881293586], "loco_mean": 0.016789323632302774, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.2862441882147689, "p": 2.3677522446575233e-10, "kendall": 0.2094893662960164, "ci95": [0.19278007884190945, 0.36980131480594247], "loco_mean": 0.03363734818312618, "n": 472}, {"predictor": "pac", "spearman": 0.2804438719231004, "p": 5.591892097984572e-10, "kendall": 0.19426751592356684, "ci95": [0.18701914871438588, 0.36463149935028044], "loco_mean": -0.007956839965853247, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.25572471258609414, "p": 1.7506347437926356e-08, "kendall": 0.1776242398071179, "ci95": [0.16331496782130528, 0.34704649902275786], "loco_mean": 0.026340683746993027, "n": 472}, {"predictor": "raptor_stability", "spearman": 0.25140485949807356, "p": 3.0840154954490405e-08, "kendall": 0.17542912663284033, "ci95": [0.15495279727696354, 0.34964140973153524], "loco_mean": -0.008636272013881119, "n": 472}, {"predictor": "fragility", "spearman": 0.20639010867268004, "p": 6.160185397919852e-06, "kendall": 0.16185440852066482, "ci95": [0.11576886871308321, 0.2946208683742071], "loco_mean": -0.1915255148436771, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.05727620177722808, "p": 0.2142077888687003, "kendall": 0.03383761372004358, "ci95": [-0.021117453461432404, 0.13897505421146655], "loco_mean": 0.07412953871153372, "n": 472}]} +{"target": "wrot", "n_configs": 472, "table": [{"predictor": "whitened_cosine_id", "spearman": 0.8514251013952239, "p": 8.05914372015269e-134, "kendall": 0.6675249873983115, "ci95": [0.8169124751649481, 0.8812713665284593], "loco_mean": 0.32081039257834737, "n": 472}, {"predictor": "raptor_stability", "spearman": 0.6983643127114038, "p": 2.8596164946924883e-70, "kendall": 0.5005019591164042, "ci95": [0.6513243738457901, 0.7396962419876157], "loco_mean": 0.1956190459132168, "n": 472}, {"predictor": "pac", "spearman": 0.5198340761382322, "p": 4.970620951228867e-34, "kendall": 0.36583210339271477, "ci95": [0.4437892430576958, 0.5877014248522723], "loco_mean": 0.23975770195516763, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.40305600280562864, "p": 7.289427684209684e-20, "kendall": 0.28328380279203086, "ci95": [0.321489729537672, 0.4776665683716137], "loco_mean": 0.2718330122689288, "n": 472}, {"predictor": "fragility", "spearman": 0.3876601425787018, "p": 2.248687932106508e-18, "kendall": 0.29273055948888754, "ci95": [0.30600219029860376, 0.4585492232132769], "loco_mean": -0.051245779921125854, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.35779386132355656, "p": 1.060277059834682e-15, "kendall": 0.2393720677505765, "ci95": [0.26486949295184636, 0.43746052453496154], "loco_mean": -0.2951519978651336, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.18559915346383918, "p": 4.973801779603059e-05, "kendall": 0.149965363343665, "ci95": [0.0892034327866862, 0.2864529588640505], "loco_mean": 0.5621560113416906, "n": 472}]} +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17575700499061594, "p": 0.00012393304884855468, "kendall": 0.11512804250988683, "ci95": [0.0931610960850263, 0.2606467550238726], "loco_mean": -0.0023408000169005913, "n": 472}, {"predictor": "pac", "spearman": 0.11658303517178874, "p": 0.011252286438128154, "kendall": 0.07565145811249811, "ci95": [0.02830630692364555, 0.1991781082073618], "loco_mean": 0.0073097962622234265, "n": 472}, {"predictor": "fragility", "spearman": 0.10334711636425067, "p": 0.02474657713128782, "kendall": 0.07809889354346514, "ci95": [0.01922910382686104, 0.19395358986124475], "loco_mean": -0.08452106568076935, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.07733050082837446, "p": 0.09332613293324303, "kendall": 0.05407788259177383, "ci95": [-0.002749244960616738, 0.16470961602267215], "loco_mean": -0.035742463048130536, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.0033606869637184205, "p": 0.9419498178902012, "kendall": 0.0035718623679457657, "ci95": [-0.08383641544625166, 0.09689432149774356], "loco_mean": 0.1228704143578274, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.06410527186395897, "p": 0.16438904827779302, "kendall": -0.04204058315610699, "ci95": [-0.1432558153374885, 0.017048804365690194], "loco_mean": -0.04871094798149104, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.08229575321966996, "p": 0.0740650171034376, "kendall": -0.054167847460500795, "ci95": [-0.16408415018695408, 0.005594174911750573], "loco_mean": -0.10642573904226603, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9013496602340944, "p": 6.357556076293926e-173, "kendall": 0.7325020691640577, "ci95": [0.877073233791799, 0.920721605087468], "loco_mean": 0.5594339280001045, "n": 472}, {"predictor": "pac", "spearman": 0.781829388284427, "p": 1.7483208849233304e-98, "kendall": 0.5956853431213789, "ci95": [0.7332270466082501, 0.8200004992710067], "loco_mean": 0.5841792346121094, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7658508161434283, "p": 3.514265058477527e-92, "kendall": 0.5720069092086796, "ci95": [0.7189178325879514, 0.8069086454775533], "loco_mean": 0.18785899064302766, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6783146205127633, "p": 6.679755301240705e-65, "kendall": 0.4958256864226852, "ci95": [0.6200355005862529, 0.7315225102963346], "loco_mean": 0.5455664216397139, "n": 472}, {"predictor": "fragility", "spearman": 0.588005881008943, "p": 3.1202424743774045e-45, "kendall": 0.4678309037688001, "ci95": [0.5174106973310436, 0.648605788949667], "loco_mean": 0.27740582829001886, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.47063768901211694, "p": 2.1656787339750425e-27, "kendall": 0.3164741444456439, "ci95": [0.383918143515142, 0.5464030664430171], "loco_mean": -0.12797015306799464, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.0765129380845355, "p": 0.09685141254312134, "kendall": -0.05025731470174274, "ci95": [-0.17367827584426898, 0.023597389380876896], "loco_mean": 0.27087300465379066, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9171318339188009, "p": 7.03697100909215e-190, "kendall": 0.754543164561517, "ci95": [0.8985646163522873, 0.9323157745567195], "loco_mean": 0.6429054444264358, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8080980526996059, "p": 4.2513628148481576e-110, "kendall": 0.6102054769872971, "ci95": [0.7732265163955616, 0.8410437538923019], "loco_mean": 0.21834023358572696, "n": 472}, {"predictor": "pac", "spearman": 0.7309003357146885, "p": 5.3030976276448176e-80, "kendall": 0.5417611285040842, "ci95": [0.6781057140322055, 0.7755613875732826], "loco_mean": 0.5353762395626722, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6087757586104348, "p": 3.3585554823196054e-49, "kendall": 0.4324013098708121, "ci95": [0.5444574758505369, 0.6677287766510341], "loco_mean": 0.47233106242534606, "n": 472}, {"predictor": "fragility", "spearman": 0.5767757877584535, "p": 3.332972431421739e-43, "kendall": 0.45543030627338144, "ci95": [0.503615838672514, 0.640064032500663], "loco_mean": 0.21437723236973916, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4661472283168003, "p": 7.75654798307929e-27, "kendall": 0.31566447155349237, "ci95": [0.3814122521483456, 0.5417964336281288], "loco_mean": -0.06285643910169526, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.06170886735023185, "p": 0.18077235809769074, "kendall": -0.03708568764779513, "ci95": [-0.15656221442170665, 0.03907977373136529], "loco_mean": 0.22444394496852818, "n": 472}]} +{"target": "excess_rotation", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.07036268552578755, "p": 0.126883408959992, "kendall": 0.043038612400590155, "ci95": [-0.015172413137460544, 0.1581937811693087], "loco_mean": 0.12860533086458137, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.020135247981740382, "p": 0.6625945023614354, "kendall": 0.01606866524477489, "ci95": [-0.06494735163655148, 0.1084776056721428], "loco_mean": 0.12268432480660452, "n": 472}, {"predictor": "pac", "spearman": -0.0003682613703979026, "p": 0.9936333708423059, "kendall": -0.00561373205225089, "ci95": [-0.08837508480373547, 0.08616045119048513], "loco_mean": 0.08205719900750737, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": -0.07707309925437059, "p": 0.09442478844209143, "kendall": -0.0561013350606355, "ci95": [-0.1694874589149742, 0.01068925084031073], "loco_mean": -0.13705407857744195, "n": 472}, {"predictor": "fragility", "spearman": -0.10057911564392039, "p": 0.02889714941347261, "kendall": -0.07539736107841664, "ci95": [-0.17717689572049977, -0.00957358352798597], "loco_mean": 0.06427193866192853, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.17701640503409988, "p": 0.00011056389280422307, "kendall": -0.11875202418223038, "ci95": [-0.26477656426580964, -0.09753277308382707], "loco_mean": -0.09582594429605816, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.2335060102948853, "p": 2.886757971471768e-07, "kendall": -0.15844398862859405, "ci95": [-0.32683586423016264, -0.1421926040161361], "loco_mean": -0.0647058036971462, "n": 472}]} diff --git a/repro_bundle/results_full_run/results/audit.jsonl b/repro_bundle/results_full_run/results/audit.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..c242a1028ed26f5f9fcaeab361d6a33e18943f33 --- /dev/null +++ b/repro_bundle/results_full_run/results/audit.jsonl @@ -0,0 +1,84 @@ +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": NaN, "n": 0} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": NaN, "n": 0} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": NaN, "n": 0} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": NaN, "n": 0} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": NaN, "n": 0} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": NaN, "n": 0} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} diff --git a/repro_bundle/results_full_run/results/eval.jsonl b/repro_bundle/results_full_run/results/eval.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..a6cae7c64e2676b08daa63ea3e26e4850d1ac8c6 --- /dev/null +++ b/repro_bundle/results_full_run/results/eval.jsonl @@ -0,0 +1,234 @@ +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7344173441734417, "drop": -0.001917344173441693, "rotation": {"subspace_principal_angle": 1.238417516343073, "mean_class_cosine": 0.32629256485000935}}, "domain": {"acc": 0.5525, "drop": 0.18000000000000005, "rotation": {"subspace_principal_angle": 1.2484850932123424, "mean_class_cosine": 0.316759623263348}}}, "iid_acc": 0.7325, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.2581986440577988, "mean_class_cosine": 0.30753146487515665}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.6910569105691057, "drop": -0.07605691056910568, "rotation": {"subspace_principal_angle": 1.4661494551492598, "mean_class_cosine": 0.6870700378326343}}, "domain": {"acc": 0.52, "drop": 0.09499999999999997, "rotation": {"subspace_principal_angle": 1.056683472699715, "mean_class_cosine": 0.5780819623595075}}}, "iid_acc": 0.615, "selectivity": 0.05999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.0231199308965173, "mean_class_cosine": 0.649658877799457}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7375, "drop": 0.0}, "paraphrase": {"acc": 0.7317073170731707, "drop": 0.005792682926829329}, "domain": {"acc": 0.5525, "drop": 0.18500000000000005}}, "iid_acc": 0.7375, "selectivity": 0.1825} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8075, "drop": 0.0}, "paraphrase": {"acc": 0.7965116279069767, "drop": 0.01098837209302328, "rotation": {"subspace_principal_angle": 0.8035537849788047, "mean_class_cosine": 0.6941529859280129}}, "domain": {"acc": 0.6875, "drop": 0.12, "rotation": {"subspace_principal_angle": 1.1686325991529884, "mean_class_cosine": 0.3914103543132651}}, "length": {"acc": 0.8, "drop": 0.007499999999999951, "rotation": {"subspace_principal_angle": 0.7897497491232404, "mean_class_cosine": 0.7040230601402793}}}, "iid_acc": 0.8075, "selectivity": 0.30374999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7674349618290983, "mean_class_cosine": 0.7196939193951748}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6414728682170543, "drop": 0.01602713178294568, "rotation": {"subspace_principal_angle": 1.288879288748167, "mean_class_cosine": 0.8485185221141581}}, "domain": {"acc": 0.6625, "drop": -0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.4368459632860788, "mean_class_cosine": 0.5607042188856695}}, "length": {"acc": 0.6621212121212121, "drop": -0.004621212121212137, "rotation": {"subspace_principal_angle": 1.2230547063204023, "mean_class_cosine": 0.9166318400068196}}}, "iid_acc": 0.6575, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.8046269051289581, "mean_class_cosine": 0.8820641372607203}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.7926356589147286, "drop": -0.005135658914728669}, "domain": {"acc": 0.69625, "drop": 0.09124999999999994}, "length": {"acc": 0.7878787878787878, "drop": -0.00037878787878786735}}, "iid_acc": 0.7875, "selectivity": 0.28374999999999995} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.8425925925925926, "drop": -0.02009259259259255, "rotation": {"subspace_principal_angle": 1.4676260741902043, "mean_class_cosine": 0.26230928570700685}}}, "iid_acc": 0.8225, "selectivity": 0.5425, "iid_split_rotation": {"subspace_principal_angle": 1.421852168249981, "mean_class_cosine": 0.27308676619000416}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8580246913580247, "drop": 0.0019753086419752597, "rotation": {"subspace_principal_angle": 1.491630450202763, "mean_class_cosine": 0.9383014448695042}}}, "iid_acc": 0.86, "selectivity": 0.58, "iid_split_rotation": {"subspace_principal_angle": 1.4774909443989137, "mean_class_cosine": 0.9260985034309768}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.4375, "drop": 0.0}, "paraphrase": {"acc": 0.41358024691358025, "drop": 0.02391975308641975}}, "iid_acc": 0.4375, "selectivity": 0.15749999999999997} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.8927680798004988, "drop": 0.04223192019950128, "rotation": {"subspace_principal_angle": 1.2106415683121374, "mean_class_cosine": 0.7066695308643387}}}, "iid_acc": 0.935, "selectivity": 0.8525, "iid_split_rotation": {"subspace_principal_angle": 1.1511700616225666, "mean_class_cosine": 0.7422879247769008}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5311720698254364, "drop": 0.02382793017456364, "rotation": {"subspace_principal_angle": 1.373149046572336, "mean_class_cosine": 0.8538155049730057}}}, "iid_acc": 0.555, "selectivity": 0.47250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3861255031014492, "mean_class_cosine": 0.8730732976173553}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.85785536159601, "drop": 0.03214463840399007}}, "iid_acc": 0.89, "selectivity": 0.8075} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5019607843137255, "drop": 0.01678921568627456, "rotation": {"subspace_principal_angle": 1.5573365448958074, "mean_class_cosine": -0.013459375494237761}}}, "iid_acc": 0.51875, "selectivity": 0.011250000000000093, "iid_split_rotation": {"subspace_principal_angle": 1.5530648223509733, "mean_class_cosine": -0.017730575309227073}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.0015604575163399304, "rotation": {"subspace_principal_angle": 1.5316164005556767, "mean_class_cosine": 0.04651835530761543}}}, "iid_acc": 0.50875, "selectivity": 0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.4740729716187824, "mean_class_cosine": 0.03841895244711871}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.5202614379084968, "drop": -0.010261437908496762}}, "iid_acc": 0.51, "selectivity": 0.0025000000000000577} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.61, "drop": 0.0}, "paraphrase": {"acc": 0.5486018641810919, "drop": 0.06139813581890807, "rotation": {"subspace_principal_angle": 1.4117139037388426, "mean_class_cosine": 0.3559915861697644}}}, "iid_acc": 0.61, "selectivity": 0.37375, "iid_split_rotation": {"subspace_principal_angle": 1.3714632310508308, "mean_class_cosine": 0.37969570224814214}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.275, "drop": 0.0}, "paraphrase": {"acc": 0.2583222370173103, "drop": 0.016677762982689748, "rotation": {"subspace_principal_angle": 1.5666753155926543, "mean_class_cosine": 0.5349177629069118}}}, "iid_acc": 0.275, "selectivity": 0.038750000000000034, "iid_split_rotation": {"subspace_principal_angle": 1.5082299186645924, "mean_class_cosine": 0.6046278432817893}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.507323568575233, "drop": -0.009823568575232977}}, "iid_acc": 0.4975, "selectivity": 0.26125} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.70875, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.0009274193548387544, "rotation": {"subspace_principal_angle": 1.3772445961922097, "mean_class_cosine": 0.19234551093090788}}}, "iid_acc": 0.70875, "selectivity": 0.18625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3584465776147807, "mean_class_cosine": 0.2107574495845495}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.5825, "drop": 0.0}, "paraphrase": {"acc": 0.5942275042444821, "drop": -0.011727504244482101, "rotation": {"subspace_principal_angle": 1.2205673531066674, "mean_class_cosine": 0.9079473240997651}}}, "iid_acc": 0.5825, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1442505653422113, "mean_class_cosine": 0.940086831782734}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.7045840407470289, "drop": 0.007915959252971172}}, "iid_acc": 0.7125, "selectivity": 0.19000000000000006} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.028266129032258003, "rotation": {"subspace_principal_angle": 1.4282418254454519, "mean_class_cosine": 0.142072164699308}}}, "iid_acc": 0.66375, "selectivity": 0.16374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3468259132949723, "mean_class_cosine": 0.2221026092043146}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5838709677419355, "drop": 0.016129032258064502, "rotation": {"subspace_principal_angle": 1.3876894491282326, "mean_class_cosine": 0.9792973864192993}}}, "iid_acc": 0.6, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.0920941984969763, "mean_class_cosine": 0.9685148140397846}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66125, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.019314516129032233}}, "iid_acc": 0.66125, "selectivity": 0.16125} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7600644122383253, "drop": -0.006314412238325295, "rotation": {"subspace_principal_angle": 1.115598956625283, "mean_class_cosine": 0.43963959517656737}}}, "iid_acc": 0.75375, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1918132960864172, "mean_class_cosine": 0.369975854351391}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.007481884057971011, "rotation": {"subspace_principal_angle": 1.431073468315769, "mean_class_cosine": 0.8636351111958729}}}, "iid_acc": 0.52875, "selectivity": -0.1037499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9984226130449826, "mean_class_cosine": 0.8215813059697017}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7342995169082126, "drop": -0.009299516908212624}}, "iid_acc": 0.725, "selectivity": 0.09250000000000003} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.879286694101509, "drop": 0.02321330589849102, "rotation": {"subspace_principal_angle": 0.9139246510391348, "mean_class_cosine": 0.6106425040490346}}}, "iid_acc": 0.9025, "selectivity": 0.44499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.884850806935217, "mean_class_cosine": 0.6334049571887388}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8491083676268861, "drop": 0.012141632373113831, "rotation": {"subspace_principal_angle": 0.783728326709098, "mean_class_cosine": 0.9410916640925551}}}, "iid_acc": 0.86125, "selectivity": 0.40374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.9780082924965329, "mean_class_cosine": 0.9403009956281769}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8710562414266118, "drop": 0.025193758573388236}}, "iid_acc": 0.89625, "selectivity": 0.43875} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9862778730703259, "drop": 0.007472126929674139, "rotation": {"subspace_principal_angle": 0.8999811261795105, "mean_class_cosine": 0.6216247525314251}}}, "iid_acc": 0.99375, "selectivity": 0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.8781549302342212, "mean_class_cosine": 0.6385721258735755}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64375, "drop": 0.0}, "paraphrase": {"acc": 0.4854202401372213, "drop": 0.15832975986277875, "rotation": {"subspace_principal_angle": 0.9223293102070813, "mean_class_cosine": 0.9921756731272653}}}, "iid_acc": 0.64375, "selectivity": -0.2024999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3103002492542979, "mean_class_cosine": 0.9993471541844559}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.97875, "drop": 0.0}, "paraphrase": {"acc": 0.9845626072041166, "drop": -0.005812607204116582}}, "iid_acc": 0.97875, "selectivity": 0.13250000000000006} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6826086956521739, "drop": -0.032608695652173836, "rotation": {"subspace_principal_angle": 1.5134217389057267, "mean_class_cosine": 0.05734311504400581}}}, "iid_acc": 0.65, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4857097607143221, "mean_class_cosine": 0.0849839360330566}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5260869565217391, "drop": 0.0001630434782609047, "rotation": {"subspace_principal_angle": 1.5681344894004097, "mean_class_cosine": 0.0027859134562919843}}}, "iid_acc": 0.52625, "selectivity": -0.10375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3461131864613545, "mean_class_cosine": -0.37599024476007736}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7202898550724638, "drop": -0.014039855072463747}}, "iid_acc": 0.70625, "selectivity": 0.07625000000000004} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.8817663817663818, "drop": -0.005516381766381806, "rotation": {"subspace_principal_angle": 1.0257225092803133, "mean_class_cosine": 0.5184812124632806}}}, "iid_acc": 0.87625, "selectivity": 0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.022872498909312, "mean_class_cosine": 0.5209161160438877}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7236467236467237, "drop": -0.0023967236467237196, "rotation": {"subspace_principal_angle": 1.5032004808784296, "mean_class_cosine": 0.9058820512519088}}}, "iid_acc": 0.72125, "selectivity": -0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.3645997436602872, "mean_class_cosine": 0.8668774947492062}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.8148148148148148, "drop": 0.002685185185185235}}, "iid_acc": 0.8175, "selectivity": 0.03375000000000006} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8019390581717452, "drop": 0.021810941828254826, "rotation": {"subspace_principal_angle": 1.2297289568018384, "mean_class_cosine": 0.33449317002319223}}, "domain": {"acc": 0.7775, "drop": 0.04625000000000001, "rotation": {"subspace_principal_angle": 1.2534806736867017, "mean_class_cosine": 0.31201735319379614}}}, "iid_acc": 0.82375, "selectivity": 0.33375, "iid_split_rotation": {"subspace_principal_angle": 1.249974932944242, "mean_class_cosine": 0.31534615054656556}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5526315789473685, "drop": -0.022631578947368447, "rotation": {"subspace_principal_angle": 1.3043937828785666, "mean_class_cosine": 0.9883475306540368}}, "domain": {"acc": 0.50125, "drop": 0.028750000000000053, "rotation": {"subspace_principal_angle": 1.4264081497308534, "mean_class_cosine": 0.495004129731469}}}, "iid_acc": 0.53, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.1840707821435512, "mean_class_cosine": 0.9761804914155537}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8060941828254847, "drop": 0.02015581717451531}, "domain": {"acc": 0.77375, "drop": 0.05249999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33625000000000005} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8275193798449613, "drop": 0.01498062015503876, "rotation": {"subspace_principal_angle": 1.095349444903557, "mean_class_cosine": 0.4577358103130952}}, "domain": {"acc": 0.71125, "drop": 0.13124999999999998, "rotation": {"subspace_principal_angle": 1.266476857896887, "mean_class_cosine": 0.29964398320533164}}, "length": {"acc": 0.843939393939394, "drop": -0.0014393939393939625, "rotation": {"subspace_principal_angle": 1.020996775517062, "mean_class_cosine": 0.5225163309800367}}}, "iid_acc": 0.8425, "selectivity": 0.38125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0185057868759257, "mean_class_cosine": 0.5246385975150215}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6996124031007752, "drop": 0.007887596899224802, "rotation": {"subspace_principal_angle": 1.5092754395213872, "mean_class_cosine": 0.8994925840610262}}, "domain": {"acc": 0.56125, "drop": 0.14625, "rotation": {"subspace_principal_angle": 1.5628669635664512, "mean_class_cosine": 0.41448835148075136}}, "length": {"acc": 0.7257575757575757, "drop": -0.018257575757575695, "rotation": {"subspace_principal_angle": 1.4863663784751628, "mean_class_cosine": 0.9364420792767987}}}, "iid_acc": 0.7075, "selectivity": 0.24625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5618431812325368, "mean_class_cosine": 0.89092989954099}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8352713178294574, "drop": 0.018478682170542604}, "domain": {"acc": 0.71875, "drop": 0.135}, "length": {"acc": 0.853030303030303, "drop": 0.0007196969696969813}}, "iid_acc": 0.85375, "selectivity": 0.3925} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8430769230769231, "drop": -0.0005769230769230305, "rotation": {"subspace_principal_angle": 1.4488633502262789, "mean_class_cosine": 0.2627189029713985}}}, "iid_acc": 0.8425, "selectivity": 0.60375, "iid_split_rotation": {"subspace_principal_angle": 1.3920742480225834, "mean_class_cosine": 0.2804084873444433}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8784615384615385, "drop": -0.019711538461538503, "rotation": {"subspace_principal_angle": 1.5157375083147102, "mean_class_cosine": 0.9772553909019803}}}, "iid_acc": 0.85875, "selectivity": 0.62, "iid_split_rotation": {"subspace_principal_angle": 1.5565345003357298, "mean_class_cosine": 0.9660708595879833}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5030769230769231, "drop": -0.0043269230769230616}}, "iid_acc": 0.49875, "selectivity": 0.26} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.027319201995012476, "rotation": {"subspace_principal_angle": 1.371071144342142, "mean_class_cosine": 0.6983572474879793}}}, "iid_acc": 0.955, "selectivity": 0.87125, "iid_split_rotation": {"subspace_principal_angle": 1.1190162175121645, "mean_class_cosine": 0.7424385333421911}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.46, "drop": 0.0}, "paraphrase": {"acc": 0.4339152119700748, "drop": 0.0260847880299252, "rotation": {"subspace_principal_angle": 1.2702079599204308, "mean_class_cosine": 0.8441489999187578}}}, "iid_acc": 0.46, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.54813616950135, "mean_class_cosine": 0.8639196384527672}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8952618453865336, "drop": 0.025988154613466374}}, "iid_acc": 0.92125, "selectivity": 0.8375} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": 0.003088235294117614, "rotation": {"subspace_principal_angle": 1.5507981147481529, "mean_class_cosine": 0.019996879097623668}}}, "iid_acc": 0.5325, "selectivity": 0.04124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4729086760692658, "mean_class_cosine": 0.09773139915411025}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.51125, "drop": 0.0}, "paraphrase": {"acc": 0.4915032679738562, "drop": 0.0197467320261438, "rotation": {"subspace_principal_angle": 1.5063411676739642, "mean_class_cosine": 0.09376607824598783}}}, "iid_acc": 0.51125, "selectivity": 0.019999999999999962, "iid_split_rotation": {"subspace_principal_angle": 1.3815718454510815, "mean_class_cosine": -0.780613294925026}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.01531045751633986}}, "iid_acc": 0.5225, "selectivity": 0.031249999999999944} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.5872170439414115, "drop": 0.0377829560585885, "rotation": {"subspace_principal_angle": 1.415219489070654, "mean_class_cosine": 0.29709377332489545}}}, "iid_acc": 0.625, "selectivity": 0.36125, "iid_split_rotation": {"subspace_principal_angle": 1.3868652000931787, "mean_class_cosine": 0.30693642876692934}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1975, "drop": 0.0}, "paraphrase": {"acc": 0.20905459387483355, "drop": -0.01155459387483354, "rotation": {"subspace_principal_angle": 1.508152156475103, "mean_class_cosine": 0.3868276136007782}}}, "iid_acc": 0.1975, "selectivity": -0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4103810864475195, "mean_class_cosine": 0.663164345963827}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.5539280958721704, "drop": 0.014821904127829577}}, "iid_acc": 0.56875, "selectivity": 0.305} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7470288624787776, "drop": 0.0017211375212223823, "rotation": {"subspace_principal_angle": 1.218944387247129, "mean_class_cosine": 0.34463687992333025}}}, "iid_acc": 0.74875, "selectivity": 0.21000000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.161541923084499, "mean_class_cosine": 0.39792541536895626}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.6281833616298812, "drop": -0.029433361629881194, "rotation": {"subspace_principal_angle": 1.4828776907746226, "mean_class_cosine": 0.8507212382776503}}}, "iid_acc": 0.59875, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5285287072761573, "mean_class_cosine": 0.889675723801062}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7589134125636672, "drop": -0.007663412563667205}}, "iid_acc": 0.75125, "selectivity": 0.21250000000000002} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6258064516129033, "drop": 0.029193548387096757, "rotation": {"subspace_principal_angle": 1.4380753701349163, "mean_class_cosine": 0.13233165634932476}}}, "iid_acc": 0.655, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.3608689467779782, "mean_class_cosine": 0.20838887470800788}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.6129032258064516, "drop": -0.009153225806451615, "rotation": {"subspace_principal_angle": 1.3501977335705537, "mean_class_cosine": 0.8788630744675302}}}, "iid_acc": 0.60375, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3042174668679, "mean_class_cosine": 0.9016543168707795}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6274193548387097, "drop": 0.01758064516129032}}, "iid_acc": 0.645, "selectivity": 0.13624999999999998} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7584541062801933, "drop": -0.015954106280193225, "rotation": {"subspace_principal_angle": 1.2412464706999458, "mean_class_cosine": 0.3236171403840489}}}, "iid_acc": 0.7425, "selectivity": 0.125, "iid_split_rotation": {"subspace_principal_angle": 1.2363583878771438, "mean_class_cosine": 0.32823830168996404}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5797101449275363, "drop": -0.013460144927536222, "rotation": {"subspace_principal_angle": 1.4328194682708153, "mean_class_cosine": 0.8163073052539073}}}, "iid_acc": 0.56625, "selectivity": -0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.393646227569837, "mean_class_cosine": 0.7747593471553436}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7471819645732689, "drop": -0.00843196457326889}}, "iid_acc": 0.73875, "selectivity": 0.12124999999999997} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.887517146776406, "drop": 0.02623285322359392, "rotation": {"subspace_principal_angle": 1.1698183926314918, "mean_class_cosine": 0.39031889296657585}}}, "iid_acc": 0.91375, "selectivity": 0.3949999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1565211457406315, "mean_class_cosine": 0.4025265308866283}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8847736625514403, "drop": 0.007726337448559661, "rotation": {"subspace_principal_angle": 0.9750908966255823, "mean_class_cosine": 0.724907117214936}}}, "iid_acc": 0.8925, "selectivity": 0.3737499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5507156284967207, "mean_class_cosine": 0.9656213717308835}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9012345679012346, "drop": 0.013765432098765462}}, "iid_acc": 0.915, "selectivity": 0.39625} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.7143429012203775, "mean_class_cosine": 0.7555238835683384}}}, "iid_acc": 0.99125, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6854415650897067, "mean_class_cosine": 0.7741396052879256}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.93625, "drop": 0.0}, "paraphrase": {"acc": 0.9468267581475128, "drop": -0.0105767581475128, "rotation": {"subspace_principal_angle": 1.309840282431262, "mean_class_cosine": 0.963893627422916}}}, "iid_acc": 0.93625, "selectivity": 0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.94741234766098, "mean_class_cosine": 0.9835024086435691}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9775, "drop": 0.0}, "paraphrase": {"acc": 0.9828473413379074, "drop": -0.005347341337907374}}, "iid_acc": 0.9775, "selectivity": 0.11250000000000004} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.006666666666666599, "rotation": {"subspace_principal_angle": 1.5316360507140625, "mean_class_cosine": 0.0391502679566816}}}, "iid_acc": 0.66, "selectivity": 0.07250000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3743529128243746, "mean_class_cosine": 0.19518239053186837}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.47246376811594204, "drop": 0.03753623188405797, "rotation": {"subspace_principal_angle": 1.118894178326917, "mean_class_cosine": -0.9771850067325109}}}, "iid_acc": 0.51, "selectivity": -0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3803133457755477, "mean_class_cosine": 0.946898836596458}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7072463768115942, "drop": -0.0009963768115941463}}, "iid_acc": 0.70625, "selectivity": 0.11875000000000002} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8703703703703703, "drop": 0.00462962962962965, "rotation": {"subspace_principal_angle": 1.2653845793904546, "mean_class_cosine": 0.30068589374987864}}}, "iid_acc": 0.875, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 1.1530609002795849, "mean_class_cosine": 0.4056916522664149}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.57875, "drop": 0.0}, "paraphrase": {"acc": 0.5911680911680912, "drop": -0.012418091168091228, "rotation": {"subspace_principal_angle": 1.5387855283082137, "mean_class_cosine": 0.9739607963186667}}}, "iid_acc": 0.57875, "selectivity": -0.17625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.0864399863171719, "mean_class_cosine": 0.8793845387817394}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8475783475783476, "drop": 0.006171652421652407}}, "iid_acc": 0.85375, "selectivity": 0.09875} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.838, "drop": 0.0}, "paraphrase": {"acc": 0.7936507936507936, "drop": 0.04434920634920636, "rotation": {"subspace_principal_angle": 1.234809606511884, "mean_class_cosine": 0.32970087848309004}}, "domain": {"acc": 0.5975, "drop": 0.24049999999999994, "rotation": {"subspace_principal_angle": 1.2854407619013895, "mean_class_cosine": 0.28149865550857306}}}, "iid_acc": 0.838, "selectivity": 0.31199999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.241338086239225, "mean_class_cosine": 0.32353045349175147}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.534, "drop": 0.0}, "paraphrase": {"acc": 0.5260770975056689, "drop": 0.007922902494331141, "rotation": {"subspace_principal_angle": 0.8710046022719048, "mean_class_cosine": 0.993134693849815}}, "domain": {"acc": 0.495, "drop": 0.039000000000000035, "rotation": {"subspace_principal_angle": 1.5563375225299891, "mean_class_cosine": 0.5611614903508646}}}, "iid_acc": 0.534, "selectivity": 0.008000000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.3346119956378737, "mean_class_cosine": -0.5121693038439772}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.838, "drop": 0.0}, "paraphrase": {"acc": 0.8435374149659864, "drop": -0.005537414965986476}, "domain": {"acc": 0.8025, "drop": 0.035499999999999976}}, "iid_acc": 0.838, "selectivity": 0.31199999999999994} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8604651162790697, "drop": 0.024534883720930267, "rotation": {"subspace_principal_angle": 1.1635939988143396, "mean_class_cosine": 0.396041968116701}}, "domain": {"acc": 0.56, "drop": 0.32499999999999996, "rotation": {"subspace_principal_angle": 1.3073928416159974, "mean_class_cosine": 0.26036815053777257}}, "length": {"acc": 0.8878787878787879, "drop": -0.002878787878787925, "rotation": {"subspace_principal_angle": 1.1306543361974652, "mean_class_cosine": 0.42606792700100266}}}, "iid_acc": 0.885, "selectivity": 0.41375, "iid_split_rotation": {"subspace_principal_angle": 1.1314507236110818, "mean_class_cosine": 0.42534730729927617}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7383720930232558, "drop": 0.03412790697674417, "rotation": {"subspace_principal_angle": 0.9299498041850935, "mean_class_cosine": 0.9017395970359101}}, "domain": {"acc": 0.5575, "drop": 0.21499999999999997, "rotation": {"subspace_principal_angle": 1.4111384197278427, "mean_class_cosine": 0.27301695590829955}}, "length": {"acc": 0.7878787878787878, "drop": -0.01537878787878788, "rotation": {"subspace_principal_angle": 1.0571900651122168, "mean_class_cosine": 0.9405678373193577}}}, "iid_acc": 0.7725, "selectivity": 0.30124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.339591955524043, "mean_class_cosine": 0.8972887886451133}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8662790697674418, "drop": 0.022470930232558195}, "domain": {"acc": 0.65, "drop": 0.23875000000000002}, "length": {"acc": 0.8878787878787879, "drop": 0.000871212121212106}}, "iid_acc": 0.88875, "selectivity": 0.41750000000000004} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8981481481481481, "drop": -0.0006481481481481755, "rotation": {"subspace_principal_angle": 1.237598916483474, "mean_class_cosine": 0.4232975213360548}}}, "iid_acc": 0.8975, "selectivity": 0.6375, "iid_split_rotation": {"subspace_principal_angle": 1.3220663076329475, "mean_class_cosine": 0.3932991756101093}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.6125, "drop": 0.0}, "paraphrase": {"acc": 0.6141975308641975, "drop": -0.0016975308641974385, "rotation": {"subspace_principal_angle": 1.129062659446162, "mean_class_cosine": 0.9448291188886624}}}, "iid_acc": 0.6125, "selectivity": 0.35250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2505018094041758, "mean_class_cosine": 0.9056129957175688}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8827160493827161, "drop": -0.005216049382716137}}, "iid_acc": 0.8775, "selectivity": 0.6174999999999999} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.9725, "drop": 0.0}, "paraphrase": {"acc": 0.9551122194513716, "drop": 0.01738778054862844, "rotation": {"subspace_principal_angle": 1.1805912863774084, "mean_class_cosine": 0.696959257792652}}}, "iid_acc": 0.9725, "selectivity": 0.875, "iid_split_rotation": {"subspace_principal_angle": 1.1819768188421824, "mean_class_cosine": 0.7412159284602584}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.4275, "drop": 0.0}, "paraphrase": {"acc": 0.42643391521197005, "drop": 0.0010660847880299418, "rotation": {"subspace_principal_angle": 1.267210976332211, "mean_class_cosine": 0.8469328938765297}}}, "iid_acc": 0.4275, "selectivity": 0.32999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2708278894045024, "mean_class_cosine": 0.8352289736541864}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.0210692019950125}}, "iid_acc": 0.94875, "selectivity": 0.85125} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5568627450980392, "drop": -0.006862745098039191, "rotation": {"subspace_principal_angle": 1.4681160564274516, "mean_class_cosine": 0.10249993503953822}}}, "iid_acc": 0.55, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.5020669180006716, "mean_class_cosine": 0.06867531169240668}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.4875816993464052, "drop": 0.00991830065359478, "rotation": {"subspace_principal_angle": 1.0583311490868843, "mean_class_cosine": 0.6025164388499016}}}, "iid_acc": 0.4975, "selectivity": -0.03250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2056755940549575, "mean_class_cosine": -0.8743607002802327}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5372549019607843, "drop": 0.0002450980392156854}}, "iid_acc": 0.5375, "selectivity": 0.007499999999999951} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6604527296937417, "drop": -0.005452729693741665, "rotation": {"subspace_principal_angle": 1.3690495088359922, "mean_class_cosine": 0.35780435447931885}}}, "iid_acc": 0.655, "selectivity": 0.39375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.418659878889111, "mean_class_cosine": 0.31040393556774787}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.18125, "drop": 0.0}, "paraphrase": {"acc": 0.1904127829560586, "drop": -0.009162782956058602, "rotation": {"subspace_principal_angle": 1.4238306284342992, "mean_class_cosine": 0.3336979016265693}}}, "iid_acc": 0.18125, "selectivity": -0.07999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3516589964423749, "mean_class_cosine": 0.6675023436977741}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5712383488681758, "drop": 0.030011651131824135}}, "iid_acc": 0.60125, "selectivity": 0.33999999999999997} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.00039685908319186947, "rotation": {"subspace_principal_angle": 1.3803752925053967, "mean_class_cosine": 0.18927233568020757}}}, "iid_acc": 0.72875, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 1.4107099285898823, "mean_class_cosine": 0.1594035006847374}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5229202037351444, "drop": -0.015420203735144411, "rotation": {"subspace_principal_angle": 1.132903800026888, "mean_class_cosine": 0.9957211105004683}}}, "iid_acc": 0.5075, "selectivity": -0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1567819921044111, "mean_class_cosine": 0.997306798497698}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7606112054329371, "drop": -0.0018612054329371075}}, "iid_acc": 0.75875, "selectivity": 0.22750000000000004} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.03556451612903222, "rotation": {"subspace_principal_angle": 1.4500907567958619, "mean_class_cosine": 0.12041267342093487}}}, "iid_acc": 0.6775, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4451716043887228, "mean_class_cosine": 0.12529455713007304}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5435483870967742, "drop": 0.002701612903225792, "rotation": {"subspace_principal_angle": 1.0905650289278745, "mean_class_cosine": 0.9919303109580393}}}, "iid_acc": 0.54625, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.435319911955165, "mean_class_cosine": 0.9822586493642307}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.02806451612903227}}, "iid_acc": 0.67, "selectivity": 0.16375000000000006} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7568438003220612, "drop": 0.014406199677938769, "rotation": {"subspace_principal_angle": 1.362823497882489, "mean_class_cosine": 0.2064768368861471}}}, "iid_acc": 0.77125, "selectivity": 0.21124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3556181289971392, "mean_class_cosine": 0.21352151995440388}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.01748188405797102, "rotation": {"subspace_principal_angle": 1.162087777749045, "mean_class_cosine": 0.4004265677277434}}}, "iid_acc": 0.51875, "selectivity": -0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5040593331311518, "mean_class_cosine": 0.6240661876988235}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.775, "drop": 0.0}, "paraphrase": {"acc": 0.7729468599033816, "drop": 0.002053140096618389}}, "iid_acc": 0.775, "selectivity": 0.21499999999999997} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9245541838134431, "drop": 0.017945816186556884, "rotation": {"subspace_principal_angle": 1.1580571429217417, "mean_class_cosine": 0.401119992002977}}}, "iid_acc": 0.9425, "selectivity": 0.44875, "iid_split_rotation": {"subspace_principal_angle": 1.0639249835392377, "mean_class_cosine": 0.48544434408974335}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5116598079561042, "drop": 0.0033401920438957955, "rotation": {"subspace_principal_angle": 1.474754823264982, "mean_class_cosine": 0.27562171118558787}}}, "iid_acc": 0.515, "selectivity": 0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6855036886417566, "mean_class_cosine": 0.9705005903895466}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.93, "drop": 0.0}, "paraphrase": {"acc": 0.934156378600823, "drop": -0.004156378600822963}}, "iid_acc": 0.93, "selectivity": 0.43625} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0006389365351628928, "rotation": {"subspace_principal_angle": 0.7922854168062247, "mean_class_cosine": 0.7022200256213631}}}, "iid_acc": 0.9925, "selectivity": 0.13250000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.7419982107629325, "mean_class_cosine": 0.7371197159686783}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9502572898799314, "drop": -0.00900728987993138, "rotation": {"subspace_principal_angle": 1.2962420205278622, "mean_class_cosine": 0.9560615750258359}}}, "iid_acc": 0.94125, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4821593641247295, "mean_class_cosine": 0.9814265656310696}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9875, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": -0.0022084048027444236}}, "iid_acc": 0.9875, "selectivity": 0.12750000000000006} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6608695652173913, "drop": 0.03413043478260869, "rotation": {"subspace_principal_angle": 1.5269251960916086, "mean_class_cosine": 0.04385705910466185}}}, "iid_acc": 0.695, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.452899790541391, "mean_class_cosine": 0.117623607019381}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.041141304347826146, "rotation": {"subspace_principal_angle": 1.2698364171990568, "mean_class_cosine": -0.7441747656964055}}}, "iid_acc": 0.52375, "selectivity": -0.06624999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.421349910061863, "mean_class_cosine": 0.7671339569758747}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7231884057971014, "drop": -0.009438405797101423}}, "iid_acc": 0.71375, "selectivity": 0.12375000000000003} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8803418803418803, "drop": 0.0009081196581196549, "rotation": {"subspace_principal_angle": 1.336870712814003, "mean_class_cosine": 0.23179799551578698}}}, "iid_acc": 0.88125, "selectivity": 0.15749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.230199866592587, "mean_class_cosine": 0.3340493484245235}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.5811965811965812, "drop": -0.004946581196581201, "rotation": {"subspace_principal_angle": 0.6698879155775275, "mean_class_cosine": 0.9725719155523136}}}, "iid_acc": 0.57625, "selectivity": -0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9538385113441675, "mean_class_cosine": 0.9585511263219829}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8974358974358975, "drop": 0.0013141025641025816}}, "iid_acc": 0.89875, "selectivity": 0.17500000000000004} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8337950138504155, "drop": 0.027454986149584415, "rotation": {"subspace_principal_angle": 1.2672102146950088, "mean_class_cosine": 0.2989442428389649}}, "domain": {"acc": 0.88, "drop": -0.018750000000000044, "rotation": {"subspace_principal_angle": 1.401321553710003, "mean_class_cosine": 0.1686646701084955}}}, "iid_acc": 0.86125, "selectivity": 0.36749999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.2210418787662296, "mean_class_cosine": 0.3426671322986696}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5346260387811634, "drop": -0.009626038781163415, "rotation": {"subspace_principal_angle": 1.1075097075912679, "mean_class_cosine": 0.9893879557921825}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.2937330372237155, "mean_class_cosine": 0.27468238367063336}}}, "iid_acc": 0.525, "selectivity": 0.03125, "iid_split_rotation": {"subspace_principal_angle": 1.474523039419839, "mean_class_cosine": 0.9693041836230525}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8476454293628809, "drop": 0.04110457063711914}, "domain": {"acc": 0.86875, "drop": 0.020000000000000018}}, "iid_acc": 0.88875, "selectivity": 0.395} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.875968992248062, "drop": 0.012781007751938023, "rotation": {"subspace_principal_angle": 1.3733967390372437, "mean_class_cosine": 0.1961200848251179}}, "domain": {"acc": 0.5525, "drop": 0.33625000000000005, "rotation": {"subspace_principal_angle": 1.427574987743908, "mean_class_cosine": 0.14273220656435498}}, "length": {"acc": 0.8924242424242425, "drop": -0.003674242424242413, "rotation": {"subspace_principal_angle": 1.32660559111955, "mean_class_cosine": 0.24177114792102647}}}, "iid_acc": 0.88875, "selectivity": 0.3975, "iid_split_rotation": {"subspace_principal_angle": 1.3721751230402406, "mean_class_cosine": 0.19731783026248}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.84875, "drop": 0.0}, "paraphrase": {"acc": 0.8197674418604651, "drop": 0.028982558139534875, "rotation": {"subspace_principal_angle": 1.0890107964995421, "mean_class_cosine": 0.9365459180728041}}, "domain": {"acc": 0.5575, "drop": 0.29125, "rotation": {"subspace_principal_angle": 1.2433163855943603, "mean_class_cosine": 0.290612600127489}}, "length": {"acc": 0.8575757575757575, "drop": -0.008825757575757542, "rotation": {"subspace_principal_angle": 0.7717021170570981, "mean_class_cosine": 0.9578281621778026}}}, "iid_acc": 0.84875, "selectivity": 0.3575, "iid_split_rotation": {"subspace_principal_angle": 1.2334456221930967, "mean_class_cosine": 0.9323952105718878}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8875968992248062, "drop": 0.00990310077519374}, "domain": {"acc": 0.695, "drop": 0.2025}, "length": {"acc": 0.9015151515151515, "drop": -0.004015151515151527}}, "iid_acc": 0.8975, "selectivity": 0.40624999999999994} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8907692307692308, "drop": -0.00951923076923078, "rotation": {"subspace_principal_angle": 1.3982668079683969, "mean_class_cosine": 0.2412385692456759}}}, "iid_acc": 0.88125, "selectivity": 0.63, "iid_split_rotation": {"subspace_principal_angle": 1.3873607135770687, "mean_class_cosine": 0.27893385434758605}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.805, "drop": 0.0}, "paraphrase": {"acc": 0.8046153846153846, "drop": 0.00038461538461542766, "rotation": {"subspace_principal_angle": 1.1223434771672867, "mean_class_cosine": 0.9797187391070092}}}, "iid_acc": 0.805, "selectivity": 0.5537500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1951168916205284, "mean_class_cosine": 0.9636873787333077}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.9092307692307692, "drop": -0.010480769230769127}}, "iid_acc": 0.89875, "selectivity": 0.6475000000000001} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.972568578553616, "drop": 0.0024314214463839745, "rotation": {"subspace_principal_angle": 1.0887911120828875, "mean_class_cosine": 0.6783458118132375}}}, "iid_acc": 0.975, "selectivity": 0.8624999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.083853464584136, "mean_class_cosine": 0.7060030624663257}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.8125, "drop": 0.0}, "paraphrase": {"acc": 0.773067331670823, "drop": 0.03943266832917702, "rotation": {"subspace_principal_angle": 1.1185256458497195, "mean_class_cosine": 0.8994165755464111}}}, "iid_acc": 0.8125, "selectivity": 0.7, "iid_split_rotation": {"subspace_principal_angle": 0.9631112899466643, "mean_class_cosine": 0.9391589025650812}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9650872817955112, "drop": 0.011162718204488775}}, "iid_acc": 0.97625, "selectivity": 0.8637499999999999} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.5830065359477125, "drop": 0.024493464052287583, "rotation": {"subspace_principal_angle": 1.4835452240288083, "mean_class_cosine": 0.08714044135255376}}}, "iid_acc": 0.6075, "selectivity": 0.08750000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.483433918298054, "mean_class_cosine": 0.08725132314037118}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.4888888888888889, "drop": 0.01611111111111113, "rotation": {"subspace_principal_angle": 1.2048851781158827, "mean_class_cosine": 0.5321986735325194}}}, "iid_acc": 0.505, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4614710386057692, "mean_class_cosine": -0.6438340541689455}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5542483660130719, "drop": 0.03450163398692807}}, "iid_acc": 0.58875, "selectivity": 0.06874999999999998} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65875, "drop": 0.0}, "paraphrase": {"acc": 0.6085219707057257, "drop": 0.05022802929427428, "rotation": {"subspace_principal_angle": 1.4274964103041523, "mean_class_cosine": 0.2819535108490442}}}, "iid_acc": 0.65875, "selectivity": 0.3837499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3942837316412728, "mean_class_cosine": 0.30680585311258146}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.18, "drop": 0.0}, "paraphrase": {"acc": 0.19440745672436752, "drop": -0.014407456724367523, "rotation": {"subspace_principal_angle": 1.42729065004506, "mean_class_cosine": 0.3233324825686333}}}, "iid_acc": 0.18, "selectivity": -0.09500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.543456747904511, "mean_class_cosine": 0.6716440420685439}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02558089214380832}}, "iid_acc": 0.64875, "selectivity": 0.37375} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.005396859083191874, "rotation": {"subspace_principal_angle": 1.4458284299287356, "mean_class_cosine": 0.1246428806664818}}}, "iid_acc": 0.73375, "selectivity": 0.23625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4100208755322998, "mean_class_cosine": 0.16008370527990506}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5314091680814941, "drop": -0.012659168081494032, "rotation": {"subspace_principal_angle": 1.5017224503953317, "mean_class_cosine": 0.9922645891758275}}}, "iid_acc": 0.51875, "selectivity": 0.021250000000000047, "iid_split_rotation": {"subspace_principal_angle": 0.5510785135387823, "mean_class_cosine": 0.994402608644392}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7504244482173175, "drop": 0.020825551782682528}}, "iid_acc": 0.77125, "selectivity": 0.27375} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.6532258064516129, "drop": 0.031774193548387175, "rotation": {"subspace_principal_angle": 1.4862916929900118, "mean_class_cosine": 0.08440409464427018}}}, "iid_acc": 0.685, "selectivity": 0.18375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.504979130468674, "mean_class_cosine": 0.06576968766218992}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5564516129032258, "drop": 0.009798387096774275, "rotation": {"subspace_principal_angle": 1.4670137647528678, "mean_class_cosine": 0.9493974522411648}}}, "iid_acc": 0.56625, "selectivity": 0.06500000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2038765745555904, "mean_class_cosine": 0.9198294280970227}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.6935483870967742, "drop": 0.01895161290322578}}, "iid_acc": 0.7125, "selectivity": 0.21125000000000005} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7439613526570048, "drop": 0.0047886473429952225, "rotation": {"subspace_principal_angle": 1.4293803795970716, "mean_class_cosine": 0.14094506785538105}}}, "iid_acc": 0.74875, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4308280823099806, "mean_class_cosine": 0.13951166975487922}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.571658615136876, "drop": -0.019158615136876, "rotation": {"subspace_principal_angle": 1.3467680040328998, "mean_class_cosine": 0.780244054310545}}}, "iid_acc": 0.5525, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.5305723014222863, "mean_class_cosine": 0.8225959610699823}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.7616747181964574, "drop": 0.01207528180354267}}, "iid_acc": 0.77375, "selectivity": 0.19625000000000004} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.02258058984910838, "rotation": {"subspace_principal_angle": 1.1992451710106438, "mean_class_cosine": 0.3630611813014619}}}, "iid_acc": 0.95125, "selectivity": 0.44125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.197550530346397, "mean_class_cosine": 0.3646396665958699}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6598079561042524, "drop": -0.0023079561042523844, "rotation": {"subspace_principal_angle": 0.8946495725521172, "mean_class_cosine": 0.6299993713824044}}}, "iid_acc": 0.6575, "selectivity": 0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7999262041956853, "mean_class_cosine": 0.9697684559622158}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.945, "drop": 0.0}, "paraphrase": {"acc": 0.9314128943758574, "drop": 0.01358710562414256}}, "iid_acc": 0.945, "selectivity": 0.43499999999999994} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99875, "drop": 0.0}, "paraphrase": {"acc": 0.9982847341337907, "drop": 0.000465265866209319, "rotation": {"subspace_principal_angle": 0.931771543682588, "mean_class_cosine": 0.5964129401770092}}}, "iid_acc": 0.99875, "selectivity": 0.18125000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9266228730293125, "mean_class_cosine": 0.600537740322992}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6375, "drop": 0.0}, "paraphrase": {"acc": 0.4065180102915952, "drop": 0.23098198970840478, "rotation": {"subspace_principal_angle": 0.4804883746103172, "mean_class_cosine": 0.9773138541986853}}}, "iid_acc": 0.6375, "selectivity": -0.18000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0322313143428026, "mean_class_cosine": 0.9999323022857498}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9948542024013722, "drop": 0.001395797598627735}}, "iid_acc": 0.99625, "selectivity": 0.17874999999999996} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.6884057971014492, "drop": 0.04284420289855073, "rotation": {"subspace_principal_angle": 1.5103706788720144, "mean_class_cosine": 0.060388883021034304}}}, "iid_acc": 0.73125, "selectivity": 0.14374999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.4200020283035577, "mean_class_cosine": 0.1502234646254889}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.4666666666666667, "drop": 0.04083333333333328, "rotation": {"subspace_principal_angle": 0.9965627610068227, "mean_class_cosine": -0.9475923907212384}}}, "iid_acc": 0.5075, "selectivity": -0.08000000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.7465288923899847, "mean_class_cosine": 0.963542850231548}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7275362318840579, "drop": 0.039963768115942044}}, "iid_acc": 0.7675, "selectivity": 0.17999999999999994} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8874643874643875, "drop": 0.005035612535612466, "rotation": {"subspace_principal_angle": 1.373498225543769, "mean_class_cosine": 0.19602056818834496}}}, "iid_acc": 0.8925, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.272801169025973, "mean_class_cosine": 0.29360431510047513}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.5883190883190883, "drop": 0.0016809116809116675, "rotation": {"subspace_principal_angle": 1.0911846628602964, "mean_class_cosine": 0.9555356213618866}}}, "iid_acc": 0.59, "selectivity": -0.125, "iid_split_rotation": {"subspace_principal_angle": 1.1224788303280901, "mean_class_cosine": 0.9089747436544128}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8846153846153846, "drop": 0.01163461538461541}}, "iid_acc": 0.89625, "selectivity": 0.18125000000000002} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7853185595567868, "drop": 0.01843144044321321, "rotation": {"subspace_principal_angle": 1.3395977046860834, "mean_class_cosine": 0.2291444175563393}}, "domain": {"acc": 0.78875, "drop": 0.015000000000000013, "rotation": {"subspace_principal_angle": 1.3075865369139061, "mean_class_cosine": 0.2601811310213691}}}, "iid_acc": 0.80375, "selectivity": 0.31124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3161987442877714, "mean_class_cosine": 0.25185598358320005}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5373961218836565, "drop": -0.017396121883656468, "rotation": {"subspace_principal_angle": 1.1243174241253076, "mean_class_cosine": 0.997655368526527}}, "domain": {"acc": 0.50125, "drop": 0.018750000000000044, "rotation": {"subspace_principal_angle": 1.434291363492328, "mean_class_cosine": 0.5041835597720421}}}, "iid_acc": 0.52, "selectivity": 0.027500000000000024, "iid_split_rotation": {"subspace_principal_angle": 1.4733549164259618, "mean_class_cosine": 0.9931712323643861}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.775623268698061, "drop": 0.050626731301939065}, "domain": {"acc": 0.7475, "drop": 0.07874999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33375000000000005} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8081395348837209, "drop": 0.01561046511627906, "rotation": {"subspace_principal_angle": 1.4878379029744278, "mean_class_cosine": 0.08286330186240415}}, "domain": {"acc": 0.63875, "drop": 0.18499999999999994, "rotation": {"subspace_principal_angle": 1.454691067906504, "mean_class_cosine": 0.11584457649805602}}, "length": {"acc": 0.8227272727272728, "drop": 0.0010227272727272307, "rotation": {"subspace_principal_angle": 1.4554754679619462, "mean_class_cosine": 0.11506542197274067}}}, "iid_acc": 0.82375, "selectivity": 0.3475, "iid_split_rotation": {"subspace_principal_angle": 1.3842160089393434, "mean_class_cosine": 0.18549965489482007}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7325581395348837, "drop": 0.03494186046511627, "rotation": {"subspace_principal_angle": 1.2562269465379385, "mean_class_cosine": 0.9025478152861728}}, "domain": {"acc": 0.5575, "drop": 0.20999999999999996, "rotation": {"subspace_principal_angle": 1.3251714868629496, "mean_class_cosine": 0.29596572601612736}}, "length": {"acc": 0.7727272727272727, "drop": -0.005227272727272747, "rotation": {"subspace_principal_angle": 1.0637760788494128, "mean_class_cosine": 0.9397971855033075}}}, "iid_acc": 0.7675, "selectivity": 0.29124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.061070800271238, "mean_class_cosine": 0.9127773418749959}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.8725, "drop": 0.0}, "paraphrase": {"acc": 0.8527131782945736, "drop": 0.01978682170542645}, "domain": {"acc": 0.57875, "drop": 0.29375000000000007}, "length": {"acc": 0.8712121212121212, "drop": 0.0012878787878788378}}, "iid_acc": 0.8725, "selectivity": 0.39625000000000005} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8661538461538462, "drop": 0.0013461538461538858, "rotation": {"subspace_principal_angle": 1.2824688472167018, "mean_class_cosine": 0.4528696194346733}}}, "iid_acc": 0.8675, "selectivity": 0.64, "iid_split_rotation": {"subspace_principal_angle": 1.267228970027625, "mean_class_cosine": 0.4341122200684181}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.47875, "drop": 0.0}, "paraphrase": {"acc": 0.42615384615384616, "drop": 0.05259615384615385, "rotation": {"subspace_principal_angle": 0.5775569724987617, "mean_class_cosine": 0.9093320332006872}}}, "iid_acc": 0.47875, "selectivity": 0.25125, "iid_split_rotation": {"subspace_principal_angle": 0.7727337369433303, "mean_class_cosine": 0.8668228792038879}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.0036538461538461187}}, "iid_acc": 0.8775, "selectivity": 0.6499999999999999} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.95375, "drop": 0.0}, "paraphrase": {"acc": 0.9451371571072319, "drop": 0.008612842892768091, "rotation": {"subspace_principal_angle": 1.198927059140121, "mean_class_cosine": 0.6054966764735671}}}, "iid_acc": 0.95375, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.1388755195423386, "mean_class_cosine": 0.6482459135365446}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6084788029925187, "drop": 0.01402119700748139, "rotation": {"subspace_principal_angle": 1.493246923874091, "mean_class_cosine": 0.8677677280952006}}}, "iid_acc": 0.6225, "selectivity": 0.5375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5091321723919393, "mean_class_cosine": 0.8880006055054167}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.9351620947630923, "drop": 0.012337905236907698}}, "iid_acc": 0.9475, "selectivity": 0.8625} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.495, "drop": 0.0}, "paraphrase": {"acc": 0.4745098039215686, "drop": 0.02049019607843139, "rotation": {"subspace_principal_angle": 1.557032384565418, "mean_class_cosine": -0.013763507646052914}}}, "iid_acc": 0.495, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.5359063301369187, "mean_class_cosine": -0.03488291842098544}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.0017892156862744923, "rotation": {"subspace_principal_angle": 1.3709234192517532, "mean_class_cosine": 0.6267242100586605}}}, "iid_acc": 0.49625, "selectivity": -0.014999999999999958, "iid_split_rotation": {"subspace_principal_angle": 1.3896477426172604, "mean_class_cosine": -0.3304774205182823}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.4825, "drop": 0.0}, "paraphrase": {"acc": 0.47320261437908495, "drop": 0.009297385620915033}}, "iid_acc": 0.4825, "selectivity": -0.028749999999999998} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5805592543275633, "drop": 0.02319074567243673, "rotation": {"subspace_principal_angle": 1.3989325795834617, "mean_class_cosine": 0.365872980266819}}}, "iid_acc": 0.60375, "selectivity": 0.335, "iid_split_rotation": {"subspace_principal_angle": 1.4035880336880546, "mean_class_cosine": 0.3488005843140671}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.1775, "drop": 0.0}, "paraphrase": {"acc": 0.16511318242343542, "drop": 0.012386817576564568, "rotation": {"subspace_principal_angle": 1.5619855127249644, "mean_class_cosine": 0.3341308312243725}}}, "iid_acc": 0.1775, "selectivity": -0.09125, "iid_split_rotation": {"subspace_principal_angle": 1.3905000233838531, "mean_class_cosine": 0.6611663679873362}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5778961384820239, "drop": 0.022103861517976053}}, "iid_acc": 0.6, "selectivity": 0.33125} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.015927419354838768, "rotation": {"subspace_principal_angle": 1.5028161180968338, "mean_class_cosine": 0.06792786120606348}}}, "iid_acc": 0.69375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4722418499371575, "mean_class_cosine": 0.09839501129737169}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.533106960950764, "drop": -0.020606960950764086, "rotation": {"subspace_principal_angle": 1.024515247435546, "mean_class_cosine": 0.9955533421541893}}}, "iid_acc": 0.5125, "selectivity": -0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.2562415527805546, "mean_class_cosine": 0.9954953689417243}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7555178268251274, "drop": 0.00448217317487265}}, "iid_acc": 0.76, "selectivity": 0.24624999999999997} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.6338709677419355, "drop": 0.017379032258064475, "rotation": {"subspace_principal_angle": 1.4673214972757795, "mean_class_cosine": 0.10329027681844688}}}, "iid_acc": 0.65125, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.5077800700720831, "mean_class_cosine": 0.06297455823321907}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.535483870967742, "drop": 0.005766129032258038, "rotation": {"subspace_principal_angle": 1.440925356190957, "mean_class_cosine": 0.9961212841602167}}}, "iid_acc": 0.54125, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3284144435763523, "mean_class_cosine": 0.9886091312692893}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6209677419354839, "drop": 0.03403225806451615}}, "iid_acc": 0.655, "selectivity": 0.17000000000000004} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.750402576489533, "drop": -0.009152576489533049, "rotation": {"subspace_principal_angle": 1.4211901787085208, "mean_class_cosine": 0.14904869151211453}}}, "iid_acc": 0.74125, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.4965103140884382, "mean_class_cosine": 0.07421770808908856}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5265700483091788, "drop": 0.03967995169082128, "rotation": {"subspace_principal_angle": 1.5565645508048995, "mean_class_cosine": 0.00760410778674959}}}, "iid_acc": 0.56625, "selectivity": 0.0025000000000000577, "iid_split_rotation": {"subspace_principal_angle": 0.9594487263253761, "mean_class_cosine": -0.4124908154959152}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7536231884057971, "drop": -0.0011231884057971708}}, "iid_acc": 0.7525, "selectivity": 0.18874999999999997} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8998628257887518, "drop": 0.02138717421124825, "rotation": {"subspace_principal_angle": 1.2563517152731103, "mean_class_cosine": 0.30928836211834454}}}, "iid_acc": 0.92125, "selectivity": 0.43375, "iid_split_rotation": {"subspace_principal_angle": 1.1864007137524348, "mean_class_cosine": 0.3749989233038349}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.545, "drop": 0.0}, "paraphrase": {"acc": 0.5569272976680384, "drop": -0.01192729766803835, "rotation": {"subspace_principal_angle": 0.6735562930427222, "mean_class_cosine": 0.3885547331126563}}}, "iid_acc": 0.545, "selectivity": 0.05750000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.3659301793739206, "mean_class_cosine": 0.9497848325707625}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.9039780521262003, "drop": 0.009771947873799647}}, "iid_acc": 0.91375, "selectivity": 0.42624999999999996} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9965694682675815, "drop": -0.0015694682675815308, "rotation": {"subspace_principal_angle": 0.9204115550682747, "mean_class_cosine": 0.605492671467479}}}, "iid_acc": 0.995, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 0.964600661915144, "mean_class_cosine": 0.5697451063410308}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7753001715265866, "drop": -0.00780017152658663, "rotation": {"subspace_principal_angle": 1.5561305399812533, "mean_class_cosine": 0.9880733179203292}}}, "iid_acc": 0.7675, "selectivity": -0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.8575713187579257, "mean_class_cosine": 0.9958656438305629}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": 0.0006110634648370805}}, "iid_acc": 0.99375, "selectivity": 0.19125000000000003} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6478260869565218, "drop": 0.0096739130434782, "rotation": {"subspace_principal_angle": 1.5203803161832818, "mean_class_cosine": 0.050394655640553065}}}, "iid_acc": 0.6575, "selectivity": 0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.4388032016199581, "mean_class_cosine": 0.13161019079523278}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4623188405797101, "drop": 0.040181159420289825, "rotation": {"subspace_principal_angle": 1.5601182686089141, "mean_class_cosine": -0.9599175649315286}}}, "iid_acc": 0.5025, "selectivity": -0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3056048703674055, "mean_class_cosine": 0.9664617957303832}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.6956521739130435, "drop": 0.018097826086956537}}, "iid_acc": 0.71375, "selectivity": 0.15500000000000003} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8831908831908832, "drop": 0.0030591168091167464, "rotation": {"subspace_principal_angle": 1.4868944039444256, "mean_class_cosine": 0.08380351910501391}}}, "iid_acc": 0.88625, "selectivity": 0.20999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3010782735174253, "mean_class_cosine": 0.26645969404516406}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.575, "drop": 0.0}, "paraphrase": {"acc": 0.5797720797720798, "drop": -0.004772079772079829, "rotation": {"subspace_principal_angle": 1.1062280403413047, "mean_class_cosine": 0.9715489933239789}}}, "iid_acc": 0.575, "selectivity": -0.10125000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.5004862855814347, "mean_class_cosine": 0.9429161811360605}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.9017094017094017, "drop": -0.0017094017094017033}}, "iid_acc": 0.9, "selectivity": 0.22375} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8490304709141274, "drop": 0.024719529085872605, "rotation": {"subspace_principal_angle": 1.176281709220393, "mean_class_cosine": 0.384360137185048}}, "domain": {"acc": 0.73625, "drop": 0.13750000000000007, "rotation": {"subspace_principal_angle": 1.1970353549367978, "mean_class_cosine": 0.36511932310217704}}}, "iid_acc": 0.87375, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.0786643668776719, "mean_class_cosine": 0.4725059154778732}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5304709141274239, "drop": -0.004220914127423869, "rotation": {"subspace_principal_angle": 1.3152812819817528, "mean_class_cosine": 0.9992665145477913}}, "domain": {"acc": 0.50125, "drop": 0.025000000000000022, "rotation": {"subspace_principal_angle": 1.308682431264261, "mean_class_cosine": 0.5566033754658513}}}, "iid_acc": 0.52625, "selectivity": 0.04249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.391200206971303, "mean_class_cosine": 0.9964068683023172}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8130193905817175, "drop": 0.018230609418282584}, "domain": {"acc": 0.82875, "drop": 0.0025000000000000577}}, "iid_acc": 0.83125, "selectivity": 0.34750000000000003} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8643410852713178, "drop": 0.023158914728682145, "rotation": {"subspace_principal_angle": 1.1410222217173775, "mean_class_cosine": 0.4166654610472472}}, "domain": {"acc": 0.815, "drop": 0.07250000000000001, "rotation": {"subspace_principal_angle": 1.250009657687724, "mean_class_cosine": 0.3153131973834549}}, "length": {"acc": 0.8878787878787879, "drop": -0.00037878787878797837, "rotation": {"subspace_principal_angle": 1.0751426744599302, "mean_class_cosine": 0.4756067433505382}}}, "iid_acc": 0.8875, "selectivity": 0.42999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.080452581028767, "mean_class_cosine": 0.4709291585448011}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7364341085271318, "drop": -0.0014341085271317722, "rotation": {"subspace_principal_angle": 1.133371334962397, "mean_class_cosine": 0.8868509368956804}}, "domain": {"acc": 0.5575, "drop": 0.1775, "rotation": {"subspace_principal_angle": 1.4962557115172213, "mean_class_cosine": 0.27127763310435205}}, "length": {"acc": 0.7651515151515151, "drop": -0.03015151515151515, "rotation": {"subspace_principal_angle": 1.3646255328037025, "mean_class_cosine": 0.9602447183571129}}}, "iid_acc": 0.735, "selectivity": 0.27749999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.9276681297932009, "mean_class_cosine": 0.8062680171943362}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8798449612403101, "drop": 0.01765503875968988}, "domain": {"acc": 0.69625, "drop": 0.20124999999999993}, "length": {"acc": 0.9, "drop": -0.0025000000000000577}}, "iid_acc": 0.8975, "selectivity": 0.43999999999999995} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.004903846153846203, "rotation": {"subspace_principal_angle": 1.3890007649660276, "mean_class_cosine": 0.284822099506686}}}, "iid_acc": 0.87875, "selectivity": 0.60625, "iid_split_rotation": {"subspace_principal_angle": 1.5702965678160945, "mean_class_cosine": 0.3020514335534847}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.5661538461538461, "drop": -0.009903846153846096, "rotation": {"subspace_principal_angle": 1.553740809284586, "mean_class_cosine": 0.9736297672635466}}}, "iid_acc": 0.55625, "selectivity": 0.28375, "iid_split_rotation": {"subspace_principal_angle": 1.4574124850080719, "mean_class_cosine": 0.9616324023244914}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8861538461538462, "drop": -0.001153846153846172}}, "iid_acc": 0.885, "selectivity": 0.6125} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9476309226932669, "drop": 0.02361907730673307, "rotation": {"subspace_principal_angle": 1.0954764445356109, "mean_class_cosine": 0.7159555553963199}}}, "iid_acc": 0.97125, "selectivity": 0.8775, "iid_split_rotation": {"subspace_principal_angle": 1.0349871771004895, "mean_class_cosine": 0.757327529417333}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.26375, "drop": 0.0}, "paraphrase": {"acc": 0.29177057356608477, "drop": -0.028020573566084783, "rotation": {"subspace_principal_angle": 1.4115196982796947, "mean_class_cosine": 0.8384434499541353}}}, "iid_acc": 0.26375, "selectivity": 0.16999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2413128275371268, "mean_class_cosine": 0.7859751886800755}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.912718204488778, "drop": 0.03603179551122193}}, "iid_acc": 0.94875, "selectivity": 0.855} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.5882352941176471, "drop": 0.02926470588235297, "rotation": {"subspace_principal_angle": 1.360311035504444, "mean_class_cosine": 0.20893450516523}}}, "iid_acc": 0.6175, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4063133781244486, "mean_class_cosine": 0.1637422801353008}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.4925, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.005539215686274523, "rotation": {"subspace_principal_angle": 1.441946064731949, "mean_class_cosine": 0.6865636818700251}}}, "iid_acc": 0.4925, "selectivity": -0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5463324868354524, "mean_class_cosine": -0.9630196496317205}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5581699346405229, "drop": -0.009419934640522931}}, "iid_acc": 0.54875, "selectivity": 0.012499999999999956} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02183089214380829, "rotation": {"subspace_principal_angle": 1.25177621884401, "mean_class_cosine": 0.5310790447646092}}}, "iid_acc": 0.645, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2608266828416124, "mean_class_cosine": 0.45773127447897305}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1825, "drop": 0.0}, "paraphrase": {"acc": 0.17709720372836218, "drop": 0.005402796271637811, "rotation": {"subspace_principal_angle": 1.4219864569031055, "mean_class_cosine": 0.3343326799788812}}}, "iid_acc": 0.1825, "selectivity": -0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4213554573473413, "mean_class_cosine": 0.6639945990861887}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.521970705725699, "drop": 0.00427929427430096}}, "iid_acc": 0.52625, "selectivity": 0.27} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7707979626485568, "drop": -0.01704796264855679, "rotation": {"subspace_principal_angle": 1.3829256711520628, "mean_class_cosine": 0.18676744322680136}}}, "iid_acc": 0.75375, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3348303902625251, "mean_class_cosine": 0.23378226387175424}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.5212224108658744, "drop": -0.00497241086587441, "rotation": {"subspace_principal_angle": 1.478236251419889, "mean_class_cosine": 0.9994667861891016}}}, "iid_acc": 0.51625, "selectivity": -0.0050000000000000044, "iid_split_rotation": {"subspace_principal_angle": 1.3448819076659233, "mean_class_cosine": 0.9996533797319738}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6960950764006791, "drop": 0.008904923599320824}}, "iid_acc": 0.705, "selectivity": 0.18374999999999997} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67625, "drop": 0.0}, "paraphrase": {"acc": 0.632258064516129, "drop": 0.043991935483871036, "rotation": {"subspace_principal_angle": 1.4188129223223702, "mean_class_cosine": 0.15139897022933}}}, "iid_acc": 0.67625, "selectivity": 0.1875, "iid_split_rotation": {"subspace_principal_angle": 1.381661689857851, "mean_class_cosine": 0.18800903417603418}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5338709677419354, "drop": 0.002379032258064573, "rotation": {"subspace_principal_angle": 1.2619016165915666, "mean_class_cosine": 0.9985457827336248}}}, "iid_acc": 0.53625, "selectivity": 0.04749999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2970420520156083, "mean_class_cosine": 0.9948538875393387}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67125, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.035766129032258065}}, "iid_acc": 0.67125, "selectivity": 0.1825} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7681159420289855, "drop": -0.008115942028985468, "rotation": {"subspace_principal_angle": 1.310808643208806, "mean_class_cosine": 0.2570686494136881}}}, "iid_acc": 0.76, "selectivity": 0.15375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3820982126149526, "mean_class_cosine": 0.18758027789963122}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.47125, "drop": 0.0}, "paraphrase": {"acc": 0.49597423510466987, "drop": -0.024724235104669867, "rotation": {"subspace_principal_angle": 1.452719223490081, "mean_class_cosine": 0.2128800431912587}}}, "iid_acc": 0.47125, "selectivity": -0.13499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.9161871593522571, "mean_class_cosine": -0.32128070846271867}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6843800322061192, "drop": 0.010619967793880747}}, "iid_acc": 0.695, "selectivity": 0.08875} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.94, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.011330589849108286, "rotation": {"subspace_principal_angle": 1.137125759529054, "mean_class_cosine": 0.4202044066034193}}}, "iid_acc": 0.94, "selectivity": 0.47624999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.0643181542191005, "mean_class_cosine": 0.4851005702018669}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5253772290809328, "drop": 0.008372770919067185, "rotation": {"subspace_principal_angle": 0.516621762104431, "mean_class_cosine": 0.21743596239076726}}}, "iid_acc": 0.53375, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.8840737623848459, "mean_class_cosine": 0.976674523046531}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9368998628257887, "drop": 0.0006001371742112571}}, "iid_acc": 0.9375, "selectivity": 0.47375} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.6634680250005783, "mean_class_cosine": 0.7878611804971933}}}, "iid_acc": 0.99125, "selectivity": 0.11874999999999991, "iid_split_rotation": {"subspace_principal_angle": 0.6348168891373653, "mean_class_cosine": 0.8051803002223438}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9777015437392796, "drop": -0.006451543739279697, "rotation": {"subspace_principal_angle": 1.3391381172924783, "mean_class_cosine": 0.9776428455730666}}}, "iid_acc": 0.97125, "selectivity": 0.0987499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4227835246448142, "mean_class_cosine": 0.9908198389945848}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": 0.0002915951972555231}}, "iid_acc": 0.99, "selectivity": 0.11749999999999994} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7, "drop": 0.02750000000000008, "rotation": {"subspace_principal_angle": 1.472522322965299, "mean_class_cosine": 0.09811589541574661}}}, "iid_acc": 0.7275, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.3920524003697496, "mean_class_cosine": 0.17779365245715506}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.03989130434782606, "rotation": {"subspace_principal_angle": 1.4319943336187273, "mean_class_cosine": -0.9921843120457909}}}, "iid_acc": 0.5225, "selectivity": -0.08374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.0962721333543264, "mean_class_cosine": 0.9947347244035363}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.717391304347826, "drop": -0.0036413043478260576}}, "iid_acc": 0.71375, "selectivity": 0.10750000000000004} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.8888888888888888, "drop": 0.0011111111111111738, "rotation": {"subspace_principal_angle": 1.346328125428775, "mean_class_cosine": 0.22258793643339736}}}, "iid_acc": 0.89, "selectivity": 0.135, "iid_split_rotation": {"subspace_principal_angle": 1.2354156575580886, "mean_class_cosine": 0.3291286539962054}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5598290598290598, "drop": 0.007670940170940166, "rotation": {"subspace_principal_angle": 1.089834172363605, "mean_class_cosine": 0.9822532101639228}}}, "iid_acc": 0.5675, "selectivity": -0.1875, "iid_split_rotation": {"subspace_principal_angle": 0.9991972798605081, "mean_class_cosine": 0.8996231510474297}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8532763532763533, "drop": 0.00672364672364667}}, "iid_acc": 0.86, "selectivity": 0.10499999999999998} diff --git a/repro_bundle/results_full_run/results/predictors.jsonl b/repro_bundle/results_full_run/results/predictors.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..e098f806e67509d0cdcfe5b2712104586243f1ef --- /dev/null +++ b/repro_bundle/results_full_run/results/predictors.jsonl @@ -0,0 +1,78 @@ +{"model": "pythia-70m", "dataset": "sst2", "seed": 0, "layer": 1, "concept": "sentiment", "predictors": {"sip_eigengap": 1.0825317547292495, "raptor_stability": 0.837236762046814, "fragility": 1.0, "augmentation_robustness": 0.7404108718939687, "xie_feature_dispersion": 1.8292699253554, "pac": 0.7888238169703914, "whitened_cosine_id": 0.8086112141609192}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 0, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998517, "raptor_stability": 0.9221990704536438, "fragility": 1.0, "augmentation_robustness": 0.9364739773993105, "xie_feature_dispersion": 1.0993059236404399, "pac": 0.9293365239264771, "whitened_cosine_id": 0.8288846611976624}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 0, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.452105016045313, "raptor_stability": 0.8456851840019226, "fragility": 0.25, "augmentation_robustness": 0.8514520350954915, "xie_feature_dispersion": 0.6587695281308451, "pac": 0.848568609548707, "whitened_cosine_id": 0.7871170043945312}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 0, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14189668585488155, "raptor_stability": 0.9446240663528442, "fragility": 1.5, "augmentation_robustness": 0.8839691673684479, "xie_feature_dispersion": 5.2630010355712145, "pac": 0.9142966168606461, "whitened_cosine_id": 0.9304912686347961}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 0, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974287, "raptor_stability": 0.7597023248672485, "fragility": 1.0, "augmentation_robustness": 0.7320095613414648, "xie_feature_dispersion": 0.4779918901319739, "pac": 0.7458559431043567, "whitened_cosine_id": 0.6635854244232178}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 0, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5953587445214507, "raptor_stability": 0.869035542011261, "fragility": 0.5, "augmentation_robustness": 0.8065598522498184, "xie_feature_dispersion": 1.5333196894740788, "pac": 0.8377976971305396, "whitened_cosine_id": 0.8452926278114319}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 0, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994247, "raptor_stability": 0.8179649114608765, "fragility": 0.5, "augmentation_robustness": 0.7221715232065848, "xie_feature_dispersion": 1.779944270255231, "pac": 0.7700682173337307, "whitened_cosine_id": 0.815054714679718}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 0, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.711632992199055, "raptor_stability": 0.8173510432243347, "fragility": 0.25, "augmentation_robustness": 0.676819096553469, "xie_feature_dispersion": 2.512801472259043, "pac": 0.7470850698889018, "whitened_cosine_id": 0.766620934009552}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989692, "raptor_stability": 0.8470741510391235, "fragility": 1.0, "augmentation_robustness": 0.8442697431706384, "xie_feature_dispersion": 1.163949469847951, "pac": 0.845671947104881, "whitened_cosine_id": 0.7620958089828491}} +{"model": "pythia-70m", "dataset": "subj", "seed": 0, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002171, "raptor_stability": 0.9096688032150269, "fragility": 1.5, "augmentation_robustness": 0.8646995541446441, "xie_feature_dispersion": 2.3756762827181857, "pac": 0.8871841786798355, "whitened_cosine_id": 0.9003387093544006}} +{"model": "pythia-70m", "dataset": "spam", "seed": 0, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004674, "raptor_stability": 0.9239181280136108, "fragility": 1.5, "augmentation_robustness": 0.884000762861815, "xie_feature_dispersion": 10.408354824969607, "pac": 0.9039594454377129, "whitened_cosine_id": 0.9501436352729797}} +{"model": "pythia-70m", "dataset": "cola", "seed": 0, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921980065, "raptor_stability": 0.7801818251609802, "fragility": 1.0, "augmentation_robustness": 0.751228525306698, "xie_feature_dispersion": 0.5820240387291664, "pac": 0.7657051752338391, "whitened_cosine_id": 0.7021960020065308}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 0, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998, "raptor_stability": 0.8848187327384949, "fragility": 1.5, "augmentation_robustness": 0.8695133740159424, "xie_feature_dispersion": 1.822712974246315, "pac": 0.8771660533772186, "whitened_cosine_id": 0.8629143834114075}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350087, "raptor_stability": 0.8475507497787476, "fragility": 0.25, "augmentation_robustness": 0.6188354080698831, "xie_feature_dispersion": 1.4101369914642357, "pac": 0.7331930789243153, "whitened_cosine_id": 0.8635326027870178}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 0, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350252, "raptor_stability": 0.878171443939209, "fragility": 1.0, "augmentation_robustness": 0.8724963955930902, "xie_feature_dispersion": 1.3391749548314755, "pac": 0.8753339197661496, "whitened_cosine_id": 0.8675085306167603}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3953680487548481, "raptor_stability": 0.8400441408157349, "fragility": 0.25, "augmentation_robustness": 0.8483093395593638, "xie_feature_dispersion": 0.9318287616318928, "pac": 0.8441767401875493, "whitened_cosine_id": 0.8931494355201721}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 0, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10457984207175601, "raptor_stability": 0.9445252418518066, "fragility": 1.5, "augmentation_robustness": 0.8857764499429269, "xie_feature_dispersion": 5.199477497190765, "pac": 0.9151508458973667, "whitened_cosine_id": 0.9166393876075745}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 0, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342227, "raptor_stability": 0.765457034111023, "fragility": 0.25, "augmentation_robustness": 0.6816650768598687, "xie_feature_dispersion": 0.25779876141354974, "pac": 0.7235610554854458, "whitened_cosine_id": 0.7145544290542603}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.574379658862361, "raptor_stability": 0.8579342365264893, "fragility": 0.25, "augmentation_robustness": 0.7682780634518579, "xie_feature_dispersion": 1.6192479136149953, "pac": 0.8131061499891736, "whitened_cosine_id": 0.847996711730957}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348304, "raptor_stability": 0.8491518497467041, "fragility": 1.0, "augmentation_robustness": 0.7851229050246351, "xie_feature_dispersion": 2.2631913339622987, "pac": 0.8171373773856696, "whitened_cosine_id": 0.8206788301467896}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 0, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934627, "raptor_stability": 0.8196693062782288, "fragility": 0.5, "augmentation_robustness": 0.7136331067294028, "xie_feature_dispersion": 1.9587240977086187, "pac": 0.7666512065038158, "whitened_cosine_id": 0.7840940952301025}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934696, "raptor_stability": 0.8339419960975647, "fragility": 1.0, "augmentation_robustness": 0.7890852919390294, "xie_feature_dispersion": 1.5342572755652866, "pac": 0.8115136440182971, "whitened_cosine_id": 0.798081636428833}} +{"model": "pythia-160m", "dataset": "subj", "seed": 0, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351844, "raptor_stability": 0.86474609375, "fragility": 1.0, "augmentation_robustness": 0.7746642653183333, "xie_feature_dispersion": 1.77134285145721, "pac": 0.8197051795341667, "whitened_cosine_id": 0.902005136013031}} +{"model": "pythia-160m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352976, "raptor_stability": 0.9428686499595642, "fragility": 3.0, "augmentation_robustness": 0.9172945300081299, "xie_feature_dispersion": 6.963799146343127, "pac": 0.930081589983847, "whitened_cosine_id": 0.9462446570396423}} +{"model": "pythia-160m", "dataset": "cola", "seed": 0, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344465, "raptor_stability": 0.7942293882369995, "fragility": 0.25, "augmentation_robustness": 0.5972243911972416, "xie_feature_dispersion": 0.7758972276639047, "pac": 0.6957268897171205, "whitened_cosine_id": 0.7800706624984741}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 0, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935011, "raptor_stability": 0.8662534952163696, "fragility": 0.5, "augmentation_robustness": 0.7943799335400991, "xie_feature_dispersion": 1.5213101554825748, "pac": 0.8303167143782344, "whitened_cosine_id": 0.8762480020523071}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 0, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 0.9882117688015961, "raptor_stability": 0.857214629650116, "fragility": 0.25, "augmentation_robustness": NaN, "augmentation_robustness__error": "Found input variables with inconsistent numbers of samples: [600, 1000]", "xie_feature_dispersion": 1.219591183270763, "pac": NaN, "pac__error": "Found input variables with inconsistent numbers of samples: [600, 1000]", "whitened_cosine_id": 0.8411453366279602}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881813, "raptor_stability": 0.8615289330482483, "fragility": 1.0, "augmentation_robustness": 0.8437283807712266, "xie_feature_dispersion": 1.923545301649778, "pac": 0.8526286569097374, "whitened_cosine_id": 0.8741693496704102}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 0, "layer": 14, "concept": "topic", "predictors": {"sip_eigengap": 0.26611413467423856, "raptor_stability": 0.8730260133743286, "fragility": 1.0, "augmentation_robustness": 0.9009292021779557, "xie_feature_dispersion": 3.6796556574761454, "pac": 0.8869776077761422, "whitened_cosine_id": 0.8328061699867249}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.1116614706117857, "raptor_stability": 0.9447072744369507, "fragility": 2.0, "augmentation_robustness": 0.8856969264519378, "xie_feature_dispersion": 7.687914979274747, "pac": 0.9152021004444442, "whitened_cosine_id": 0.8943790197372437}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 0, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879029, "raptor_stability": 0.7828097343444824, "fragility": 0.25, "augmentation_robustness": 0.7337481325439938, "xie_feature_dispersion": 0.12206039095754939, "pac": 0.7582789334442381, "whitened_cosine_id": 0.7742199897766113}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 0, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.4882081459032719, "raptor_stability": 0.8603615164756775, "fragility": 0.25, "augmentation_robustness": 0.7884647975902315, "xie_feature_dispersion": 2.173352441480384, "pac": 0.8244131570329545, "whitened_cosine_id": 0.8452408909797668}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 0, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880458, "raptor_stability": 0.8067511320114136, "fragility": 0.25, "augmentation_robustness": 0.7034424602081344, "xie_feature_dispersion": 4.485569514407045, "pac": 0.755096796109774, "whitened_cosine_id": 0.8220662474632263}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 0, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879928, "raptor_stability": 0.7928941249847412, "fragility": 0.25, "augmentation_robustness": 0.6753132678040145, "xie_feature_dispersion": 2.5329861784126293, "pac": 0.7341036963943779, "whitened_cosine_id": 0.804050624370575}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 0, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880347, "raptor_stability": 0.8111140727996826, "fragility": 0.25, "augmentation_robustness": 0.6923253341445008, "xie_feature_dispersion": 0.7524282237703519, "pac": 0.7517197034720917, "whitened_cosine_id": 0.8138276934623718}} +{"model": "pythia-410m", "dataset": "subj", "seed": 0, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882566, "raptor_stability": 0.8834404945373535, "fragility": 0.5, "augmentation_robustness": 0.7988711466010014, "xie_feature_dispersion": 1.428972872571398, "pac": 0.8411558205691775, "whitened_cosine_id": 0.8940153121948242}} +{"model": "pythia-410m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882899, "raptor_stability": 0.9429098963737488, "fragility": 3.0, "augmentation_robustness": 0.9052157104985206, "xie_feature_dispersion": 6.72893787243834, "pac": 0.9240628034361347, "whitened_cosine_id": 0.9344553351402283}} +{"model": "pythia-410m", "dataset": "cola", "seed": 0, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879473, "raptor_stability": 0.7795286178588867, "fragility": 0.25, "augmentation_robustness": 0.5039294079390225, "xie_feature_dispersion": 1.1149557488538653, "pac": 0.6417290128989546, "whitened_cosine_id": 0.7961053848266602}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 0, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881513, "raptor_stability": 0.8518741130828857, "fragility": 0.5, "augmentation_robustness": 0.7625019501536046, "xie_feature_dispersion": 1.5630966219328584, "pac": 0.8071880316182451, "whitened_cosine_id": 0.8650529980659485}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.8530784249305725, "fragility": 0.25, "augmentation_robustness": 0.5848209179858852, "xie_feature_dispersion": 2.122662038410848, "pac": 0.7189496714582289, "whitened_cosine_id": 0.8195644617080688}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8263017535209656, "fragility": 1.0, "augmentation_robustness": 0.7804717428621477, "xie_feature_dispersion": 2.576726583434504, "pac": 0.8033867481915566, "whitened_cosine_id": 0.8429552912712097}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 0, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22041139949027433, "raptor_stability": 0.8416460156440735, "fragility": 1.0, "augmentation_robustness": 0.8650683659984831, "xie_feature_dispersion": 5.917035974988243, "pac": 0.8533571908212783, "whitened_cosine_id": 0.8176273703575134}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06228035557873579, "raptor_stability": 0.939380943775177, "fragility": 3.0, "augmentation_robustness": 0.8779983715906402, "xie_feature_dispersion": 9.493553367505593, "pac": 0.9086896576829087, "whitened_cosine_id": 0.8264285922050476}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 0, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.7794453501701355, "fragility": 0.25, "augmentation_robustness": 0.7408315387590054, "xie_feature_dispersion": 0.2653617366201949, "pac": 0.7601384444645705, "whitened_cosine_id": 0.8138558268547058}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.019767617442798265, "raptor_stability": 0.8648917078971863, "fragility": 0.25, "augmentation_robustness": 0.755285317721517, "xie_feature_dispersion": 2.8877683311134636, "pac": 0.8100885128093516, "whitened_cosine_id": 0.8282346129417419}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 0, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010617, "raptor_stability": 0.8094132542610168, "fragility": 0.25, "augmentation_robustness": 0.6718229230312025, "xie_feature_dispersion": 5.342310320027777, "pac": 0.7406180886461097, "whitened_cosine_id": 0.8205950260162354}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 0, "layer": 22, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.7882674336433411, "fragility": 0.25, "augmentation_robustness": 0.6470101589682584, "xie_feature_dispersion": 2.217429056075072, "pac": 0.7176387963057997, "whitened_cosine_id": 0.8153953552246094}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 0, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8043517470359802, "fragility": 0.5, "augmentation_robustness": 0.7119613985713426, "xie_feature_dispersion": 1.5482508686542642, "pac": 0.7581565728036614, "whitened_cosine_id": 0.8166601061820984}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 0, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010872, "raptor_stability": 0.8618161082267761, "fragility": 1.0, "augmentation_robustness": 0.7720944652439713, "xie_feature_dispersion": 2.5476390335776844, "pac": 0.8169552867353738, "whitened_cosine_id": 0.829445481300354}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 0, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010434, "raptor_stability": 0.9222065806388855, "fragility": 1.0, "augmentation_robustness": 0.895838461689926, "xie_feature_dispersion": 23.432722674772293, "pac": 0.9090225211644057, "whitened_cosine_id": 0.8605679869651794}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 0, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010818, "raptor_stability": 0.8025708794593811, "fragility": 0.25, "augmentation_robustness": 0.5249275775049415, "xie_feature_dispersion": 1.549092757263006, "pac": 0.6637492284821613, "whitened_cosine_id": 0.8153374195098877}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8472058773040771, "fragility": 0.5, "augmentation_robustness": 0.7513937676225657, "xie_feature_dispersion": 2.045523026748139, "pac": 0.7992998224633214, "whitened_cosine_id": 0.8325246572494507}} +{"model": "gpt2", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349679, "raptor_stability": 0.8157426118850708, "fragility": 0.25, "augmentation_robustness": 0.5123065291062753, "xie_feature_dispersion": 1.6815976407755642, "pac": 0.6640245704956731, "whitened_cosine_id": 0.855627715587616}} +{"model": "gpt2", "dataset": "imdb", "seed": 0, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350731, "raptor_stability": 0.7823665738105774, "fragility": 0.25, "augmentation_robustness": 0.6517344633023527, "xie_feature_dispersion": 1.6579993440971876, "pac": 0.7170505185564651, "whitened_cosine_id": 0.8720094561576843}} +{"model": "gpt2", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.34635368740834577, "raptor_stability": 0.8811239004135132, "fragility": 0.25, "augmentation_robustness": 0.9187097662639041, "xie_feature_dispersion": 0.8678892394256513, "pac": 0.8999168333387086, "whitened_cosine_id": 0.9088287949562073}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 0, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.12115672881499416, "raptor_stability": 0.9285115003585815, "fragility": 1.5, "augmentation_robustness": 0.8321901130650252, "xie_feature_dispersion": 6.121815304004312, "pac": 0.8803508067118033, "whitened_cosine_id": 0.9021502733230591}} +{"model": "gpt2", "dataset": "counterfact", "seed": 0, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859340033, "raptor_stability": 0.7618192434310913, "fragility": 0.25, "augmentation_robustness": 0.809669912657523, "xie_feature_dispersion": 0.1095429604074235, "pac": 0.7857445780443071, "whitened_cosine_id": 0.6977522969245911}} +{"model": "gpt2", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.7366388390202255, "raptor_stability": 0.8703832626342773, "fragility": 0.25, "augmentation_robustness": 0.8145765018284364, "xie_feature_dispersion": 1.77009486868241, "pac": 0.8424798822313568, "whitened_cosine_id": 0.8424334526062012}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347128, "raptor_stability": 0.759522557258606, "fragility": 0.25, "augmentation_robustness": 0.551602767544549, "xie_feature_dispersion": 3.694561283935865, "pac": 0.6555626624015775, "whitened_cosine_id": 0.812667965888977}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 0, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346073, "raptor_stability": 0.7821463346481323, "fragility": 0.25, "augmentation_robustness": 0.6861444856119838, "xie_feature_dispersion": 2.309585917543921, "pac": 0.7341454101300581, "whitened_cosine_id": 0.8025559782981873}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 0, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346306, "raptor_stability": 0.7760246396064758, "fragility": 0.25, "augmentation_robustness": 0.657242871403201, "xie_feature_dispersion": 0.25866709333672444, "pac": 0.7166337555048384, "whitened_cosine_id": 0.7979826331138611}} +{"model": "gpt2", "dataset": "subj", "seed": 0, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352106, "raptor_stability": 0.8525124788284302, "fragility": 0.25, "augmentation_robustness": 0.7295931094481585, "xie_feature_dispersion": 0.951143558340801, "pac": 0.7910527941382943, "whitened_cosine_id": 0.9076346755027771}} +{"model": "gpt2", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593527, "raptor_stability": 0.9162791967391968, "fragility": 2.0, "augmentation_robustness": 0.9103539019844458, "xie_feature_dispersion": 6.497801646724056, "pac": 0.9133165493618213, "whitened_cosine_id": 0.9396259784698486}} +{"model": "gpt2", "dataset": "cola", "seed": 0, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859346528, "raptor_stability": 0.784694492816925, "fragility": 0.25, "augmentation_robustness": 0.49778314433790527, "xie_feature_dispersion": 0.9649847502240455, "pac": 0.6412388185774152, "whitened_cosine_id": 0.8002498149871826}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935004, "raptor_stability": 0.8261088728904724, "fragility": 0.25, "augmentation_robustness": 0.6972854536891078, "xie_feature_dispersion": 1.1399162300669312, "pac": 0.7616971632897901, "whitened_cosine_id": 0.8749386668205261}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648414, "raptor_stability": 0.8763455152511597, "fragility": 0.25, "augmentation_robustness": 0.67470015678154, "xie_feature_dispersion": 1.8204152059660144, "pac": 0.7755228360163499, "whitened_cosine_id": 0.877658486366272}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 0, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.29387292376502, "raptor_stability": 0.877450704574585, "fragility": 1.0, "augmentation_robustness": 0.8720676344266011, "xie_feature_dispersion": 2.218615090585379, "pac": 0.874759169500593, "whitened_cosine_id": 0.8857330083847046}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.33450754332538074, "raptor_stability": 0.8494884967803955, "fragility": 1.0, "augmentation_robustness": 0.8744251513008834, "xie_feature_dispersion": 3.7561628051141023, "pac": 0.8619568240406394, "whitened_cosine_id": 0.9021056890487671}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07399960414320991, "raptor_stability": 0.9478516578674316, "fragility": 1.5, "augmentation_robustness": 0.891918719098521, "xie_feature_dispersion": 7.327642739764713, "pac": 0.9198851884829764, "whitened_cosine_id": 0.9197250604629517}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 0, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237645932, "raptor_stability": 0.8008507490158081, "fragility": 0.25, "augmentation_robustness": 0.8373520059306747, "xie_feature_dispersion": 0.06081582657176685, "pac": 0.8191013774732414, "whitened_cosine_id": 0.7541150450706482}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6131208141710327, "raptor_stability": 0.8854432106018066, "fragility": 0.25, "augmentation_robustness": 0.8904501661922035, "xie_feature_dispersion": 1.992273238599221, "pac": 0.8879466883970051, "whitened_cosine_id": 0.8487815260887146}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645996, "raptor_stability": 0.8148406744003296, "fragility": 0.25, "augmentation_robustness": 0.7607385932812474, "xie_feature_dispersion": 4.466765724760196, "pac": 0.7877896338407885, "whitened_cosine_id": 0.8100382685661316}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 0, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647149, "raptor_stability": 0.8064932823181152, "fragility": 0.25, "augmentation_robustness": 0.7276199921424138, "xie_feature_dispersion": 2.8843177774025395, "pac": 0.7670566372302645, "whitened_cosine_id": 0.7997785806655884}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 0, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647015, "raptor_stability": 0.8129029273986816, "fragility": 0.25, "augmentation_robustness": 0.7939897357140582, "xie_feature_dispersion": 0.15381271196675575, "pac": 0.8034463315563699, "whitened_cosine_id": 0.799782931804657}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 0, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651123, "raptor_stability": 0.8869607448577881, "fragility": 0.5, "augmentation_robustness": 0.8243175866171568, "xie_feature_dispersion": 0.9955751494789675, "pac": 0.8556391657374725, "whitened_cosine_id": 0.912450909614563}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 0, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923765166, "raptor_stability": 0.9521724581718445, "fragility": 3.0, "augmentation_robustness": 0.94099314226979, "xie_feature_dispersion": 9.854132791030557, "pac": 0.9465828002208172, "whitened_cosine_id": 0.9525545835494995}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 0, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764728, "raptor_stability": 0.7995859384536743, "fragility": 0.25, "augmentation_robustness": 0.5589292088936479, "xie_feature_dispersion": 1.0666210514085848, "pac": 0.6792575736736611, "whitened_cosine_id": 0.8146979808807373}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 0, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649436, "raptor_stability": 0.8498495817184448, "fragility": 0.25, "augmentation_robustness": 0.7821972078583704, "xie_feature_dispersion": 1.1742203784808831, "pac": 0.8160233947884077, "whitened_cosine_id": 0.8703920841217041}} diff --git a/repro_bundle/results_full_run/results/stats_ranking.jsonl b/repro_bundle/results_full_run/results/stats_ranking.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..18e31b3f982a51f51ae9e2ccd28629901d6ed6a4 --- /dev/null +++ b/repro_bundle/results_full_run/results/stats_ranking.jsonl @@ -0,0 +1,4 @@ +{"target": "acc_drop", "n_configs": 78, "table": [{"predictor": "augmentation_robustness", "spearman": 0.15157310829700824, "p": 0.18820379891952224, "kendall": 0.10152110898573491, "ci95": [-0.07544146172027451, 0.35515914059827824], "loco_mean": 0.10282411873306434, "n": 77}, {"predictor": "sip_eigengap", "spearman": 0.11885733073674094, "p": 0.29999170761632016, "kendall": 0.08161225849433709, "ci95": [-0.10849988929815374, 0.3351540095893337], "loco_mean": 0.35275792755738855, "n": 78}, {"predictor": "fragility", "spearman": 0.09330610593561703, "p": 0.4164890423188147, "kendall": 0.07234738375624572, "ci95": [-0.13327134005431285, 0.2974848834552163], "loco_mean": -0.02716885733329969, "n": 78}, {"predictor": "pac", "spearman": 0.08692680298049739, "p": 0.4522147170892051, "kendall": 0.06050247909250868, "ci95": [-0.14081012171909665, 0.3011441571305958], "loco_mean": 0.06441388526936352, "n": 77}, {"predictor": "xie_feature_dispersion", "spearman": -0.010786749875339702, "p": 0.9253231455510073, "kendall": -0.01265612007557011, "ci95": [-0.22842499654534829, 0.2082521348228227], "loco_mean": -0.2549160517074272, "n": 78}, {"predictor": "raptor_stability", "spearman": -0.0720676290030023, "p": 0.5306434477662242, "kendall": -0.04796003397058147, "ci95": [-0.2796759142605101, 0.1513610831777458], "loco_mean": -0.04396847652201657, "n": 78}, {"predictor": "whitened_cosine_id", "spearman": -0.1250276623886092, "p": 0.2754167430795452, "kendall": -0.08059950153389386, "ci95": [-0.34331034561421037, 0.10470724958566129], "loco_mean": -0.24843992150147443, "n": 78}]} +{"target": "rotation", "n_configs": 78, "table": [{"predictor": "raptor_stability", "spearman": 0.9506063556696468, "p": 2.330553596159084e-40, "kendall": 0.8215118215118216, "ci95": [0.9089046238709781, 0.9700975672695751], "loco_mean": 0.8580510398692218, "n": 78}, {"predictor": "pac", "spearman": 0.8953152111046847, "p": 4.658961543202584e-28, "kendall": 0.732740943267259, "ci95": [0.8179189833991094, 0.9414542969755223], "loco_mean": 0.7593588229951869, "n": 77}, {"predictor": "augmentation_robustness", "spearman": 0.8108733371891266, "p": 3.9824372883504834e-19, "kendall": 0.6288448393711551, "ci95": [0.6977449617081606, 0.8819133930707614], "loco_mean": 0.6829897375351921, "n": 77}, {"predictor": "whitened_cosine_id", "spearman": 0.741423133828197, "p": 8.286280338933175e-15, "kendall": 0.5624375624375625, "ci95": [0.6064569266539844, 0.8460456408461438], "loco_mean": 0.23948778494233036, "n": 78}, {"predictor": "fragility", "spearman": 0.684812375739998, "p": 4.710024811739493e-12, "kendall": 0.5658230797653422, "ci95": [0.5068536867004124, 0.8106460075587721], "loco_mean": 0.4087392628838959, "n": 78}, {"predictor": "xie_feature_dispersion", "spearman": 0.5015111470807673, "p": 2.9077681366557542e-06, "kendall": 0.3493173493173493, "ci95": [0.2850284364599902, 0.6782866866024942], "loco_mean": 0.032694578149123604, "n": 78}, {"predictor": "sip_eigengap", "spearman": -0.15063512838809676, "p": 0.18803308276210945, "kendall": -0.09991673743871139, "ci95": [-0.39004985654054014, 0.11212512209734807], "loco_mean": 0.24411951684678962, "n": 78}]} +{"target": "iid_split_rotation", "n_configs": 78, "table": [{"predictor": "raptor_stability", "spearman": 0.9754675704042792, "p": 1.0380217149155117e-51, "kendall": 0.8654678654678655, "ci95": [0.9514150344464546, 0.9837566498330318], "loco_mean": 0.8891108891108892, "n": 78}, {"predictor": "pac", "spearman": 0.8587465166412533, "p": 1.7694398856482523e-23, "kendall": 0.6746411483253587, "ci95": [0.763925704827811, 0.9143156610296186], "loco_mean": 0.7037235491780947, "n": 77}, {"predictor": "whitened_cosine_id", "spearman": 0.7901592078807268, "p": 7.957935903755063e-18, "kendall": 0.5944055944055944, "ci95": [0.6822898987988187, 0.8627434021954578], "loco_mean": 0.1695213877032059, "n": 78}, {"predictor": "augmentation_robustness", "spearman": 0.7517482517482518, "p": 3.321238175252685e-15, "kendall": 0.5584415584415584, "ci95": [0.6223030616915662, 0.840503620514955], "loco_mean": 0.605558078285351, "n": 77}, {"predictor": "fragility", "spearman": 0.6846762030663571, "p": 4.7741898605683166e-12, "kendall": 0.5569820941440088, "ci95": [0.5096799996531299, 0.8060224031839003], "loco_mean": 0.3636334217788528, "n": 78}, {"predictor": "xie_feature_dispersion", "spearman": 0.49149584592622564, "p": 4.890837262117578e-06, "kendall": 0.3413253413253413, "ci95": [0.27232520406676164, 0.6682673406588026], "loco_mean": 0.015620742893470173, "n": 78}, {"predictor": "sip_eigengap", "spearman": -0.09199719266482573, "p": 0.42308521702002105, "kendall": -0.0546211497998289, "ci95": [-0.32381688992618785, 0.1644384857413355], "loco_mean": 0.21603850694759788, "n": 78}]} +{"target": "excess_rotation", "n_configs": 78, "table": [{"predictor": "augmentation_robustness", "spearman": 0.22488038277511957, "p": 0.049263364388599594, "kendall": 0.14490772385509226, "ci95": [0.0005950821127730556, 0.4127188540008681], "loco_mean": 0.08501498501498503, "n": 77}, {"predictor": "pac", "spearman": 0.12448078237551921, "p": 0.2807437305277193, "kendall": 0.07245386192754613, "ci95": [-0.0940283271892315, 0.3303878104170904], "loco_mean": -0.02678230860049041, "n": 77}, {"predictor": "fragility", "spearman": -0.015019845902589337, "p": 0.8961571306606886, "kendall": -0.008840985621333472, "ci95": [-0.2376863906837564, 0.20038844877008058], "loco_mean": -0.03976781626476142, "n": 78}, {"predictor": "xie_feature_dispersion", "spearman": -0.052947052947052944, "p": 0.645236909077995, "kendall": -0.02763902763902764, "ci95": [-0.2773630557877433, 0.18098235775774268], "loco_mean": -0.012732721823630904, "n": 78}, {"predictor": "raptor_stability", "spearman": -0.07487449259601157, "p": 0.5147155544076958, "kendall": -0.053613053613053616, "ci95": [-0.2771725780628797, 0.1507945447829959], "loco_mean": -0.22889837435291985, "n": 78}, {"predictor": "sip_eigengap", "spearman": -0.10064682562465264, "p": 0.38061338065690825, "kendall": -0.05528726138275364, "ci95": [-0.3074424291704091, 0.10337141574594425], "loco_mean": 0.048587775860503134, "n": 78}, {"predictor": "whitened_cosine_id", "spearman": -0.18640852818068004, "p": 0.10223932124369983, "kendall": -0.11022311022311024, "ci95": [-0.3981586286334883, 0.03776191770410821], "loco_mean": -0.13517391699209885, "n": 78}]} diff --git a/repro_bundle/results_tier2/results/audit.jsonl b/repro_bundle/results_tier2/results/audit.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..99c8fe80782a55fd34925cf2bb2ef75f3e39f080 --- /dev/null +++ b/repro_bundle/results_tier2/results/audit.jsonl @@ -0,0 +1,155 @@ +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": 0.15384615384615385, "n": 208} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 1, "flip_rate": 0.10124999999999995, "n": 800} +{"dataset": "imdb", "seed": 1, "flip_rate": 0.38749999999999996, "n": 800} +{"dataset": "ag_news", "seed": 1, "flip_rate": 0.20750000000000002, "n": 800} +{"dataset": "dbpedia", "seed": 1, "flip_rate": 0.5137499999999999, "n": 800} +{"dataset": "counterfact", "seed": 1, "flip_rate": 0.03249999999999997, "n": 800} +{"dataset": "emotion", "seed": 1, "flip_rate": 0.07625000000000004, "n": 800} +{"dataset": "tweet_hate", "seed": 1, "flip_rate": 0.26, "n": 800} +{"dataset": "tweet_irony", "seed": 1, "flip_rate": 0.23124999999999996, "n": 800} +{"dataset": "tweet_offensive", "seed": 1, "flip_rate": 0.23375, "n": 800} +{"dataset": "subj", "seed": 1, "flip_rate": 0.08250000000000002, "n": 800} +{"dataset": "spam", "seed": 1, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "cola", "seed": 1, "flip_rate": 0.13624999999999998, "n": 800} +{"dataset": "stance", "seed": 1, "flip_rate": 0.1826923076923077, "n": 208} +{"dataset": "amazon_cf", "seed": 1, "flip_rate": 0.11375000000000002, "n": 800} +{"dataset": "sst2", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "imdb", "seed": 2, "flip_rate": 0.35624999999999996, "n": 800} +{"dataset": "ag_news", "seed": 2, "flip_rate": 0.20625000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 2, "flip_rate": 0.5037499999999999, "n": 800} +{"dataset": "counterfact", "seed": 2, "flip_rate": 0.04125000000000001, "n": 800} +{"dataset": "emotion", "seed": 2, "flip_rate": 0.05874999999999997, "n": 800} +{"dataset": "tweet_hate", "seed": 2, "flip_rate": 0.22750000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 2, "flip_rate": 0.22624999999999995, "n": 800} +{"dataset": "tweet_offensive", "seed": 2, "flip_rate": 0.23875000000000002, "n": 800} +{"dataset": "subj", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "spam", "seed": 2, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 2, "flip_rate": 0.10875000000000001, "n": 800} +{"dataset": "stance", "seed": 2, "flip_rate": 0.14903846153846156, "n": 208} +{"dataset": "amazon_cf", "seed": 2, "flip_rate": 0.08750000000000002, "n": 800} +{"dataset": "sst2", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "imdb", "seed": 3, "flip_rate": 0.3425, "n": 800} +{"dataset": "ag_news", "seed": 3, "flip_rate": 0.19874999999999998, "n": 800} +{"dataset": "dbpedia", "seed": 3, "flip_rate": 0.47875, "n": 800} +{"dataset": "counterfact", "seed": 3, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 3, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 3, "flip_rate": 0.25125, "n": 800} +{"dataset": "tweet_irony", "seed": 3, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "tweet_offensive", "seed": 3, "flip_rate": 0.2025, "n": 800} +{"dataset": "subj", "seed": 3, "flip_rate": 0.10375000000000001, "n": 800} +{"dataset": "spam", "seed": 3, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "cola", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "stance", "seed": 3, "flip_rate": 0.16826923076923073, "n": 208} +{"dataset": "amazon_cf", "seed": 3, "flip_rate": 0.09875, "n": 800} +{"dataset": "sst2", "seed": 4, "flip_rate": 0.12124999999999997, "n": 800} +{"dataset": "imdb", "seed": 4, "flip_rate": 0.36250000000000004, "n": 800} +{"dataset": "ag_news", "seed": 4, "flip_rate": 0.17500000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 4, "flip_rate": 0.48624999999999996, "n": 800} +{"dataset": "counterfact", "seed": 4, "flip_rate": 0.03874999999999995, "n": 800} +{"dataset": "emotion", "seed": 4, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 4, "flip_rate": 0.24750000000000005, "n": 800} +{"dataset": "tweet_irony", "seed": 4, "flip_rate": 0.24624999999999997, "n": 800} +{"dataset": "tweet_offensive", "seed": 4, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "subj", "seed": 4, "flip_rate": 0.08125000000000004, "n": 800} +{"dataset": "spam", "seed": 4, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 4, "flip_rate": 0.11499999999999999, "n": 800} +{"dataset": "stance", "seed": 4, "flip_rate": 0.1875, "n": 208} +{"dataset": "amazon_cf", "seed": 4, "flip_rate": 0.13, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": 0.15384615384615385, "n": 208} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 1, "flip_rate": 0.10124999999999995, "n": 800} +{"dataset": "sst2", "seed": 0, "flip_rate": 0.09750000000000003, "n": 800} +{"dataset": "imdb", "seed": 0, "flip_rate": 0.355, "n": 800} +{"dataset": "ag_news", "seed": 0, "flip_rate": 0.1875, "n": 800} +{"dataset": "dbpedia", "seed": 0, "flip_rate": 0.49875, "n": 800} +{"dataset": "counterfact", "seed": 0, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 0, "flip_rate": 0.06125000000000003, "n": 800} +{"dataset": "tweet_hate", "seed": 0, "flip_rate": 0.26375000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 0, "flip_rate": 0.22499999999999998, "n": 800} +{"dataset": "tweet_offensive", "seed": 0, "flip_rate": 0.22375, "n": 800} +{"dataset": "subj", "seed": 0, "flip_rate": 0.08875, "n": 800} +{"dataset": "spam", "seed": 0, "flip_rate": 0.27125, "n": 800} +{"dataset": "cola", "seed": 0, "flip_rate": 0.13749999999999996, "n": 800} +{"dataset": "stance", "seed": 0, "flip_rate": 0.15384615384615385, "n": 208} +{"dataset": "amazon_cf", "seed": 0, "flip_rate": 0.12250000000000005, "n": 800} +{"dataset": "sst2", "seed": 1, "flip_rate": 0.10124999999999995, "n": 800} +{"dataset": "imdb", "seed": 1, "flip_rate": 0.38749999999999996, "n": 800} +{"dataset": "ag_news", "seed": 1, "flip_rate": 0.20750000000000002, "n": 800} +{"dataset": "dbpedia", "seed": 1, "flip_rate": 0.5137499999999999, "n": 800} +{"dataset": "counterfact", "seed": 1, "flip_rate": 0.03249999999999997, "n": 800} +{"dataset": "emotion", "seed": 1, "flip_rate": 0.07625000000000004, "n": 800} +{"dataset": "tweet_hate", "seed": 1, "flip_rate": 0.26, "n": 800} +{"dataset": "tweet_irony", "seed": 1, "flip_rate": 0.23124999999999996, "n": 800} +{"dataset": "tweet_offensive", "seed": 1, "flip_rate": 0.23375, "n": 800} +{"dataset": "subj", "seed": 1, "flip_rate": 0.08250000000000002, "n": 800} +{"dataset": "spam", "seed": 1, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "cola", "seed": 1, "flip_rate": 0.13624999999999998, "n": 800} +{"dataset": "stance", "seed": 1, "flip_rate": 0.1826923076923077, "n": 208} +{"dataset": "amazon_cf", "seed": 1, "flip_rate": 0.11375000000000002, "n": 800} +{"dataset": "sst2", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "imdb", "seed": 2, "flip_rate": 0.35624999999999996, "n": 800} +{"dataset": "ag_news", "seed": 2, "flip_rate": 0.20625000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 2, "flip_rate": 0.5037499999999999, "n": 800} +{"dataset": "counterfact", "seed": 2, "flip_rate": 0.04125000000000001, "n": 800} +{"dataset": "emotion", "seed": 2, "flip_rate": 0.05874999999999997, "n": 800} +{"dataset": "tweet_hate", "seed": 2, "flip_rate": 0.22750000000000004, "n": 800} +{"dataset": "tweet_irony", "seed": 2, "flip_rate": 0.22624999999999995, "n": 800} +{"dataset": "tweet_offensive", "seed": 2, "flip_rate": 0.23875000000000002, "n": 800} +{"dataset": "subj", "seed": 2, "flip_rate": 0.09875, "n": 800} +{"dataset": "spam", "seed": 2, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 2, "flip_rate": 0.10875000000000001, "n": 800} +{"dataset": "stance", "seed": 2, "flip_rate": 0.14903846153846156, "n": 208} +{"dataset": "amazon_cf", "seed": 2, "flip_rate": 0.08750000000000002, "n": 800} +{"dataset": "sst2", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "imdb", "seed": 3, "flip_rate": 0.3425, "n": 800} +{"dataset": "ag_news", "seed": 3, "flip_rate": 0.19874999999999998, "n": 800} +{"dataset": "dbpedia", "seed": 3, "flip_rate": 0.47875, "n": 800} +{"dataset": "counterfact", "seed": 3, "flip_rate": 0.043749999999999956, "n": 800} +{"dataset": "emotion", "seed": 3, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 3, "flip_rate": 0.25125, "n": 800} +{"dataset": "tweet_irony", "seed": 3, "flip_rate": 0.24250000000000005, "n": 800} +{"dataset": "tweet_offensive", "seed": 3, "flip_rate": 0.2025, "n": 800} +{"dataset": "subj", "seed": 3, "flip_rate": 0.10375000000000001, "n": 800} +{"dataset": "spam", "seed": 3, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "cola", "seed": 3, "flip_rate": 0.11250000000000004, "n": 800} +{"dataset": "stance", "seed": 3, "flip_rate": 0.16826923076923073, "n": 208} +{"dataset": "amazon_cf", "seed": 3, "flip_rate": 0.09875, "n": 800} +{"dataset": "sst2", "seed": 4, "flip_rate": 0.12124999999999997, "n": 800} +{"dataset": "imdb", "seed": 4, "flip_rate": 0.36250000000000004, "n": 800} +{"dataset": "ag_news", "seed": 4, "flip_rate": 0.17500000000000004, "n": 800} +{"dataset": "dbpedia", "seed": 4, "flip_rate": 0.48624999999999996, "n": 800} +{"dataset": "counterfact", "seed": 4, "flip_rate": 0.03874999999999995, "n": 800} +{"dataset": "emotion", "seed": 4, "flip_rate": 0.07250000000000001, "n": 800} +{"dataset": "tweet_hate", "seed": 4, "flip_rate": 0.24750000000000005, "n": 800} +{"dataset": "tweet_irony", "seed": 4, "flip_rate": 0.24624999999999997, "n": 800} +{"dataset": "tweet_offensive", "seed": 4, "flip_rate": 0.23624999999999996, "n": 800} +{"dataset": "subj", "seed": 4, "flip_rate": 0.08125000000000004, "n": 800} +{"dataset": "spam", "seed": 4, "flip_rate": 0.24124999999999996, "n": 800} +{"dataset": "cola", "seed": 4, "flip_rate": 0.11499999999999999, "n": 800} +{"dataset": "stance", "seed": 4, "flip_rate": 0.1875, "n": 208} +{"dataset": "amazon_cf", "seed": 4, "flip_rate": 0.13, "n": 800} diff --git a/repro_bundle/results_tier2/results/eval.jsonl b/repro_bundle/results_tier2/results/eval.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..7e4432ce4e65d30c8926336d070290bed59c0c5e --- /dev/null +++ b/repro_bundle/results_tier2/results/eval.jsonl @@ -0,0 +1,1416 @@ +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.78875, "drop": 0.0}, "paraphrase": {"acc": 0.7645429362880887, "drop": 0.02420706371191128, "rotation": {"subspace_principal_angle": 1.2173428997633269, "mean_class_cosine": 0.34613981140180783}}, "domain": {"acc": 0.66375, "drop": 0.125, "rotation": {"subspace_principal_angle": 1.2327061371099897, "mean_class_cosine": 0.3316860024743622}}}, "iid_acc": 0.78875, "selectivity": 0.29624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.287801427158086, "mean_class_cosine": 0.2792326694699942}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5415512465373962, "drop": -0.005301246537396165, "rotation": {"subspace_principal_angle": 1.4826873586765374, "mean_class_cosine": 0.8742178904437565}}, "domain": {"acc": 0.50125, "drop": 0.03500000000000003, "rotation": {"subspace_principal_angle": 1.4901649913072454, "mean_class_cosine": 0.6563624615789627}}}, "iid_acc": 0.53625, "selectivity": 0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.9335911744705346, "mean_class_cosine": 0.9103338651934845}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.7562326869806094, "drop": 0.017517313019390635}, "domain": {"acc": 0.7725, "drop": 0.0012500000000000844}}, "iid_acc": 0.77375, "selectivity": 0.28125000000000006} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8075, "drop": 0.0}, "paraphrase": {"acc": 0.7965116279069767, "drop": 0.01098837209302328, "rotation": {"subspace_principal_angle": 0.8035537849788047, "mean_class_cosine": 0.6941529859280129}}, "domain": {"acc": 0.6875, "drop": 0.12, "rotation": {"subspace_principal_angle": 1.1686325991529884, "mean_class_cosine": 0.3914103543132651}}, "length": {"acc": 0.8, "drop": 0.007499999999999951, "rotation": {"subspace_principal_angle": 0.7897497491232404, "mean_class_cosine": 0.7040230601402793}}}, "iid_acc": 0.8075, "selectivity": 0.30374999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7674349618290983, "mean_class_cosine": 0.7196939193951748}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6414728682170543, "drop": 0.01602713178294568, "rotation": {"subspace_principal_angle": 1.288879288748167, "mean_class_cosine": 0.8485185221141581}}, "domain": {"acc": 0.6625, "drop": -0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.4368459632860788, "mean_class_cosine": 0.5607042188856695}}, "length": {"acc": 0.6621212121212121, "drop": -0.004621212121212137, "rotation": {"subspace_principal_angle": 1.2230547063204023, "mean_class_cosine": 0.9166318400068196}}}, "iid_acc": 0.6575, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.8046269051289581, "mean_class_cosine": 0.8820641372607203}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.7926356589147286, "drop": -0.005135658914728669}, "domain": {"acc": 0.69625, "drop": 0.09124999999999994}, "length": {"acc": 0.7878787878787878, "drop": -0.00037878787878786735}}, "iid_acc": 0.7875, "selectivity": 0.28374999999999995} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8307692307692308, "drop": 0.015480769230769131, "rotation": {"subspace_principal_angle": 1.1461706135196499, "mean_class_cosine": 0.5254429176325506}}}, "iid_acc": 0.84625, "selectivity": 0.58, "iid_split_rotation": {"subspace_principal_angle": 1.1688799127850784, "mean_class_cosine": 0.5261001194927581}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8, "drop": 0.0}, "paraphrase": {"acc": 0.796923076923077, "drop": 0.003076923076923088, "rotation": {"subspace_principal_angle": 1.3242841060180541, "mean_class_cosine": 0.9641264652586653}}}, "iid_acc": 0.8, "selectivity": 0.5337500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3981863180782648, "mean_class_cosine": 0.9370201686458131}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.8415384615384616, "drop": 0.003461538461538405}}, "iid_acc": 0.845, "selectivity": 0.57875} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.8927680798004988, "drop": 0.04223192019950128, "rotation": {"subspace_principal_angle": 1.2106415683121374, "mean_class_cosine": 0.7066695308643387}}}, "iid_acc": 0.935, "selectivity": 0.8525, "iid_split_rotation": {"subspace_principal_angle": 1.1511700616225666, "mean_class_cosine": 0.7422879247769008}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5311720698254364, "drop": 0.02382793017456364, "rotation": {"subspace_principal_angle": 1.373149046572336, "mean_class_cosine": 0.8538155049730057}}}, "iid_acc": 0.555, "selectivity": 0.47250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3861255031014492, "mean_class_cosine": 0.8730732976173553}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.85785536159601, "drop": 0.03214463840399007}}, "iid_acc": 0.89, "selectivity": 0.8075} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5019607843137255, "drop": 0.01678921568627456, "rotation": {"subspace_principal_angle": 1.5573365448958074, "mean_class_cosine": -0.013459375494237761}}}, "iid_acc": 0.51875, "selectivity": 0.011250000000000093, "iid_split_rotation": {"subspace_principal_angle": 1.5530648223509733, "mean_class_cosine": -0.017730575309227073}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.0015604575163399304, "rotation": {"subspace_principal_angle": 1.5316164005556767, "mean_class_cosine": 0.04651835530761543}}}, "iid_acc": 0.50875, "selectivity": 0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.4740729716187824, "mean_class_cosine": 0.03841895244711871}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.5202614379084968, "drop": -0.010261437908496762}}, "iid_acc": 0.51, "selectivity": 0.0025000000000000577} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.61, "drop": 0.0}, "paraphrase": {"acc": 0.5486018641810919, "drop": 0.06139813581890807, "rotation": {"subspace_principal_angle": 1.4117139037388426, "mean_class_cosine": 0.3559915861697644}}}, "iid_acc": 0.61, "selectivity": 0.37375, "iid_split_rotation": {"subspace_principal_angle": 1.3714632310508308, "mean_class_cosine": 0.37969570224814214}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.275, "drop": 0.0}, "paraphrase": {"acc": 0.2583222370173103, "drop": 0.016677762982689748, "rotation": {"subspace_principal_angle": 1.5666753155926543, "mean_class_cosine": 0.5349177629069118}}}, "iid_acc": 0.275, "selectivity": 0.038750000000000034, "iid_split_rotation": {"subspace_principal_angle": 1.5082299186645924, "mean_class_cosine": 0.6046278432817893}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.507323568575233, "drop": -0.009823568575232977}}, "iid_acc": 0.4975, "selectivity": 0.26125} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.70875, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.0009274193548387544, "rotation": {"subspace_principal_angle": 1.3772445961922097, "mean_class_cosine": 0.19234551093090788}}}, "iid_acc": 0.70875, "selectivity": 0.18625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3584465776147807, "mean_class_cosine": 0.2107574495845495}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.5825, "drop": 0.0}, "paraphrase": {"acc": 0.5942275042444821, "drop": -0.011727504244482101, "rotation": {"subspace_principal_angle": 1.2205673531066674, "mean_class_cosine": 0.9079473240997651}}}, "iid_acc": 0.5825, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1442505653422113, "mean_class_cosine": 0.940086831782734}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.7045840407470289, "drop": 0.007915959252971172}}, "iid_acc": 0.7125, "selectivity": 0.19000000000000006} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.028266129032258003, "rotation": {"subspace_principal_angle": 1.4282418254454519, "mean_class_cosine": 0.142072164699308}}}, "iid_acc": 0.66375, "selectivity": 0.16374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3468259132949723, "mean_class_cosine": 0.2221026092043146}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5838709677419355, "drop": 0.016129032258064502, "rotation": {"subspace_principal_angle": 1.3876894491282326, "mean_class_cosine": 0.9792973864192993}}}, "iid_acc": 0.6, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.0920941984969763, "mean_class_cosine": 0.9685148140397846}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.66125, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.019314516129032233}}, "iid_acc": 0.66125, "selectivity": 0.16125} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7600644122383253, "drop": -0.006314412238325295, "rotation": {"subspace_principal_angle": 1.115598956625283, "mean_class_cosine": 0.43963959517656737}}}, "iid_acc": 0.75375, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1918132960864172, "mean_class_cosine": 0.369975854351391}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.007481884057971011, "rotation": {"subspace_principal_angle": 1.431073468315769, "mean_class_cosine": 0.8636351111958729}}}, "iid_acc": 0.52875, "selectivity": -0.1037499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9984226130449826, "mean_class_cosine": 0.8215813059697017}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7342995169082126, "drop": -0.009299516908212624}}, "iid_acc": 0.725, "selectivity": 0.09250000000000003} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.879286694101509, "drop": 0.02321330589849102, "rotation": {"subspace_principal_angle": 0.9139246510391348, "mean_class_cosine": 0.6106425040490346}}}, "iid_acc": 0.9025, "selectivity": 0.44499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.884850806935217, "mean_class_cosine": 0.6334049571887388}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8491083676268861, "drop": 0.012141632373113831, "rotation": {"subspace_principal_angle": 0.783728326709098, "mean_class_cosine": 0.9410916640925551}}}, "iid_acc": 0.86125, "selectivity": 0.40374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.9780082924965329, "mean_class_cosine": 0.9403009956281769}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8710562414266118, "drop": 0.025193758573388236}}, "iid_acc": 0.89625, "selectivity": 0.43875} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9862778730703259, "drop": 0.007472126929674139, "rotation": {"subspace_principal_angle": 0.8999811261795105, "mean_class_cosine": 0.6216247525314251}}}, "iid_acc": 0.99375, "selectivity": 0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.8781549302342212, "mean_class_cosine": 0.6385721258735755}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64375, "drop": 0.0}, "paraphrase": {"acc": 0.4854202401372213, "drop": 0.15832975986277875, "rotation": {"subspace_principal_angle": 0.9223293102070813, "mean_class_cosine": 0.9921756731272653}}}, "iid_acc": 0.64375, "selectivity": -0.2024999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3103002492542979, "mean_class_cosine": 0.9993471541844559}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.97875, "drop": 0.0}, "paraphrase": {"acc": 0.9845626072041166, "drop": -0.005812607204116582}}, "iid_acc": 0.97875, "selectivity": 0.13250000000000006} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6826086956521739, "drop": -0.032608695652173836, "rotation": {"subspace_principal_angle": 1.5134217389057267, "mean_class_cosine": 0.05734311504400581}}}, "iid_acc": 0.65, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4857097607143221, "mean_class_cosine": 0.0849839360330566}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5260869565217391, "drop": 0.0001630434782609047, "rotation": {"subspace_principal_angle": 1.5681344894004097, "mean_class_cosine": 0.0027859134562919843}}}, "iid_acc": 0.52625, "selectivity": -0.10375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3461131864613545, "mean_class_cosine": -0.37599024476007736}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7202898550724638, "drop": -0.014039855072463747}}, "iid_acc": 0.70625, "selectivity": 0.07625000000000004} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6306818181818182, "drop": 0.023164335664335622, "rotation": {"subspace_principal_angle": 1.3319284714979758, "mean_class_cosine": 0.23716017540167048}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.21634615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.372005152667984, "mean_class_cosine": 0.24328801012012044}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.5384615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.5511363636363636, "drop": -0.01267482517482521, "rotation": {"subspace_principal_angle": 1.3661942544847507, "mean_class_cosine": 0.680691975446298}}}, "iid_acc": 0.5384615384615384, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.326965973200558, "mean_class_cosine": 0.5637407887003096}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6363636363636364, "drop": -0.011363636363636354}}, "iid_acc": 0.625, "selectivity": 0.1875} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.8817663817663818, "drop": -0.005516381766381806, "rotation": {"subspace_principal_angle": 1.0257225092803133, "mean_class_cosine": 0.5184812124632806}}}, "iid_acc": 0.87625, "selectivity": 0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.022872498909312, "mean_class_cosine": 0.5209161160438877}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7236467236467237, "drop": -0.0023967236467237196, "rotation": {"subspace_principal_angle": 1.5032004808784296, "mean_class_cosine": 0.9058820512519088}}}, "iid_acc": 0.72125, "selectivity": -0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.3645997436602872, "mean_class_cosine": 0.8668774947492062}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.8148148148148148, "drop": 0.002685185185185235}}, "iid_acc": 0.8175, "selectivity": 0.03375000000000006} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8019390581717452, "drop": 0.021810941828254826, "rotation": {"subspace_principal_angle": 1.2297289568018384, "mean_class_cosine": 0.33449317002319223}}, "domain": {"acc": 0.7775, "drop": 0.04625000000000001, "rotation": {"subspace_principal_angle": 1.2534806736867017, "mean_class_cosine": 0.31201735319379614}}}, "iid_acc": 0.82375, "selectivity": 0.33375, "iid_split_rotation": {"subspace_principal_angle": 1.249974932944242, "mean_class_cosine": 0.31534615054656556}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5526315789473685, "drop": -0.022631578947368447, "rotation": {"subspace_principal_angle": 1.3043937828785666, "mean_class_cosine": 0.9883475306540368}}, "domain": {"acc": 0.50125, "drop": 0.028750000000000053, "rotation": {"subspace_principal_angle": 1.4264081497308534, "mean_class_cosine": 0.495004129731469}}}, "iid_acc": 0.53, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.1840707821435512, "mean_class_cosine": 0.9761804914155537}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8060941828254847, "drop": 0.02015581717451531}, "domain": {"acc": 0.77375, "drop": 0.05249999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33625000000000005} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8275193798449613, "drop": 0.01498062015503876, "rotation": {"subspace_principal_angle": 1.095349444903557, "mean_class_cosine": 0.4577358103130952}}, "domain": {"acc": 0.71125, "drop": 0.13124999999999998, "rotation": {"subspace_principal_angle": 1.266476857896887, "mean_class_cosine": 0.29964398320533164}}, "length": {"acc": 0.843939393939394, "drop": -0.0014393939393939625, "rotation": {"subspace_principal_angle": 1.020996775517062, "mean_class_cosine": 0.5225163309800367}}}, "iid_acc": 0.8425, "selectivity": 0.38125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0185057868759257, "mean_class_cosine": 0.5246385975150215}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6996124031007752, "drop": 0.007887596899224802, "rotation": {"subspace_principal_angle": 1.5092754395213872, "mean_class_cosine": 0.8994925840610262}}, "domain": {"acc": 0.56125, "drop": 0.14625, "rotation": {"subspace_principal_angle": 1.5628669635664512, "mean_class_cosine": 0.41448835148075136}}, "length": {"acc": 0.7257575757575757, "drop": -0.018257575757575695, "rotation": {"subspace_principal_angle": 1.4863663784751628, "mean_class_cosine": 0.9364420792767987}}}, "iid_acc": 0.7075, "selectivity": 0.24625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5618431812325368, "mean_class_cosine": 0.89092989954099}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8352713178294574, "drop": 0.018478682170542604}, "domain": {"acc": 0.71875, "drop": 0.135}, "length": {"acc": 0.853030303030303, "drop": 0.0007196969696969813}}, "iid_acc": 0.85375, "selectivity": 0.3925} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8430769230769231, "drop": -0.0005769230769230305, "rotation": {"subspace_principal_angle": 1.4488633502262789, "mean_class_cosine": 0.2627189029713985}}}, "iid_acc": 0.8425, "selectivity": 0.60375, "iid_split_rotation": {"subspace_principal_angle": 1.3920742480225834, "mean_class_cosine": 0.2804084873444433}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8784615384615385, "drop": -0.019711538461538503, "rotation": {"subspace_principal_angle": 1.5157375083147102, "mean_class_cosine": 0.9772553909019803}}}, "iid_acc": 0.85875, "selectivity": 0.62, "iid_split_rotation": {"subspace_principal_angle": 1.5565345003357298, "mean_class_cosine": 0.9660708595879833}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5030769230769231, "drop": -0.0043269230769230616}}, "iid_acc": 0.49875, "selectivity": 0.26} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.027319201995012476, "rotation": {"subspace_principal_angle": 1.371071144342142, "mean_class_cosine": 0.6983572474879793}}}, "iid_acc": 0.955, "selectivity": 0.87125, "iid_split_rotation": {"subspace_principal_angle": 1.1190162175121645, "mean_class_cosine": 0.7424385333421911}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.46, "drop": 0.0}, "paraphrase": {"acc": 0.4339152119700748, "drop": 0.0260847880299252, "rotation": {"subspace_principal_angle": 1.2702079599204308, "mean_class_cosine": 0.8441489999187578}}}, "iid_acc": 0.46, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.54813616950135, "mean_class_cosine": 0.8639196384527672}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8952618453865336, "drop": 0.025988154613466374}}, "iid_acc": 0.92125, "selectivity": 0.8375} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": 0.003088235294117614, "rotation": {"subspace_principal_angle": 1.5507981147481529, "mean_class_cosine": 0.019996879097623668}}}, "iid_acc": 0.5325, "selectivity": 0.04124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4729086760692658, "mean_class_cosine": 0.09773139915411025}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.51125, "drop": 0.0}, "paraphrase": {"acc": 0.4915032679738562, "drop": 0.0197467320261438, "rotation": {"subspace_principal_angle": 1.5063411676739642, "mean_class_cosine": 0.09376607824598783}}}, "iid_acc": 0.51125, "selectivity": 0.019999999999999962, "iid_split_rotation": {"subspace_principal_angle": 1.3815718454510815, "mean_class_cosine": -0.780613294925026}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.5071895424836601, "drop": 0.01531045751633986}}, "iid_acc": 0.5225, "selectivity": 0.031249999999999944} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.5872170439414115, "drop": 0.0377829560585885, "rotation": {"subspace_principal_angle": 1.415219489070654, "mean_class_cosine": 0.29709377332489545}}}, "iid_acc": 0.625, "selectivity": 0.36125, "iid_split_rotation": {"subspace_principal_angle": 1.3868652000931787, "mean_class_cosine": 0.30693642876692934}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1975, "drop": 0.0}, "paraphrase": {"acc": 0.20905459387483355, "drop": -0.01155459387483354, "rotation": {"subspace_principal_angle": 1.508152156475103, "mean_class_cosine": 0.3868276136007782}}}, "iid_acc": 0.1975, "selectivity": -0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4103810864475195, "mean_class_cosine": 0.663164345963827}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.5539280958721704, "drop": 0.014821904127829577}}, "iid_acc": 0.56875, "selectivity": 0.305} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7470288624787776, "drop": 0.0017211375212223823, "rotation": {"subspace_principal_angle": 1.218944387247129, "mean_class_cosine": 0.34463687992333025}}}, "iid_acc": 0.74875, "selectivity": 0.21000000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.161541923084499, "mean_class_cosine": 0.39792541536895626}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.6281833616298812, "drop": -0.029433361629881194, "rotation": {"subspace_principal_angle": 1.4828776907746226, "mean_class_cosine": 0.8507212382776503}}}, "iid_acc": 0.59875, "selectivity": 0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5285287072761573, "mean_class_cosine": 0.889675723801062}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7589134125636672, "drop": -0.007663412563667205}}, "iid_acc": 0.75125, "selectivity": 0.21250000000000002} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6258064516129033, "drop": 0.029193548387096757, "rotation": {"subspace_principal_angle": 1.4380753701349163, "mean_class_cosine": 0.13233165634932476}}}, "iid_acc": 0.655, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.3608689467779782, "mean_class_cosine": 0.20838887470800788}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.6129032258064516, "drop": -0.009153225806451615, "rotation": {"subspace_principal_angle": 1.3501977335705537, "mean_class_cosine": 0.8788630744675302}}}, "iid_acc": 0.60375, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3042174668679, "mean_class_cosine": 0.9016543168707795}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6274193548387097, "drop": 0.01758064516129032}}, "iid_acc": 0.645, "selectivity": 0.13624999999999998} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7584541062801933, "drop": -0.015954106280193225, "rotation": {"subspace_principal_angle": 1.2412464706999458, "mean_class_cosine": 0.3236171403840489}}}, "iid_acc": 0.7425, "selectivity": 0.125, "iid_split_rotation": {"subspace_principal_angle": 1.2363583878771438, "mean_class_cosine": 0.32823830168996404}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5797101449275363, "drop": -0.013460144927536222, "rotation": {"subspace_principal_angle": 1.4328194682708153, "mean_class_cosine": 0.8163073052539073}}}, "iid_acc": 0.56625, "selectivity": -0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.393646227569837, "mean_class_cosine": 0.7747593471553436}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7471819645732689, "drop": -0.00843196457326889}}, "iid_acc": 0.73875, "selectivity": 0.12124999999999997} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.887517146776406, "drop": 0.02623285322359392, "rotation": {"subspace_principal_angle": 1.1698183926314918, "mean_class_cosine": 0.39031889296657585}}}, "iid_acc": 0.91375, "selectivity": 0.3949999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1565211457406315, "mean_class_cosine": 0.4025265308866283}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8847736625514403, "drop": 0.007726337448559661, "rotation": {"subspace_principal_angle": 0.9750908966255823, "mean_class_cosine": 0.724907117214936}}}, "iid_acc": 0.8925, "selectivity": 0.3737499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5507156284967207, "mean_class_cosine": 0.9656213717308835}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9012345679012346, "drop": 0.013765432098765462}}, "iid_acc": 0.915, "selectivity": 0.39625} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.7143429012203775, "mean_class_cosine": 0.7555238835683384}}}, "iid_acc": 0.99125, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6854415650897067, "mean_class_cosine": 0.7741396052879256}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.93625, "drop": 0.0}, "paraphrase": {"acc": 0.9468267581475128, "drop": -0.0105767581475128, "rotation": {"subspace_principal_angle": 1.309840282431262, "mean_class_cosine": 0.963893627422916}}}, "iid_acc": 0.93625, "selectivity": 0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.94741234766098, "mean_class_cosine": 0.9835024086435691}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9775, "drop": 0.0}, "paraphrase": {"acc": 0.9828473413379074, "drop": -0.005347341337907374}}, "iid_acc": 0.9775, "selectivity": 0.11250000000000004} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.006666666666666599, "rotation": {"subspace_principal_angle": 1.5316360507140625, "mean_class_cosine": 0.0391502679566816}}}, "iid_acc": 0.66, "selectivity": 0.07250000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3743529128243746, "mean_class_cosine": 0.19518239053186837}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.47246376811594204, "drop": 0.03753623188405797, "rotation": {"subspace_principal_angle": 1.118894178326917, "mean_class_cosine": -0.9771850067325109}}}, "iid_acc": 0.51, "selectivity": -0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3803133457755477, "mean_class_cosine": 0.946898836596458}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7072463768115942, "drop": -0.0009963768115941463}}, "iid_acc": 0.70625, "selectivity": 0.11875000000000002} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6079545454545454, "drop": -0.0021853146853146876, "rotation": {"subspace_principal_angle": 1.5660532537843979, "mean_class_cosine": 0.23593286698387292}}}, "iid_acc": 0.6057692307692307, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.3587755369509562, "mean_class_cosine": 0.22622489885469554}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6022727272727273, "drop": -0.02054195804195802, "rotation": {"subspace_principal_angle": 1.425664296220229, "mean_class_cosine": 0.6549701520168614}}}, "iid_acc": 0.5817307692307693, "selectivity": 0.1538461538461539, "iid_split_rotation": {"subspace_principal_angle": 1.3242514768412414, "mean_class_cosine": 0.5857958842617208}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.0021853146853146876}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2163461538461539} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8703703703703703, "drop": 0.00462962962962965, "rotation": {"subspace_principal_angle": 1.2653845793904546, "mean_class_cosine": 0.30068589374987864}}}, "iid_acc": 0.875, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 1.1530609002795849, "mean_class_cosine": 0.4056916522664149}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.57875, "drop": 0.0}, "paraphrase": {"acc": 0.5911680911680912, "drop": -0.012418091168091228, "rotation": {"subspace_principal_angle": 1.5387855283082137, "mean_class_cosine": 0.9739607963186667}}}, "iid_acc": 0.57875, "selectivity": -0.17625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.0864399863171719, "mean_class_cosine": 0.8793845387817394}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8475783475783476, "drop": 0.006171652421652407}}, "iid_acc": 0.85375, "selectivity": 0.09875} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8213296398891967, "drop": 0.03117036011080332, "rotation": {"subspace_principal_angle": 1.2728072178283032, "mean_class_cosine": 0.2935985328813959}}, "domain": {"acc": 0.6475, "drop": 0.20500000000000007, "rotation": {"subspace_principal_angle": 1.2548682229869004, "mean_class_cosine": 0.31069877540648505}}}, "iid_acc": 0.8525, "selectivity": 0.35625, "iid_split_rotation": {"subspace_principal_angle": 1.1836995794199188, "mean_class_cosine": 0.37750157196014655}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5318559556786704, "drop": -0.006855955678670367, "rotation": {"subspace_principal_angle": 1.183174660406376, "mean_class_cosine": 0.9978979106979409}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.4296731020342406, "mean_class_cosine": 0.41513571955567125}}}, "iid_acc": 0.525, "selectivity": 0.028749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9398355095136322, "mean_class_cosine": 0.9889864954707369}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8421052631578947, "drop": 0.004144736842105257}, "domain": {"acc": 0.8225, "drop": 0.023749999999999938}}, "iid_acc": 0.84625, "selectivity": 0.3499999999999999} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8604651162790697, "drop": 0.024534883720930267, "rotation": {"subspace_principal_angle": 1.1635939988143396, "mean_class_cosine": 0.396041968116701}}, "domain": {"acc": 0.56, "drop": 0.32499999999999996, "rotation": {"subspace_principal_angle": 1.3073928416159974, "mean_class_cosine": 0.26036815053777257}}, "length": {"acc": 0.8878787878787879, "drop": -0.002878787878787925, "rotation": {"subspace_principal_angle": 1.1306543361974652, "mean_class_cosine": 0.42606792700100266}}}, "iid_acc": 0.885, "selectivity": 0.41375, "iid_split_rotation": {"subspace_principal_angle": 1.1314507236110818, "mean_class_cosine": 0.42534730729927617}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7383720930232558, "drop": 0.03412790697674417, "rotation": {"subspace_principal_angle": 0.9299498041850935, "mean_class_cosine": 0.9017395970359101}}, "domain": {"acc": 0.5575, "drop": 0.21499999999999997, "rotation": {"subspace_principal_angle": 1.4111384197278427, "mean_class_cosine": 0.27301695590829955}}, "length": {"acc": 0.7878787878787878, "drop": -0.01537878787878788, "rotation": {"subspace_principal_angle": 1.0571900651122168, "mean_class_cosine": 0.9405678373193577}}}, "iid_acc": 0.7725, "selectivity": 0.30124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.339591955524043, "mean_class_cosine": 0.8972887886451133}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8662790697674418, "drop": 0.022470930232558195}, "domain": {"acc": 0.65, "drop": 0.23875000000000002}, "length": {"acc": 0.8878787878787879, "drop": 0.000871212121212106}}, "iid_acc": 0.88875, "selectivity": 0.41750000000000004} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.88, "drop": -0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.4381874155007441, "mean_class_cosine": 0.2360974231087299}}}, "iid_acc": 0.875, "selectivity": 0.5974999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3889398949790899, "mean_class_cosine": 0.28986888876348327}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8123076923076923, "drop": 0.013942307692307754, "rotation": {"subspace_principal_angle": 1.4726936861262263, "mean_class_cosine": 0.9825256512058247}}}, "iid_acc": 0.82625, "selectivity": 0.5487500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.413087776634539, "mean_class_cosine": 0.9710401150655834}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8861538461538462, "drop": 0.005096153846153806}}, "iid_acc": 0.89125, "selectivity": 0.61375} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.9725, "drop": 0.0}, "paraphrase": {"acc": 0.9551122194513716, "drop": 0.01738778054862844, "rotation": {"subspace_principal_angle": 1.1805912863774084, "mean_class_cosine": 0.696959257792652}}}, "iid_acc": 0.9725, "selectivity": 0.875, "iid_split_rotation": {"subspace_principal_angle": 1.1819768188421824, "mean_class_cosine": 0.7412159284602584}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.4275, "drop": 0.0}, "paraphrase": {"acc": 0.42643391521197005, "drop": 0.0010660847880299418, "rotation": {"subspace_principal_angle": 1.267210976332211, "mean_class_cosine": 0.8469328938765297}}}, "iid_acc": 0.4275, "selectivity": 0.32999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2708278894045024, "mean_class_cosine": 0.8352289736541864}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.9276807980049875, "drop": 0.0210692019950125}}, "iid_acc": 0.94875, "selectivity": 0.85125} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5568627450980392, "drop": -0.006862745098039191, "rotation": {"subspace_principal_angle": 1.4681160564274516, "mean_class_cosine": 0.10249993503953822}}}, "iid_acc": 0.55, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.5020669180006716, "mean_class_cosine": 0.06867531169240668}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.4875816993464052, "drop": 0.00991830065359478, "rotation": {"subspace_principal_angle": 1.0583311490868843, "mean_class_cosine": 0.6025164388499016}}}, "iid_acc": 0.4975, "selectivity": -0.03250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2056755940549575, "mean_class_cosine": -0.8743607002802327}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5372549019607843, "drop": 0.0002450980392156854}}, "iid_acc": 0.5375, "selectivity": 0.007499999999999951} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6604527296937417, "drop": -0.005452729693741665, "rotation": {"subspace_principal_angle": 1.3690495088359922, "mean_class_cosine": 0.35780435447931885}}}, "iid_acc": 0.655, "selectivity": 0.39375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.418659878889111, "mean_class_cosine": 0.31040393556774787}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.18125, "drop": 0.0}, "paraphrase": {"acc": 0.1904127829560586, "drop": -0.009162782956058602, "rotation": {"subspace_principal_angle": 1.4238306284342992, "mean_class_cosine": 0.3336979016265693}}}, "iid_acc": 0.18125, "selectivity": -0.07999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3516589964423749, "mean_class_cosine": 0.6675023436977741}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5712383488681758, "drop": 0.030011651131824135}}, "iid_acc": 0.60125, "selectivity": 0.33999999999999997} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.00039685908319186947, "rotation": {"subspace_principal_angle": 1.3803752925053967, "mean_class_cosine": 0.18927233568020757}}}, "iid_acc": 0.72875, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 1.4107099285898823, "mean_class_cosine": 0.1594035006847374}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5229202037351444, "drop": -0.015420203735144411, "rotation": {"subspace_principal_angle": 1.132903800026888, "mean_class_cosine": 0.9957211105004683}}}, "iid_acc": 0.5075, "selectivity": -0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1567819921044111, "mean_class_cosine": 0.997306798497698}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7606112054329371, "drop": -0.0018612054329371075}}, "iid_acc": 0.75875, "selectivity": 0.22750000000000004} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.03556451612903222, "rotation": {"subspace_principal_angle": 1.4500907567958619, "mean_class_cosine": 0.12041267342093487}}}, "iid_acc": 0.6775, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4451716043887228, "mean_class_cosine": 0.12529455713007304}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5435483870967742, "drop": 0.002701612903225792, "rotation": {"subspace_principal_angle": 1.0905650289278745, "mean_class_cosine": 0.9919303109580393}}}, "iid_acc": 0.54625, "selectivity": 0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.435319911955165, "mean_class_cosine": 0.9822586493642307}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6419354838709678, "drop": 0.02806451612903227}}, "iid_acc": 0.67, "selectivity": 0.16375000000000006} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7568438003220612, "drop": 0.014406199677938769, "rotation": {"subspace_principal_angle": 1.362823497882489, "mean_class_cosine": 0.2064768368861471}}}, "iid_acc": 0.77125, "selectivity": 0.21124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3556181289971392, "mean_class_cosine": 0.21352151995440388}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5362318840579711, "drop": -0.01748188405797102, "rotation": {"subspace_principal_angle": 1.162087777749045, "mean_class_cosine": 0.4004265677277434}}}, "iid_acc": 0.51875, "selectivity": -0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5040593331311518, "mean_class_cosine": 0.6240661876988235}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.775, "drop": 0.0}, "paraphrase": {"acc": 0.7729468599033816, "drop": 0.002053140096618389}}, "iid_acc": 0.775, "selectivity": 0.21499999999999997} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9245541838134431, "drop": 0.017945816186556884, "rotation": {"subspace_principal_angle": 1.1580571429217417, "mean_class_cosine": 0.401119992002977}}}, "iid_acc": 0.9425, "selectivity": 0.44875, "iid_split_rotation": {"subspace_principal_angle": 1.0639249835392377, "mean_class_cosine": 0.48544434408974335}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5116598079561042, "drop": 0.0033401920438957955, "rotation": {"subspace_principal_angle": 1.474754823264982, "mean_class_cosine": 0.27562171118558787}}}, "iid_acc": 0.515, "selectivity": 0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6855036886417566, "mean_class_cosine": 0.9705005903895466}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.93, "drop": 0.0}, "paraphrase": {"acc": 0.934156378600823, "drop": -0.004156378600822963}}, "iid_acc": 0.93, "selectivity": 0.43625} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0006389365351628928, "rotation": {"subspace_principal_angle": 0.7922854168062247, "mean_class_cosine": 0.7022200256213631}}}, "iid_acc": 0.9925, "selectivity": 0.13250000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.7419982107629325, "mean_class_cosine": 0.7371197159686783}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9502572898799314, "drop": -0.00900728987993138, "rotation": {"subspace_principal_angle": 1.2962420205278622, "mean_class_cosine": 0.9560615750258359}}}, "iid_acc": 0.94125, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4821593641247295, "mean_class_cosine": 0.9814265656310696}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.9875, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": -0.0022084048027444236}}, "iid_acc": 0.9875, "selectivity": 0.12750000000000006} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6608695652173913, "drop": 0.03413043478260869, "rotation": {"subspace_principal_angle": 1.5269251960916086, "mean_class_cosine": 0.04385705910466185}}}, "iid_acc": 0.695, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.452899790541391, "mean_class_cosine": 0.117623607019381}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.041141304347826146, "rotation": {"subspace_principal_angle": 1.2698364171990568, "mean_class_cosine": -0.7441747656964055}}}, "iid_acc": 0.52375, "selectivity": -0.06624999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.421349910061863, "mean_class_cosine": 0.7671339569758747}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 23, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7231884057971014, "drop": -0.009438405797101423}}, "iid_acc": 0.71375, "selectivity": 0.12375000000000003} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6826923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.6363636363636364, "drop": 0.046328671328671356, "rotation": {"subspace_principal_angle": 1.389965032090286, "mean_class_cosine": 0.19817848562152038}}}, "iid_acc": 0.6826923076923077, "selectivity": 0.25480769230769235, "iid_split_rotation": {"subspace_principal_angle": 1.5631105240216077, "mean_class_cosine": 0.18163197534662465}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6105769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.625, "drop": -0.014423076923076872, "rotation": {"subspace_principal_angle": 1.3818731643580437, "mean_class_cosine": 0.6330079224250492}}}, "iid_acc": 0.6105769230769231, "selectivity": 0.18269230769230776, "iid_split_rotation": {"subspace_principal_angle": 1.188799274694121, "mean_class_cosine": 0.5596495884045671}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.006993006993006978}}, "iid_acc": 0.6490384615384616, "selectivity": 0.2211538461538462} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8803418803418803, "drop": 0.0009081196581196549, "rotation": {"subspace_principal_angle": 1.336870712814003, "mean_class_cosine": 0.23179799551578698}}}, "iid_acc": 0.88125, "selectivity": 0.15749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.230199866592587, "mean_class_cosine": 0.3340493484245235}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.5811965811965812, "drop": -0.004946581196581201, "rotation": {"subspace_principal_angle": 0.6698879155775275, "mean_class_cosine": 0.9725719155523136}}}, "iid_acc": 0.57625, "selectivity": -0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9538385113441675, "mean_class_cosine": 0.9585511263219829}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8974358974358975, "drop": 0.0013141025641025816}}, "iid_acc": 0.89875, "selectivity": 0.17500000000000004} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8337950138504155, "drop": 0.027454986149584415, "rotation": {"subspace_principal_angle": 1.2672102146950088, "mean_class_cosine": 0.2989442428389649}}, "domain": {"acc": 0.88, "drop": -0.018750000000000044, "rotation": {"subspace_principal_angle": 1.401321553710003, "mean_class_cosine": 0.1686646701084955}}}, "iid_acc": 0.86125, "selectivity": 0.36749999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.2210418787662296, "mean_class_cosine": 0.3426671322986696}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5346260387811634, "drop": -0.009626038781163415, "rotation": {"subspace_principal_angle": 1.1075097075912679, "mean_class_cosine": 0.9893879557921825}}, "domain": {"acc": 0.50125, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.2937330372237155, "mean_class_cosine": 0.27468238367063336}}}, "iid_acc": 0.525, "selectivity": 0.03125, "iid_split_rotation": {"subspace_principal_angle": 1.474523039419839, "mean_class_cosine": 0.9693041836230525}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 15, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8476454293628809, "drop": 0.04110457063711914}, "domain": {"acc": 0.86875, "drop": 0.020000000000000018}}, "iid_acc": 0.88875, "selectivity": 0.395} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.875968992248062, "drop": 0.012781007751938023, "rotation": {"subspace_principal_angle": 1.3733967390372437, "mean_class_cosine": 0.1961200848251179}}, "domain": {"acc": 0.5525, "drop": 0.33625000000000005, "rotation": {"subspace_principal_angle": 1.427574987743908, "mean_class_cosine": 0.14273220656435498}}, "length": {"acc": 0.8924242424242425, "drop": -0.003674242424242413, "rotation": {"subspace_principal_angle": 1.32660559111955, "mean_class_cosine": 0.24177114792102647}}}, "iid_acc": 0.88875, "selectivity": 0.3975, "iid_split_rotation": {"subspace_principal_angle": 1.3721751230402406, "mean_class_cosine": 0.19731783026248}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.84875, "drop": 0.0}, "paraphrase": {"acc": 0.8197674418604651, "drop": 0.028982558139534875, "rotation": {"subspace_principal_angle": 1.0890107964995421, "mean_class_cosine": 0.9365459180728041}}, "domain": {"acc": 0.5575, "drop": 0.29125, "rotation": {"subspace_principal_angle": 1.2433163855943603, "mean_class_cosine": 0.290612600127489}}, "length": {"acc": 0.8575757575757575, "drop": -0.008825757575757542, "rotation": {"subspace_principal_angle": 0.7717021170570981, "mean_class_cosine": 0.9578281621778026}}}, "iid_acc": 0.84875, "selectivity": 0.3575, "iid_split_rotation": {"subspace_principal_angle": 1.2334456221930967, "mean_class_cosine": 0.9323952105718878}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8875968992248062, "drop": 0.00990310077519374}, "domain": {"acc": 0.695, "drop": 0.2025}, "length": {"acc": 0.9015151515151515, "drop": -0.004015151515151527}}, "iid_acc": 0.8975, "selectivity": 0.40624999999999994} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8907692307692308, "drop": -0.00951923076923078, "rotation": {"subspace_principal_angle": 1.3982668079683969, "mean_class_cosine": 0.2412385692456759}}}, "iid_acc": 0.88125, "selectivity": 0.63, "iid_split_rotation": {"subspace_principal_angle": 1.3873607135770687, "mean_class_cosine": 0.27893385434758605}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.805, "drop": 0.0}, "paraphrase": {"acc": 0.8046153846153846, "drop": 0.00038461538461542766, "rotation": {"subspace_principal_angle": 1.1223434771672867, "mean_class_cosine": 0.9797187391070092}}}, "iid_acc": 0.805, "selectivity": 0.5537500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1951168916205284, "mean_class_cosine": 0.9636873787333077}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.9092307692307692, "drop": -0.010480769230769127}}, "iid_acc": 0.89875, "selectivity": 0.6475000000000001} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.972568578553616, "drop": 0.0024314214463839745, "rotation": {"subspace_principal_angle": 1.0887911120828875, "mean_class_cosine": 0.6783458118132375}}}, "iid_acc": 0.975, "selectivity": 0.8624999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.083853464584136, "mean_class_cosine": 0.7060030624663257}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.8125, "drop": 0.0}, "paraphrase": {"acc": 0.773067331670823, "drop": 0.03943266832917702, "rotation": {"subspace_principal_angle": 1.1185256458497195, "mean_class_cosine": 0.8994165755464111}}}, "iid_acc": 0.8125, "selectivity": 0.7, "iid_split_rotation": {"subspace_principal_angle": 0.9631112899466643, "mean_class_cosine": 0.9391589025650812}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9650872817955112, "drop": 0.011162718204488775}}, "iid_acc": 0.97625, "selectivity": 0.8637499999999999} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.5830065359477125, "drop": 0.024493464052287583, "rotation": {"subspace_principal_angle": 1.4835452240288083, "mean_class_cosine": 0.08714044135255376}}}, "iid_acc": 0.6075, "selectivity": 0.08750000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.483433918298054, "mean_class_cosine": 0.08725132314037118}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.4888888888888889, "drop": 0.01611111111111113, "rotation": {"subspace_principal_angle": 1.2048851781158827, "mean_class_cosine": 0.5321986735325194}}}, "iid_acc": 0.505, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4614710386057692, "mean_class_cosine": -0.6438340541689455}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5542483660130719, "drop": 0.03450163398692807}}, "iid_acc": 0.58875, "selectivity": 0.06874999999999998} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65875, "drop": 0.0}, "paraphrase": {"acc": 0.6085219707057257, "drop": 0.05022802929427428, "rotation": {"subspace_principal_angle": 1.4274964103041523, "mean_class_cosine": 0.2819535108490442}}}, "iid_acc": 0.65875, "selectivity": 0.3837499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3942837316412728, "mean_class_cosine": 0.30680585311258146}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.18, "drop": 0.0}, "paraphrase": {"acc": 0.19440745672436752, "drop": -0.014407456724367523, "rotation": {"subspace_principal_angle": 1.42729065004506, "mean_class_cosine": 0.3233324825686333}}}, "iid_acc": 0.18, "selectivity": -0.09500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.543456747904511, "mean_class_cosine": 0.6716440420685439}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02558089214380832}}, "iid_acc": 0.64875, "selectivity": 0.37375} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7283531409168081, "drop": 0.005396859083191874, "rotation": {"subspace_principal_angle": 1.4458284299287356, "mean_class_cosine": 0.1246428806664818}}}, "iid_acc": 0.73375, "selectivity": 0.23625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4100208755322998, "mean_class_cosine": 0.16008370527990506}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5314091680814941, "drop": -0.012659168081494032, "rotation": {"subspace_principal_angle": 1.5017224503953317, "mean_class_cosine": 0.9922645891758275}}}, "iid_acc": 0.51875, "selectivity": 0.021250000000000047, "iid_split_rotation": {"subspace_principal_angle": 0.5510785135387823, "mean_class_cosine": 0.994402608644392}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.77125, "drop": 0.0}, "paraphrase": {"acc": 0.7504244482173175, "drop": 0.020825551782682528}}, "iid_acc": 0.77125, "selectivity": 0.27375} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.6532258064516129, "drop": 0.031774193548387175, "rotation": {"subspace_principal_angle": 1.4862916929900118, "mean_class_cosine": 0.08440409464427018}}}, "iid_acc": 0.685, "selectivity": 0.18375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.504979130468674, "mean_class_cosine": 0.06576968766218992}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5564516129032258, "drop": 0.009798387096774275, "rotation": {"subspace_principal_angle": 1.4670137647528678, "mean_class_cosine": 0.9493974522411648}}}, "iid_acc": 0.56625, "selectivity": 0.06500000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2038765745555904, "mean_class_cosine": 0.9198294280970227}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.6935483870967742, "drop": 0.01895161290322578}}, "iid_acc": 0.7125, "selectivity": 0.21125000000000005} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.7439613526570048, "drop": 0.0047886473429952225, "rotation": {"subspace_principal_angle": 1.4293803795970716, "mean_class_cosine": 0.14094506785538105}}}, "iid_acc": 0.74875, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4308280823099806, "mean_class_cosine": 0.13951166975487922}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.571658615136876, "drop": -0.019158615136876, "rotation": {"subspace_principal_angle": 1.3467680040328998, "mean_class_cosine": 0.780244054310545}}}, "iid_acc": 0.5525, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.5305723014222863, "mean_class_cosine": 0.8225959610699823}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.7616747181964574, "drop": 0.01207528180354267}}, "iid_acc": 0.77375, "selectivity": 0.19625000000000004} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.02258058984910838, "rotation": {"subspace_principal_angle": 1.1992451710106438, "mean_class_cosine": 0.3630611813014619}}}, "iid_acc": 0.95125, "selectivity": 0.44125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.197550530346397, "mean_class_cosine": 0.3646396665958699}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6598079561042524, "drop": -0.0023079561042523844, "rotation": {"subspace_principal_angle": 0.8946495725521172, "mean_class_cosine": 0.6299993713824044}}}, "iid_acc": 0.6575, "selectivity": 0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.7999262041956853, "mean_class_cosine": 0.9697684559622158}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.945, "drop": 0.0}, "paraphrase": {"acc": 0.9314128943758574, "drop": 0.01358710562414256}}, "iid_acc": 0.945, "selectivity": 0.43499999999999994} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99875, "drop": 0.0}, "paraphrase": {"acc": 0.9982847341337907, "drop": 0.000465265866209319, "rotation": {"subspace_principal_angle": 0.931771543682588, "mean_class_cosine": 0.5964129401770092}}}, "iid_acc": 0.99875, "selectivity": 0.18125000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9266228730293125, "mean_class_cosine": 0.600537740322992}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6375, "drop": 0.0}, "paraphrase": {"acc": 0.4065180102915952, "drop": 0.23098198970840478, "rotation": {"subspace_principal_angle": 0.4804883746103172, "mean_class_cosine": 0.9773138541986853}}}, "iid_acc": 0.6375, "selectivity": -0.18000000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0322313143428026, "mean_class_cosine": 0.9999323022857498}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9948542024013722, "drop": 0.001395797598627735}}, "iid_acc": 0.99625, "selectivity": 0.17874999999999996} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.6884057971014492, "drop": 0.04284420289855073, "rotation": {"subspace_principal_angle": 1.5103706788720144, "mean_class_cosine": 0.060388883021034304}}}, "iid_acc": 0.73125, "selectivity": 0.14374999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.4200020283035577, "mean_class_cosine": 0.1502234646254889}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.4666666666666667, "drop": 0.04083333333333328, "rotation": {"subspace_principal_angle": 0.9965627610068227, "mean_class_cosine": -0.9475923907212384}}}, "iid_acc": 0.5075, "selectivity": -0.08000000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.7465288923899847, "mean_class_cosine": 0.963542850231548}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7275362318840579, "drop": 0.039963768115942044}}, "iid_acc": 0.7675, "selectivity": 0.17999999999999994} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6647727272727273, "drop": 0.008304195804195835, "rotation": {"subspace_principal_angle": 1.4238317227425443, "mean_class_cosine": 0.16606688749587134}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.18269230769230776, "iid_split_rotation": {"subspace_principal_angle": 1.4493637206220726, "mean_class_cosine": 0.18022085560237225}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.4423076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.4090909090909091, "drop": 0.033216783216783174, "rotation": {"subspace_principal_angle": 1.314437097913778, "mean_class_cosine": 0.6492731252980364}}}, "iid_acc": 0.4423076923076923, "selectivity": -0.04807692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.0082539944993898, "mean_class_cosine": 0.81084397247279}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6761363636363636, "drop": -0.0030594405594405183}}, "iid_acc": 0.6730769230769231, "selectivity": 0.18269230769230776} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8874643874643875, "drop": 0.005035612535612466, "rotation": {"subspace_principal_angle": 1.373498225543769, "mean_class_cosine": 0.19602056818834496}}}, "iid_acc": 0.8925, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.272801169025973, "mean_class_cosine": 0.29360431510047513}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.5883190883190883, "drop": 0.0016809116809116675, "rotation": {"subspace_principal_angle": 1.0911846628602964, "mean_class_cosine": 0.9555356213618866}}}, "iid_acc": 0.59, "selectivity": -0.125, "iid_split_rotation": {"subspace_principal_angle": 1.1224788303280901, "mean_class_cosine": 0.9089747436544128}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8846153846153846, "drop": 0.01163461538461541}}, "iid_acc": 0.89625, "selectivity": 0.18125000000000002} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7853185595567868, "drop": 0.01843144044321321, "rotation": {"subspace_principal_angle": 1.3395977046860834, "mean_class_cosine": 0.2291444175563393}}, "domain": {"acc": 0.78875, "drop": 0.015000000000000013, "rotation": {"subspace_principal_angle": 1.3075865369139061, "mean_class_cosine": 0.2601811310213691}}}, "iid_acc": 0.80375, "selectivity": 0.31124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3161987442877714, "mean_class_cosine": 0.25185598358320005}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5373961218836565, "drop": -0.017396121883656468, "rotation": {"subspace_principal_angle": 1.1243174241253076, "mean_class_cosine": 0.997655368526527}}, "domain": {"acc": 0.50125, "drop": 0.018750000000000044, "rotation": {"subspace_principal_angle": 1.434291363492328, "mean_class_cosine": 0.5041835597720421}}}, "iid_acc": 0.52, "selectivity": 0.027500000000000024, "iid_split_rotation": {"subspace_principal_angle": 1.4733549164259618, "mean_class_cosine": 0.9931712323643861}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.775623268698061, "drop": 0.050626731301939065}, "domain": {"acc": 0.7475, "drop": 0.07874999999999999}}, "iid_acc": 0.82625, "selectivity": 0.33375000000000005} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.8081395348837209, "drop": 0.01561046511627906, "rotation": {"subspace_principal_angle": 1.4878379029744278, "mean_class_cosine": 0.08286330186240415}}, "domain": {"acc": 0.63875, "drop": 0.18499999999999994, "rotation": {"subspace_principal_angle": 1.454691067906504, "mean_class_cosine": 0.11584457649805602}}, "length": {"acc": 0.8227272727272728, "drop": 0.0010227272727272307, "rotation": {"subspace_principal_angle": 1.4554754679619462, "mean_class_cosine": 0.11506542197274067}}}, "iid_acc": 0.82375, "selectivity": 0.3475, "iid_split_rotation": {"subspace_principal_angle": 1.3842160089393434, "mean_class_cosine": 0.18549965489482007}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7325581395348837, "drop": 0.03494186046511627, "rotation": {"subspace_principal_angle": 1.2562269465379385, "mean_class_cosine": 0.9025478152861728}}, "domain": {"acc": 0.5575, "drop": 0.20999999999999996, "rotation": {"subspace_principal_angle": 1.3251714868629496, "mean_class_cosine": 0.29596572601612736}}, "length": {"acc": 0.7727272727272727, "drop": -0.005227272727272747, "rotation": {"subspace_principal_angle": 1.0637760788494128, "mean_class_cosine": 0.9397971855033075}}}, "iid_acc": 0.7675, "selectivity": 0.29124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.061070800271238, "mean_class_cosine": 0.9127773418749959}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.8725, "drop": 0.0}, "paraphrase": {"acc": 0.8527131782945736, "drop": 0.01978682170542645}, "domain": {"acc": 0.57875, "drop": 0.29375000000000007}, "length": {"acc": 0.8712121212121212, "drop": 0.0012878787878788378}}, "iid_acc": 0.8725, "selectivity": 0.39625000000000005} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8661538461538462, "drop": 0.0013461538461538858, "rotation": {"subspace_principal_angle": 1.2824688472167018, "mean_class_cosine": 0.4528696194346733}}}, "iid_acc": 0.8675, "selectivity": 0.64, "iid_split_rotation": {"subspace_principal_angle": 1.267228970027625, "mean_class_cosine": 0.4341122200684181}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.47875, "drop": 0.0}, "paraphrase": {"acc": 0.42615384615384616, "drop": 0.05259615384615385, "rotation": {"subspace_principal_angle": 0.5775569724987617, "mean_class_cosine": 0.9093320332006872}}}, "iid_acc": 0.47875, "selectivity": 0.25125, "iid_split_rotation": {"subspace_principal_angle": 0.7727337369433303, "mean_class_cosine": 0.8668228792038879}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.0036538461538461187}}, "iid_acc": 0.8775, "selectivity": 0.6499999999999999} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.95375, "drop": 0.0}, "paraphrase": {"acc": 0.9451371571072319, "drop": 0.008612842892768091, "rotation": {"subspace_principal_angle": 1.198927059140121, "mean_class_cosine": 0.6054966764735671}}}, "iid_acc": 0.95375, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.1388755195423386, "mean_class_cosine": 0.6482459135365446}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6084788029925187, "drop": 0.01402119700748139, "rotation": {"subspace_principal_angle": 1.493246923874091, "mean_class_cosine": 0.8677677280952006}}}, "iid_acc": 0.6225, "selectivity": 0.5375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5091321723919393, "mean_class_cosine": 0.8880006055054167}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 10, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.9351620947630923, "drop": 0.012337905236907698}}, "iid_acc": 0.9475, "selectivity": 0.8625} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.495, "drop": 0.0}, "paraphrase": {"acc": 0.4745098039215686, "drop": 0.02049019607843139, "rotation": {"subspace_principal_angle": 1.557032384565418, "mean_class_cosine": -0.013763507646052914}}}, "iid_acc": 0.495, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.5359063301369187, "mean_class_cosine": -0.03488291842098544}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.0017892156862744923, "rotation": {"subspace_principal_angle": 1.3709234192517532, "mean_class_cosine": 0.6267242100586605}}}, "iid_acc": 0.49625, "selectivity": -0.014999999999999958, "iid_split_rotation": {"subspace_principal_angle": 1.3896477426172604, "mean_class_cosine": -0.3304774205182823}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.4825, "drop": 0.0}, "paraphrase": {"acc": 0.47320261437908495, "drop": 0.009297385620915033}}, "iid_acc": 0.4825, "selectivity": -0.028749999999999998} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5805592543275633, "drop": 0.02319074567243673, "rotation": {"subspace_principal_angle": 1.3989325795834617, "mean_class_cosine": 0.365872980266819}}}, "iid_acc": 0.60375, "selectivity": 0.335, "iid_split_rotation": {"subspace_principal_angle": 1.4035880336880546, "mean_class_cosine": 0.3488005843140671}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.1775, "drop": 0.0}, "paraphrase": {"acc": 0.16511318242343542, "drop": 0.012386817576564568, "rotation": {"subspace_principal_angle": 1.5619855127249644, "mean_class_cosine": 0.3341308312243725}}}, "iid_acc": 0.1775, "selectivity": -0.09125, "iid_split_rotation": {"subspace_principal_angle": 1.3905000233838531, "mean_class_cosine": 0.6611663679873362}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5778961384820239, "drop": 0.022103861517976053}}, "iid_acc": 0.6, "selectivity": 0.33125} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.7096774193548387, "drop": -0.015927419354838768, "rotation": {"subspace_principal_angle": 1.5028161180968338, "mean_class_cosine": 0.06792786120606348}}}, "iid_acc": 0.69375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4722418499371575, "mean_class_cosine": 0.09839501129737169}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.533106960950764, "drop": -0.020606960950764086, "rotation": {"subspace_principal_angle": 1.024515247435546, "mean_class_cosine": 0.9955533421541893}}}, "iid_acc": 0.5125, "selectivity": -0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.2562415527805546, "mean_class_cosine": 0.9954953689417243}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7555178268251274, "drop": 0.00448217317487265}}, "iid_acc": 0.76, "selectivity": 0.24624999999999997} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.6338709677419355, "drop": 0.017379032258064475, "rotation": {"subspace_principal_angle": 1.4673214972757795, "mean_class_cosine": 0.10329027681844688}}}, "iid_acc": 0.65125, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.5077800700720831, "mean_class_cosine": 0.06297455823321907}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.535483870967742, "drop": 0.005766129032258038, "rotation": {"subspace_principal_angle": 1.440925356190957, "mean_class_cosine": 0.9961212841602167}}}, "iid_acc": 0.54125, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3284144435763523, "mean_class_cosine": 0.9886091312692893}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 4, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6209677419354839, "drop": 0.03403225806451615}}, "iid_acc": 0.655, "selectivity": 0.17000000000000004} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.750402576489533, "drop": -0.009152576489533049, "rotation": {"subspace_principal_angle": 1.4211901787085208, "mean_class_cosine": 0.14904869151211453}}}, "iid_acc": 0.74125, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.4965103140884382, "mean_class_cosine": 0.07421770808908856}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5265700483091788, "drop": 0.03967995169082128, "rotation": {"subspace_principal_angle": 1.5565645508048995, "mean_class_cosine": 0.00760410778674959}}}, "iid_acc": 0.56625, "selectivity": 0.0025000000000000577, "iid_split_rotation": {"subspace_principal_angle": 0.9594487263253761, "mean_class_cosine": -0.4124908154959152}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7536231884057971, "drop": -0.0011231884057971708}}, "iid_acc": 0.7525, "selectivity": 0.18874999999999997} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.8998628257887518, "drop": 0.02138717421124825, "rotation": {"subspace_principal_angle": 1.2563517152731103, "mean_class_cosine": 0.30928836211834454}}}, "iid_acc": 0.92125, "selectivity": 0.43375, "iid_split_rotation": {"subspace_principal_angle": 1.1864007137524348, "mean_class_cosine": 0.3749989233038349}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.545, "drop": 0.0}, "paraphrase": {"acc": 0.5569272976680384, "drop": -0.01192729766803835, "rotation": {"subspace_principal_angle": 0.6735562930427222, "mean_class_cosine": 0.3885547331126563}}}, "iid_acc": 0.545, "selectivity": 0.05750000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.3659301793739206, "mean_class_cosine": 0.9497848325707625}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 6, "dists": {"iid": {"acc": 0.91375, "drop": 0.0}, "paraphrase": {"acc": 0.9039780521262003, "drop": 0.009771947873799647}}, "iid_acc": 0.91375, "selectivity": 0.42624999999999996} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9965694682675815, "drop": -0.0015694682675815308, "rotation": {"subspace_principal_angle": 0.9204115550682747, "mean_class_cosine": 0.605492671467479}}}, "iid_acc": 0.995, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 0.964600661915144, "mean_class_cosine": 0.5697451063410308}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.7675, "drop": 0.0}, "paraphrase": {"acc": 0.7753001715265866, "drop": -0.00780017152658663, "rotation": {"subspace_principal_angle": 1.5561305399812533, "mean_class_cosine": 0.9880733179203292}}}, "iid_acc": 0.7675, "selectivity": -0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.8575713187579257, "mean_class_cosine": 0.9958656438305629}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": 0.0006110634648370805}}, "iid_acc": 0.99375, "selectivity": 0.19125000000000003} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6478260869565218, "drop": 0.0096739130434782, "rotation": {"subspace_principal_angle": 1.5203803161832818, "mean_class_cosine": 0.050394655640553065}}}, "iid_acc": 0.6575, "selectivity": 0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.4388032016199581, "mean_class_cosine": 0.13161019079523278}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4623188405797101, "drop": 0.040181159420289825, "rotation": {"subspace_principal_angle": 1.5601182686089141, "mean_class_cosine": -0.9599175649315286}}}, "iid_acc": 0.5025, "selectivity": -0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3056048703674055, "mean_class_cosine": 0.9664617957303832}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 9, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.6956521739130435, "drop": 0.018097826086956537}}, "iid_acc": 0.71375, "selectivity": 0.15500000000000003} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6477272727272727, "drop": -0.008304195804195835, "rotation": {"subspace_principal_angle": 1.4577549081283396, "mean_class_cosine": 0.13478050883475684}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.24999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.5438885641755995, "mean_class_cosine": 0.12661575417990478}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.40865384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.3693181818181818, "drop": 0.03933566433566432, "rotation": {"subspace_principal_angle": 1.3024704772930913, "mean_class_cosine": 0.751592937345538}}}, "iid_acc": 0.40865384615384615, "selectivity": 0.019230769230769218, "iid_split_rotation": {"subspace_principal_angle": 1.0672377902275059, "mean_class_cosine": 0.7053655504479837}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": 0.0021853146853146876}}, "iid_acc": 0.6442307692307693, "selectivity": 0.25480769230769235} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8831908831908832, "drop": 0.0030591168091167464, "rotation": {"subspace_principal_angle": 1.4868944039444256, "mean_class_cosine": 0.08380351910501391}}}, "iid_acc": 0.88625, "selectivity": 0.20999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3010782735174253, "mean_class_cosine": 0.26645969404516406}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.575, "drop": 0.0}, "paraphrase": {"acc": 0.5797720797720798, "drop": -0.004772079772079829, "rotation": {"subspace_principal_angle": 1.1062280403413047, "mean_class_cosine": 0.9715489933239789}}}, "iid_acc": 0.575, "selectivity": -0.10125000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.5004862855814347, "mean_class_cosine": 0.9429161811360605}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.9017094017094017, "drop": -0.0017094017094017033}}, "iid_acc": 0.9, "selectivity": 0.22375} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.8448753462603878, "drop": 0.010124653739612133, "rotation": {"subspace_principal_angle": 1.2793309538353963, "mean_class_cosine": 0.28735610225096786}}, "domain": {"acc": 0.835, "drop": 0.020000000000000018, "rotation": {"subspace_principal_angle": 1.4170648249450182, "mean_class_cosine": 0.15312668462263485}}}, "iid_acc": 0.855, "selectivity": 0.36374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2898741682012294, "mean_class_cosine": 0.27724177646677317}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5373961218836565, "drop": -0.017396121883656468, "rotation": {"subspace_principal_angle": 1.5469965306207487, "mean_class_cosine": 0.995777039241573}}, "domain": {"acc": 0.50125, "drop": 0.018750000000000044, "rotation": {"subspace_principal_angle": 1.4241429560997638, "mean_class_cosine": 0.3829223526550779}}}, "iid_acc": 0.52, "selectivity": 0.028749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9408750735467649, "mean_class_cosine": 0.9869080660276348}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8365650969529086, "drop": 0.04593490304709136}, "domain": {"acc": 0.68125, "drop": 0.20124999999999993}}, "iid_acc": 0.8825, "selectivity": 0.39124999999999993} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5241830065359477, "drop": -0.009183006535947724, "rotation": {"subspace_principal_angle": 1.534491721385843, "mean_class_cosine": 0.03629663087546754}}}, "iid_acc": 0.515, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.5355851741198854, "mean_class_cosine": 0.03520387717988177}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5215686274509804, "drop": -0.006568627450980413, "rotation": {"subspace_principal_angle": 1.4962498880827226, "mean_class_cosine": -0.003140814081629747}}}, "iid_acc": 0.515, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.3928150255004763, "mean_class_cosine": -0.22397488753911812}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 3, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.5098039215686274, "drop": -0.0035539215686274384}}, "iid_acc": 0.50625, "selectivity": 0.0012499999999999734} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.6625, "drop": 0.0}, "paraphrase": {"acc": 0.6351531291611185, "drop": 0.027346870838881432, "rotation": {"subspace_principal_angle": 1.3792866566009878, "mean_class_cosine": 0.32365198757370595}}}, "iid_acc": 0.6625, "selectivity": 0.41625, "iid_split_rotation": {"subspace_principal_angle": 1.3919845995237112, "mean_class_cosine": 0.34481299862959075}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.30625, "drop": 0.0}, "paraphrase": {"acc": 0.2969374167776298, "drop": 0.009312583222370219, "rotation": {"subspace_principal_angle": 1.1990223569073335, "mean_class_cosine": 0.6137390921956974}}}, "iid_acc": 0.30625, "selectivity": 0.060000000000000026, "iid_split_rotation": {"subspace_principal_angle": 1.4316896651831825, "mean_class_cosine": 0.6279344942694856}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.64375, "drop": 0.0}, "paraphrase": {"acc": 0.6378162450066578, "drop": 0.005933754993342255}}, "iid_acc": 0.64375, "selectivity": 0.3975000000000001} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.7028862478777589, "drop": -0.012886247877758938, "rotation": {"subspace_principal_angle": 1.4427027917966888, "mean_class_cosine": 0.12774353013272075}}}, "iid_acc": 0.69, "selectivity": 0.1612499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3294573165238053, "mean_class_cosine": 0.23900304446924894}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5517826825127334, "drop": -0.0017826825127333912, "rotation": {"subspace_principal_angle": 0.8282909656023721, "mean_class_cosine": 0.9050477286017182}}}, "iid_acc": 0.55, "selectivity": 0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5117065619049865, "mean_class_cosine": 0.9515453007674062}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7623089983022071, "drop": -0.008558998302207077}}, "iid_acc": 0.75375, "selectivity": 0.22499999999999998} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5887096774193549, "drop": 0.015040322580645138, "rotation": {"subspace_principal_angle": 1.5014698349556506, "mean_class_cosine": 0.06927097278538774}}}, "iid_acc": 0.60375, "selectivity": 0.08875, "iid_split_rotation": {"subspace_principal_angle": 1.505561040213484, "mean_class_cosine": 0.0651890267481754}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5419354838709678, "drop": -0.0006854838709677846, "rotation": {"subspace_principal_angle": 1.2893002371450315, "mean_class_cosine": 0.9894024792146818}}}, "iid_acc": 0.54125, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.295158193254292, "mean_class_cosine": 0.9686636169079809}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.6675, "drop": 0.0}, "paraphrase": {"acc": 0.6580645161290323, "drop": 0.009435483870967709}}, "iid_acc": 0.6675, "selectivity": 0.15249999999999997} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7053140096618358, "drop": 0.019685990338164205, "rotation": {"subspace_principal_angle": 1.48218154327834, "mean_class_cosine": 0.08849885326614075}}}, "iid_acc": 0.725, "selectivity": 0.18874999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4683469244564828, "mean_class_cosine": 0.10227028026027415}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.5942028985507246, "drop": 0.0007971014492753614, "rotation": {"subspace_principal_angle": 1.4158562736410767, "mean_class_cosine": 0.12335819756367886}}}, "iid_acc": 0.595, "selectivity": 0.05874999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6688260501385498, "mean_class_cosine": 0.15014143544853242}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7568438003220612, "drop": 0.015656199677938742}}, "iid_acc": 0.7725, "selectivity": 0.23624999999999996} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9259259259259259, "drop": 0.016574074074074074, "rotation": {"subspace_principal_angle": 1.157452401071877, "mean_class_cosine": 0.4016738775889428}}}, "iid_acc": 0.9425, "selectivity": 0.42000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.104624499684144, "mean_class_cosine": 0.4494698976663737}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.585, "drop": 0.0}, "paraphrase": {"acc": 0.5624142661179699, "drop": 0.022585733882030112, "rotation": {"subspace_principal_angle": 1.2005636922035423, "mean_class_cosine": 0.8139615746809903}}}, "iid_acc": 0.585, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 0.549247413832069, "mean_class_cosine": 0.30809866212710807}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 24, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9272976680384087, "drop": 0.003952331961591282}}, "iid_acc": 0.93125, "selectivity": 0.40875000000000006} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.6623188405797101, "drop": 0.0339311594202899, "rotation": {"subspace_principal_angle": 1.496893033147877, "mean_class_cosine": 0.07383603911862277}}}, "iid_acc": 0.69625, "selectivity": 0.11625000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4214587189180432, "mean_class_cosine": 0.1487831455638894}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.51, "drop": 0.0}, "paraphrase": {"acc": 0.4797101449275362, "drop": 0.03028985507246379, "rotation": {"subspace_principal_angle": 1.1094284833982664, "mean_class_cosine": -0.7834732078378617}}}, "iid_acc": 0.51, "selectivity": -0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2652565110598484, "mean_class_cosine": 0.8601341288733604}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 22, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.6420289855072464, "drop": 0.050471014492753596}}, "iid_acc": 0.6925, "selectivity": 0.11250000000000004} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6704545454545454, "drop": -0.02622377622377614, "rotation": {"subspace_principal_angle": 1.4296664682007603, "mean_class_cosine": 0.16574943658385013}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2163461538461539, "iid_split_rotation": {"subspace_principal_angle": 1.551374852609326, "mean_class_cosine": 0.11936510654614986}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.4230769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.4147727272727273, "drop": 0.00830419580419578, "rotation": {"subspace_principal_angle": 1.3518141128300416, "mean_class_cosine": 0.666584729764562}}}, "iid_acc": 0.4230769230769231, "selectivity": -0.004807692307692291, "iid_split_rotation": {"subspace_principal_angle": 1.4575770760056468, "mean_class_cosine": 0.7420250665679694}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6761363636363636, "drop": -0.00786713286713292}}, "iid_acc": 0.6682692307692307, "selectivity": 0.24038461538461536} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8846153846153846, "drop": 0.0066346153846154055, "rotation": {"subspace_principal_angle": 1.4268427703533795, "mean_class_cosine": 0.1434568887054392}}}, "iid_acc": 0.89125, "selectivity": 0.20499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.368747840402915, "mean_class_cosine": 0.20067656560132463}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5826210826210826, "drop": -0.0026210826210826266, "rotation": {"subspace_principal_angle": 0.9151023657985915, "mean_class_cosine": 0.9645849728606082}}}, "iid_acc": 0.58, "selectivity": -0.10625000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.47399007231295404, "mean_class_cosine": 0.9356570924866734}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.9002849002849003, "drop": 0.0009650997150997265}}, "iid_acc": 0.90125, "selectivity": 0.21499999999999997} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8490304709141274, "drop": 0.024719529085872605, "rotation": {"subspace_principal_angle": 1.176281709220393, "mean_class_cosine": 0.384360137185048}}, "domain": {"acc": 0.73625, "drop": 0.13750000000000007, "rotation": {"subspace_principal_angle": 1.1970353549367978, "mean_class_cosine": 0.36511932310217704}}}, "iid_acc": 0.87375, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.0786643668776719, "mean_class_cosine": 0.4725059154778732}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5304709141274239, "drop": -0.004220914127423869, "rotation": {"subspace_principal_angle": 1.3152812819817528, "mean_class_cosine": 0.9992665145477913}}, "domain": {"acc": 0.50125, "drop": 0.025000000000000022, "rotation": {"subspace_principal_angle": 1.308682431264261, "mean_class_cosine": 0.5566033754658513}}}, "iid_acc": 0.52625, "selectivity": 0.04249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.391200206971303, "mean_class_cosine": 0.9964068683023172}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8130193905817175, "drop": 0.018230609418282584}, "domain": {"acc": 0.82875, "drop": 0.0025000000000000577}}, "iid_acc": 0.83125, "selectivity": 0.34750000000000003} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8643410852713178, "drop": 0.023158914728682145, "rotation": {"subspace_principal_angle": 1.1410222217173775, "mean_class_cosine": 0.4166654610472472}}, "domain": {"acc": 0.815, "drop": 0.07250000000000001, "rotation": {"subspace_principal_angle": 1.250009657687724, "mean_class_cosine": 0.3153131973834549}}, "length": {"acc": 0.8878787878787879, "drop": -0.00037878787878797837, "rotation": {"subspace_principal_angle": 1.0751426744599302, "mean_class_cosine": 0.4756067433505382}}}, "iid_acc": 0.8875, "selectivity": 0.42999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.080452581028767, "mean_class_cosine": 0.4709291585448011}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7364341085271318, "drop": -0.0014341085271317722, "rotation": {"subspace_principal_angle": 1.133371334962397, "mean_class_cosine": 0.8868509368956804}}, "domain": {"acc": 0.5575, "drop": 0.1775, "rotation": {"subspace_principal_angle": 1.4962557115172213, "mean_class_cosine": 0.27127763310435205}}, "length": {"acc": 0.7651515151515151, "drop": -0.03015151515151515, "rotation": {"subspace_principal_angle": 1.3646255328037025, "mean_class_cosine": 0.9602447183571129}}}, "iid_acc": 0.735, "selectivity": 0.27749999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.9276681297932009, "mean_class_cosine": 0.8062680171943362}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 0, "layer": 14, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8798449612403101, "drop": 0.01765503875968988}, "domain": {"acc": 0.69625, "drop": 0.20124999999999993}, "length": {"acc": 0.9, "drop": -0.0025000000000000577}}, "iid_acc": 0.8975, "selectivity": 0.43999999999999995} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8738461538461538, "drop": 0.004903846153846203, "rotation": {"subspace_principal_angle": 1.3890007649660276, "mean_class_cosine": 0.284822099506686}}}, "iid_acc": 0.87875, "selectivity": 0.60625, "iid_split_rotation": {"subspace_principal_angle": 1.5702965678160945, "mean_class_cosine": 0.3020514335534847}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.5661538461538461, "drop": -0.009903846153846096, "rotation": {"subspace_principal_angle": 1.553740809284586, "mean_class_cosine": 0.9736297672635466}}}, "iid_acc": 0.55625, "selectivity": 0.28375, "iid_split_rotation": {"subspace_principal_angle": 1.4574124850080719, "mean_class_cosine": 0.9616324023244914}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8861538461538462, "drop": -0.001153846153846172}}, "iid_acc": 0.885, "selectivity": 0.6125} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9476309226932669, "drop": 0.02361907730673307, "rotation": {"subspace_principal_angle": 1.0954764445356109, "mean_class_cosine": 0.7159555553963199}}}, "iid_acc": 0.97125, "selectivity": 0.8775, "iid_split_rotation": {"subspace_principal_angle": 1.0349871771004895, "mean_class_cosine": 0.757327529417333}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.26375, "drop": 0.0}, "paraphrase": {"acc": 0.29177057356608477, "drop": -0.028020573566084783, "rotation": {"subspace_principal_angle": 1.4115196982796947, "mean_class_cosine": 0.8384434499541353}}}, "iid_acc": 0.26375, "selectivity": 0.16999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2413128275371268, "mean_class_cosine": 0.7859751886800755}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 0, "layer": 16, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.912718204488778, "drop": 0.03603179551122193}}, "iid_acc": 0.94875, "selectivity": 0.855} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.5882352941176471, "drop": 0.02926470588235297, "rotation": {"subspace_principal_angle": 1.360311035504444, "mean_class_cosine": 0.20893450516523}}}, "iid_acc": 0.6175, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4063133781244486, "mean_class_cosine": 0.1637422801353008}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.4925, "drop": 0.0}, "paraphrase": {"acc": 0.4980392156862745, "drop": -0.005539215686274523, "rotation": {"subspace_principal_angle": 1.441946064731949, "mean_class_cosine": 0.6865636818700251}}}, "iid_acc": 0.4925, "selectivity": -0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5463324868354524, "mean_class_cosine": -0.9630196496317205}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5581699346405229, "drop": -0.009419934640522931}}, "iid_acc": 0.54875, "selectivity": 0.012499999999999956} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6231691078561917, "drop": 0.02183089214380829, "rotation": {"subspace_principal_angle": 1.25177621884401, "mean_class_cosine": 0.5310790447646092}}}, "iid_acc": 0.645, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2608266828416124, "mean_class_cosine": 0.45773127447897305}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.1825, "drop": 0.0}, "paraphrase": {"acc": 0.17709720372836218, "drop": 0.005402796271637811, "rotation": {"subspace_principal_angle": 1.4219864569031055, "mean_class_cosine": 0.3343326799788812}}}, "iid_acc": 0.1825, "selectivity": -0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4213554573473413, "mean_class_cosine": 0.6639945990861887}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 0, "layer": 7, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.521970705725699, "drop": 0.00427929427430096}}, "iid_acc": 0.52625, "selectivity": 0.27} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7707979626485568, "drop": -0.01704796264855679, "rotation": {"subspace_principal_angle": 1.3829256711520628, "mean_class_cosine": 0.18676744322680136}}}, "iid_acc": 0.75375, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3348303902625251, "mean_class_cosine": 0.23378226387175424}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.5212224108658744, "drop": -0.00497241086587441, "rotation": {"subspace_principal_angle": 1.478236251419889, "mean_class_cosine": 0.9994667861891016}}}, "iid_acc": 0.51625, "selectivity": -0.0050000000000000044, "iid_split_rotation": {"subspace_principal_angle": 1.3448819076659233, "mean_class_cosine": 0.9996533797319738}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 0, "layer": 8, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6960950764006791, "drop": 0.008904923599320824}}, "iid_acc": 0.705, "selectivity": 0.18374999999999997} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67625, "drop": 0.0}, "paraphrase": {"acc": 0.632258064516129, "drop": 0.043991935483871036, "rotation": {"subspace_principal_angle": 1.4188129223223702, "mean_class_cosine": 0.15139897022933}}}, "iid_acc": 0.67625, "selectivity": 0.1875, "iid_split_rotation": {"subspace_principal_angle": 1.381661689857851, "mean_class_cosine": 0.18800903417603418}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5338709677419354, "drop": 0.002379032258064573, "rotation": {"subspace_principal_angle": 1.2619016165915666, "mean_class_cosine": 0.9985457827336248}}}, "iid_acc": 0.53625, "selectivity": 0.04749999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2970420520156083, "mean_class_cosine": 0.9948538875393387}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.67125, "drop": 0.0}, "paraphrase": {"acc": 0.635483870967742, "drop": 0.035766129032258065}}, "iid_acc": 0.67125, "selectivity": 0.1825} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7681159420289855, "drop": -0.008115942028985468, "rotation": {"subspace_principal_angle": 1.310808643208806, "mean_class_cosine": 0.2570686494136881}}}, "iid_acc": 0.76, "selectivity": 0.15375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3820982126149526, "mean_class_cosine": 0.18758027789963122}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.47125, "drop": 0.0}, "paraphrase": {"acc": 0.49597423510466987, "drop": -0.024724235104669867, "rotation": {"subspace_principal_angle": 1.452719223490081, "mean_class_cosine": 0.2128800431912587}}}, "iid_acc": 0.47125, "selectivity": -0.13499999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.9161871593522571, "mean_class_cosine": -0.32128070846271867}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 0, "layer": 5, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6843800322061192, "drop": 0.010619967793880747}}, "iid_acc": 0.695, "selectivity": 0.08875} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.94, "drop": 0.0}, "paraphrase": {"acc": 0.9286694101508917, "drop": 0.011330589849108286, "rotation": {"subspace_principal_angle": 1.137125759529054, "mean_class_cosine": 0.4202044066034193}}}, "iid_acc": 0.94, "selectivity": 0.47624999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.0643181542191005, "mean_class_cosine": 0.4851005702018669}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5253772290809328, "drop": 0.008372770919067185, "rotation": {"subspace_principal_angle": 0.516621762104431, "mean_class_cosine": 0.21743596239076726}}}, "iid_acc": 0.53375, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 0.8840737623848459, "mean_class_cosine": 0.976674523046531}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 0, "layer": 13, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9368998628257887, "drop": 0.0006001371742112571}}, "iid_acc": 0.9375, "selectivity": 0.47375} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9931389365351629, "drop": -0.0018889365351629772, "rotation": {"subspace_principal_angle": 0.6634680250005783, "mean_class_cosine": 0.7878611804971933}}}, "iid_acc": 0.99125, "selectivity": 0.11874999999999991, "iid_split_rotation": {"subspace_principal_angle": 0.6348168891373653, "mean_class_cosine": 0.8051803002223438}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9777015437392796, "drop": -0.006451543739279697, "rotation": {"subspace_principal_angle": 1.3391381172924783, "mean_class_cosine": 0.9776428455730666}}}, "iid_acc": 0.97125, "selectivity": 0.0987499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4227835246448142, "mean_class_cosine": 0.9908198389945848}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 0, "layer": 2, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9897084048027445, "drop": 0.0002915951972555231}}, "iid_acc": 0.99, "selectivity": 0.11749999999999994} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7, "drop": 0.02750000000000008, "rotation": {"subspace_principal_angle": 1.472522322965299, "mean_class_cosine": 0.09811589541574661}}}, "iid_acc": 0.7275, "selectivity": 0.12125000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.3920524003697496, "mean_class_cosine": 0.17779365245715506}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.4826086956521739, "drop": 0.03989130434782606, "rotation": {"subspace_principal_angle": 1.4319943336187273, "mean_class_cosine": -0.9921843120457909}}}, "iid_acc": 0.5225, "selectivity": -0.08374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.0962721333543264, "mean_class_cosine": 0.9947347244035363}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 0, "layer": 11, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.717391304347826, "drop": -0.0036413043478260576}}, "iid_acc": 0.71375, "selectivity": 0.10750000000000004} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.6153846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.6136363636363636, "drop": 0.0017482517482517723, "rotation": {"subspace_principal_angle": 1.4739338903736143, "mean_class_cosine": 0.13152150958787132}}}, "iid_acc": 0.6153846153846154, "selectivity": 0.17307692307692313, "iid_split_rotation": {"subspace_principal_angle": 1.45720032842477, "mean_class_cosine": 0.19720918409507146}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.41346153846153844, "drop": 0.0}, "paraphrase": {"acc": 0.4147727272727273, "drop": -0.001311188811188857, "rotation": {"subspace_principal_angle": 1.550847759851561, "mean_class_cosine": 0.7251546533887714}}}, "iid_acc": 0.41346153846153844, "selectivity": -0.028846153846153855, "iid_split_rotation": {"subspace_principal_angle": 0.8940696505827914, "mean_class_cosine": 0.8831178368398586}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 0, "layer": 20, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6420454545454546, "drop": -0.017045454545454586}}, "iid_acc": 0.625, "selectivity": 0.1826923076923077} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.8888888888888888, "drop": 0.0011111111111111738, "rotation": {"subspace_principal_angle": 1.346328125428775, "mean_class_cosine": 0.22258793643339736}}}, "iid_acc": 0.89, "selectivity": 0.135, "iid_split_rotation": {"subspace_principal_angle": 1.2354156575580886, "mean_class_cosine": 0.3291286539962054}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5598290598290598, "drop": 0.007670940170940166, "rotation": {"subspace_principal_angle": 1.089834172363605, "mean_class_cosine": 0.9822532101639228}}}, "iid_acc": 0.5675, "selectivity": -0.1875, "iid_split_rotation": {"subspace_principal_angle": 0.9991972798605081, "mean_class_cosine": 0.8996231510474297}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 0, "layer": 12, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8532763532763533, "drop": 0.00672364672364667}}, "iid_acc": 0.86, "selectivity": 0.10499999999999998} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7649513212795549, "drop": -0.0074513212795549455, "rotation": {"subspace_principal_angle": 1.260155827346341, "mean_class_cosine": 0.30566854295453516}}, "domain": {"acc": 0.72625, "drop": 0.03125, "rotation": {"subspace_principal_angle": 1.2669695481971335, "mean_class_cosine": 0.2991738951377597}}}, "iid_acc": 0.7575, "selectivity": 0.24874999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3245794968247167, "mean_class_cosine": 0.24373663709592575}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5521557719054242, "drop": -0.017155771905424166, "rotation": {"subspace_principal_angle": 1.4523160206390673, "mean_class_cosine": 0.9818329782722122}}, "domain": {"acc": 0.495, "drop": 0.040000000000000036, "rotation": {"subspace_principal_angle": 1.4451431888735655, "mean_class_cosine": 0.5803352104264222}}}, "iid_acc": 0.535, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3059026276224222, "mean_class_cosine": 0.9628009093282439}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.741307371349096, "drop": -0.008807371349095927}, "domain": {"acc": 0.75125, "drop": -0.018749999999999933}}, "iid_acc": 0.7325, "selectivity": 0.22375} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.8020408163265306, "drop": 0.015459183673469434, "rotation": {"subspace_principal_angle": 0.9261920049968357, "mean_class_cosine": 0.6008822051011515}}, "domain": {"acc": 0.68375, "drop": 0.13375000000000004, "rotation": {"subspace_principal_angle": 1.1716370438461978, "mean_class_cosine": 0.38864385309548416}}, "length": {"acc": 0.8115501519756839, "drop": 0.0059498480243160845, "rotation": {"subspace_principal_angle": 0.8775406092988467, "mean_class_cosine": 0.6390447633833902}}}, "iid_acc": 0.8175, "selectivity": 0.31625000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.957610160630722, "mean_class_cosine": 0.5754760826890111}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6877551020408164, "drop": 0.0034948979591836737, "rotation": {"subspace_principal_angle": 1.2806469825658773, "mean_class_cosine": 0.8910748524092273}}, "domain": {"acc": 0.56875, "drop": 0.12250000000000005, "rotation": {"subspace_principal_angle": 1.438815881455066, "mean_class_cosine": 0.41273195854934175}}, "length": {"acc": 0.6975683890577508, "drop": -0.0063183890577507285, "rotation": {"subspace_principal_angle": 1.5481828698411682, "mean_class_cosine": 0.9192162272707554}}}, "iid_acc": 0.69125, "selectivity": 0.19000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3565470315186383, "mean_class_cosine": 0.8817941693849485}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.78875, "drop": 0.0}, "paraphrase": {"acc": 0.7755102040816326, "drop": 0.013239795918367325}, "domain": {"acc": 0.69125, "drop": 0.09749999999999992}, "length": {"acc": 0.7781155015197568, "drop": 0.010634498480243115}}, "iid_acc": 0.78875, "selectivity": 0.2875} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8454258675078864, "drop": 0.008324132492113612, "rotation": {"subspace_principal_angle": 1.4872713948684544, "mean_class_cosine": 0.5396680342819231}}}, "iid_acc": 0.85375, "selectivity": 0.62375, "iid_split_rotation": {"subspace_principal_angle": 1.1001196196694507, "mean_class_cosine": 0.5629295021677052}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.78, "drop": 0.0}, "paraphrase": {"acc": 0.7744479495268138, "drop": 0.0055520504731861875, "rotation": {"subspace_principal_angle": 1.3957068228612415, "mean_class_cosine": 0.9585242100327607}}}, "iid_acc": 0.78, "selectivity": 0.55, "iid_split_rotation": {"subspace_principal_angle": 1.0534755137152318, "mean_class_cosine": 0.9472616950218877}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8249211356466877, "drop": 0.006328864353312369}}, "iid_acc": 0.83125, "selectivity": 0.6012500000000001} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.9177377892030848, "drop": 0.0022622107969152427, "rotation": {"subspace_principal_angle": 1.5481489232177876, "mean_class_cosine": 0.7602277235959066}}}, "iid_acc": 0.92, "selectivity": 0.8500000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.0913499627869498, "mean_class_cosine": 0.8064226222572987}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.81375, "drop": 0.0}, "paraphrase": {"acc": 0.7583547557840618, "drop": 0.05539524421593822, "rotation": {"subspace_principal_angle": 1.434488039184557, "mean_class_cosine": 0.8699474461477923}}}, "iid_acc": 0.81375, "selectivity": 0.7437499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4670162663468989, "mean_class_cosine": 0.9406629826516945}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8174807197943444, "drop": 0.03626928020565556}}, "iid_acc": 0.85375, "selectivity": 0.78375} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.53875, "drop": 0.0}, "paraphrase": {"acc": 0.5297157622739018, "drop": 0.009034237726098127, "rotation": {"subspace_principal_angle": 1.506602262713435, "mean_class_cosine": 0.06414998384863341}}}, "iid_acc": 0.53875, "selectivity": 0.062499999999999944, "iid_split_rotation": {"subspace_principal_angle": 1.5141486385951506, "mean_class_cosine": 0.056617396360705646}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.47674418604651164, "drop": 0.022005813953488385, "rotation": {"subspace_principal_angle": 1.5018907330856117, "mean_class_cosine": 0.3385630136055}}}, "iid_acc": 0.49875, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5223535908671477, "mean_class_cosine": 0.3337844099038709}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5335917312661499, "drop": 0.017658268733850124}}, "iid_acc": 0.55125, "selectivity": 0.07500000000000001} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5115020297699594, "drop": 0.028497970230040637, "rotation": {"subspace_principal_angle": 1.3837974822380827, "mean_class_cosine": 0.2853919153689364}}}, "iid_acc": 0.54, "selectivity": 0.3175, "iid_split_rotation": {"subspace_principal_angle": 1.4430652239408488, "mean_class_cosine": 0.32302299850041155}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.2075, "drop": 0.0}, "paraphrase": {"acc": 0.21380243572395127, "drop": -0.006302435723951283, "rotation": {"subspace_principal_angle": 1.4121359208224795, "mean_class_cosine": 0.761742118163654}}}, "iid_acc": 0.2075, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4351405486147384, "mean_class_cosine": 0.6192669179592563}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.4438430311231394, "drop": 0.03115696887686059}}, "iid_acc": 0.475, "selectivity": 0.25249999999999995} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7043918918918919, "drop": 0.021858108108108065, "rotation": {"subspace_principal_angle": 1.2557206455227148, "mean_class_cosine": 0.3098884278198152}}}, "iid_acc": 0.72625, "selectivity": 0.20999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2635908058672536, "mean_class_cosine": 0.30239617255824447}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5506756756756757, "drop": -0.02317567567567569, "rotation": {"subspace_principal_angle": 1.563519056112211, "mean_class_cosine": 0.9678685291584537}}}, "iid_acc": 0.5275, "selectivity": 0.011249999999999982, "iid_split_rotation": {"subspace_principal_angle": 1.4948991382902788, "mean_class_cosine": 0.9835336511500095}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.70625, "drop": 0.0}, "paraphrase": {"acc": 0.7010135135135135, "drop": 0.005236486486486558}}, "iid_acc": 0.70625, "selectivity": 0.19000000000000006} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6715447154471544, "drop": 0.01595528455284556, "rotation": {"subspace_principal_angle": 1.3527346728817384, "mean_class_cosine": 0.21633758729886746}}}, "iid_acc": 0.6875, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.396125798434941, "mean_class_cosine": 0.17378368868184255}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.59125, "drop": 0.0}, "paraphrase": {"acc": 0.583739837398374, "drop": 0.007510162601626069, "rotation": {"subspace_principal_angle": 1.533612671019165, "mean_class_cosine": 0.9008068153198737}}}, "iid_acc": 0.59125, "selectivity": 0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.260460650130829, "mean_class_cosine": 0.890537070105323}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6520325203252032, "drop": 0.007967479674796829}}, "iid_acc": 0.66, "selectivity": 0.19125000000000003} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7079934747145188, "drop": 0.013256525285481158, "rotation": {"subspace_principal_angle": 1.4022375641064082, "mean_class_cosine": 0.16776171230922843}}}, "iid_acc": 0.72125, "selectivity": 0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4265826075864037, "mean_class_cosine": 0.14371435563850823}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5432300163132137, "drop": -0.005730016313213682, "rotation": {"subspace_principal_angle": 1.1709151024773818, "mean_class_cosine": 0.6775048474955148}}}, "iid_acc": 0.5375, "selectivity": -0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1992395749899722, "mean_class_cosine": 0.7778731026109806}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7259380097879282, "drop": 0.0003119902120717333}}, "iid_acc": 0.72625, "selectivity": 0.14749999999999996} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8446866485013624, "drop": 0.041563351498637546, "rotation": {"subspace_principal_angle": 0.9763955114091522, "mean_class_cosine": 0.5600124400720721}}}, "iid_acc": 0.88625, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 0.901229402467838, "mean_class_cosine": 0.620646474717384}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.84, "drop": 0.0}, "paraphrase": {"acc": 0.8079019073569482, "drop": 0.032098092643051745, "rotation": {"subspace_principal_angle": 1.5194268659444015, "mean_class_cosine": 0.9313227017091655}}}, "iid_acc": 0.84, "selectivity": 0.36, "iid_split_rotation": {"subspace_principal_angle": 1.4080821249653852, "mean_class_cosine": 0.9589713457056335}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8337874659400545, "drop": 0.037462534059945485}}, "iid_acc": 0.87125, "selectivity": 0.39125} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9818481848184818, "drop": 0.006901815181518223, "rotation": {"subspace_principal_angle": 0.66528493719011, "mean_class_cosine": 0.7867409300908236}}}, "iid_acc": 0.98875, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6314500555079575, "mean_class_cosine": 0.8071723665064887}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8861386138613861, "drop": 0.0026113861386138915, "rotation": {"subspace_principal_angle": 1.4453777964256416, "mean_class_cosine": 0.9697563041315995}}}, "iid_acc": 0.88875, "selectivity": 0.026249999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9974066490876098, "mean_class_cosine": 0.9768283657309265}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98, "drop": 0.0}, "paraphrase": {"acc": 0.9752475247524752, "drop": 0.004752475247524757}}, "iid_acc": 0.98, "selectivity": 0.11749999999999994} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.703328509406657, "drop": -0.013328509406657085, "rotation": {"subspace_principal_angle": 1.4989704535588646, "mean_class_cosine": 0.07176413141027983}}}, "iid_acc": 0.69, "selectivity": 0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.5357515116941332, "mean_class_cosine": 0.03503764222349777}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.54375, "drop": 0.0}, "paraphrase": {"acc": 0.5470332850940666, "drop": -0.0032832850940666125, "rotation": {"subspace_principal_angle": 1.5166062020639395, "mean_class_cosine": 0.39412823203921477}}}, "iid_acc": 0.54375, "selectivity": -0.14, "iid_split_rotation": {"subspace_principal_angle": 1.5290148158529324, "mean_class_cosine": -0.0372845302255322}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": -0.009247467438494916}}, "iid_acc": 0.71, "selectivity": 0.026249999999999996} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5961538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.6647058823529411, "drop": -0.068552036199095, "rotation": {"subspace_principal_angle": 1.4788344627937355, "mean_class_cosine": 0.29399670434967345}}}, "iid_acc": 0.5961538461538461, "selectivity": 0.15384615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.3726634159821796, "mean_class_cosine": 0.2274413627536632}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5048076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": -0.03636877828054297, "rotation": {"subspace_principal_angle": 0.913621189609579, "mean_class_cosine": 0.6670655608930605}}}, "iid_acc": 0.5048076923076923, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.0837588978590438, "mean_class_cosine": 0.5902856300722443}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6009615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6235294117647059, "drop": -0.02256787330316745}}, "iid_acc": 0.6009615384615384, "selectivity": 0.15865384615384615} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": -0.0021985190409027355, "rotation": {"subspace_principal_angle": 0.9713008818157643, "mean_class_cosine": 0.564225974312156}}}, "iid_acc": 0.89625, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 0.9979017449204138, "mean_class_cosine": 0.5420667359536389}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7362482369534555, "drop": -0.009998236953455586, "rotation": {"subspace_principal_angle": 1.3253351141042944, "mean_class_cosine": 0.8910372103240486}}}, "iid_acc": 0.72625, "selectivity": -0.030000000000000027, "iid_split_rotation": {"subspace_principal_angle": 1.5153708369050465, "mean_class_cosine": 0.9019435549852257}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.785, "drop": 0.0}, "paraphrase": {"acc": 0.7870239774330042, "drop": -0.002023977433004198}}, "iid_acc": 0.785, "selectivity": 0.028750000000000053} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.8, "drop": 0.0}, "paraphrase": {"acc": 0.8052851182197497, "drop": -0.0052851182197496405, "rotation": {"subspace_principal_angle": 1.289669049843915, "mean_class_cosine": 0.2774388483889756}}, "domain": {"acc": 0.71, "drop": 0.09000000000000008, "rotation": {"subspace_principal_angle": 1.2978069882976488, "mean_class_cosine": 0.2696112780665089}}}, "iid_acc": 0.8, "selectivity": 0.2875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2350581884793985, "mean_class_cosine": 0.32946618568105485}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5549374130737135, "drop": 0.005062586926286583, "rotation": {"subspace_principal_angle": 1.3268922292489094, "mean_class_cosine": 0.9883969165833375}}, "domain": {"acc": 0.495, "drop": 0.06500000000000006, "rotation": {"subspace_principal_angle": 1.131095069142722, "mean_class_cosine": 0.7133167140261951}}}, "iid_acc": 0.56, "selectivity": 0.0475000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2611138196752891, "mean_class_cosine": 0.9895181201439}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.815, "drop": 0.0}, "paraphrase": {"acc": 0.780250347705146, "drop": 0.03474965229485394}, "domain": {"acc": 0.76625, "drop": 0.04874999999999996}}, "iid_acc": 0.815, "selectivity": 0.3025} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.0027295918367347127, "rotation": {"subspace_principal_angle": 1.1073395630387064, "mean_class_cosine": 0.4470428901807506}}, "domain": {"acc": 0.71375, "drop": 0.14, "rotation": {"subspace_principal_angle": 1.310897035728456, "mean_class_cosine": 0.2569832264845267}}, "length": {"acc": 0.8556231003039514, "drop": -0.0018731003039513805, "rotation": {"subspace_principal_angle": 1.106355203321873, "mean_class_cosine": 0.44792319554846216}}}, "iid_acc": 0.85375, "selectivity": 0.3375, "iid_split_rotation": {"subspace_principal_angle": 1.1106698792892344, "mean_class_cosine": 0.4440614070193354}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6795918367346939, "drop": 0.002908163265306074, "rotation": {"subspace_principal_angle": 1.0125111213870783, "mean_class_cosine": 0.9329519454684184}}, "domain": {"acc": 0.56625, "drop": 0.11624999999999996, "rotation": {"subspace_principal_angle": 1.1283485572455334, "mean_class_cosine": 0.5142940873907857}}, "length": {"acc": 0.6960486322188449, "drop": -0.013548632218844947, "rotation": {"subspace_principal_angle": 1.3998779208397523, "mean_class_cosine": 0.9308749733569011}}}, "iid_acc": 0.6825, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 0.8902790332464676, "mean_class_cosine": 0.8762482918205301}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.000229591836734655}, "domain": {"acc": 0.7775, "drop": 0.07374999999999998}, "length": {"acc": 0.8541033434650456, "drop": -0.0028533434650456213}}, "iid_acc": 0.85125, "selectivity": 0.33499999999999996} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8470031545741324, "drop": 0.0154968454258676, "rotation": {"subspace_principal_angle": 1.3265156223459154, "mean_class_cosine": 0.357128289219289}}}, "iid_acc": 0.8625, "selectivity": 0.6375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.347641540682651, "mean_class_cosine": 0.38870803168480306}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6876971608832808, "drop": -0.00019716088328081138, "rotation": {"subspace_principal_angle": 1.49331931022075, "mean_class_cosine": 0.9421571273887641}}}, "iid_acc": 0.6875, "selectivity": 0.4625, "iid_split_rotation": {"subspace_principal_angle": 0.8842746339269641, "mean_class_cosine": 0.9544218315516564}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8517350157728707, "drop": 0.010764984227129348}}, "iid_acc": 0.8625, "selectivity": 0.6375000000000001} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9280205655526992, "drop": 0.01447943444730082, "rotation": {"subspace_principal_angle": 1.543953519865052, "mean_class_cosine": 0.6356132585262798}}}, "iid_acc": 0.9425, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.1199023344741124, "mean_class_cosine": 0.6818253733919893}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8791773778920309, "drop": -0.007927377892030907, "rotation": {"subspace_principal_angle": 1.347264998699133, "mean_class_cosine": 0.909582314970496}}}, "iid_acc": 0.87125, "selectivity": 0.7975, "iid_split_rotation": {"subspace_principal_angle": 1.4779951280492787, "mean_class_cosine": 0.9618009806772534}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94375, "drop": 0.0}, "paraphrase": {"acc": 0.922879177377892, "drop": 0.020870822622107932}}, "iid_acc": 0.94375, "selectivity": 0.87} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.49483204134366926, "drop": 0.010167958656330744, "rotation": {"subspace_principal_angle": 1.524398385950912, "mean_class_cosine": -0.046381295294995115}}}, "iid_acc": 0.505, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5107404805094091, "mean_class_cosine": 0.06001975217823446}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.4909560723514212, "drop": 0.015293927648578787, "rotation": {"subspace_principal_angle": 1.5668349405587232, "mean_class_cosine": -0.03841414374121909}}}, "iid_acc": 0.50625, "selectivity": 0.023749999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.5352103349454524, "mean_class_cosine": -0.02088599099775313}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5167958656330749, "drop": 0.0007041343669250688}}, "iid_acc": 0.5175, "selectivity": 0.034999999999999976} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.5764546684709067, "drop": 0.026045331529093385, "rotation": {"subspace_principal_angle": 1.5428367399341, "mean_class_cosine": 0.37101745351040355}}}, "iid_acc": 0.6025, "selectivity": 0.33875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4285270805830252, "mean_class_cosine": 0.3347078246464954}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.21, "drop": 0.0}, "paraphrase": {"acc": 0.19079837618403248, "drop": 0.019201623815967517, "rotation": {"subspace_principal_angle": 1.4456365520979793, "mean_class_cosine": 0.6187678845759252}}}, "iid_acc": 0.21, "selectivity": -0.05374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5049088148460505, "mean_class_cosine": 0.472889652110318}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 2, "dists": {"iid": {"acc": 0.50625, "drop": 0.0}, "paraphrase": {"acc": 0.496617050067659, "drop": 0.009632949932340984}}, "iid_acc": 0.50625, "selectivity": 0.2425} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7331081081081081, "drop": 0.009391891891891935, "rotation": {"subspace_principal_angle": 1.3068079885037773, "mean_class_cosine": 0.2609327871606552}}}, "iid_acc": 0.7425, "selectivity": 0.23625000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.2782285316613433, "mean_class_cosine": 0.28841185363838734}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5929054054054054, "drop": -0.020405405405405364, "rotation": {"subspace_principal_angle": 1.4497936937308118, "mean_class_cosine": 0.9158504322082752}}}, "iid_acc": 0.5725, "selectivity": 0.06625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3799098692242713, "mean_class_cosine": 0.8877578773800847}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7364864864864865, "drop": -0.0027364864864865}}, "iid_acc": 0.73375, "selectivity": 0.22750000000000004} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.67875, "drop": 0.0}, "paraphrase": {"acc": 0.6764227642276422, "drop": 0.0023272357723577164, "rotation": {"subspace_principal_angle": 1.3819592931432023, "mean_class_cosine": 0.18771672963966035}}}, "iid_acc": 0.67875, "selectivity": 0.19374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.416763883086912, "mean_class_cosine": 0.15342407040448366}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.5788617886178862, "drop": -0.010111788617886197, "rotation": {"subspace_principal_angle": 0.8085931323629197, "mean_class_cosine": 0.9884789144089932}}}, "iid_acc": 0.56875, "selectivity": 0.08374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3601702468050045, "mean_class_cosine": 0.9887573457930535}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.6455284552845528, "drop": 0.03947154471544723}}, "iid_acc": 0.685, "selectivity": 0.20000000000000007} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.732463295269168, "drop": -0.003713295269167971, "rotation": {"subspace_principal_angle": 1.4514696427072735, "mean_class_cosine": 0.11904370635094103}}}, "iid_acc": 0.72875, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4175736807895711, "mean_class_cosine": 0.1526238101323753}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.4893964110929853, "drop": -0.011896411092985337, "rotation": {"subspace_principal_angle": 1.3731868842609583, "mean_class_cosine": 0.8675782704012913}}}, "iid_acc": 0.4775, "selectivity": -0.10625000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4543529756472604, "mean_class_cosine": 0.7758454557886542}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7259380097879282, "drop": 0.006561990212071822}}, "iid_acc": 0.7325, "selectivity": 0.14875000000000005} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.8991825613079019, "drop": 0.019567438692098094, "rotation": {"subspace_principal_angle": 1.17550589332711, "mean_class_cosine": 0.38507624166782645}}}, "iid_acc": 0.91875, "selectivity": 0.41625, "iid_split_rotation": {"subspace_principal_angle": 1.1563275504871824, "mean_class_cosine": 0.40270374204802895}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.6771117166212534, "drop": 0.11038828337874662, "rotation": {"subspace_principal_angle": 0.9740972022042724, "mean_class_cosine": 0.9590361076960369}}}, "iid_acc": 0.7875, "selectivity": 0.28500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4700765532293367, "mean_class_cosine": 0.8827661242087979}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.888283378746594, "drop": 0.012966621253405974}}, "iid_acc": 0.90125, "selectivity": 0.39875000000000005} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9867986798679867, "drop": 0.003201320132013241, "rotation": {"subspace_principal_angle": 0.7435989691909886, "mean_class_cosine": 0.7360370400174403}}}, "iid_acc": 0.99, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.7520555858872627, "mean_class_cosine": 0.7302861569945356}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9422442244224423, "drop": -0.0009942244224422359, "rotation": {"subspace_principal_angle": 1.441143445268482, "mean_class_cosine": 0.9558286303461359}}}, "iid_acc": 0.94125, "selectivity": 0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5613733244607428, "mean_class_cosine": 0.973426291781849}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.98125, "drop": 0.0}, "paraphrase": {"acc": 0.9818481848184818, "drop": -0.0005981848184818395}}, "iid_acc": 0.98125, "selectivity": 0.125} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.662807525325615, "drop": 0.023442474674384983, "rotation": {"subspace_principal_angle": 1.5311195648072153, "mean_class_cosine": 0.03966635264677357}}}, "iid_acc": 0.68625, "selectivity": 0.07874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.46679518399481, "mean_class_cosine": 0.10381376065362727}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5151953690303908, "drop": 0.01980463096960927, "rotation": {"subspace_principal_angle": 1.1930287896465346, "mean_class_cosine": 0.911295872781087}}}, "iid_acc": 0.535, "selectivity": -0.07250000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.6904458872144239, "mean_class_cosine": -0.9709984265270024}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": -0.007997467438494832}}, "iid_acc": 0.71125, "selectivity": 0.10375000000000001} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6294117647058823, "drop": -0.004411764705882337, "rotation": {"subspace_principal_angle": 1.5554803313899561, "mean_class_cosine": 0.31267859264684567}}}, "iid_acc": 0.625, "selectivity": 0.1875, "iid_split_rotation": {"subspace_principal_angle": 1.48134181230664, "mean_class_cosine": 0.16283662737976473}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5647058823529412, "drop": -0.01182126696832575, "rotation": {"subspace_principal_angle": 1.0131680189558905, "mean_class_cosine": 0.6680893935892259}}}, "iid_acc": 0.5528846153846154, "selectivity": 0.11538461538461542, "iid_split_rotation": {"subspace_principal_angle": 1.4160182659695404, "mean_class_cosine": 0.5680667839120414}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6346153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.6411764705882353, "drop": -0.0065610859728507664}}, "iid_acc": 0.6346153846153846, "selectivity": 0.19711538461538458} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8899858956276445, "drop": 0.00876410437235553, "rotation": {"subspace_principal_angle": 1.2045865557395712, "mean_class_cosine": 0.3580791088837819}}}, "iid_acc": 0.89875, "selectivity": 0.16000000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2021828264056524, "mean_class_cosine": 0.3603224132960571}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.61625, "drop": 0.0}, "paraphrase": {"acc": 0.5994358251057827, "drop": 0.016814174894217215, "rotation": {"subspace_principal_angle": 1.1806325446572814, "mean_class_cosine": 0.8922635781825321}}}, "iid_acc": 0.61625, "selectivity": -0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3837719103631978, "mean_class_cosine": 0.7905423303610484}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.835, "drop": 0.0}, "paraphrase": {"acc": 0.8349788434414669, "drop": 2.1156558533075298e-05}}, "iid_acc": 0.835, "selectivity": 0.09624999999999995} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8247566063977747, "drop": 0.012743393602225317, "rotation": {"subspace_principal_angle": 1.2798965624456102, "mean_class_cosine": 0.28681430291560384}}, "domain": {"acc": 0.805, "drop": 0.03249999999999997, "rotation": {"subspace_principal_angle": 1.2851613063016167, "mean_class_cosine": 0.28176679938187554}}}, "iid_acc": 0.8375, "selectivity": 0.33625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1765812661272652, "mean_class_cosine": 0.3840835740297509}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.56884561891516, "drop": -0.012595618915159923, "rotation": {"subspace_principal_angle": 1.2901370394335034, "mean_class_cosine": 0.9977925223579588}}, "domain": {"acc": 0.495, "drop": 0.06125000000000003, "rotation": {"subspace_principal_angle": 1.4599392249622891, "mean_class_cosine": 0.6915007566423281}}}, "iid_acc": 0.55625, "selectivity": 0.05500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4453053092946215, "mean_class_cosine": 0.998233597525252}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.8122392211404729, "drop": -0.008489221140472902}, "domain": {"acc": 0.79625, "drop": 0.007499999999999951}}, "iid_acc": 0.80375, "selectivity": 0.3025} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8816326530612245, "drop": 0.0021173469387755217, "rotation": {"subspace_principal_angle": 1.190715066843778, "mean_class_cosine": 0.3709959314311648}}, "domain": {"acc": 0.60625, "drop": 0.2775000000000001, "rotation": {"subspace_principal_angle": 1.4172475398253837, "mean_class_cosine": 0.1529461220229834}}, "length": {"acc": 0.8890577507598785, "drop": -0.005307750759878438, "rotation": {"subspace_principal_angle": 1.2010441268654077, "mean_class_cosine": 0.3613843900828707}}}, "iid_acc": 0.88375, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2079982748541473, "mean_class_cosine": 0.35489153878867186}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.79, "drop": 0.0}, "paraphrase": {"acc": 0.7816326530612245, "drop": 0.0083673469387755, "rotation": {"subspace_principal_angle": 1.3105422369786384, "mean_class_cosine": 0.9556516177420615}}, "domain": {"acc": 0.56625, "drop": 0.22375, "rotation": {"subspace_principal_angle": 1.5060453332017627, "mean_class_cosine": 0.35625753388954995}}, "length": {"acc": 0.8069908814589666, "drop": -0.016990881458966545, "rotation": {"subspace_principal_angle": 1.5564351284094222, "mean_class_cosine": 0.9479993237414271}}}, "iid_acc": 0.79, "selectivity": 0.29500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3575732701756293, "mean_class_cosine": 0.922075109814884}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8877551020408163, "drop": 0.004744897959183647}, "domain": {"acc": 0.6275, "drop": 0.265}, "length": {"acc": 0.8890577507598785, "drop": 0.0034422492401214866}}, "iid_acc": 0.8925, "selectivity": 0.39749999999999996} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8485804416403786, "drop": 0.018919558359621447, "rotation": {"subspace_principal_angle": 1.4399059560351106, "mean_class_cosine": 0.2556332542951133}}}, "iid_acc": 0.8675, "selectivity": 0.64875, "iid_split_rotation": {"subspace_principal_angle": 1.3788930048065602, "mean_class_cosine": 0.2888745735494024}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.7996845425867508, "drop": 0.011565457413249214, "rotation": {"subspace_principal_angle": 1.5548543754884978, "mean_class_cosine": 0.9775272980914157}}}, "iid_acc": 0.81125, "selectivity": 0.5925, "iid_split_rotation": {"subspace_principal_angle": 1.3348767857160242, "mean_class_cosine": 0.979268775558342}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8596214511041009, "drop": 0.020378548895899073}}, "iid_acc": 0.88, "selectivity": 0.66125} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.9525, "drop": 0.0}, "paraphrase": {"acc": 0.9434447300771208, "drop": 0.009055269922879194, "rotation": {"subspace_principal_angle": 1.134461645570175, "mean_class_cosine": 0.6354679842524055}}}, "iid_acc": 0.9525, "selectivity": 0.875, "iid_split_rotation": {"subspace_principal_angle": 1.1124246218092984, "mean_class_cosine": 0.678012477701743}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7712082262210797, "drop": 0.03254177377892031, "rotation": {"subspace_principal_angle": 1.5560406223970316, "mean_class_cosine": 0.8899947966184091}}}, "iid_acc": 0.80375, "selectivity": 0.72625, "iid_split_rotation": {"subspace_principal_angle": 1.503910501910838, "mean_class_cosine": 0.9534530359529768}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9511568123393316, "drop": 0.005093187660668463}}, "iid_acc": 0.95625, "selectivity": 0.87875} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5258397932816538, "drop": 0.015410206718346253, "rotation": {"subspace_principal_angle": 1.496424070388461, "mean_class_cosine": 0.07430371365823527}}}, "iid_acc": 0.54125, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4808569146777435, "mean_class_cosine": 0.08981820636557197}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5025839793281653, "drop": 0.012416020671834671, "rotation": {"subspace_principal_angle": 1.5496948617311233, "mean_class_cosine": 0.9662275396228005}}}, "iid_acc": 0.515, "selectivity": -0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.1782486440521842, "mean_class_cosine": 0.8742419568492337}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.5012919896640827, "drop": 0.01870801033591729}}, "iid_acc": 0.52, "selectivity": -0.0012499999999999734} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.6425, "drop": 0.0}, "paraphrase": {"acc": 0.5872801082543978, "drop": 0.055219891745602157, "rotation": {"subspace_principal_angle": 1.36919244552489, "mean_class_cosine": 0.29038444079442377}}}, "iid_acc": 0.6425, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3882673901543217, "mean_class_cosine": 0.28989125551170686}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.32125, "drop": 0.0}, "paraphrase": {"acc": 0.3071718538565629, "drop": 0.014078146143437054, "rotation": {"subspace_principal_angle": 1.3170533049461681, "mean_class_cosine": 0.6594180780518364}}}, "iid_acc": 0.32125, "selectivity": 0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3815192780699195, "mean_class_cosine": 0.5713567606372191}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5453315290933695, "drop": 0.019668470906630486}}, "iid_acc": 0.565, "selectivity": 0.30999999999999994} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7162162162162162, "drop": 0.017533783783783785, "rotation": {"subspace_principal_angle": 1.3990618518124023, "mean_class_cosine": 0.1708915658468793}}}, "iid_acc": 0.73375, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.3739741567595274, "mean_class_cosine": 0.19555384795811886}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.527027027027027, "drop": -0.014527027027027017, "rotation": {"subspace_principal_angle": 1.2081420951650945, "mean_class_cosine": 0.994931982276282}}}, "iid_acc": 0.5125, "selectivity": -0.007500000000000062, "iid_split_rotation": {"subspace_principal_angle": 1.4812482829852658, "mean_class_cosine": 0.9985002258751091}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7297297297297297, "drop": 0.02902027027027032}}, "iid_acc": 0.75875, "selectivity": 0.23875000000000002} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.656910569105691, "drop": 0.02058943089430898, "rotation": {"subspace_principal_angle": 1.4402611653004675, "mean_class_cosine": 0.13016476977575708}}}, "iid_acc": 0.6775, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.4323787501656056, "mean_class_cosine": 0.1379759996513438}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5495934959349593, "drop": -0.0033434959349593063, "rotation": {"subspace_principal_angle": 1.1049873185467238, "mean_class_cosine": 0.9918135119807125}}}, "iid_acc": 0.54625, "selectivity": 0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2358796490004342, "mean_class_cosine": 0.9310981395287434}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6731707317073171, "drop": 0.00932926829268288}}, "iid_acc": 0.6825, "selectivity": 0.17125} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7308319738988581, "drop": 0.011668026101141926, "rotation": {"subspace_principal_angle": 1.4422948230704635, "mean_class_cosine": 0.12814814582491812}}}, "iid_acc": 0.7425, "selectivity": 0.17500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4341137765391143, "mean_class_cosine": 0.13625736099808608}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.46875, "drop": 0.0}, "paraphrase": {"acc": 0.499184339314845, "drop": -0.030434339314845016, "rotation": {"subspace_principal_angle": 0.5863090291944898, "mean_class_cosine": 0.9879298243670935}}}, "iid_acc": 0.46875, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.1523867074914143, "mean_class_cosine": 0.9911941124762024}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7487765089722676, "drop": 0.003723491027732395}}, "iid_acc": 0.7525, "selectivity": 0.18499999999999994} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9209809264305178, "drop": 0.016519073569482234, "rotation": {"subspace_principal_angle": 1.1107561694947945, "mean_class_cosine": 0.44398408965660313}}}, "iid_acc": 0.9375, "selectivity": 0.43125, "iid_split_rotation": {"subspace_principal_angle": 1.134035745754033, "mean_class_cosine": 0.42300636563354255}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5926430517711172, "drop": -0.025143051771117197, "rotation": {"subspace_principal_angle": 0.950772040164655, "mean_class_cosine": 0.9618998317960137}}}, "iid_acc": 0.5675, "selectivity": 0.06125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2078832510638606, "mean_class_cosine": 0.4895900188006161}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9073569482288828, "drop": 0.010143051771117184}}, "iid_acc": 0.9175, "selectivity": 0.41125} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9587458745874587, "drop": 0.0337541254125413, "rotation": {"subspace_principal_angle": 0.9486144857930342, "mean_class_cosine": 0.5828095295275909}}}, "iid_acc": 0.9925, "selectivity": 0.16375000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.9074513638691858, "mean_class_cosine": 0.6157559117610208}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.14026402640264027, "drop": 0.5609859735973598, "rotation": {"subspace_principal_angle": 1.1024334053734877, "mean_class_cosine": 0.5757723487754411}}}, "iid_acc": 0.70125, "selectivity": -0.12749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2565568409948566, "mean_class_cosine": 0.9967803523356993}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.976897689768977, "drop": 0.01685231023102307}}, "iid_acc": 0.99375, "selectivity": 0.16500000000000004} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6613603473227206, "drop": 0.026139652677279357, "rotation": {"subspace_principal_angle": 1.5419209125002344, "mean_class_cosine": 0.0288714017921598}}}, "iid_acc": 0.6875, "selectivity": 0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4675689654827935, "mean_class_cosine": 0.10304412909555002}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5050651230101303, "drop": 0.021184876989869705, "rotation": {"subspace_principal_angle": 1.5385757961576287, "mean_class_cosine": 0.48096021437512015}}}, "iid_acc": 0.52625, "selectivity": -0.05249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1269562907021227, "mean_class_cosine": -0.8913469139771255}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.723589001447178, "drop": -0.006089001447177944}}, "iid_acc": 0.7175, "selectivity": 0.13875000000000004} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": 0.014253393665158431, "rotation": {"subspace_principal_angle": 1.380845212313945, "mean_class_cosine": 0.22774336814303378}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.3028846153846154, "iid_split_rotation": {"subspace_principal_angle": 1.4371716903391298, "mean_class_cosine": 0.16228231461949003}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.47115384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.4411764705882353, "drop": 0.029977375565610864, "rotation": {"subspace_principal_angle": 1.077660750569373, "mean_class_cosine": 0.5488122950260849}}}, "iid_acc": 0.47115384615384615, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.3611080124878545, "mean_class_cosine": 0.7221103135486793}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 14, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": -0.014592760180995423}}, "iid_acc": 0.6442307692307693, "selectivity": 0.27403846153846156} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8885754583921015, "drop": 0.006424541607898515, "rotation": {"subspace_principal_angle": 1.1803089373724376, "mean_class_cosine": 0.38063916084124166}}}, "iid_acc": 0.895, "selectivity": 0.16500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.198391794536694, "mean_class_cosine": 0.36385619559330473}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.8208744710860366, "drop": -0.017124471086036652, "rotation": {"subspace_principal_angle": 1.4701060816070168, "mean_class_cosine": 0.8930227036508337}}}, "iid_acc": 0.80375, "selectivity": 0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.1336871573868261, "mean_class_cosine": 0.8902119338649208}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 4, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8716502115655853, "drop": 0.005849788434414638}}, "iid_acc": 0.8775, "selectivity": 0.14749999999999996} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8344923504867872, "drop": 0.03300764951321289, "rotation": {"subspace_principal_angle": 1.32725054312888, "mean_class_cosine": 0.24114527926927334}}, "domain": {"acc": 0.855, "drop": 0.012500000000000067, "rotation": {"subspace_principal_angle": 1.4511968441379492, "mean_class_cosine": 0.11931456062079314}}}, "iid_acc": 0.8675, "selectivity": 0.38125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3310106874774308, "mean_class_cosine": 0.23749440429619095}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5827538247566064, "drop": -0.02025382475660642, "rotation": {"subspace_principal_angle": 0.7190626485497094, "mean_class_cosine": 0.9647786618667977}}, "domain": {"acc": 0.495, "drop": 0.0675, "rotation": {"subspace_principal_angle": 1.1078756653311121, "mean_class_cosine": 0.28125598087122977}}}, "iid_acc": 0.5625, "selectivity": 0.07624999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.8988412922378591, "mean_class_cosine": 0.9806205060335059}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.827538247566064, "drop": 0.027461752433936004}, "domain": {"acc": 0.84875, "drop": 0.006249999999999978}}, "iid_acc": 0.855, "selectivity": 0.36874999999999997} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.8959183673469387, "drop": 0.009081632653061278, "rotation": {"subspace_principal_angle": 1.329136111358117, "mean_class_cosine": 0.2393149283948771}}, "domain": {"acc": 0.615, "drop": 0.29000000000000004, "rotation": {"subspace_principal_angle": 1.4430058206254768, "mean_class_cosine": 0.12744297808753202}}, "length": {"acc": 0.9072948328267477, "drop": -0.0022948328267476947, "rotation": {"subspace_principal_angle": 1.3188768354543379, "mean_class_cosine": 0.2492633213424405}}}, "iid_acc": 0.905, "selectivity": 0.43625, "iid_split_rotation": {"subspace_principal_angle": 1.2936213190349724, "mean_class_cosine": 0.2736395752811731}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.84375, "drop": 0.0}, "paraphrase": {"acc": 0.8448979591836735, "drop": -0.001147959183673497, "rotation": {"subspace_principal_angle": 1.2791968807329361, "mean_class_cosine": 0.9478558905818975}}, "domain": {"acc": 0.56625, "drop": 0.27749999999999997, "rotation": {"subspace_principal_angle": 1.5624247261675832, "mean_class_cosine": 0.3303097598518465}}, "length": {"acc": 0.8556231003039514, "drop": -0.01187310030395139, "rotation": {"subspace_principal_angle": 1.3538731412108984, "mean_class_cosine": 0.9594238444634923}}}, "iid_acc": 0.84375, "selectivity": 0.375, "iid_split_rotation": {"subspace_principal_angle": 1.0790921789981887, "mean_class_cosine": 0.9412243630037906}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9142857142857143, "drop": 0.003214285714285725}, "domain": {"acc": 0.8325, "drop": 0.08499999999999996}, "length": {"acc": 0.9133738601823708, "drop": 0.004126139817629215}}, "iid_acc": 0.9175, "selectivity": 0.44875} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8533123028391167, "drop": 0.016687697160883253, "rotation": {"subspace_principal_angle": 1.351979607774418, "mean_class_cosine": 0.3225426038590604}}}, "iid_acc": 0.87, "selectivity": 0.61625, "iid_split_rotation": {"subspace_principal_angle": 1.3508402767430103, "mean_class_cosine": 0.3455994411857771}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.81875, "drop": 0.0}, "paraphrase": {"acc": 0.804416403785489, "drop": 0.014333596214511024, "rotation": {"subspace_principal_angle": 1.2203955464014318, "mean_class_cosine": 0.9625472880536573}}}, "iid_acc": 0.81875, "selectivity": 0.565, "iid_split_rotation": {"subspace_principal_angle": 1.38855296445251, "mean_class_cosine": 0.9564765506858832}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8659305993690851, "drop": 0.017819400630914917}}, "iid_acc": 0.88375, "selectivity": 0.6300000000000001} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.96875, "drop": 0.0}, "paraphrase": {"acc": 0.9640102827763496, "drop": 0.004739717223650408, "rotation": {"subspace_principal_angle": 1.0857448901953828, "mean_class_cosine": 0.6613254867828217}}}, "iid_acc": 0.96875, "selectivity": 0.89375, "iid_split_rotation": {"subspace_principal_angle": 1.076414289934142, "mean_class_cosine": 0.7082374564146615}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6838046272493573, "drop": 0.026195372750642654, "rotation": {"subspace_principal_angle": 1.5350687578200197, "mean_class_cosine": 0.8594207995140349}}}, "iid_acc": 0.71, "selectivity": 0.635, "iid_split_rotation": {"subspace_principal_angle": 1.0895700706545426, "mean_class_cosine": 0.9341302332271046}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9665809768637532, "drop": 0.0009190231362468104}}, "iid_acc": 0.9675, "selectivity": 0.8925000000000001} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.60625, "drop": 0.0}, "paraphrase": {"acc": 0.5930232558139535, "drop": 0.013226744186046413, "rotation": {"subspace_principal_angle": 1.4617197666146766, "mean_class_cosine": 0.10886039552142644}}}, "iid_acc": 0.60625, "selectivity": 0.11499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.466798042788668, "mean_class_cosine": 0.10381091730611129}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": 0.005998062015503902, "rotation": {"subspace_principal_angle": 1.5323338067846937, "mean_class_cosine": 0.9209037611997781}}}, "iid_acc": 0.51375, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.7482139187600273, "mean_class_cosine": 0.7359681788398151}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.57875, "drop": 0.0}, "paraphrase": {"acc": 0.5671834625322998, "drop": 0.0115665374677002}}, "iid_acc": 0.57875, "selectivity": 0.08749999999999997} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.6125, "drop": 0.0}, "paraphrase": {"acc": 0.5926928281461434, "drop": 0.01980717185385661, "rotation": {"subspace_principal_angle": 1.445337558644254, "mean_class_cosine": 0.2135394258196436}}}, "iid_acc": 0.6125, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.42444703173012, "mean_class_cosine": 0.21657072763653504}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.2075, "drop": 0.0}, "paraphrase": {"acc": 0.21109607577807848, "drop": -0.0035960757780784947, "rotation": {"subspace_principal_angle": 1.2059558141673055, "mean_class_cosine": 0.8778793379671459}}}, "iid_acc": 0.2075, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4627224949895972, "mean_class_cosine": 0.5073250273765634}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.6102841677943166, "drop": 0.007215832205683448}}, "iid_acc": 0.6175, "selectivity": 0.395} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7094594594594594, "drop": 0.018040540540540606, "rotation": {"subspace_principal_angle": 1.439408162778464, "mean_class_cosine": 0.13101046781611214}}}, "iid_acc": 0.7275, "selectivity": 0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.440617717267834, "mean_class_cosine": 0.12981124295188987}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5726351351351351, "drop": -0.04263513513513506, "rotation": {"subspace_principal_angle": 1.1956589180448767, "mean_class_cosine": 0.9074769300513319}}}, "iid_acc": 0.53, "selectivity": 0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.4496764494101746, "mean_class_cosine": 0.9668956137286093}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 23, "dists": {"iid": {"acc": 0.76875, "drop": 0.0}, "paraphrase": {"acc": 0.7635135135135135, "drop": 0.005236486486486558}}, "iid_acc": 0.76875, "selectivity": 0.26375000000000004} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6634146341463415, "drop": 0.019085365853658498, "rotation": {"subspace_principal_angle": 1.4585106839043807, "mean_class_cosine": 0.11204984079812053}}}, "iid_acc": 0.6825, "selectivity": 0.2025, "iid_split_rotation": {"subspace_principal_angle": 1.4658854601898588, "mean_class_cosine": 0.10471852591428352}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5560975609756098, "drop": -0.004847560975609788, "rotation": {"subspace_principal_angle": 0.7177846273187343, "mean_class_cosine": 0.9945223665026368}}}, "iid_acc": 0.55125, "selectivity": 0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2365217302923708, "mean_class_cosine": 0.9628480684104785}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6731707317073171, "drop": 0.0343292682926829}}, "iid_acc": 0.7075, "selectivity": 0.22750000000000004} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.7210440456769984, "drop": -0.008544045676998402, "rotation": {"subspace_principal_angle": 1.5017342011669776, "mean_class_cosine": 0.06900723919555485}}}, "iid_acc": 0.7125, "selectivity": 0.12375000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4884858123978102, "mean_class_cosine": 0.08221760330215533}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.47125, "drop": 0.0}, "paraphrase": {"acc": 0.5024469820554649, "drop": -0.031196982055464895, "rotation": {"subspace_principal_angle": 0.9746986945864435, "mean_class_cosine": 0.9741912343048922}}}, "iid_acc": 0.47125, "selectivity": -0.1175, "iid_split_rotation": {"subspace_principal_angle": 0.721596565453364, "mean_class_cosine": 0.9782502269601113}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.76625, "drop": 0.0}, "paraphrase": {"acc": 0.7846655791190864, "drop": -0.018415579119086423}}, "iid_acc": 0.76625, "selectivity": 0.1775} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.94375, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.024131471389645776, "rotation": {"subspace_principal_angle": 1.1654703724010276, "mean_class_cosine": 0.39431832489873625}}}, "iid_acc": 0.94375, "selectivity": 0.46499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1557667157993363, "mean_class_cosine": 0.40321702771416507}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.67625, "drop": 0.0}, "paraphrase": {"acc": 0.6743869209809265, "drop": 0.0018630790190735658, "rotation": {"subspace_principal_angle": 1.141242730507284, "mean_class_cosine": 0.9694783192528957}}}, "iid_acc": 0.67625, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 1.4347294745233075, "mean_class_cosine": 0.7429243512254556}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.9318801089918256, "drop": 0.015619891008174402}}, "iid_acc": 0.9475, "selectivity": 0.46875} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9900990099009901, "drop": 0.0024009900990099586, "rotation": {"subspace_principal_angle": 0.9638933203750397, "mean_class_cosine": 0.5703262722196243}}}, "iid_acc": 0.9925, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9412550884957132, "mean_class_cosine": 0.5887740038848737}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.43399339933993397, "drop": 0.223506600660066, "rotation": {"subspace_principal_angle": 1.0823516673301268, "mean_class_cosine": 0.9745159076170016}}}, "iid_acc": 0.6575, "selectivity": -0.13875000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.4071181854905652, "mean_class_cosine": 0.9998877741118155}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 13, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9884488448844885, "drop": 0.00655115511551152}}, "iid_acc": 0.995, "selectivity": 0.19874999999999998} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.7125, "drop": 0.0}, "paraphrase": {"acc": 0.6772793053545586, "drop": 0.03522069464544142, "rotation": {"subspace_principal_angle": 1.5464207554561848, "mean_class_cosine": 0.02437315754440208}}}, "iid_acc": 0.7125, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 1.4523838978678731, "mean_class_cosine": 0.11813590281951224}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5123010130246021, "drop": 0.016448986975397983, "rotation": {"subspace_principal_angle": 1.5271447554808601, "mean_class_cosine": 0.5274519249488747}}}, "iid_acc": 0.52875, "selectivity": -0.043749999999999956, "iid_split_rotation": {"subspace_principal_angle": 0.7321498396093691, "mean_class_cosine": -0.8691134472727473}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7337192474674384, "drop": 0.007530752532561524}}, "iid_acc": 0.74125, "selectivity": 0.16874999999999996} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6634615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6764705882352942, "drop": -0.013009049773755721, "rotation": {"subspace_principal_angle": 1.4059071468679976, "mean_class_cosine": 0.20250091764866876}}}, "iid_acc": 0.6634615384615384, "selectivity": 0.3125, "iid_split_rotation": {"subspace_principal_angle": 1.4463777868534364, "mean_class_cosine": 0.14772015722986645}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.4951923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.4411764705882353, "drop": 0.05401583710407243, "rotation": {"subspace_principal_angle": 1.155152788254596, "mean_class_cosine": 0.6509669876420695}}}, "iid_acc": 0.4951923076923077, "selectivity": 0.14423076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.3178705376329367, "mean_class_cosine": 0.6023058035884701}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6705882352941176, "drop": -0.026357466063348323}}, "iid_acc": 0.6442307692307693, "selectivity": 0.29326923076923084} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": 0.01030148095909722, "rotation": {"subspace_principal_angle": 1.3664411608119795, "mean_class_cosine": 0.20293578579819216}}}, "iid_acc": 0.90875, "selectivity": 0.2024999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3058033713210209, "mean_class_cosine": 0.2619024695236912}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.7249647390691114, "drop": -0.0012147390691114035, "rotation": {"subspace_principal_angle": 1.3860419944865392, "mean_class_cosine": 0.9342763772250853}}}, "iid_acc": 0.72375, "selectivity": 0.01749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5390421123713445, "mean_class_cosine": 0.8051652349687315}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.9097320169252469, "drop": -0.003482016925246856}}, "iid_acc": 0.90625, "selectivity": 0.19999999999999996} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.78125, "drop": 0.0}, "paraphrase": {"acc": 0.7858136300417247, "drop": -0.0045636300417246645, "rotation": {"subspace_principal_angle": 1.325361507791648, "mean_class_cosine": 0.24297813599017176}}, "domain": {"acc": 0.78, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.4795657935318307, "mean_class_cosine": 0.09110403380754725}}}, "iid_acc": 0.78125, "selectivity": 0.26875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3436786283410562, "mean_class_cosine": 0.22517018028571784}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.55875, "drop": 0.0}, "paraphrase": {"acc": 0.5744089012517385, "drop": -0.015658901251738522, "rotation": {"subspace_principal_angle": 0.6230331668549962, "mean_class_cosine": 0.9910755485040825}}, "domain": {"acc": 0.495, "drop": 0.06374999999999997, "rotation": {"subspace_principal_angle": 1.4491819836983844, "mean_class_cosine": 0.5529294755790912}}}, "iid_acc": 0.55875, "selectivity": 0.04625000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4276392742434922, "mean_class_cosine": 0.9917849416289807}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.7746870653685675, "drop": 0.047812934631432547}, "domain": {"acc": 0.80625, "drop": 0.016249999999999987}}, "iid_acc": 0.8225, "selectivity": 0.31000000000000005} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.8551020408163266, "drop": -0.00010204081632658735, "rotation": {"subspace_principal_angle": 1.3692001923291004, "mean_class_cosine": 0.20023339510507437}}, "domain": {"acc": 0.6, "drop": 0.255, "rotation": {"subspace_principal_angle": 1.468428009163586, "mean_class_cosine": 0.10218962037267862}}, "length": {"acc": 0.8556231003039514, "drop": -0.0006231003039514071, "rotation": {"subspace_principal_angle": 1.3923528233365947, "mean_class_cosine": 0.17749800787298398}}}, "iid_acc": 0.855, "selectivity": 0.37374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3944016978212828, "mean_class_cosine": 0.17548129599139065}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7510204081632653, "drop": 0.0014795918367346284, "rotation": {"subspace_principal_angle": 1.3529173793192326, "mean_class_cosine": 0.9403316707188716}}, "domain": {"acc": 0.56625, "drop": 0.18624999999999992, "rotation": {"subspace_principal_angle": 1.553813034487346, "mean_class_cosine": 0.4365397449834313}}, "length": {"acc": 0.7613981762917933, "drop": -0.008898176291793347, "rotation": {"subspace_principal_angle": 1.484045792925862, "mean_class_cosine": 0.9449458799225073}}}, "iid_acc": 0.7525, "selectivity": 0.27124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.102117298311522, "mean_class_cosine": 0.9232016388495949}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8510204081632653, "drop": 0.02272959183673473}, "domain": {"acc": 0.67125, "drop": 0.2025}, "length": {"acc": 0.8723404255319149, "drop": 0.0014095744680850952}}, "iid_acc": 0.87375, "selectivity": 0.3925} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8280757097791798, "drop": 0.024424290220820266, "rotation": {"subspace_principal_angle": 1.4366178069899613, "mean_class_cosine": 0.2843840411694384}}}, "iid_acc": 0.8525, "selectivity": 0.6175, "iid_split_rotation": {"subspace_principal_angle": 1.3589778090630555, "mean_class_cosine": 0.3305635161629984}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5930599369085173, "drop": 0.018190063091482633, "rotation": {"subspace_principal_angle": 0.6501933782850093, "mean_class_cosine": 0.9238960788618794}}}, "iid_acc": 0.61125, "selectivity": 0.37625, "iid_split_rotation": {"subspace_principal_angle": 0.576594648277471, "mean_class_cosine": 0.9553403579368684}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 3, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.861198738170347, "drop": 0.028801261829653035}}, "iid_acc": 0.89, "selectivity": 0.655} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9383033419023136, "drop": 0.007946658097686465, "rotation": {"subspace_principal_angle": 1.2309707985530614, "mean_class_cosine": 0.5687671045365533}}}, "iid_acc": 0.94625, "selectivity": 0.85875, "iid_split_rotation": {"subspace_principal_angle": 1.1443944465717892, "mean_class_cosine": 0.6294410202468741}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7095115681233933, "drop": 0.03423843187660669, "rotation": {"subspace_principal_angle": 1.3105209680737453, "mean_class_cosine": 0.8754616020973991}}}, "iid_acc": 0.74375, "selectivity": 0.65625, "iid_split_rotation": {"subspace_principal_angle": 1.3306223102176358, "mean_class_cosine": 0.927772151836221}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 11, "dists": {"iid": {"acc": 0.94, "drop": 0.0}, "paraphrase": {"acc": 0.9203084832904884, "drop": 0.019691516709511525}}, "iid_acc": 0.94, "selectivity": 0.8524999999999999} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5116279069767442, "drop": 0.019622093023255793, "rotation": {"subspace_principal_angle": 1.4907936041998628, "mean_class_cosine": 0.07991740785625166}}}, "iid_acc": 0.53125, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.494326896717623, "mean_class_cosine": 0.07639492509245417}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": -0.01025193798449614, "rotation": {"subspace_principal_angle": 1.4876221863427175, "mean_class_cosine": 0.8807712318567114}}}, "iid_acc": 0.4975, "selectivity": 0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.2199667575281368, "mean_class_cosine": 0.8996775514646336}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.47375, "drop": 0.0}, "paraphrase": {"acc": 0.5064599483204134, "drop": -0.03270994832041341}}, "iid_acc": 0.47375, "selectivity": -0.00874999999999998} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6275, "drop": 0.0}, "paraphrase": {"acc": 0.5723951285520974, "drop": 0.05510487144790255, "rotation": {"subspace_principal_angle": 1.4343786178569309, "mean_class_cosine": 0.2763953601626957}}}, "iid_acc": 0.6275, "selectivity": 0.37625, "iid_split_rotation": {"subspace_principal_angle": 1.3550101339445768, "mean_class_cosine": 0.31016139289436095}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.24125, "drop": 0.0}, "paraphrase": {"acc": 0.2449255751014885, "drop": -0.0036755751014884985, "rotation": {"subspace_principal_angle": 1.2918205373490343, "mean_class_cosine": 0.7766803935782208}}}, "iid_acc": 0.24125, "selectivity": -0.009999999999999981, "iid_split_rotation": {"subspace_principal_angle": 1.5684032048848013, "mean_class_cosine": 0.5981109246694539}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.60875, "drop": 0.0}, "paraphrase": {"acc": 0.5683355886332883, "drop": 0.040414411366711755}}, "iid_acc": 0.60875, "selectivity": 0.35750000000000004} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7398648648648649, "drop": 0.00013513513513507824, "rotation": {"subspace_principal_angle": 1.2673314142482994, "mean_class_cosine": 0.2988285834755996}}}, "iid_acc": 0.74, "selectivity": 0.24, "iid_split_rotation": {"subspace_principal_angle": 1.296110954087143, "mean_class_cosine": 0.27124411828692846}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.6199324324324325, "drop": -0.01993243243243248, "rotation": {"subspace_principal_angle": 1.5658191148065415, "mean_class_cosine": 0.9653627770010756}}}, "iid_acc": 0.6, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.684621553461914, "mean_class_cosine": 0.9768155693913088}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7246621621621622, "drop": 0.010337837837837816}}, "iid_acc": 0.735, "selectivity": 0.235} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6308943089430894, "drop": 0.014105691056910619, "rotation": {"subspace_principal_angle": 1.4810005396497288, "mean_class_cosine": 0.08967516097444077}}}, "iid_acc": 0.645, "selectivity": 0.16875, "iid_split_rotation": {"subspace_principal_angle": 1.4383531851019704, "mean_class_cosine": 0.13205627952425741}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.551219512195122, "drop": -0.008719512195122014, "rotation": {"subspace_principal_angle": 1.5478870220336094, "mean_class_cosine": 0.9974275968935058}}}, "iid_acc": 0.5425, "selectivity": 0.06624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3106732220928752, "mean_class_cosine": 0.9809909328157596}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6861788617886179, "drop": -0.011178861788617822}}, "iid_acc": 0.675, "selectivity": 0.19875000000000004} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.7225, "drop": 0.0}, "paraphrase": {"acc": 0.7177814029363785, "drop": 0.0047185970636215435, "rotation": {"subspace_principal_angle": 1.3869179562010008, "mean_class_cosine": 0.18284392784898534}}}, "iid_acc": 0.7225, "selectivity": 0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3863158843348609, "mean_class_cosine": 0.18343581678343168}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.46492659053833607, "drop": 0.040073409461663934, "rotation": {"subspace_principal_angle": 1.3638991068226356, "mean_class_cosine": 0.5772026508098606}}}, "iid_acc": 0.505, "selectivity": -0.12, "iid_split_rotation": {"subspace_principal_angle": 1.0426083209720547, "mean_class_cosine": 0.9895063250474988}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7210440456769984, "drop": 0.003955954323001554}}, "iid_acc": 0.725, "selectivity": 0.09999999999999998} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8950953678474114, "drop": 0.013654632152588553, "rotation": {"subspace_principal_angle": 1.1926328016497498, "mean_class_cosine": 0.36921437562940024}}}, "iid_acc": 0.90875, "selectivity": 0.43124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2197817002343454, "mean_class_cosine": 0.34385074329564713}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5994550408719346, "drop": -0.019455040871934615, "rotation": {"subspace_principal_angle": 1.0711301307338577, "mean_class_cosine": 0.9676646451280264}}}, "iid_acc": 0.58, "selectivity": 0.10249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4692738400107856, "mean_class_cosine": 0.5961110196205139}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.9059945504087193, "drop": -0.003494550408719377}}, "iid_acc": 0.9025, "selectivity": 0.425} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9851485148514851, "drop": 0.0073514851485149135, "rotation": {"subspace_principal_angle": 0.9650258351359023, "mean_class_cosine": 0.5693956383770296}}}, "iid_acc": 0.9925, "selectivity": 0.15250000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9566139505713168, "mean_class_cosine": 0.5762905161188008}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.7838283828382838, "drop": -0.10132838283828383, "rotation": {"subspace_principal_angle": 0.7837036206433732, "mean_class_cosine": 0.9616722641195252}}}, "iid_acc": 0.6825, "selectivity": -0.15749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2673965735923087, "mean_class_cosine": 0.9865395355198125}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.9875, "drop": 0.0}, "paraphrase": {"acc": 0.9834983498349835, "drop": 0.004001650165016524}}, "iid_acc": 0.9875, "selectivity": 0.14750000000000008} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.6801736613603473, "drop": 0.009826338639652654, "rotation": {"subspace_principal_angle": 1.446957515536241, "mean_class_cosine": 0.12352252083060832}}}, "iid_acc": 0.69, "selectivity": 0.0724999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.482921903323322, "mean_class_cosine": 0.08776137400044447}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5354558610709117, "drop": 0.022044138929088297, "rotation": {"subspace_principal_angle": 1.0513666721812371, "mean_class_cosine": 0.686624560357272}}}, "iid_acc": 0.5575, "selectivity": -0.06000000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.6796913400547621, "mean_class_cosine": 0.5471521743827539}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7308248914616498, "drop": -0.01207489146164975}}, "iid_acc": 0.71875, "selectivity": 0.10124999999999995} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": 0.013744343891402777, "rotation": {"subspace_principal_angle": 1.4301231189422874, "mean_class_cosine": 0.21099799551640833}}}, "iid_acc": 0.6490384615384616, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.480503372414954, "mean_class_cosine": 0.14932363173924582}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.5288461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": -0.0005656108597285048, "rotation": {"subspace_principal_angle": 1.2751790898102846, "mean_class_cosine": 0.6777353047064866}}}, "iid_acc": 0.5288461538461539, "selectivity": 0.13942307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.437011284720561, "mean_class_cosine": 0.6376580778974325}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": -0.010294117647058787}}, "iid_acc": 0.625, "selectivity": 0.23557692307692307} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8899858956276445, "drop": 0.011264104372355477, "rotation": {"subspace_principal_angle": 1.2097432263622687, "mean_class_cosine": 0.3532596311031677}}}, "iid_acc": 0.90125, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.2597303970258766, "mean_class_cosine": 0.306073583648639}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5077574047954866, "drop": 0.017242595204513433, "rotation": {"subspace_principal_angle": 1.5668153896789387, "mean_class_cosine": 0.821628018438601}}}, "iid_acc": 0.525, "selectivity": -0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.244997779584516, "mean_class_cosine": 0.9282425176629551}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8561354019746121, "drop": 0.01886459802538787}}, "iid_acc": 0.875, "selectivity": 0.12749999999999995} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.82375, "drop": 0.0}, "paraphrase": {"acc": 0.7927677329624478, "drop": 0.030982267037552136, "rotation": {"subspace_principal_angle": 1.3441679611646846, "mean_class_cosine": 0.22469338685922752}}, "domain": {"acc": 0.68875, "drop": 0.135, "rotation": {"subspace_principal_angle": 1.4619007929505858, "mean_class_cosine": 0.10868044323365433}}}, "iid_acc": 0.82375, "selectivity": 0.2925, "iid_split_rotation": {"subspace_principal_angle": 1.3382551814356736, "mean_class_cosine": 0.23045101264055012}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.55875, "drop": 0.0}, "paraphrase": {"acc": 0.5744089012517385, "drop": -0.015658901251738522, "rotation": {"subspace_principal_angle": 1.1096712314062809, "mean_class_cosine": 0.9710046862155046}}, "domain": {"acc": 0.495, "drop": 0.06374999999999997, "rotation": {"subspace_principal_angle": 1.3394837230563983, "mean_class_cosine": 0.4067141442483526}}}, "iid_acc": 0.55875, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.0571042364623964, "mean_class_cosine": 0.9732840432471883}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 19, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.7955493741307371, "drop": 0.05820062586926289}, "domain": {"acc": 0.8325, "drop": 0.02124999999999999}}, "iid_acc": 0.85375, "selectivity": 0.3225} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5297157622739018, "drop": 0.02403423772609814, "rotation": {"subspace_principal_angle": 1.4849996721834149, "mean_class_cosine": 0.08569143420695843}}}, "iid_acc": 0.55375, "selectivity": 0.04125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4079692233979924, "mean_class_cosine": 0.1621085598267091}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5077519379844961, "drop": -0.009001937984496111, "rotation": {"subspace_principal_angle": 1.5312281927446767, "mean_class_cosine": 0.9121731321951317}}}, "iid_acc": 0.49875, "selectivity": -0.013749999999999929, "iid_split_rotation": {"subspace_principal_angle": 1.414487810468904, "mean_class_cosine": 0.9131370420534566}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5129198966408268, "drop": -0.004169896640826787}}, "iid_acc": 0.50875, "selectivity": -0.00374999999999992} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.6575, "drop": 0.0}, "paraphrase": {"acc": 0.6224627875507442, "drop": 0.03503721244925573, "rotation": {"subspace_principal_angle": 1.377806750699231, "mean_class_cosine": 0.3147008679395653}}}, "iid_acc": 0.6575, "selectivity": 0.40375, "iid_split_rotation": {"subspace_principal_angle": 1.3221669271577434, "mean_class_cosine": 0.32518311136584716}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.27375, "drop": 0.0}, "paraphrase": {"acc": 0.2665764546684709, "drop": 0.007173545331529085, "rotation": {"subspace_principal_angle": 1.3088731708244634, "mean_class_cosine": 0.746477870474282}}}, "iid_acc": 0.27375, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.4761219391324176, "mean_class_cosine": 0.6090428280673176}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5737483085250338, "drop": 0.025001691474966226}}, "iid_acc": 0.59875, "selectivity": 0.34500000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.722972972972973, "drop": 0.018277027027026937, "rotation": {"subspace_principal_angle": 1.3556838479482727, "mean_class_cosine": 0.21345731612896818}}}, "iid_acc": 0.74125, "selectivity": 0.24, "iid_split_rotation": {"subspace_principal_angle": 1.3660213343678467, "mean_class_cosine": 0.20334685861759932}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5827702702702703, "drop": 0.00597972972972971, "rotation": {"subspace_principal_angle": 0.7662797422733312, "mean_class_cosine": 0.8372309786422542}}}, "iid_acc": 0.58875, "selectivity": 0.08750000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.19745934756960443, "mean_class_cosine": 0.9910109223557726}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7466216216216216, "drop": -0.007871621621621583}}, "iid_acc": 0.73875, "selectivity": 0.23750000000000004} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.65625, "drop": 0.0}, "paraphrase": {"acc": 0.6113821138211382, "drop": 0.04486788617886184, "rotation": {"subspace_principal_angle": 1.5244430780993927, "mean_class_cosine": 0.046336651197513255}}}, "iid_acc": 0.65625, "selectivity": 0.14, "iid_split_rotation": {"subspace_principal_angle": 1.48060525921707, "mean_class_cosine": 0.09006884183258959}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5560975609756098, "drop": -0.01609756097560977, "rotation": {"subspace_principal_angle": 1.4447759012617603, "mean_class_cosine": 0.9966733139859822}}}, "iid_acc": 0.54, "selectivity": 0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5053011307485271, "mean_class_cosine": 0.9733537861349781}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.6682926829268293, "drop": 0.024207317073170698}}, "iid_acc": 0.6925, "selectivity": 0.17625000000000002} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7047308319738989, "drop": 0.014019168026101148, "rotation": {"subspace_principal_angle": 1.4273253943610407, "mean_class_cosine": 0.14297923999787693}}}, "iid_acc": 0.71875, "selectivity": 0.12624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.410754254834671, "mean_class_cosine": 0.15935974106023298}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.49375, "drop": 0.0}, "paraphrase": {"acc": 0.46492659053833607, "drop": 0.02882340946166395, "rotation": {"subspace_principal_angle": 1.2475152228835678, "mean_class_cosine": -0.2873781232568553}}}, "iid_acc": 0.49375, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 0.3595110102120592, "mean_class_cosine": 0.9955684555352899}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 24, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7373572593800979, "drop": 0.005142740619902164}}, "iid_acc": 0.7425, "selectivity": 0.15000000000000002} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.01163147138964582, "rotation": {"subspace_principal_angle": 1.1462155217150516, "mean_class_cosine": 0.4119388427182844}}}, "iid_acc": 0.93125, "selectivity": 0.4525, "iid_split_rotation": {"subspace_principal_angle": 1.2036153547134156, "mean_class_cosine": 0.3589857417663745}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.57375, "drop": 0.0}, "paraphrase": {"acc": 0.5885558583106267, "drop": -0.014805858310626752, "rotation": {"subspace_principal_angle": 0.6949506609048397, "mean_class_cosine": 0.9687824706589729}}}, "iid_acc": 0.57375, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4852957312103852, "mean_class_cosine": 0.7009380316990886}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 10, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.9155313351498637, "drop": -0.00428133514986373}}, "iid_acc": 0.91125, "selectivity": 0.4325} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6555716353111433, "drop": 0.0544283646888567, "rotation": {"subspace_principal_angle": 1.530036517597367, "mean_class_cosine": 0.04074852400167777}}}, "iid_acc": 0.71, "selectivity": 0.12249999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4332650590425584, "mean_class_cosine": 0.1370981137331397}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5180897250361794, "drop": 0.008160274963820546, "rotation": {"subspace_principal_angle": 1.3868122762825261, "mean_class_cosine": 0.5674558871637044}}}, "iid_acc": 0.52625, "selectivity": -0.06125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3698230368243622, "mean_class_cosine": -0.902684531716938}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7192474674384949, "drop": 0.03450253256150515}}, "iid_acc": 0.75375, "selectivity": 0.16625} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6588235294117647, "drop": -0.004977375565610842, "rotation": {"subspace_principal_angle": 1.4918296501472632, "mean_class_cosine": 0.14481202131126455}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.2548076923076923, "iid_split_rotation": {"subspace_principal_angle": 1.4905675320918412, "mean_class_cosine": 0.12376594768867437}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.5576923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.5235294117647059, "drop": 0.0341628959276018, "rotation": {"subspace_principal_angle": 1.1976344371529, "mean_class_cosine": 0.7296312684418567}}}, "iid_acc": 0.5576923076923077, "selectivity": 0.15865384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.5298301184380643, "mean_class_cosine": 0.5946390438469394}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.6298076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6352941176470588, "drop": -0.005486425339366496}}, "iid_acc": 0.6298076923076923, "selectivity": 0.23076923076923073} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.9012693935119888, "drop": -0.0012693935119887367, "rotation": {"subspace_principal_angle": 1.3718118378240132, "mean_class_cosine": 0.1976739601069373}}}, "iid_acc": 0.9, "selectivity": 0.21999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.320617301683332, "mean_class_cosine": 0.24757741494334223}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7475317348377997, "drop": -0.015031734837799626, "rotation": {"subspace_principal_angle": 0.7004143627777456, "mean_class_cosine": 0.8791149950634916}}}, "iid_acc": 0.7325, "selectivity": 0.05249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.298216893295834, "mean_class_cosine": 0.687969391669262}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.9055007052186178, "drop": -0.0005007052186177807}}, "iid_acc": 0.905, "selectivity": 0.22499999999999998} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8150208623087621, "drop": 0.031229137691237807, "rotation": {"subspace_principal_angle": 1.2085181784579044, "mean_class_cosine": 0.3544054290791189}}, "domain": {"acc": 0.84375, "drop": 0.0024999999999999467, "rotation": {"subspace_principal_angle": 1.2026486146003328, "mean_class_cosine": 0.35988787403384076}}}, "iid_acc": 0.84625, "selectivity": 0.3499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1571656029830548, "mean_class_cosine": 0.40193650581631374}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5660639777468707, "drop": -0.014813977746870655, "rotation": {"subspace_principal_angle": 1.245440017394346, "mean_class_cosine": 0.9990012115847782}}, "domain": {"acc": 0.495, "drop": 0.05625000000000002, "rotation": {"subspace_principal_angle": 1.122537065823632, "mean_class_cosine": 0.8260255526449155}}}, "iid_acc": 0.55125, "selectivity": 0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.503465091230446, "mean_class_cosine": 0.9992316721985006}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.79375, "drop": 0.0}, "paraphrase": {"acc": 0.8108484005563282, "drop": -0.017098400556328275}, "domain": {"acc": 0.81875, "drop": -0.025000000000000022}}, "iid_acc": 0.79375, "selectivity": 0.29749999999999993} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.889795918367347, "drop": 0.0089540816326531, "rotation": {"subspace_principal_angle": 1.1767967051679535, "mean_class_cosine": 0.38388465061115085}}, "domain": {"acc": 0.62875, "drop": 0.27, "rotation": {"subspace_principal_angle": 1.383856999934436, "mean_class_cosine": 0.18585242109581535}}, "length": {"acc": 0.9057750759878419, "drop": -0.007025075987841856, "rotation": {"subspace_principal_angle": 1.1410061611227862, "mean_class_cosine": 0.41668006103594923}}}, "iid_acc": 0.89875, "selectivity": 0.37875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2153568944914745, "mean_class_cosine": 0.3480023638189218}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.67875, "drop": 0.0}, "paraphrase": {"acc": 0.6959183673469388, "drop": -0.01716836734693883, "rotation": {"subspace_principal_angle": 1.4280844473025907, "mean_class_cosine": 0.9628713037828026}}, "domain": {"acc": 0.56625, "drop": 0.11249999999999993, "rotation": {"subspace_principal_angle": 1.3603869108648967, "mean_class_cosine": 0.44643897304207514}}, "length": {"acc": 0.6990881458966566, "drop": -0.020338145896656612, "rotation": {"subspace_principal_angle": 1.3643554774870585, "mean_class_cosine": 0.9296532611805729}}}, "iid_acc": 0.67875, "selectivity": 0.15874999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4371116325199274, "mean_class_cosine": 0.8728452884225864}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.8959183673469387, "drop": 0.012831632653061198}, "domain": {"acc": 0.7025, "drop": 0.20624999999999993}, "length": {"acc": 0.9103343465045592, "drop": -0.0015843465045592975}}, "iid_acc": 0.90875, "selectivity": 0.38874999999999993} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8690851735015773, "drop": 0.013414826498422627, "rotation": {"subspace_principal_angle": 1.3839824797303943, "mean_class_cosine": 0.31779551037707243}}}, "iid_acc": 0.8825, "selectivity": 0.6425, "iid_split_rotation": {"subspace_principal_angle": 1.3316581856764766, "mean_class_cosine": 0.3540098873003071}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.64625, "drop": 0.0}, "paraphrase": {"acc": 0.6403785488958991, "drop": 0.005871451104100922, "rotation": {"subspace_principal_angle": 0.9413392447354468, "mean_class_cosine": 0.9542917678949543}}}, "iid_acc": 0.64625, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 1.3936985587921362, "mean_class_cosine": 0.957153329410988}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 1, "layer": 15, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8706624605678234, "drop": 0.014337539432176638}}, "iid_acc": 0.885, "selectivity": 0.645} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9588688946015425, "drop": -0.00011889460154246301, "rotation": {"subspace_principal_angle": 1.1552531570580125, "mean_class_cosine": 0.6468507558692805}}}, "iid_acc": 0.95875, "selectivity": 0.8825, "iid_split_rotation": {"subspace_principal_angle": 1.135852999209662, "mean_class_cosine": 0.6934419744150427}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.6349614395886889, "drop": 0.055038560411311055, "rotation": {"subspace_principal_angle": 0.9810281506834297, "mean_class_cosine": 0.8971151689162156}}}, "iid_acc": 0.69, "selectivity": 0.6137499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9002976361653219, "mean_class_cosine": 0.9196584919183077}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 1, "layer": 21, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9588688946015425, "drop": -0.00011889460154246301}}, "iid_acc": 0.95875, "selectivity": 0.8825} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.6253229974160207, "drop": -0.01782299741602067, "rotation": {"subspace_principal_angle": 1.2853027316760957, "mean_class_cosine": 0.28163110133345426}}}, "iid_acc": 0.6075, "selectivity": 0.12625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3656019976170561, "mean_class_cosine": 0.20375741617602003}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.50375, "drop": 0.0}, "paraphrase": {"acc": 0.5258397932816538, "drop": -0.022089793281653725, "rotation": {"subspace_principal_angle": 1.415941389813765, "mean_class_cosine": 0.989806795103265}}}, "iid_acc": 0.50375, "selectivity": 0.02250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.1847539410134869, "mean_class_cosine": 0.9708123793493897}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 1, "layer": 8, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5658914728682171, "drop": -0.008391472868217065}}, "iid_acc": 0.5575, "selectivity": 0.07624999999999998} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5520974289580515, "drop": 0.04665257104194853, "rotation": {"subspace_principal_angle": 1.2698836003378577, "mean_class_cosine": 0.41354171313299243}}}, "iid_acc": 0.59875, "selectivity": 0.36125, "iid_split_rotation": {"subspace_principal_angle": 1.279435072679175, "mean_class_cosine": 0.4311997525753071}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.1775, "drop": 0.0}, "paraphrase": {"acc": 0.18132611637347767, "drop": -0.0038261163734776837, "rotation": {"subspace_principal_angle": 1.2291192051808308, "mean_class_cosine": 0.9975774044324508}}}, "iid_acc": 0.1775, "selectivity": -0.06, "iid_split_rotation": {"subspace_principal_angle": 1.4905226932759024, "mean_class_cosine": 0.6650981545648424}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5128552097428958, "drop": 0.013394790257104217}}, "iid_acc": 0.52625, "selectivity": 0.28875} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7398648648648649, "drop": 0.0038851351351351093, "rotation": {"subspace_principal_angle": 1.2773857388810559, "mean_class_cosine": 0.289218730636968}}}, "iid_acc": 0.74375, "selectivity": 0.23625000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.283587337214898, "mean_class_cosine": 0.2832766461897106}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5202702702702703, "drop": -0.012770270270270334, "rotation": {"subspace_principal_angle": 1.416822487851161, "mean_class_cosine": 0.9990366010707538}}}, "iid_acc": 0.5075, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.172452398513104, "mean_class_cosine": 0.9997365426935431}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 1, "layer": 5, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7212837837837838, "drop": 0.009966216216216184}}, "iid_acc": 0.73125, "selectivity": 0.22375} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6260162601626016, "drop": 0.04773373983739837, "rotation": {"subspace_principal_angle": 1.4713221046099612, "mean_class_cosine": 0.09931025175584188}}}, "iid_acc": 0.67375, "selectivity": 0.18624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4688407807398294, "mean_class_cosine": 0.10177900098741591}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5495934959349593, "drop": -0.007093495934959337, "rotation": {"subspace_principal_angle": 1.1126001122879392, "mean_class_cosine": 0.9966307622920645}}}, "iid_acc": 0.5425, "selectivity": 0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2227358373040218, "mean_class_cosine": 0.9830887042288201}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 1, "layer": 20, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6325203252032521, "drop": 0.044979674796747915}}, "iid_acc": 0.6775, "selectivity": 0.19} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.734094616639478, "drop": 0.0009053833605220385, "rotation": {"subspace_principal_angle": 1.3384619923865202, "mean_class_cosine": 0.23024976330131475}}}, "iid_acc": 0.735, "selectivity": 0.14874999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3807433555450965, "mean_class_cosine": 0.1889109127031547}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.46375, "drop": 0.0}, "paraphrase": {"acc": 0.4926590538336052, "drop": -0.028909053833605203, "rotation": {"subspace_principal_angle": 1.240506779082087, "mean_class_cosine": 0.9979825325820296}}}, "iid_acc": 0.46375, "selectivity": -0.12250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4130621801133338, "mean_class_cosine": 0.9984335729440443}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 1, "layer": 6, "dists": {"iid": {"acc": 0.755, "drop": 0.0}, "paraphrase": {"acc": 0.7504078303425775, "drop": 0.004592169657422485}}, "iid_acc": 0.755, "selectivity": 0.16874999999999996} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.93, "drop": 0.0}, "paraphrase": {"acc": 0.9196185286103542, "drop": 0.010381471389645847, "rotation": {"subspace_principal_angle": 1.1842664562934508, "mean_class_cosine": 0.37697657827864284}}}, "iid_acc": 0.93, "selectivity": 0.42125, "iid_split_rotation": {"subspace_principal_angle": 1.1417564076798083, "mean_class_cosine": 0.4159979297955525}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.5575, "drop": 0.0}, "paraphrase": {"acc": 0.5722070844686649, "drop": -0.014707084468664866, "rotation": {"subspace_principal_angle": 0.9890820205173421, "mean_class_cosine": 0.9786448762488624}}}, "iid_acc": 0.5575, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4310370860587216, "mean_class_cosine": 0.6707511189916638}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 1, "layer": 16, "dists": {"iid": {"acc": 0.93125, "drop": 0.0}, "paraphrase": {"acc": 0.9291553133514986, "drop": 0.0020946866485014315}}, "iid_acc": 0.93125, "selectivity": 0.4225} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9900990099009901, "drop": 0.0024009900990099586, "rotation": {"subspace_principal_angle": 0.9546523855786807, "mean_class_cosine": 0.5778924853874827}}}, "iid_acc": 0.9925, "selectivity": 0.15500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.9113621026036162, "mean_class_cosine": 0.6126697953735774}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.6815181518151815, "drop": -0.017768151815181543, "rotation": {"subspace_principal_angle": 0.7769812943933525, "mean_class_cosine": 0.9999668724105906}}}, "iid_acc": 0.66375, "selectivity": -0.17375000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.4808794246553239, "mean_class_cosine": 0.999987060235844}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 1, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9867986798679867, "drop": 0.005701320132013299}}, "iid_acc": 0.9925, "selectivity": 0.15500000000000003} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.6975397973950795, "drop": 0.047460202604920454, "rotation": {"subspace_principal_angle": 1.4955638990865465, "mean_class_cosine": 0.07516147955849134}}}, "iid_acc": 0.745, "selectivity": 0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.378258975516904, "mean_class_cosine": 0.19134997405306095}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.53875, "drop": 0.0}, "paraphrase": {"acc": 0.5340086830680174, "drop": 0.004741316931982542, "rotation": {"subspace_principal_angle": 1.021497555457415, "mean_class_cosine": 0.9732173089300729}}}, "iid_acc": 0.53875, "selectivity": -0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5200003895350536, "mean_class_cosine": -0.9852001616272563}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 1, "layer": 12, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7264833574529667, "drop": -0.016483357452966696}}, "iid_acc": 0.71, "selectivity": 0.08624999999999994} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6634615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.6764705882352942, "drop": -0.013009049773755721, "rotation": {"subspace_principal_angle": 1.411172607075842, "mean_class_cosine": 0.18933079531066535}}}, "iid_acc": 0.6634615384615384, "selectivity": 0.2740384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.4720863809909137, "mean_class_cosine": 0.12696001888551056}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.5764705882352941, "drop": 0.0052601809954752055, "rotation": {"subspace_principal_angle": 1.0998469877442718, "mean_class_cosine": 0.6836666330824704}}}, "iid_acc": 0.5817307692307693, "selectivity": 0.19230769230769235, "iid_split_rotation": {"subspace_principal_angle": 1.540737942845596, "mean_class_cosine": 0.6936262779934396}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 1, "layer": 22, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.6470588235294118, "drop": 0.006787330316742057}}, "iid_acc": 0.6538461538461539, "selectivity": 0.2644230769230769} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8984485190409027, "drop": -0.0009485190409027622, "rotation": {"subspace_principal_angle": 1.2279252167452932, "mean_class_cosine": 0.3361924663049618}}}, "iid_acc": 0.8975, "selectivity": 0.15499999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.2122444869762357, "mean_class_cosine": 0.3509185358458786}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.6175, "drop": 0.0}, "paraphrase": {"acc": 0.6234132581100141, "drop": -0.005913258110014086, "rotation": {"subspace_principal_angle": 0.8729067617905422, "mean_class_cosine": 0.7807952771130344}}}, "iid_acc": 0.6175, "selectivity": -0.125, "iid_split_rotation": {"subspace_principal_angle": 0.5477116558377627, "mean_class_cosine": 0.21905717301122568}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 1, "layer": 9, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.9026798307475318, "drop": 0.003570169252468225}}, "iid_acc": 0.90625, "selectivity": 0.16374999999999995} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7503467406380028, "drop": -0.007846740638002725, "rotation": {"subspace_principal_angle": 1.1359120026803335, "mean_class_cosine": 0.4213054955523352}}, "domain": {"acc": 0.75, "drop": -0.007499999999999951, "rotation": {"subspace_principal_angle": 1.2165320665064534, "mean_class_cosine": 0.34690040746322265}}}, "iid_acc": 0.7425, "selectivity": 0.18375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1365226365220666, "mean_class_cosine": 0.4207516217391289}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6324549237170597, "drop": -0.009954923717059638, "rotation": {"subspace_principal_angle": 1.4631222901212197, "mean_class_cosine": 0.8231564279245471}}, "domain": {"acc": 0.5025, "drop": 0.1200000000000001, "rotation": {"subspace_principal_angle": 1.3032516443365474, "mean_class_cosine": 0.7093055206790797}}}, "iid_acc": 0.6225, "selectivity": 0.06375000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.0549484337600161, "mean_class_cosine": 0.8394584109810561}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.7475, "drop": 0.0}, "paraphrase": {"acc": 0.7517337031900139, "drop": -0.0042337031900138156}, "domain": {"acc": 0.7575, "drop": -0.009999999999999898}}, "iid_acc": 0.7475, "selectivity": 0.18875000000000008} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.7747572815533981, "drop": 0.02774271844660192, "rotation": {"subspace_principal_angle": 1.0704211059136206, "mean_class_cosine": 0.4797547921494105}}, "domain": {"acc": 0.5775, "drop": 0.22499999999999998, "rotation": {"subspace_principal_angle": 1.3307188056276382, "mean_class_cosine": 0.23777792496643144}}, "length": {"acc": 0.8054298642533937, "drop": -0.002929864253393699, "rotation": {"subspace_principal_angle": 1.0251786604711886, "mean_class_cosine": 0.5189461749096541}}}, "iid_acc": 0.8025, "selectivity": 0.26, "iid_split_rotation": {"subspace_principal_angle": 1.1020990971387528, "mean_class_cosine": 0.4517243926605984}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.76125, "drop": 0.0}, "paraphrase": {"acc": 0.7475728155339806, "drop": 0.013677184466019399, "rotation": {"subspace_principal_angle": 1.3667128970253954, "mean_class_cosine": 0.8708735415100642}}, "domain": {"acc": 0.6175, "drop": 0.14374999999999993, "rotation": {"subspace_principal_angle": 1.3448216837274964, "mean_class_cosine": 0.549993690079513}}, "length": {"acc": 0.751131221719457, "drop": 0.010118778280542973, "rotation": {"subspace_principal_angle": 1.5333491169148545, "mean_class_cosine": 0.8840723107383672}}}, "iid_acc": 0.76125, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.4498931415198635, "mean_class_cosine": 0.8774451765117859}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.79, "drop": 0.0}, "paraphrase": {"acc": 0.7708737864077669, "drop": 0.0191262135922331}, "domain": {"acc": 0.6, "drop": 0.19000000000000006}, "length": {"acc": 0.799396681749623, "drop": -0.009396681749622937}}, "iid_acc": 0.79, "selectivity": 0.24750000000000005} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8440944881889764, "drop": 0.014655511811023647, "rotation": {"subspace_principal_angle": 1.420395850894294, "mean_class_cosine": 0.3177960201739077}}}, "iid_acc": 0.85875, "selectivity": 0.6012500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.457737166249678, "mean_class_cosine": 0.2964930332276186}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.83, "drop": 0.0}, "paraphrase": {"acc": 0.8125984251968504, "drop": 0.017401574803149522, "rotation": {"subspace_principal_angle": 1.4786133211134285, "mean_class_cosine": 0.9691458832694383}}}, "iid_acc": 0.83, "selectivity": 0.5725, "iid_split_rotation": {"subspace_principal_angle": 1.5646208839230873, "mean_class_cosine": 0.9660745167896441}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.84251968503937, "drop": 0.02748031496062997}}, "iid_acc": 0.87, "selectivity": 0.6125} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9193954659949622, "drop": 0.019354534005037727, "rotation": {"subspace_principal_angle": 1.2541440798612775, "mean_class_cosine": 0.6719234888555385}}}, "iid_acc": 0.93875, "selectivity": 0.8875, "iid_split_rotation": {"subspace_principal_angle": 1.3380418703494907, "mean_class_cosine": 0.7071156351962342}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.84125, "drop": 0.0}, "paraphrase": {"acc": 0.8312342569269522, "drop": 0.010015743073047867, "rotation": {"subspace_principal_angle": 1.4201631598283133, "mean_class_cosine": 0.8910464593425866}}}, "iid_acc": 0.84125, "selectivity": 0.79, "iid_split_rotation": {"subspace_principal_angle": 1.5685103886275429, "mean_class_cosine": 0.9592733462262945}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.9017632241813602, "drop": 0.01698677581863972}}, "iid_acc": 0.91875, "selectivity": 0.8674999999999999} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.51875, "drop": 0.0}, "paraphrase": {"acc": 0.5058670143415906, "drop": 0.012882985658409396, "rotation": {"subspace_principal_angle": 1.4825642540594317, "mean_class_cosine": 0.08811763766369723}}}, "iid_acc": 0.51875, "selectivity": 0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.514014896298524, "mean_class_cosine": -0.056750923621074825}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.49375, "drop": 0.0}, "paraphrase": {"acc": 0.4980443285528031, "drop": -0.004294328552803095, "rotation": {"subspace_principal_angle": 1.5186826580706572, "mean_class_cosine": -0.08658206386049154}}}, "iid_acc": 0.49375, "selectivity": -0.0012499999999999734, "iid_split_rotation": {"subspace_principal_angle": 1.5356626656582177, "mean_class_cosine": 0.14967252329717184}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.516297262059974, "drop": -0.001297262059973936}}, "iid_acc": 0.515, "selectivity": 0.020000000000000018} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5391766268260292, "drop": 0.02707337317397085, "rotation": {"subspace_principal_angle": 1.3218665508999698, "mean_class_cosine": 0.40140129389793905}}}, "iid_acc": 0.56625, "selectivity": 0.30500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.4544563315042423, "mean_class_cosine": 0.3616549205752137}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.2425, "drop": 0.0}, "paraphrase": {"acc": 0.1952191235059761, "drop": 0.047280876494023905, "rotation": {"subspace_principal_angle": 1.5288782648781811, "mean_class_cosine": 0.49201628880852194}}}, "iid_acc": 0.2425, "selectivity": -0.01874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4051891188356136, "mean_class_cosine": 0.5694708832697903}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.4675, "drop": 0.0}, "paraphrase": {"acc": 0.44754316069057104, "drop": 0.019956839309428986}}, "iid_acc": 0.4675, "selectivity": 0.20625000000000004} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7249190938511327, "drop": 0.001330906148867217, "rotation": {"subspace_principal_angle": 1.1511849070923794, "mean_class_cosine": 0.40740561391162183}}}, "iid_acc": 0.72625, "selectivity": 0.16374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1756361449389974, "mean_class_cosine": 0.3849560311735334}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6075, "drop": 0.0}, "paraphrase": {"acc": 0.63915857605178, "drop": -0.03165857605177991, "rotation": {"subspace_principal_angle": 0.8931518264214847, "mean_class_cosine": 0.793022261839432}}}, "iid_acc": 0.6075, "selectivity": 0.04500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.0684485107486943, "mean_class_cosine": 0.6944069959278446}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.7313915857605178, "drop": -0.015141585760517717}}, "iid_acc": 0.71625, "selectivity": 0.15375000000000005} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6494345718901454, "drop": 0.020565428109854622, "rotation": {"subspace_principal_angle": 1.326914373237704, "mean_class_cosine": 0.2414715148282628}}}, "iid_acc": 0.67, "selectivity": 0.19250000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.297386449224978, "mean_class_cosine": 0.27001622040531514}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.6187399030694669, "drop": 0.007510096930533061, "rotation": {"subspace_principal_angle": 1.1476051561016243, "mean_class_cosine": 0.909033474781661}}}, "iid_acc": 0.62625, "selectivity": 0.14875, "iid_split_rotation": {"subspace_principal_angle": 1.4081586003059547, "mean_class_cosine": 0.8941316972733946}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6625, "drop": 0.0}, "paraphrase": {"acc": 0.6235864297253635, "drop": 0.03891357027463649}}, "iid_acc": 0.6625, "selectivity": 0.185} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7405582922824302, "drop": -0.011808292282430188, "rotation": {"subspace_principal_angle": 1.1923774338443787, "mean_class_cosine": 0.36945168819099083}}}, "iid_acc": 0.72875, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2706314343701106, "mean_class_cosine": 0.29567773039327916}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.54375, "drop": 0.0}, "paraphrase": {"acc": 0.5500821018062397, "drop": -0.006332101806239776, "rotation": {"subspace_principal_angle": 1.539326603180651, "mean_class_cosine": 0.8691103147755386}}}, "iid_acc": 0.54375, "selectivity": -0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3002024261544294, "mean_class_cosine": 0.7081293912566515}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7274220032840722, "drop": -0.00867200328407225}}, "iid_acc": 0.71875, "selectivity": 0.12375000000000003} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8363384188626907, "drop": 0.05491158113730932, "rotation": {"subspace_principal_angle": 0.9678914596829231, "mean_class_cosine": 0.5670375780110272}}}, "iid_acc": 0.89125, "selectivity": 0.435, "iid_split_rotation": {"subspace_principal_angle": 0.9419088602911864, "mean_class_cosine": 0.5882454358575304}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.02100728155339804, "rotation": {"subspace_principal_angle": 1.2291895890809326, "mean_class_cosine": 0.9098769842554898}}}, "iid_acc": 0.84625, "selectivity": 0.38999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3955051971962031, "mean_class_cosine": 0.9616189776316897}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8626907073509015, "drop": 0.016059292649098555}}, "iid_acc": 0.87875, "selectivity": 0.42250000000000004} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0034102141680395492, "rotation": {"subspace_principal_angle": 0.8491797480583015, "mean_class_cosine": 0.6605991630042665}}}, "iid_acc": 0.99, "selectivity": 0.15374999999999994, "iid_split_rotation": {"subspace_principal_angle": 0.7965198420896995, "mean_class_cosine": 0.69919899770492}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.5420098846787479, "drop": 0.1567401153212521, "rotation": {"subspace_principal_angle": 1.4157293467420775, "mean_class_cosine": 0.9954481825771943}}}, "iid_acc": 0.69875, "selectivity": -0.13750000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.4943111294129614, "mean_class_cosine": 0.9994355209688002}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.98625, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": -0.0005704283360791207}}, "iid_acc": 0.98625, "selectivity": 0.1499999999999999} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6816269284712483, "drop": -0.006626928471248239, "rotation": {"subspace_principal_angle": 1.493369684917862, "mean_class_cosine": 0.07734930442811239}}}, "iid_acc": 0.675, "selectivity": 0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.4475486480920703, "mean_class_cosine": 0.12293589375679194}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.517531556802244, "drop": -0.005031556802244097, "rotation": {"subspace_principal_angle": 1.407007437734732, "mean_class_cosine": 0.1636007713519359}}}, "iid_acc": 0.5125, "selectivity": -0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4687784126101222, "mean_class_cosine": 0.14581035402784961}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.7110799438990182, "drop": -0.009829943899018145}}, "iid_acc": 0.70125, "selectivity": 0.04125000000000001} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6214689265536724, "drop": 0.01795415036940451, "rotation": {"subspace_principal_angle": 1.4099267955197121, "mean_class_cosine": 0.21703022459825547}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.19230769230769224, "iid_split_rotation": {"subspace_principal_angle": 1.4925996250553264, "mean_class_cosine": 0.20125792019838662}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.5769230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.519774011299435, "drop": 0.05714906562364186, "rotation": {"subspace_principal_angle": 1.5438703377286263, "mean_class_cosine": 0.6081265892996393}}}, "iid_acc": 0.5769230769230769, "selectivity": 0.12980769230769224, "iid_split_rotation": {"subspace_principal_angle": 1.367531053578908, "mean_class_cosine": 0.5070016095884643}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6101694915254238, "drop": -0.004400260756193042}}, "iid_acc": 0.6057692307692307, "selectivity": 0.1586538461538461} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8849315068493151, "drop": 0.017568493150684872, "rotation": {"subspace_principal_angle": 1.171653684236408, "mean_class_cosine": 0.3886285207848383}}}, "iid_acc": 0.9025, "selectivity": 0.13249999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1155850933836575, "mean_class_cosine": 0.43965204674130126}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.6, "drop": 0.0025000000000000577, "rotation": {"subspace_principal_angle": 1.1040454036261176, "mean_class_cosine": 0.827992833300117}}}, "iid_acc": 0.6025, "selectivity": -0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.126081986855619, "mean_class_cosine": 0.6844596631580138}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8479452054794521, "drop": 0.01955479452054798}}, "iid_acc": 0.8675, "selectivity": 0.09750000000000003} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.76625, "drop": 0.0}, "paraphrase": {"acc": 0.7503467406380028, "drop": 0.015903259361997213, "rotation": {"subspace_principal_angle": 1.2612626710199533, "mean_class_cosine": 0.3046144880272891}}, "domain": {"acc": 0.72875, "drop": 0.03749999999999998, "rotation": {"subspace_principal_angle": 1.3058544057767159, "mean_class_cosine": 0.2618532161180114}}}, "iid_acc": 0.76625, "selectivity": 0.2875, "iid_split_rotation": {"subspace_principal_angle": 1.2715880951662406, "mean_class_cosine": 0.294763708893907}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.49930651872399445, "drop": 0.0031934812760054943, "rotation": {"subspace_principal_angle": 1.1711657362498487, "mean_class_cosine": 0.9874129557052536}}, "domain": {"acc": 0.49875, "drop": 0.00374999999999992, "rotation": {"subspace_principal_angle": 1.5016858906817072, "mean_class_cosine": 0.3434093433034723}}}, "iid_acc": 0.5025, "selectivity": 0.023749999999999938, "iid_split_rotation": {"subspace_principal_angle": 1.2751956782954972, "mean_class_cosine": 0.972007963226671}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.765, "drop": 0.0}, "paraphrase": {"acc": 0.7600554785020804, "drop": 0.004944521497919574}, "domain": {"acc": 0.72375, "drop": 0.04125000000000001}}, "iid_acc": 0.765, "selectivity": 0.28625} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.829126213592233, "drop": 0.02212378640776691, "rotation": {"subspace_principal_angle": 1.1254573688815634, "mean_class_cosine": 0.43076380259456154}}, "domain": {"acc": 0.53625, "drop": 0.31499999999999995, "rotation": {"subspace_principal_angle": 1.3011899623729255, "mean_class_cosine": 0.2663520415198698}}, "length": {"acc": 0.8597285067873304, "drop": -0.00847850678733042, "rotation": {"subspace_principal_angle": 1.0844876622709578, "mean_class_cosine": 0.4673657026245072}}}, "iid_acc": 0.85125, "selectivity": 0.36624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1313772551394048, "mean_class_cosine": 0.4254137973371145}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7320388349514563, "drop": -0.0007888349514563187, "rotation": {"subspace_principal_angle": 1.3889114115978614, "mean_class_cosine": 0.9101913863100519}}, "domain": {"acc": 0.58, "drop": 0.15125, "rotation": {"subspace_principal_angle": 1.2835193898835013, "mean_class_cosine": 0.5072723519659894}}, "length": {"acc": 0.7450980392156863, "drop": -0.013848039215686336, "rotation": {"subspace_principal_angle": 1.474745277649587, "mean_class_cosine": 0.9218320832330005}}}, "iid_acc": 0.73125, "selectivity": 0.24624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4693373559237952, "mean_class_cosine": 0.8923141074617422}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.01725728155339812}, "domain": {"acc": 0.5775, "drop": 0.265}, "length": {"acc": 0.8476621417797888, "drop": -0.005162141779788798}}, "iid_acc": 0.8425, "selectivity": 0.35750000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8677165354330708, "drop": -0.00021653543307076362, "rotation": {"subspace_principal_angle": 1.3138102642722838, "mean_class_cosine": 0.4178222215166151}}}, "iid_acc": 0.8675, "selectivity": 0.6100000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.269543032505647, "mean_class_cosine": 0.41884051584646115}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.59375, "drop": 0.0}, "paraphrase": {"acc": 0.5748031496062992, "drop": 0.01894685039370081, "rotation": {"subspace_principal_angle": 1.462647268843256, "mean_class_cosine": 0.9428914975925955}}}, "iid_acc": 0.59375, "selectivity": 0.33625, "iid_split_rotation": {"subspace_principal_angle": 1.5667150391990796, "mean_class_cosine": 0.7928532264394269}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8566929133858268, "drop": 0.01455708661417321}}, "iid_acc": 0.87125, "selectivity": 0.61375} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.965, "drop": 0.0}, "paraphrase": {"acc": 0.9319899244332494, "drop": 0.03301007556675062, "rotation": {"subspace_principal_angle": 1.2176081199133377, "mean_class_cosine": 0.6442873004051297}}}, "iid_acc": 0.965, "selectivity": 0.9037499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1385695404879772, "mean_class_cosine": 0.7109378375596532}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.6801007556675063, "drop": 0.04114924433249367, "rotation": {"subspace_principal_angle": 1.5679042974829251, "mean_class_cosine": 0.8621074777728631}}}, "iid_acc": 0.72125, "selectivity": 0.6599999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2212236304536004, "mean_class_cosine": 0.9460677922131192}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9395465994962217, "drop": 0.02170340050377839}}, "iid_acc": 0.96125, "selectivity": 0.9} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5071707953063885, "drop": 0.02907920469361147, "rotation": {"subspace_principal_angle": 1.4838566858436246, "mean_class_cosine": 0.08683016010539553}}}, "iid_acc": 0.53625, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.465396612413356, "mean_class_cosine": 0.10520467342535929}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.49674054758800523, "drop": -0.021740547588005255, "rotation": {"subspace_principal_angle": 1.5469172283064518, "mean_class_cosine": -0.5397626183074191}}}, "iid_acc": 0.475, "selectivity": -0.03375000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.408080647568939, "mean_class_cosine": 0.4860859905388361}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4876140808344198, "drop": 0.01488591916558013}}, "iid_acc": 0.5025, "selectivity": -0.006250000000000089} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5524568393094289, "drop": 0.058793160690571056, "rotation": {"subspace_principal_angle": 1.3850534905313705, "mean_class_cosine": 0.28991503737744245}}}, "iid_acc": 0.61125, "selectivity": 0.37749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.419189771224309, "mean_class_cosine": 0.29718954050314617}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.0925, "drop": 0.0}, "paraphrase": {"acc": 0.1049136786188579, "drop": -0.012413678618857907, "rotation": {"subspace_principal_angle": 1.540744396600694, "mean_class_cosine": 0.06532846024974738}}}, "iid_acc": 0.0925, "selectivity": -0.14125000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4223489889510221, "mean_class_cosine": 0.5690159019187561}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.535, "drop": 0.0}, "paraphrase": {"acc": 0.5232403718459495, "drop": 0.011759628154050517}}, "iid_acc": 0.535, "selectivity": 0.30125} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.73875, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": -0.0007322006472492149, "rotation": {"subspace_principal_angle": 1.3881466203882136, "mean_class_cosine": 0.18163583883336032}}}, "iid_acc": 0.73875, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3959757692399888, "mean_class_cosine": 0.17393143305355055}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5906148867313916, "drop": -0.0393648867313916, "rotation": {"subspace_principal_angle": 1.4683816158112597, "mean_class_cosine": 0.9483792840431817}}}, "iid_acc": 0.55125, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.3971642089849379, "mean_class_cosine": 0.9161271445117759}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7540453074433657, "drop": -0.0027953074433657576}}, "iid_acc": 0.75125, "selectivity": 0.20875} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6397415185783522, "drop": 0.044008481421647816, "rotation": {"subspace_principal_angle": 1.3758917619051532, "mean_class_cosine": 0.19367290768765433}}}, "iid_acc": 0.68375, "selectivity": 0.18374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.345477898881277, "mean_class_cosine": 0.22341675246913884}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.5993537964458805, "drop": 0.011896203554119467, "rotation": {"subspace_principal_angle": 1.4573232614008496, "mean_class_cosine": 0.870794766106819}}}, "iid_acc": 0.61125, "selectivity": 0.11124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.385445998953203, "mean_class_cosine": 0.8535614192902857}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.6397415185783522, "drop": 0.046508481421647874}}, "iid_acc": 0.68625, "selectivity": 0.18625000000000003} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7405582922824302, "drop": -0.013058292282430162, "rotation": {"subspace_principal_angle": 1.3879649771162461, "mean_class_cosine": 0.1818144576289447}}}, "iid_acc": 0.7275, "selectivity": 0.1312500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4079457648518725, "mean_class_cosine": 0.16213170804038407}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.50375, "drop": 0.0}, "paraphrase": {"acc": 0.5320197044334976, "drop": -0.028269704433497522, "rotation": {"subspace_principal_angle": 1.5607306280375703, "mean_class_cosine": 0.9876176460724118}}}, "iid_acc": 0.50375, "selectivity": -0.09249999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.166861107011101, "mean_class_cosine": 0.9302660401353273}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7422003284072249, "drop": 0.010299671592775006}}, "iid_acc": 0.7525, "selectivity": 0.15625} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8793342579750347, "drop": 0.020665742024965295, "rotation": {"subspace_principal_angle": 1.1063584135999445, "mean_class_cosine": 0.4479203253262932}}}, "iid_acc": 0.9, "selectivity": 0.46125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0249583822746835, "mean_class_cosine": 0.5191344576441722}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.5728155339805825, "drop": 0.14468446601941753, "rotation": {"subspace_principal_angle": 1.284951624986405, "mean_class_cosine": 0.8542882000899992}}}, "iid_acc": 0.7175, "selectivity": 0.27875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0904185384770684, "mean_class_cosine": 0.9172822881817656}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8904299583911235, "drop": 0.00832004160887656}}, "iid_acc": 0.89875, "selectivity": 0.4600000000000001} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9967051070840197, "drop": -0.00045510708401974576, "rotation": {"subspace_principal_angle": 0.8815453404218065, "mean_class_cosine": 0.6359593299455631}}}, "iid_acc": 0.99625, "selectivity": 0.16874999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.829603647498299, "mean_class_cosine": 0.6751681879987259}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.5634266886326195, "drop": 0.12157331136738059, "rotation": {"subspace_principal_angle": 1.1147774126986099, "mean_class_cosine": 0.9975604664853437}}}, "iid_acc": 0.685, "selectivity": -0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.351613895560496, "mean_class_cosine": 0.9997571774292637}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9950576606260296, "drop": -0.005057660626029636}}, "iid_acc": 0.99, "selectivity": 0.16249999999999998} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6633941093969145, "drop": 0.01910589060308554, "rotation": {"subspace_principal_angle": 1.5456837867774578, "mean_class_cosine": 0.02510990060669071}}}, "iid_acc": 0.6825, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4420552406446043, "mean_class_cosine": 0.12838574922234897}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.46875, "drop": 0.0}, "paraphrase": {"acc": 0.4950911640953717, "drop": -0.026341164095371683, "rotation": {"subspace_principal_angle": 0.8450834965822174, "mean_class_cosine": 0.8344612624127071}}}, "iid_acc": 0.46875, "selectivity": -0.14375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.447600013734126, "mean_class_cosine": -0.6148431774148773}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7237026647966339, "drop": -0.009952664796633925}}, "iid_acc": 0.71375, "selectivity": 0.10124999999999995} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.615819209039548, "drop": 0.00918079096045199, "rotation": {"subspace_principal_angle": 1.344146750268554, "mean_class_cosine": 0.24255494243559642}}}, "iid_acc": 0.625, "selectivity": 0.16346153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.4369180768778205, "mean_class_cosine": 0.21822584946678156}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5423728813559322, "drop": 0.0201271186440678, "rotation": {"subspace_principal_angle": 1.5227921354385472, "mean_class_cosine": 0.6841529664773013}}}, "iid_acc": 0.5625, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.2545192468784838, "mean_class_cosine": 0.5447114834286542}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.632768361581921, "drop": -0.02699913081269023}}, "iid_acc": 0.6057692307692307, "selectivity": 0.14423076923076916} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8945205479452055, "drop": 0.004229452054794525, "rotation": {"subspace_principal_angle": 1.248973078003772, "mean_class_cosine": 0.3162967291391359}}}, "iid_acc": 0.89875, "selectivity": 0.15250000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.2122862189044603, "mean_class_cosine": 0.35087945751268557}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5767123287671233, "drop": -0.009212328767123301, "rotation": {"subspace_principal_angle": 1.21580262592107, "mean_class_cosine": 0.8869377958216829}}}, "iid_acc": 0.5675, "selectivity": -0.17874999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.839488040678943, "mean_class_cosine": 0.7546602597663057}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8767123287671232, "drop": 0.012037671232876801}}, "iid_acc": 0.88875, "selectivity": 0.14250000000000007} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7739251040221914, "drop": 0.029824895977808574, "rotation": {"subspace_principal_angle": 1.2659109438320224, "mean_class_cosine": 0.30018384612544047}}, "domain": {"acc": 0.76875, "drop": 0.03499999999999992, "rotation": {"subspace_principal_angle": 1.34258356435988, "mean_class_cosine": 0.22623698719981744}}}, "iid_acc": 0.80375, "selectivity": 0.32499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2506624679779867, "mean_class_cosine": 0.3146936213275413}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5034674063800277, "drop": -0.004717406380027711, "rotation": {"subspace_principal_angle": 1.239282670172901, "mean_class_cosine": 0.9986892573730854}}, "domain": {"acc": 0.5, "drop": -0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.3097283989710364, "mean_class_cosine": 0.1603346495077691}}}, "iid_acc": 0.49875, "selectivity": 0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.420934784951417, "mean_class_cosine": 0.9934847558506945}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.8030513176144244, "drop": 0.019448682385575622}, "domain": {"acc": 0.81625, "drop": 0.006249999999999978}}, "iid_acc": 0.8225, "selectivity": 0.34375} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.8524271844660194, "drop": 0.02382281553398058, "rotation": {"subspace_principal_angle": 1.220154188841714, "mean_class_cosine": 0.34350094358966443}}, "domain": {"acc": 0.4975, "drop": 0.37875, "rotation": {"subspace_principal_angle": 1.438605482929702, "mean_class_cosine": 0.13180618706503908}}, "length": {"acc": 0.8808446455505279, "drop": -0.004594645550527909, "rotation": {"subspace_principal_angle": 1.1893943688634518, "mean_class_cosine": 0.37222205321854485}}}, "iid_acc": 0.87625, "selectivity": 0.38625, "iid_split_rotation": {"subspace_principal_angle": 1.1874268592484474, "mean_class_cosine": 0.3740474632573129}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.8425, "drop": 0.0}, "paraphrase": {"acc": 0.829126213592233, "drop": 0.013373786407766985, "rotation": {"subspace_principal_angle": 1.525376087221384, "mean_class_cosine": 0.942182847156327}}, "domain": {"acc": 0.57875, "drop": 0.26375000000000004, "rotation": {"subspace_principal_angle": 1.436822367907557, "mean_class_cosine": 0.27136585793969137}}, "length": {"acc": 0.8536953242835595, "drop": -0.011195324283559516, "rotation": {"subspace_principal_angle": 1.4983468510851918, "mean_class_cosine": 0.9461442600007871}}}, "iid_acc": 0.8425, "selectivity": 0.35250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.287421066876229, "mean_class_cosine": 0.910695802907292}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8640776699029126, "drop": 0.015922330097087434}, "domain": {"acc": 0.505, "drop": 0.375}, "length": {"acc": 0.8883861236802413, "drop": -0.008386123680241275}}, "iid_acc": 0.88, "selectivity": 0.39} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8582677165354331, "drop": 0.015482283464566926, "rotation": {"subspace_principal_angle": 1.2827164275902705, "mean_class_cosine": 0.4163674814452723}}}, "iid_acc": 0.87375, "selectivity": 0.6000000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.26135377780834, "mean_class_cosine": 0.4442699111544142}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7975, "drop": 0.0}, "paraphrase": {"acc": 0.7968503937007874, "drop": 0.0006496062992126239, "rotation": {"subspace_principal_angle": 1.424694638770457, "mean_class_cosine": 0.9469278805103363}}}, "iid_acc": 0.7975, "selectivity": 0.5237499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4837220741637616, "mean_class_cosine": 0.9518191309197193}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8598425196850393, "drop": 0.017657480314960616}}, "iid_acc": 0.8775, "selectivity": 0.60375} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9672544080604534, "drop": 0.00024559193954665215, "rotation": {"subspace_principal_angle": 1.1145296779462641, "mean_class_cosine": 0.7020277810816539}}}, "iid_acc": 0.9675, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.5624189489347664, "mean_class_cosine": 0.7403838209375735}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.44125, "drop": 0.0}, "paraphrase": {"acc": 0.4256926952141058, "drop": 0.015557304785894177, "rotation": {"subspace_principal_angle": 1.033444992470805, "mean_class_cosine": 0.7765287771414837}}}, "iid_acc": 0.44125, "selectivity": 0.38, "iid_split_rotation": {"subspace_principal_angle": 1.2853261455962783, "mean_class_cosine": 0.8790932160427343}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.96375, "drop": 0.0}, "paraphrase": {"acc": 0.9571788413098237, "drop": 0.006571158690176326}}, "iid_acc": 0.96375, "selectivity": 0.9025} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5397653194263363, "drop": 0.013984680573663644, "rotation": {"subspace_principal_angle": 1.5274443365773367, "mean_class_cosine": 0.0433384122408627}}}, "iid_acc": 0.55375, "selectivity": 0.06999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5117604681674472, "mean_class_cosine": 0.05900157231956833}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.5071707953063885, "drop": -0.029670795306388553, "rotation": {"subspace_principal_angle": 0.8759468408895317, "mean_class_cosine": -0.6781424224487453}}}, "iid_acc": 0.4775, "selectivity": -0.006250000000000033, "iid_split_rotation": {"subspace_principal_angle": 1.0844620330806314, "mean_class_cosine": -0.4769926683134672}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.49282920469361147, "drop": 0.02717079530638855}}, "iid_acc": 0.52, "selectivity": 0.036250000000000004} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.62875, "drop": 0.0}, "paraphrase": {"acc": 0.5803452855245684, "drop": 0.0484047144754316, "rotation": {"subspace_principal_angle": 1.4145413502440674, "mean_class_cosine": 0.27663015100707133}}}, "iid_acc": 0.62875, "selectivity": 0.37125, "iid_split_rotation": {"subspace_principal_angle": 1.4392990909748595, "mean_class_cosine": 0.2769217490323135}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.09, "drop": 0.0}, "paraphrase": {"acc": 0.07436918990703852, "drop": 0.01563081009296148, "rotation": {"subspace_principal_angle": 1.3350767969885171, "mean_class_cosine": -0.06735755908616899}}}, "iid_acc": 0.09, "selectivity": -0.1675, "iid_split_rotation": {"subspace_principal_angle": 1.2649936899279066, "mean_class_cosine": 0.6127869206852473}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5391766268260292, "drop": 0.01332337317397081}}, "iid_acc": 0.5525, "selectivity": 0.295} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.7425, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": 0.0030177993527508162, "rotation": {"subspace_principal_angle": 1.3800650372385295, "mean_class_cosine": 0.18957697385454825}}}, "iid_acc": 0.7425, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3745352340470285, "mean_class_cosine": 0.19500357265711427}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.64125, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.025416666666666643, "rotation": {"subspace_principal_angle": 1.2209077466423848, "mean_class_cosine": 0.8479639619526143}}}, "iid_acc": 0.64125, "selectivity": 0.09124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.563353933733748, "mean_class_cosine": 0.8055681849044787}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.76, "drop": 0.0}, "paraphrase": {"acc": 0.7572815533980582, "drop": 0.0027184466019417597}}, "iid_acc": 0.76, "selectivity": 0.20999999999999996} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.65375, "drop": 0.0}, "paraphrase": {"acc": 0.6155088852988692, "drop": 0.038241114701130896, "rotation": {"subspace_principal_angle": 1.4613251262739329, "mean_class_cosine": 0.10925268204629429}}}, "iid_acc": 0.65375, "selectivity": 0.1462500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.467586611767328, "mean_class_cosine": 0.10302657672993193}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.6125, "drop": 0.0}, "paraphrase": {"acc": 0.5912762520193862, "drop": 0.021223747980613883, "rotation": {"subspace_principal_angle": 1.2485040409587074, "mean_class_cosine": 0.866292700063821}}}, "iid_acc": 0.6125, "selectivity": 0.1050000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5556750038869702, "mean_class_cosine": 0.8382504448687405}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6510500807754442, "drop": 0.03269991922455573}}, "iid_acc": 0.68375, "selectivity": 0.17625000000000002} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.722495894909688, "drop": -0.0024958949096880456, "rotation": {"subspace_principal_angle": 1.4020686555705464, "mean_class_cosine": 0.1679282246076349}}}, "iid_acc": 0.72, "selectivity": 0.13374999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3616761916383724, "mean_class_cosine": 0.20759928424855742}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.56375, "drop": 0.0}, "paraphrase": {"acc": 0.5369458128078818, "drop": 0.02680418719211819, "rotation": {"subspace_principal_angle": 1.525739498161781, "mean_class_cosine": 0.8790506170094416}}}, "iid_acc": 0.56375, "selectivity": -0.022500000000000075, "iid_split_rotation": {"subspace_principal_angle": 1.355838296904698, "mean_class_cosine": 0.7623898156919341}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7504105090311987, "drop": 0.0008394909688013152}}, "iid_acc": 0.75125, "selectivity": 0.16499999999999992} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.897364771151179, "drop": 0.02388522884882105, "rotation": {"subspace_principal_angle": 1.2009122822029357, "mean_class_cosine": 0.3615073211363401}}}, "iid_acc": 0.92125, "selectivity": 0.45375, "iid_split_rotation": {"subspace_principal_angle": 1.1365147935104427, "mean_class_cosine": 0.42075873671714054}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7253814147018031, "drop": 0.008368585298196951, "rotation": {"subspace_principal_angle": 1.382859019749326, "mean_class_cosine": 0.9518475699667834}}}, "iid_acc": 0.73375, "selectivity": 0.26625, "iid_split_rotation": {"subspace_principal_angle": 0.8342063363329182, "mean_class_cosine": 0.8834390160996848}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.92625, "drop": 0.0}, "paraphrase": {"acc": 0.9140083217753121, "drop": 0.012241678224687913}}, "iid_acc": 0.92625, "selectivity": 0.45875} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.0007372322899505956, "rotation": {"subspace_principal_angle": 0.9022113956514256, "mean_class_cosine": 0.6198762040410312}}}, "iid_acc": 0.9925, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 0.7803136404078311, "mean_class_cosine": 0.7106929262056456}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.69, "drop": 0.0}, "paraphrase": {"acc": 0.4744645799011532, "drop": 0.21553542009884674, "rotation": {"subspace_principal_angle": 1.5013958950641693, "mean_class_cosine": 0.997742196187771}}}, "iid_acc": 0.69, "selectivity": -0.1362500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3648075725684445, "mean_class_cosine": 0.9999469425446367}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": -0.00011532125205926391}}, "iid_acc": 0.99, "selectivity": 0.16374999999999995} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6900420757363254, "drop": 0.019957924263674598, "rotation": {"subspace_principal_angle": 1.4053917466501165, "mean_class_cosine": 0.16465140281778454}}}, "iid_acc": 0.71, "selectivity": 0.10249999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.4198215714680307, "mean_class_cosine": 0.15040187120207954}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.511921458625526, "drop": -0.03442145862552598, "rotation": {"subspace_principal_angle": 1.088765737718087, "mean_class_cosine": -0.46901391020350736}}}, "iid_acc": 0.4775, "selectivity": -0.13000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4869971047307045, "mean_class_cosine": 0.19635370929712698}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7223001402524544, "drop": -0.012300140252454406}}, "iid_acc": 0.71, "selectivity": 0.10249999999999992} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6538461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.672316384180791, "drop": -0.018470230334637128, "rotation": {"subspace_principal_angle": 1.4095901786238862, "mean_class_cosine": 0.19654959439481243}}}, "iid_acc": 0.6538461538461539, "selectivity": 0.27403846153846156, "iid_split_rotation": {"subspace_principal_angle": 1.5567476235509987, "mean_class_cosine": 0.20024194309517288}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.423728813559322, "drop": 0.03780964797913955, "rotation": {"subspace_principal_angle": 1.318674137405094, "mean_class_cosine": 0.5319901746483242}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.08173076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.1328152149633164, "mean_class_cosine": 0.6908347513167752}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6201923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": -0.01822577140373749}}, "iid_acc": 0.6201923076923077, "selectivity": 0.24038461538461542} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8780821917808219, "drop": 0.02066780821917813, "rotation": {"subspace_principal_angle": 1.4022671366129482, "mean_class_cosine": 0.16773255884346316}}}, "iid_acc": 0.89875, "selectivity": 0.17500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3156344568426237, "mean_class_cosine": 0.25240204094053126}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.6025, "drop": 0.0}, "paraphrase": {"acc": 0.584931506849315, "drop": 0.017568493150684983, "rotation": {"subspace_principal_angle": 1.2909886826134447, "mean_class_cosine": 0.8167709778669217}}}, "iid_acc": 0.6025, "selectivity": -0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.181444552949185, "mean_class_cosine": 0.8742409810564655}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 18, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.8972602739726028, "drop": 0.013989726027397231}}, "iid_acc": 0.91125, "selectivity": 0.1875} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.7988904299583911, "drop": 0.047359570041608845, "rotation": {"subspace_principal_angle": 1.307852975470961, "mean_class_cosine": 0.25992385942235496}}, "domain": {"acc": 0.865, "drop": -0.018750000000000044, "rotation": {"subspace_principal_angle": 1.4101720618034508, "mean_class_cosine": 0.15993446696213712}}}, "iid_acc": 0.84625, "selectivity": 0.3374999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2803570242694347, "mean_class_cosine": 0.2863731564533277}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5034674063800277, "drop": -0.004717406380027711, "rotation": {"subspace_principal_angle": 1.4967653281587188, "mean_class_cosine": 0.9956396889374676}}, "domain": {"acc": 0.5, "drop": -0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5195064626946009, "mean_class_cosine": 0.17206983639586906}}}, "iid_acc": 0.49875, "selectivity": -0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.53305740516774, "mean_class_cosine": 0.9726130574764482}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.855, "drop": 0.0}, "paraphrase": {"acc": 0.812760055478502, "drop": 0.04223994452149793}, "domain": {"acc": 0.84375, "drop": 0.011249999999999982}}, "iid_acc": 0.855, "selectivity": 0.34624999999999995} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8815533980582524, "drop": 0.020946601941747578, "rotation": {"subspace_principal_angle": 1.3231535337909477, "mean_class_cosine": 0.24511934689015158}}, "domain": {"acc": 0.56, "drop": 0.3424999999999999, "rotation": {"subspace_principal_angle": 1.4299398396010543, "mean_class_cosine": 0.14039117067809936}}, "length": {"acc": 0.9140271493212669, "drop": -0.011527149321266972, "rotation": {"subspace_principal_angle": 1.2714271775831436, "mean_class_cosine": 0.2949174731196204}}}, "iid_acc": 0.9025, "selectivity": 0.43749999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.3312504683770023, "mean_class_cosine": 0.23726147696520083}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8524271844660194, "drop": 0.03507281553398056, "rotation": {"subspace_principal_angle": 1.1382054241911463, "mean_class_cosine": 0.9322577478917091}}, "domain": {"acc": 0.57875, "drop": 0.30874999999999997, "rotation": {"subspace_principal_angle": 1.5632845119627006, "mean_class_cosine": 0.34081581535856803}}, "length": {"acc": 0.8883861236802413, "drop": -0.0008861236802413242, "rotation": {"subspace_principal_angle": 1.098044803005494, "mean_class_cosine": 0.9434252502763861}}}, "iid_acc": 0.8875, "selectivity": 0.42249999999999993, "iid_split_rotation": {"subspace_principal_angle": 0.7127743899001, "mean_class_cosine": 0.9390001647602704}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.90375, "drop": 0.0}, "paraphrase": {"acc": 0.8990291262135922, "drop": 0.004720873786407842}, "domain": {"acc": 0.73125, "drop": 0.1725000000000001}, "length": {"acc": 0.9155354449472096, "drop": -0.011785444947209567}}, "iid_acc": 0.90375, "selectivity": 0.43875000000000003} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8866141732283465, "drop": 0.005885826771653502, "rotation": {"subspace_principal_angle": 1.3590153860798901, "mean_class_cosine": 0.3384795168112319}}}, "iid_acc": 0.8925, "selectivity": 0.67375, "iid_split_rotation": {"subspace_principal_angle": 1.5137713950823037, "mean_class_cosine": 0.34878144427948704}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6661417322834645, "drop": 0.021358267716535484, "rotation": {"subspace_principal_angle": 1.2049572219735756, "mean_class_cosine": 0.9516923360218306}}}, "iid_acc": 0.6875, "selectivity": 0.46875, "iid_split_rotation": {"subspace_principal_angle": 1.161004465491632, "mean_class_cosine": 0.8826822088496902}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 10, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8771653543307086, "drop": 0.017834645669291382}}, "iid_acc": 0.895, "selectivity": 0.67625} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.964735516372796, "drop": 0.010264483627204002, "rotation": {"subspace_principal_angle": 1.110589965959208, "mean_class_cosine": 0.671095672439944}}}, "iid_acc": 0.975, "selectivity": 0.92375, "iid_split_rotation": {"subspace_principal_angle": 1.0922410354486654, "mean_class_cosine": 0.7234153961945159}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.84375, "drop": 0.0}, "paraphrase": {"acc": 0.8488664987405542, "drop": -0.005116498740554198, "rotation": {"subspace_principal_angle": 1.4098145943553153, "mean_class_cosine": 0.8856785823869348}}}, "iid_acc": 0.84375, "selectivity": 0.7925, "iid_split_rotation": {"subspace_principal_angle": 0.8542184898946922, "mean_class_cosine": 0.9572548062279118}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.975, "drop": 0.0}, "paraphrase": {"acc": 0.9596977329974811, "drop": 0.01530226700251891}}, "iid_acc": 0.975, "selectivity": 0.92375} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.58, "drop": 0.0}, "paraphrase": {"acc": 0.5736636245110821, "drop": 0.006336375488917856, "rotation": {"subspace_principal_angle": 1.4824473599970098, "mean_class_cosine": 0.08823407641471295}}}, "iid_acc": 0.58, "selectivity": 0.05499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4950406885229706, "mean_class_cosine": 0.07568319984431088}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.5058670143415906, "drop": -0.02961701434159064, "rotation": {"subspace_principal_angle": 1.1493594930263484, "mean_class_cosine": -0.4424939882567903}}}, "iid_acc": 0.47625, "selectivity": -0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.551473168662097, "mean_class_cosine": 0.01234111003206595}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5867014341590613, "drop": -0.00045143415906123696}}, "iid_acc": 0.58625, "selectivity": 0.06125000000000003} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6148738379814077, "drop": 0.03512616201859231, "rotation": {"subspace_principal_angle": 1.4073412354609784, "mean_class_cosine": 0.26349298302241686}}}, "iid_acc": 0.65, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.413824435567323, "mean_class_cosine": 0.24889219745848057}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.095, "drop": 0.0}, "paraphrase": {"acc": 0.10092961487383798, "drop": -0.005929614873837974, "rotation": {"subspace_principal_angle": 1.331943855465791, "mean_class_cosine": 0.03732221069944747}}}, "iid_acc": 0.095, "selectivity": -0.165, "iid_split_rotation": {"subspace_principal_angle": 1.5577947437159871, "mean_class_cosine": 0.5674187824840483}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.64, "drop": 0.0}, "paraphrase": {"acc": 0.6347941567065073, "drop": 0.005205843293492718}}, "iid_acc": 0.64, "selectivity": 0.38} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.77625, "drop": 0.0}, "paraphrase": {"acc": 0.7427184466019418, "drop": 0.033531553398058245, "rotation": {"subspace_principal_angle": 1.4430864610974254, "mean_class_cosine": 0.12736299475185067}}}, "iid_acc": 0.77625, "selectivity": 0.22499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4358406552905756, "mean_class_cosine": 0.13454638570974817}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5598705501618123, "drop": -0.02237055016181233, "rotation": {"subspace_principal_angle": 1.1405278653750983, "mean_class_cosine": 0.804713745342145}}}, "iid_acc": 0.5375, "selectivity": -0.01375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.509364684034334, "mean_class_cosine": 0.9432691877610848}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.7925, "drop": 0.0}, "paraphrase": {"acc": 0.7944983818770227, "drop": -0.0019983818770227257}}, "iid_acc": 0.7925, "selectivity": 0.24124999999999996} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.6494345718901454, "drop": 0.03681542810985461, "rotation": {"subspace_principal_angle": 1.4705871216508313, "mean_class_cosine": 0.10004157445035826}}}, "iid_acc": 0.68625, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.4090091592636642, "mean_class_cosine": 0.16108229179329214}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5282714054927302, "drop": 0.00922859450726976, "rotation": {"subspace_principal_angle": 0.8130735690110286, "mean_class_cosine": 0.9877317674481327}}}, "iid_acc": 0.5375, "selectivity": 0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 0.6815839977232387, "mean_class_cosine": 0.9939458887778236}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6591276252019386, "drop": 0.04587237479806139}}, "iid_acc": 0.705, "selectivity": 0.18999999999999995} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.6962233169129721, "drop": 0.011276683087027894, "rotation": {"subspace_principal_angle": 1.49979082124605, "mean_class_cosine": 0.07094585487697073}}}, "iid_acc": 0.7075, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.482620368933285, "mean_class_cosine": 0.08806174093377613}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.4875, "drop": 0.0}, "paraphrase": {"acc": 0.49589490968801314, "drop": -0.008394909688013152, "rotation": {"subspace_principal_angle": 1.524068667642953, "mean_class_cosine": 0.9875864078448118}}}, "iid_acc": 0.4875, "selectivity": -0.07500000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8171205192766501, "mean_class_cosine": 0.770949459806412}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7684729064039408, "drop": -0.004722906403940796}}, "iid_acc": 0.76375, "selectivity": 0.20125000000000004} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.93375, "drop": 0.0}, "paraphrase": {"acc": 0.9181692094313454, "drop": 0.015580790568654579, "rotation": {"subspace_principal_angle": 1.245296425269521, "mean_class_cosine": 0.3197824779830509}}}, "iid_acc": 0.93375, "selectivity": 0.46624999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.1642335870118337, "mean_class_cosine": 0.3954545965394525}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.855755894590846, "drop": -0.004505894590846049, "rotation": {"subspace_principal_angle": 0.9569293407079221, "mean_class_cosine": 0.948995043172376}}}, "iid_acc": 0.85125, "selectivity": 0.3837499999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5155592875644858, "mean_class_cosine": 0.9503525832749367}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.9292649098474342, "drop": 0.003235090152565845}}, "iid_acc": 0.9325, "selectivity": 0.46499999999999997} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.001987232289950569, "rotation": {"subspace_principal_angle": 0.9228456065715872, "mean_class_cosine": 0.6035537376948626}}}, "iid_acc": 0.99375, "selectivity": 0.19000000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.836801876997636, "mean_class_cosine": 0.6698408686733577}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.42174629324546953, "drop": 0.2632537067545305, "rotation": {"subspace_principal_angle": 1.1528917670379706, "mean_class_cosine": 0.9973529147197886}}}, "iid_acc": 0.685, "selectivity": -0.11874999999999991, "iid_split_rotation": {"subspace_principal_angle": 1.1220283287635184, "mean_class_cosine": 0.999929258322932}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9917627677100495, "drop": 0.0007372322899505956}}, "iid_acc": 0.9925, "selectivity": 0.18875000000000008} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.74875, "drop": 0.0}, "paraphrase": {"acc": 0.699859747545582, "drop": 0.048890252454418026, "rotation": {"subspace_principal_angle": 1.4762812316903078, "mean_class_cosine": 0.09437443909422669}}}, "iid_acc": 0.74875, "selectivity": 0.16500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4435800079379149, "mean_class_cosine": 0.12687345177932846}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.5063113604488079, "drop": -0.03131136044880789, "rotation": {"subspace_principal_angle": 1.4269418061859824, "mean_class_cosine": -0.10748417912085814}}}, "iid_acc": 0.475, "selectivity": -0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4676928130838114, "mean_class_cosine": 0.13776060857811906}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 15, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.726507713884993, "drop": 0.004742286115006933}}, "iid_acc": 0.73125, "selectivity": 0.14749999999999996} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.6923076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6836158192090396, "drop": 0.008691873098652714, "rotation": {"subspace_principal_angle": 1.414178459512195, "mean_class_cosine": 0.1826323602732557}}}, "iid_acc": 0.6923076923076923, "selectivity": 0.2932692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.4350751526080228, "mean_class_cosine": 0.1805559724526645}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.4855769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.4180790960451977, "drop": 0.06749782703172535, "rotation": {"subspace_principal_angle": 1.2602978332522063, "mean_class_cosine": 0.6807601457540157}}}, "iid_acc": 0.4855769230769231, "selectivity": 0.08653846153846151, "iid_split_rotation": {"subspace_principal_angle": 1.1362894639616432, "mean_class_cosine": 0.6888773765938021}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 11, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6836158192090396, "drop": -0.010538896132116449}}, "iid_acc": 0.6730769230769231, "selectivity": 0.27403846153846156} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8945205479452055, "drop": 0.01297945205479445, "rotation": {"subspace_principal_angle": 1.411004019350323, "mean_class_cosine": 0.15911316342225837}}}, "iid_acc": 0.9075, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.3881521587997803, "mean_class_cosine": 0.1816303925457004}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.647945205479452, "drop": 0.020804794520547953, "rotation": {"subspace_principal_angle": 1.1198503241914632, "mean_class_cosine": 0.8163975702558504}}}, "iid_acc": 0.66875, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.401866352414971, "mean_class_cosine": 0.9059315149970117}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9125, "drop": 0.0}, "paraphrase": {"acc": 0.9082191780821918, "drop": 0.0042808219178082085}}, "iid_acc": 0.9125, "selectivity": 0.21875} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.78375, "drop": 0.0}, "paraphrase": {"acc": 0.7961165048543689, "drop": -0.012366504854368965, "rotation": {"subspace_principal_angle": 1.2928467517215603, "mean_class_cosine": 0.2743844969854279}}, "domain": {"acc": 0.725, "drop": 0.05874999999999997, "rotation": {"subspace_principal_angle": 1.3363808809411806, "mean_class_cosine": 0.23227445847260966}}}, "iid_acc": 0.78375, "selectivity": 0.26625, "iid_split_rotation": {"subspace_principal_angle": 1.252105311514529, "mean_class_cosine": 0.3133237568177669}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.007765256588072145, "rotation": {"subspace_principal_angle": 0.5795075771494037, "mean_class_cosine": 0.9984633717830367}}, "domain": {"acc": 0.5, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5279705890862962, "mean_class_cosine": 0.23923966769590235}}}, "iid_acc": 0.50125, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.2020721809074764, "mean_class_cosine": 0.9962696884890198}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.805, "drop": 0.0}, "paraphrase": {"acc": 0.7364771151178918, "drop": 0.06852288488210823}, "domain": {"acc": 0.69625, "drop": 0.10875000000000001}}, "iid_acc": 0.805, "selectivity": 0.2875000000000001} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8271844660194175, "drop": 0.035315533980582514, "rotation": {"subspace_principal_angle": 1.1721917031350069, "mean_class_cosine": 0.38813273689356925}}, "domain": {"acc": 0.59, "drop": 0.2725000000000001, "rotation": {"subspace_principal_angle": 1.3310637787422674, "mean_class_cosine": 0.23744283169580327}}, "length": {"acc": 0.8748114630467572, "drop": -0.01231146304675712, "rotation": {"subspace_principal_angle": 1.0758385777551667, "mean_class_cosine": 0.4749944713239639}}}, "iid_acc": 0.8625, "selectivity": 0.33125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.16650891029775, "mean_class_cosine": 0.3933637231840712}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5805825242718446, "drop": -0.014332524271844616, "rotation": {"subspace_principal_angle": 1.390314107154939, "mean_class_cosine": 0.9207722901404116}}, "domain": {"acc": 0.57875, "drop": -0.012499999999999956, "rotation": {"subspace_principal_angle": 1.0701854814469178, "mean_class_cosine": 0.8833474305608653}}, "length": {"acc": 0.5641025641025641, "drop": 0.002147435897435934, "rotation": {"subspace_principal_angle": 0.9369329512012231, "mean_class_cosine": 0.9177680143320355}}}, "iid_acc": 0.56625, "selectivity": 0.03500000000000003, "iid_split_rotation": {"subspace_principal_angle": 0.9871644210171315, "mean_class_cosine": 0.4073480943439788}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8252427184466019, "drop": 0.006007281553398136}, "domain": {"acc": 0.75375, "drop": 0.07750000000000001}, "length": {"acc": 0.8355957767722474, "drop": -0.004345776772247345}}, "iid_acc": 0.83125, "selectivity": 0.30000000000000004} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.8725, "drop": 0.0}, "paraphrase": {"acc": 0.8614173228346457, "drop": 0.011082677165354382, "rotation": {"subspace_principal_angle": 1.3840540885530221, "mean_class_cosine": 0.2934795135845933}}}, "iid_acc": 0.8725, "selectivity": 0.59125, "iid_split_rotation": {"subspace_principal_angle": 1.4099364101494178, "mean_class_cosine": 0.29058275145224166}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.5925, "drop": 0.0}, "paraphrase": {"acc": 0.5669291338582677, "drop": 0.025570866141732318, "rotation": {"subspace_principal_angle": 0.998890821047691, "mean_class_cosine": 0.94716741691378}}}, "iid_acc": 0.5925, "selectivity": 0.31125, "iid_split_rotation": {"subspace_principal_angle": 0.9488762179483594, "mean_class_cosine": 0.7707278199904818}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 3, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8866141732283465, "drop": -0.0016141732283464494}}, "iid_acc": 0.885, "selectivity": 0.60375} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9370277078085643, "drop": 0.02422229219143579, "rotation": {"subspace_principal_angle": 1.2359971876297358, "mean_class_cosine": 0.6039952840126505}}}, "iid_acc": 0.96125, "selectivity": 0.86875, "iid_split_rotation": {"subspace_principal_angle": 1.2037218056275953, "mean_class_cosine": 0.6646147092871253}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.2575, "drop": 0.0}, "paraphrase": {"acc": 0.24181360201511334, "drop": 0.015686397984886663, "rotation": {"subspace_principal_angle": 1.3871046502979878, "mean_class_cosine": 0.6602864355874835}}}, "iid_acc": 0.2575, "selectivity": 0.165, "iid_split_rotation": {"subspace_principal_angle": 1.5631281003472528, "mean_class_cosine": 0.7824016096532195}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9168765743073047, "drop": 0.024373425692695294}}, "iid_acc": 0.94125, "selectivity": 0.84875} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5580182529335072, "drop": -0.006768252933507135, "rotation": {"subspace_principal_angle": 1.4094629345382517, "mean_class_cosine": 0.16063442581070056}}}, "iid_acc": 0.55125, "selectivity": 0.05375000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4680870349990278, "mean_class_cosine": 0.10252880356925093}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.4915254237288136, "drop": -0.015275423728813575, "rotation": {"subspace_principal_angle": 1.4399498453238624, "mean_class_cosine": -0.9176241029118328}}}, "iid_acc": 0.47625, "selectivity": -0.02124999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4611313378914106, "mean_class_cosine": 0.7185000764227502}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5032594524119948, "drop": 0.009240547588005188}}, "iid_acc": 0.5125, "selectivity": 0.014999999999999958} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.6325, "drop": 0.0}, "paraphrase": {"acc": 0.6228419654714475, "drop": 0.009658034528552406, "rotation": {"subspace_principal_angle": 1.348983488807089, "mean_class_cosine": 0.32809803411682364}}}, "iid_acc": 0.6325, "selectivity": 0.39249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3929305981197904, "mean_class_cosine": 0.31905211151253626}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.17, "drop": 0.0}, "paraphrase": {"acc": 0.17795484727755645, "drop": -0.007954847277556437, "rotation": {"subspace_principal_angle": 1.4531704709040898, "mean_class_cosine": 0.4279120214634104}}}, "iid_acc": 0.17, "selectivity": -0.06999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.5500505520500567, "mean_class_cosine": 0.6189009860707383}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.602921646746348, "drop": -0.004171646746347957}}, "iid_acc": 0.59875, "selectivity": 0.35875} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.72125, "drop": 0.0}, "paraphrase": {"acc": 0.7168284789644013, "drop": 0.0044215210355986745, "rotation": {"subspace_principal_angle": 1.378307022373732, "mean_class_cosine": 0.19130281479130162}}}, "iid_acc": 0.72125, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4153437530917778, "mean_class_cosine": 0.1548272315027917}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6585760517799353, "drop": -0.036076051779935225, "rotation": {"subspace_principal_angle": 1.0555425497782547, "mean_class_cosine": 0.88892568318339}}}, "iid_acc": 0.6225, "selectivity": 0.06375000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9396491512167356, "mean_class_cosine": 0.32052389745129145}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7394822006472492, "drop": -0.008232200647249277}}, "iid_acc": 0.73125, "selectivity": 0.1725} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6316639741518578, "drop": 0.017086025848142228, "rotation": {"subspace_principal_angle": 1.3946146188040096, "mean_class_cosine": 0.17527167497740181}}}, "iid_acc": 0.64875, "selectivity": 0.16125000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4780475383542893, "mean_class_cosine": 0.09261586955669032}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.56875, "drop": 0.0}, "paraphrase": {"acc": 0.568659127625202, "drop": 9.087237479799004e-05, "rotation": {"subspace_principal_angle": 1.042022638358511, "mean_class_cosine": 0.9908204267164294}}}, "iid_acc": 0.56875, "selectivity": 0.08124999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.4747132575045126, "mean_class_cosine": 0.9899105548993554}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.65, "drop": 0.0}, "paraphrase": {"acc": 0.6381260096930533, "drop": 0.011873990306946691}}, "iid_acc": 0.65, "selectivity": 0.16250000000000003} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7075, "drop": 0.0}, "paraphrase": {"acc": 0.7241379310344828, "drop": -0.016637931034482745, "rotation": {"subspace_principal_angle": 1.4200343797614488, "mean_class_cosine": 0.15019148021114892}}}, "iid_acc": 0.7075, "selectivity": 0.13124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.382096362448142, "mean_class_cosine": 0.18758209522431973}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5024630541871922, "drop": -0.01246305418719218, "rotation": {"subspace_principal_angle": 1.534087120470578, "mean_class_cosine": 0.9835950818873114}}}, "iid_acc": 0.49, "selectivity": -0.08625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.0584655028541943, "mean_class_cosine": 0.7127138345764553}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7323481116584565, "drop": 0.00015188834154356012}}, "iid_acc": 0.7325, "selectivity": 0.15625} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8918169209431346, "drop": 0.0031830790568654344, "rotation": {"subspace_principal_angle": 1.2670757577041036, "mean_class_cosine": 0.29907254848700277}}}, "iid_acc": 0.895, "selectivity": 0.435, "iid_split_rotation": {"subspace_principal_angle": 1.1636722638264427, "mean_class_cosine": 0.3959701014331015}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.64125, "drop": 0.0}, "paraphrase": {"acc": 0.6310679611650486, "drop": 0.01018203883495139, "rotation": {"subspace_principal_angle": 1.2408104805968017, "mean_class_cosine": 0.9660598087989238}}}, "iid_acc": 0.64125, "selectivity": 0.18124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1355294988893803, "mean_class_cosine": 0.9278898529249071}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.9195561719833565, "drop": -0.01830617198335649}}, "iid_acc": 0.90125, "selectivity": 0.44125} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915, "rotation": {"subspace_principal_angle": 0.9875978535885324, "mean_class_cosine": 0.5506965324007853}}}, "iid_acc": 0.9925, "selectivity": 0.1975, "iid_split_rotation": {"subspace_principal_angle": 0.858018142885935, "mean_class_cosine": 0.6539381215477612}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.685, "drop": 0.0}, "paraphrase": {"acc": 0.7232289950576606, "drop": -0.03822899505766053, "rotation": {"subspace_principal_angle": 0.5872782478443009, "mean_class_cosine": 0.9996739997061842}}}, "iid_acc": 0.685, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5575352393582083, "mean_class_cosine": 0.999883744660425}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915}}, "iid_acc": 0.9925, "selectivity": 0.1975} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.66375, "drop": 0.0}, "paraphrase": {"acc": 0.6465638148667602, "drop": 0.017186185133239773, "rotation": {"subspace_principal_angle": 1.519163670668585, "mean_class_cosine": 0.05160971766617552}}}, "iid_acc": 0.66375, "selectivity": 0.0774999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4801296825592345, "mean_class_cosine": 0.0905424753256214}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.5077138849929874, "drop": -0.03021388499298744, "rotation": {"subspace_principal_angle": 1.43231424996757, "mean_class_cosine": -0.6739732724428789}}}, "iid_acc": 0.4775, "selectivity": -0.10875000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.2395084599326798, "mean_class_cosine": 0.2320322894243369}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6830294530154277, "drop": 0.011970546984572228}}, "iid_acc": 0.695, "selectivity": 0.1087499999999999} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": 0.0058126901347240745, "rotation": {"subspace_principal_angle": 1.429726953139992, "mean_class_cosine": 0.17310900825401718}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.4510331206422105, "mean_class_cosine": 0.16038164209334843}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.47115384615384615, "drop": 0.0}, "paraphrase": {"acc": 0.423728813559322, "drop": 0.04742503259452413, "rotation": {"subspace_principal_angle": 1.1889023495577067, "mean_class_cosine": 0.5428679561508623}}}, "iid_acc": 0.47115384615384615, "selectivity": 0.08653846153846151, "iid_split_rotation": {"subspace_principal_angle": 1.560544322206414, "mean_class_cosine": 0.7853248916815403}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 8, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": -0.027243589743589758}}, "iid_acc": 0.6394230769230769, "selectivity": 0.25480769230769224} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8835616438356164, "drop": 0.005188356164383623, "rotation": {"subspace_principal_angle": 1.4311921432320234, "mean_class_cosine": 0.13915115995570887}}}, "iid_acc": 0.88875, "selectivity": 0.21500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.3924741674351642, "mean_class_cosine": 0.17737858927148076}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.59125, "drop": 0.0}, "paraphrase": {"acc": 0.589041095890411, "drop": 0.0022089041095890716, "rotation": {"subspace_principal_angle": 1.3624580504069597, "mean_class_cosine": 0.7634106223781487}}}, "iid_acc": 0.59125, "selectivity": -0.0824999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5428225535260213, "mean_class_cosine": 0.8607299633579588}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 6, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8958904109589041, "drop": 0.010359589041095907}}, "iid_acc": 0.90625, "selectivity": 0.23250000000000004} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8224687933425797, "drop": 0.015031206657420304, "rotation": {"subspace_principal_angle": 1.33328453850938, "mean_class_cosine": 0.23528499856171853}}, "domain": {"acc": 0.66, "drop": 0.1775, "rotation": {"subspace_principal_angle": 1.451106758229533, "mean_class_cosine": 0.1194040025165285}}}, "iid_acc": 0.8375, "selectivity": 0.31625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2770908389865925, "mean_class_cosine": 0.28950101482596285}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.007765256588072145, "rotation": {"subspace_principal_angle": 1.3142819239703394, "mean_class_cosine": 0.9932995750720417}}, "domain": {"acc": 0.5, "drop": 0.0012499999999999734, "rotation": {"subspace_principal_angle": 1.5693408077430426, "mean_class_cosine": 0.210032160652116}}}, "iid_acc": 0.50125, "selectivity": -0.020000000000000018, "iid_split_rotation": {"subspace_principal_angle": 1.0510108236306581, "mean_class_cosine": 0.9827971130549489}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.84625, "drop": 0.0}, "paraphrase": {"acc": 0.8377253814147018, "drop": 0.008524618585298183}, "domain": {"acc": 0.83625, "drop": 0.009999999999999898}}, "iid_acc": 0.84625, "selectivity": 0.32499999999999996} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8803149606299212, "drop": 0.0009350393700787718, "rotation": {"subspace_principal_angle": 1.3448528768477728, "mean_class_cosine": 0.3368814094412147}}}, "iid_acc": 0.88125, "selectivity": 0.6575, "iid_split_rotation": {"subspace_principal_angle": 1.4525237248303589, "mean_class_cosine": 0.271959475761732}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5181102362204725, "drop": 0.024389763779527507, "rotation": {"subspace_principal_angle": 1.0916569684946897, "mean_class_cosine": 0.9392101199033003}}}, "iid_acc": 0.5425, "selectivity": 0.31875, "iid_split_rotation": {"subspace_principal_angle": 1.5156324848496125, "mean_class_cosine": 0.7159087166162079}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8724409448818897, "drop": 0.022559055118110294}}, "iid_acc": 0.895, "selectivity": 0.67125} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5384615384615384, "drop": -0.01221153846153844, "rotation": {"subspace_principal_angle": 1.4851659522030354, "mean_class_cosine": 0.08552576462723123}}}, "iid_acc": 0.52625, "selectivity": 0.03249999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.491288991251595, "mean_class_cosine": 0.07942359551958941}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.47625, "drop": 0.0}, "paraphrase": {"acc": 0.4771838331160365, "drop": -0.0009338331160365088, "rotation": {"subspace_principal_angle": 1.3966119018215262, "mean_class_cosine": -0.17561134987197125}}}, "iid_acc": 0.47625, "selectivity": -0.017500000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.5601173526903696, "mean_class_cosine": 0.2908977808248975}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5149934810951761, "drop": -0.014993481095176064}}, "iid_acc": 0.5, "selectivity": 0.006249999999999978} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.6162018592297477, "drop": 0.05254814077025227, "rotation": {"subspace_principal_angle": 1.3864574345230025, "mean_class_cosine": 0.33562554763173674}}}, "iid_acc": 0.66875, "selectivity": 0.40499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3371373671231837, "mean_class_cosine": 0.3302011789609844}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.185, "drop": 0.0}, "paraphrase": {"acc": 0.20185922974767595, "drop": -0.016859229747675952, "rotation": {"subspace_principal_angle": 1.5059232736861428, "mean_class_cosine": 0.4846447217345316}}}, "iid_acc": 0.185, "selectivity": -0.07874999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1753986125618363, "mean_class_cosine": 0.6615083924419292}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 1, "dists": {"iid": {"acc": 0.645, "drop": 0.0}, "paraphrase": {"acc": 0.6254980079681275, "drop": 0.019501992031872528}}, "iid_acc": 0.645, "selectivity": 0.38125000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7184466019417476, "drop": 0.022803398058252378, "rotation": {"subspace_principal_angle": 1.5041417980246978, "mean_class_cosine": 0.06660518398513274}}}, "iid_acc": 0.74125, "selectivity": 0.22124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4864371370902933, "mean_class_cosine": 0.08425916865336666}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5485436893203883, "drop": 0.00020631067961163208, "rotation": {"subspace_principal_angle": 1.3708104020371576, "mean_class_cosine": 0.9683596041166356}}}, "iid_acc": 0.54875, "selectivity": 0.028749999999999942, "iid_split_rotation": {"subspace_principal_angle": 1.515659944278173, "mean_class_cosine": 0.9203876968475591}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.77875, "drop": 0.0}, "paraphrase": {"acc": 0.7880258899676376, "drop": -0.009275889967637507}}, "iid_acc": 0.77875, "selectivity": 0.25875000000000004} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6155088852988692, "drop": 0.03949111470113087, "rotation": {"subspace_principal_angle": 1.5045399067665934, "mean_class_cosine": 0.06620795401218435}}}, "iid_acc": 0.655, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4647655973683582, "mean_class_cosine": 0.10583216572254955}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5201938610662359, "drop": 0.017306138933764093, "rotation": {"subspace_principal_angle": 1.402582557195184, "mean_class_cosine": 0.9741348383409788}}}, "iid_acc": 0.5375, "selectivity": 0.018749999999999933, "iid_split_rotation": {"subspace_principal_angle": 1.518487153242731, "mean_class_cosine": 0.9849570196369872}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6720516962843296, "drop": 0.021698303715670386}}, "iid_acc": 0.69375, "selectivity": 0.17499999999999993} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.7027914614121511, "drop": -0.00904146141215112, "rotation": {"subspace_principal_angle": 1.4423830102309243, "mean_class_cosine": 0.12806068526617434}}}, "iid_acc": 0.69375, "selectivity": 0.16125, "iid_split_rotation": {"subspace_principal_angle": 1.380371124548562, "mean_class_cosine": 0.1892764282980448}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.545, "drop": 0.0}, "paraphrase": {"acc": 0.5500821018062397, "drop": -0.005082101806239692, "rotation": {"subspace_principal_angle": 1.0329294851079376, "mean_class_cosine": 0.9281244210082812}}}, "iid_acc": 0.545, "selectivity": 0.012500000000000067, "iid_split_rotation": {"subspace_principal_angle": 1.1987031561182855, "mean_class_cosine": 0.701876474533382}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 2, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7323481116584565, "drop": -0.0010981116584565243}}, "iid_acc": 0.73125, "selectivity": 0.19874999999999998} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.9292649098474342, "drop": 0.005735090152565903, "rotation": {"subspace_principal_angle": 1.229191428617486, "mean_class_cosine": 0.3349996872547975}}}, "iid_acc": 0.935, "selectivity": 0.43875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0937898079150548, "mean_class_cosine": 0.4591219079682709}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6907073509015257, "drop": 0.000542649098474346, "rotation": {"subspace_principal_angle": 1.2452578371926764, "mean_class_cosine": 0.9675450369885805}}}, "iid_acc": 0.69125, "selectivity": 0.195, "iid_split_rotation": {"subspace_principal_angle": 1.2562849442009068, "mean_class_cosine": 0.9533786218456732}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9251040221914009, "drop": -0.0026040221914008788}}, "iid_acc": 0.9225, "selectivity": 0.42624999999999996} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.7, "drop": 0.0}, "paraphrase": {"acc": 0.6690042075736325, "drop": 0.030995792426367408, "rotation": {"subspace_principal_angle": 1.4968125785391955, "mean_class_cosine": 0.0739162738788901}}}, "iid_acc": 0.7, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.424774943726807, "mean_class_cosine": 0.14550301874390104}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.48625, "drop": 0.0}, "paraphrase": {"acc": 0.5133239831697055, "drop": -0.027073983169705496, "rotation": {"subspace_principal_angle": 1.473176890758434, "mean_class_cosine": -0.2335794158894724}}}, "iid_acc": 0.48625, "selectivity": -0.06749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2575303716090405, "mean_class_cosine": 0.3250973513251493}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 19, "dists": {"iid": {"acc": 0.6975, "drop": 0.0}, "paraphrase": {"acc": 0.6760168302945302, "drop": 0.021483169705469818}}, "iid_acc": 0.6975, "selectivity": 0.14375000000000004} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6384180790960452, "drop": 0.010620382442416365, "rotation": {"subspace_principal_angle": 1.4206287662194153, "mean_class_cosine": 0.19628077809764047}}}, "iid_acc": 0.6490384615384616, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.455372039968678, "mean_class_cosine": 0.19590134511595567}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.39903846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.3785310734463277, "drop": 0.02050738809213387, "rotation": {"subspace_principal_angle": 0.9199382488566391, "mean_class_cosine": 0.9264117531113749}}}, "iid_acc": 0.39903846153846156, "selectivity": 0.009615384615384637, "iid_split_rotation": {"subspace_principal_angle": 1.366185399901852, "mean_class_cosine": 0.35869369801616585}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 24, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6271186440677966, "drop": 0.04115058670143412}}, "iid_acc": 0.6682692307692307, "selectivity": 0.2788461538461538} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8876712328767123, "drop": 0.01357876712328765, "rotation": {"subspace_principal_angle": 1.431717330740105, "mean_class_cosine": 0.13863106275045062}}}, "iid_acc": 0.90125, "selectivity": 0.21999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4181086678178447, "mean_class_cosine": 0.15209506900442057}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5794520547945206, "drop": 0.006797945205479494, "rotation": {"subspace_principal_angle": 0.8966790530917582, "mean_class_cosine": 0.766361309893145}}}, "iid_acc": 0.58625, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.7483984334756586, "mean_class_cosine": 0.8336256970933262}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 7, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8890410958904109, "drop": 0.018458904109589058}}, "iid_acc": 0.9075, "selectivity": 0.22624999999999995} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.835, "drop": 0.0}, "paraphrase": {"acc": 0.7877947295423023, "drop": 0.04720527045769762, "rotation": {"subspace_principal_angle": 1.2702602526749749, "mean_class_cosine": 0.29603229534182074}}, "domain": {"acc": 0.78375, "drop": 0.05125000000000002, "rotation": {"subspace_principal_angle": 1.2958495701467163, "mean_class_cosine": 0.2714956938278416}}}, "iid_acc": 0.835, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 1.1803611728107277, "mean_class_cosine": 0.38059085697680345}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.5090152565880721, "drop": -0.001515256588072167, "rotation": {"subspace_principal_angle": 1.2132834342045293, "mean_class_cosine": 0.9993075605412058}}, "domain": {"acc": 0.5, "drop": 0.007499999999999951, "rotation": {"subspace_principal_angle": 1.3982125329965656, "mean_class_cosine": 0.036308404316974255}}}, "iid_acc": 0.5075, "selectivity": 0.0012499999999999734, "iid_split_rotation": {"subspace_principal_angle": 1.4113229091197326, "mean_class_cosine": 0.9975285200002872}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.8175, "drop": 0.0}, "paraphrase": {"acc": 0.7877947295423023, "drop": 0.029705270457697663}, "domain": {"acc": 0.855, "drop": -0.03749999999999998}}, "iid_acc": 0.8175, "selectivity": 0.31125} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.8893203883495145, "drop": 0.013179611650485423, "rotation": {"subspace_principal_angle": 1.067178302018369, "mean_class_cosine": 0.48259750926799183}}, "domain": {"acc": 0.5675, "drop": 0.33499999999999996, "rotation": {"subspace_principal_angle": 1.2593770522331844, "mean_class_cosine": 0.3064099515537772}}, "length": {"acc": 0.9095022624434389, "drop": -0.0070022624434389336, "rotation": {"subspace_principal_angle": 1.0045870056146216, "mean_class_cosine": 0.536436803133892}}}, "iid_acc": 0.9025, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.0653451564032532, "mean_class_cosine": 0.4842022438803039}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.7611650485436893, "drop": -0.06491504854368924, "rotation": {"subspace_principal_angle": 1.4069108146369018, "mean_class_cosine": 0.9395668351834272}}, "domain": {"acc": 0.57875, "drop": 0.11750000000000005, "rotation": {"subspace_principal_angle": 1.2932791063335667, "mean_class_cosine": 0.34457030192799165}}, "length": {"acc": 0.7134238310708899, "drop": -0.017173831070889878, "rotation": {"subspace_principal_angle": 1.5504433084962168, "mean_class_cosine": 0.9246357737712342}}}, "iid_acc": 0.69625, "selectivity": 0.18125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3250489130035334, "mean_class_cosine": 0.7575530733718268}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8699029126213592, "drop": 0.030097087378640808}, "domain": {"acc": 0.515, "drop": 0.385}, "length": {"acc": 0.9034690799396682, "drop": -0.0034690799396681626}}, "iid_acc": 0.9, "selectivity": 0.385} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8881889763779528, "drop": 0.006811023622047219, "rotation": {"subspace_principal_angle": 1.2507965877486584, "mean_class_cosine": 0.4519091647049873}}}, "iid_acc": 0.895, "selectivity": 0.635, "iid_split_rotation": {"subspace_principal_angle": 1.2538562323332731, "mean_class_cosine": 0.4436394901742749}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.3725, "drop": 0.0}, "paraphrase": {"acc": 0.36220472440944884, "drop": 0.010295275590551156, "rotation": {"subspace_principal_angle": 1.1149135781556565, "mean_class_cosine": 0.9771677078579994}}}, "iid_acc": 0.3725, "selectivity": 0.11249999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1275037300836335, "mean_class_cosine": 0.5054561634322786}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.88625, "drop": 0.0}, "paraphrase": {"acc": 0.8771653543307086, "drop": 0.009084645669291347}}, "iid_acc": 0.88625, "selectivity": 0.62625} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.9675, "drop": 0.0}, "paraphrase": {"acc": 0.9571788413098237, "drop": 0.010321158690176357, "rotation": {"subspace_principal_angle": 1.1895811374243113, "mean_class_cosine": 0.6930412270179037}}}, "iid_acc": 0.9675, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.1204246248516365, "mean_class_cosine": 0.734504756933933}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.28125, "drop": 0.0}, "paraphrase": {"acc": 0.3047858942065491, "drop": -0.023535894206549113, "rotation": {"subspace_principal_angle": 1.566917995749759, "mean_class_cosine": 0.7334608100976151}}}, "iid_acc": 0.28125, "selectivity": 0.22, "iid_split_rotation": {"subspace_principal_angle": 1.5646555108195546, "mean_class_cosine": 0.8084161094056437}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 2, "layer": 17, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9445843828715366, "drop": 0.010415617128463395}}, "iid_acc": 0.955, "selectivity": 0.8937499999999999} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.60625, "drop": 0.0}, "paraphrase": {"acc": 0.6375488917861799, "drop": -0.03129889178617995, "rotation": {"subspace_principal_angle": 1.2380910520249748, "mean_class_cosine": 0.3266011440007731}}}, "iid_acc": 0.60625, "selectivity": 0.08499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2935899170666836, "mean_class_cosine": 0.2736697785737916}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.4775, "drop": 0.0}, "paraphrase": {"acc": 0.4915254237288136, "drop": -0.014025423728813602, "rotation": {"subspace_principal_angle": 1.481666093252446, "mean_class_cosine": -0.9406980138297385}}}, "iid_acc": 0.4775, "selectivity": -0.04375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2591046492876659, "mean_class_cosine": 0.7431471438900024}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5019556714471969, "drop": 0.011794328552803157}}, "iid_acc": 0.51375, "selectivity": -0.007499999999999951} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.5962815405046481, "drop": -0.006281540504648131, "rotation": {"subspace_principal_angle": 1.3616817945475932, "mean_class_cosine": 0.34929473551176465}}}, "iid_acc": 0.59, "selectivity": 0.33499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3738625012283205, "mean_class_cosine": 0.31435666695467485}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.07625, "drop": 0.0}, "paraphrase": {"acc": 0.07702523240371846, "drop": -0.0007752324037184621, "rotation": {"subspace_principal_angle": 1.4014772128715691, "mean_class_cosine": -0.26880588026423574}}}, "iid_acc": 0.07625, "selectivity": -0.17875000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.2922523705095328, "mean_class_cosine": 0.5906582844419805}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 2, "layer": 16, "dists": {"iid": {"acc": 0.49125, "drop": 0.0}, "paraphrase": {"acc": 0.4913678618857902, "drop": -0.00011786188579016033}}, "iid_acc": 0.49125, "selectivity": 0.23625000000000002} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.765, "drop": 0.0}, "paraphrase": {"acc": 0.7686084142394822, "drop": -0.003608414239482216, "rotation": {"subspace_principal_angle": 1.267703414249137, "mean_class_cosine": 0.2984735606750531}}}, "iid_acc": 0.765, "selectivity": 0.21250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.297520243254995, "mean_class_cosine": 0.26988739363466807}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5355987055016181, "drop": 0.0019012944983818336, "rotation": {"subspace_principal_angle": 1.4289319500576079, "mean_class_cosine": 0.9995585763899515}}}, "iid_acc": 0.5375, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.095765507578661, "mean_class_cosine": 0.9992212510316336}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.6828478964401294, "drop": 0.012152103559870508}}, "iid_acc": 0.695, "selectivity": 0.14249999999999996} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6865912762520194, "drop": 0.007158723747980544, "rotation": {"subspace_principal_angle": 1.3189864885611238, "mean_class_cosine": 0.2491571278556936}}}, "iid_acc": 0.69375, "selectivity": 0.19374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3741028601612726, "mean_class_cosine": 0.19542762781810974}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5137318255250404, "drop": 0.015018174474959678, "rotation": {"subspace_principal_angle": 1.410712709591301, "mean_class_cosine": 0.9996250431090286}}}, "iid_acc": 0.52875, "selectivity": 0.028750000000000053, "iid_split_rotation": {"subspace_principal_angle": 0.9117211516634655, "mean_class_cosine": 0.9997105404220764}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 2, "layer": 4, "dists": {"iid": {"acc": 0.61125, "drop": 0.0}, "paraphrase": {"acc": 0.592891760904685, "drop": 0.018358239095314977}}, "iid_acc": 0.61125, "selectivity": 0.11124999999999996} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.6995073891625616, "drop": 0.016742610837438443, "rotation": {"subspace_principal_angle": 1.398637777187281, "mean_class_cosine": 0.17130938688822317}}}, "iid_acc": 0.71625, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4203666415350267, "mean_class_cosine": 0.14986297902276266}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.4975369458128079, "drop": -0.0012869458128078604, "rotation": {"subspace_principal_angle": 1.5334169157311857, "mean_class_cosine": 0.9992809010722258}}}, "iid_acc": 0.49625, "selectivity": -0.08625, "iid_split_rotation": {"subspace_principal_angle": 1.523586797466333, "mean_class_cosine": 0.9461897010476042}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 2, "layer": 9, "dists": {"iid": {"acc": 0.69125, "drop": 0.0}, "paraphrase": {"acc": 0.6765188834154351, "drop": 0.01473111658456494}}, "iid_acc": 0.69125, "selectivity": 0.10875000000000001} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.9275, "drop": 0.0}, "paraphrase": {"acc": 0.9181692094313454, "drop": 0.009330790568654601, "rotation": {"subspace_principal_angle": 1.0867093987105656, "mean_class_cosine": 0.4654003933713694}}}, "iid_acc": 0.9275, "selectivity": 0.45875, "iid_split_rotation": {"subspace_principal_angle": 1.0076905925780175, "mean_class_cosine": 0.533814981080676}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5312066574202496, "drop": 4.33425797503606e-05, "rotation": {"subspace_principal_angle": 0.9422006877595401, "mean_class_cosine": 0.9720210108018141}}}, "iid_acc": 0.53125, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 0.9637054547388073, "mean_class_cosine": 0.8521486536674154}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 2, "layer": 12, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.8987517337031901, "drop": 0.016248266296809977}}, "iid_acc": 0.915, "selectivity": 0.44625000000000004} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9967051070840197, "drop": -0.0017051070840197191, "rotation": {"subspace_principal_angle": 1.004243983718731, "mean_class_cosine": 0.5367262616365928}}}, "iid_acc": 0.995, "selectivity": 0.1725, "iid_split_rotation": {"subspace_principal_angle": 0.8818699720917821, "mean_class_cosine": 0.6357087708516111}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.695, "drop": 0.0}, "paraphrase": {"acc": 0.7166392092257001, "drop": -0.021639209225700173, "rotation": {"subspace_principal_angle": 1.3903038602930748, "mean_class_cosine": 0.999942812375732}}}, "iid_acc": 0.695, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2573644298943418, "mean_class_cosine": 0.9999792008680795}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 2, "layer": 13, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934102141680395, "drop": -0.0009102141680394915}}, "iid_acc": 0.9925, "selectivity": 0.17000000000000004} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6451612903225806, "drop": 0.03733870967741937, "rotation": {"subspace_principal_angle": 1.534693889353481, "mean_class_cosine": 0.03609459538397165}}}, "iid_acc": 0.6825, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.402974256327293, "mean_class_cosine": 0.16703541534575633}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.47375, "drop": 0.0}, "paraphrase": {"acc": 0.49789621318373073, "drop": -0.024146213183730725, "rotation": {"subspace_principal_angle": 1.4387989610109857, "mean_class_cosine": -0.9029123388757152}}}, "iid_acc": 0.47375, "selectivity": -0.12749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5531483539033937, "mean_class_cosine": -0.16026020693951765}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.7124824684431977, "drop": -0.011232468443197696}}, "iid_acc": 0.70125, "selectivity": 0.10000000000000009} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.6682692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6779661016949152, "drop": -0.009696870925684498, "rotation": {"subspace_principal_angle": 1.3728662084585508, "mean_class_cosine": 0.23801349724373286}}}, "iid_acc": 0.6682692307692307, "selectivity": 0.23076923076923073, "iid_split_rotation": {"subspace_principal_angle": 1.3852570180024775, "mean_class_cosine": 0.240802390739838}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.49038461538461536, "drop": 0.0}, "paraphrase": {"acc": 0.4124293785310734, "drop": 0.07795523685354194, "rotation": {"subspace_principal_angle": 1.093285198742622, "mean_class_cosine": 0.5340692801147456}}}, "iid_acc": 0.49038461538461536, "selectivity": 0.05288461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.1837899565316878, "mean_class_cosine": 0.4155974505600208}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 2, "layer": 5, "dists": {"iid": {"acc": 0.6298076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.6497175141242938, "drop": -0.019909821816601503}}, "iid_acc": 0.6298076923076923, "selectivity": 0.1923076923076923} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.873972602739726, "drop": 0.026027397260274032, "rotation": {"subspace_principal_angle": 1.2965339594699503, "mean_class_cosine": 0.2708369468728177}}}, "iid_acc": 0.9, "selectivity": 0.17125, "iid_split_rotation": {"subspace_principal_angle": 1.2862423961352993, "mean_class_cosine": 0.28072934769702657}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.575, "drop": 0.0}, "paraphrase": {"acc": 0.5671232876712329, "drop": 0.007876712328767077, "rotation": {"subspace_principal_angle": 1.0126162968840653, "mean_class_cosine": 0.687860663967105}}}, "iid_acc": 0.575, "selectivity": -0.15375000000000005, "iid_split_rotation": {"subspace_principal_angle": 0.5178918149259684, "mean_class_cosine": 0.7969393733558711}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 2, "layer": 14, "dists": {"iid": {"acc": 0.86125, "drop": 0.0}, "paraphrase": {"acc": 0.8575342465753425, "drop": 0.003715753424657464}}, "iid_acc": 0.86125, "selectivity": 0.13249999999999995} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7098591549295775, "drop": 0.04639084507042246, "rotation": {"subspace_principal_angle": 1.285187527092102, "mean_class_cosine": 0.2817416408844383}}, "domain": {"acc": 0.71375, "drop": 0.04249999999999998, "rotation": {"subspace_principal_angle": 1.3127911813868958, "mean_class_cosine": 0.25515223416644334}}}, "iid_acc": 0.75625, "selectivity": 0.22124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.2283794043097545, "mean_class_cosine": 0.33576468083719035}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5253521126760563, "drop": -0.026602112676056278, "rotation": {"subspace_principal_angle": 1.214807789184398, "mean_class_cosine": 0.9831397387284473}}, "domain": {"acc": 0.4875, "drop": 0.011250000000000038, "rotation": {"subspace_principal_angle": 1.5537344695259054, "mean_class_cosine": 0.34478575371993236}}}, "iid_acc": 0.49875, "selectivity": -0.036250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1073273476869623, "mean_class_cosine": 0.9951775418509261}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7126760563380282, "drop": 0.018573943661971737}, "domain": {"acc": 0.7225, "drop": 0.008749999999999925}}, "iid_acc": 0.73125, "selectivity": 0.19624999999999992} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.82, "drop": 0.0}, "paraphrase": {"acc": 0.8269961977186312, "drop": -0.006996197718631247, "rotation": {"subspace_principal_angle": 1.0415904118081376, "mean_class_cosine": 0.5048480397255118}}, "domain": {"acc": 0.59375, "drop": 0.22624999999999995, "rotation": {"subspace_principal_angle": 1.2834570977197297, "mean_class_cosine": 0.28340154844279125}}, "length": {"acc": 0.8264840182648402, "drop": -0.006484018264840241, "rotation": {"subspace_principal_angle": 0.999883755381926, "mean_class_cosine": 0.5404001186906672}}}, "iid_acc": 0.82, "selectivity": 0.3549999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.9949686563230938, "mean_class_cosine": 0.5445291790193694}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6863117870722434, "drop": 0.018688212927756576, "rotation": {"subspace_principal_angle": 1.3283778591084485, "mean_class_cosine": 0.9327609071818541}}, "domain": {"acc": 0.565, "drop": 0.14, "rotation": {"subspace_principal_angle": 1.5278891311014122, "mean_class_cosine": 0.3806568220146318}}, "length": {"acc": 0.7062404870624048, "drop": -0.0012404870624048714, "rotation": {"subspace_principal_angle": 1.3125076791898616, "mean_class_cosine": 0.9017445724833599}}}, "iid_acc": 0.705, "selectivity": 0.23999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4606505742125795, "mean_class_cosine": 0.9081384351848174}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.7975, "drop": 0.0}, "paraphrase": {"acc": 0.7984790874524715, "drop": -0.0009790874524715498}, "domain": {"acc": 0.645, "drop": 0.15249999999999997}, "length": {"acc": 0.7975646879756468, "drop": -6.46879756468488e-05}}, "iid_acc": 0.7975, "selectivity": 0.33249999999999996} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.8625, "drop": 0.0}, "paraphrase": {"acc": 0.8611544461778471, "drop": 0.0013455538221529562, "rotation": {"subspace_principal_angle": 0.9939269073783296, "mean_class_cosine": 0.6453866739528615}}}, "iid_acc": 0.8625, "selectivity": 0.5362500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4271388898728552, "mean_class_cosine": 0.6524369913147846}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7316692667706708, "drop": 0.00458073322932917, "rotation": {"subspace_principal_angle": 1.3724063354724922, "mean_class_cosine": 0.9469017405917457}}}, "iid_acc": 0.73625, "selectivity": 0.41, "iid_split_rotation": {"subspace_principal_angle": 1.0603683518688838, "mean_class_cosine": 0.950801823774931}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.81625, "drop": 0.0}, "paraphrase": {"acc": 0.8081123244929798, "drop": 0.00813767550702027}}, "iid_acc": 0.81625, "selectivity": 0.49000000000000005} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9160671462829736, "drop": 0.021432853717026412, "rotation": {"subspace_principal_angle": 1.1847135524986785, "mean_class_cosine": 0.706584730242114}}}, "iid_acc": 0.9375, "selectivity": 0.87625, "iid_split_rotation": {"subspace_principal_angle": 1.11150260490418, "mean_class_cosine": 0.7556327352536992}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.4460431654676259, "drop": 0.05520683453237407, "rotation": {"subspace_principal_angle": 1.199717400716062, "mean_class_cosine": 0.8611479821783318}}}, "iid_acc": 0.50125, "selectivity": 0.43999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4104637914876992, "mean_class_cosine": 0.8005163008883979}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8465227817745803, "drop": 0.03222721822541974}}, "iid_acc": 0.87875, "selectivity": 0.8175} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.515032679738562, "drop": 0.009967320261437984, "rotation": {"subspace_principal_angle": 1.566988880756832, "mean_class_cosine": 0.003807436838872146}}}, "iid_acc": 0.525, "selectivity": 0.03750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.515059781737158, "mean_class_cosine": 0.055707691363518934}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5189542483660131, "drop": 0.00729575163398688, "rotation": {"subspace_principal_angle": 1.565436854816222, "mean_class_cosine": 0.11469077846809743}}}, "iid_acc": 0.52625, "selectivity": 0.03875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.519484137603478, "mean_class_cosine": 0.056727235397258546}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5124183006535947, "drop": 0.023831699346405277}}, "iid_acc": 0.53625, "selectivity": 0.048750000000000016} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5485175202156334, "drop": 0.023982479784366628, "rotation": {"subspace_principal_angle": 1.4303692527758758, "mean_class_cosine": 0.35644155459058985}}}, "iid_acc": 0.5725, "selectivity": 0.2575, "iid_split_rotation": {"subspace_principal_angle": 1.4149642770924389, "mean_class_cosine": 0.3618340196725102}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.28125, "drop": 0.0}, "paraphrase": {"acc": 0.28975741239892183, "drop": -0.008507412398921832, "rotation": {"subspace_principal_angle": 1.5249290421568074, "mean_class_cosine": 0.644382717112205}}}, "iid_acc": 0.28125, "selectivity": -0.03375, "iid_split_rotation": {"subspace_principal_angle": 1.5382442039631525, "mean_class_cosine": 0.5744283723940081}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5053908355795148, "drop": 0.009609164420485206}}, "iid_acc": 0.515, "selectivity": 0.2} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.7345575959933222, "drop": -0.01080759599332215, "rotation": {"subspace_principal_angle": 1.2514538362385088, "mean_class_cosine": 0.31394236141299214}}}, "iid_acc": 0.72375, "selectivity": 0.22499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3629405243520136, "mean_class_cosine": 0.20636233074920893}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.6060100166944908, "drop": -0.0297600166944908, "rotation": {"subspace_principal_angle": 1.4254734942092195, "mean_class_cosine": 0.930851220197922}}}, "iid_acc": 0.57625, "selectivity": 0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5027930622930314, "mean_class_cosine": 0.8705083526841852}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7312186978297162, "drop": 0.003781302170283829}}, "iid_acc": 0.735, "selectivity": 0.23624999999999996} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6825, "drop": 0.0}, "paraphrase": {"acc": 0.6237623762376238, "drop": 0.058737623762376234, "rotation": {"subspace_principal_angle": 1.3701462267495714, "mean_class_cosine": 0.19930643009504329}}}, "iid_acc": 0.6825, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3138345624755305, "mean_class_cosine": 0.25414324935789645}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.6072607260726073, "drop": -0.007260726072607304, "rotation": {"subspace_principal_angle": 1.420901469694498, "mean_class_cosine": 0.9343142755680263}}}, "iid_acc": 0.6, "selectivity": 0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2228313385316762, "mean_class_cosine": 0.8726714354786228}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.66625, "drop": 0.0}, "paraphrase": {"acc": 0.6254125412541254, "drop": 0.04083745874587463}}, "iid_acc": 0.66625, "selectivity": 0.17625000000000002} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7648902821316614, "drop": -0.03364028213166148, "rotation": {"subspace_principal_angle": 1.1748420004300262, "mean_class_cosine": 0.38568885340073633}}}, "iid_acc": 0.73125, "selectivity": 0.10624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1987833070239917, "mean_class_cosine": 0.3634914913995175}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5094043887147336, "drop": 0.020595611285266435, "rotation": {"subspace_principal_angle": 1.3931515700293393, "mean_class_cosine": 0.8341203196581115}}}, "iid_acc": 0.53, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.5409782777316359, "mean_class_cosine": 0.7395207496379319}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7523510971786834, "drop": -0.026101097178683474}}, "iid_acc": 0.72625, "selectivity": 0.10124999999999995} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.8688981868898187, "drop": 0.046101813110181356, "rotation": {"subspace_principal_angle": 1.0974938475903255, "mean_class_cosine": 0.45582819608799463}}}, "iid_acc": 0.915, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 0.9804764137968498, "mean_class_cosine": 0.5566268231619027}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.8125, "drop": 0.0}, "paraphrase": {"acc": 0.6443514644351465, "drop": 0.16814853556485354, "rotation": {"subspace_principal_angle": 1.3274674971100646, "mean_class_cosine": 0.8179461473320861}}}, "iid_acc": 0.8125, "selectivity": 0.25125, "iid_split_rotation": {"subspace_principal_angle": 1.311349815093557, "mean_class_cosine": 0.9540198561738099}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8521617852161785, "drop": 0.04783821478382155}}, "iid_acc": 0.9, "selectivity": 0.33875} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9852700490998363, "drop": 0.0059799509001636775, "rotation": {"subspace_principal_angle": 0.6333378965948789, "mean_class_cosine": 0.8060565063456264}}}, "iid_acc": 0.99125, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 0.6062508845786333, "mean_class_cosine": 0.8217899986183517}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8772504091653028, "drop": 0.0027495908346971687, "rotation": {"subspace_principal_angle": 1.13183836457679, "mean_class_cosine": 0.9746417417399029}}}, "iid_acc": 0.88, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.0774674925239829, "mean_class_cosine": 0.989696915490704}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9885433715220949, "drop": 0.00020662847790509087}}, "iid_acc": 0.98875, "selectivity": 0.11750000000000005} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.7, "drop": -0.022499999999999964, "rotation": {"subspace_principal_angle": 1.493482753933245, "mean_class_cosine": 0.0772365736671137}}}, "iid_acc": 0.6775, "selectivity": 0.012499999999999956, "iid_split_rotation": {"subspace_principal_angle": 1.5233694038561607, "mean_class_cosine": 0.0474091452723575}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.523943661971831, "drop": 0.003556338028169015, "rotation": {"subspace_principal_angle": 1.364642797562692, "mean_class_cosine": 0.26221087675633176}}}, "iid_acc": 0.5275, "selectivity": -0.13750000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.5645617307273036, "mean_class_cosine": 0.0046736400727034775}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7492957746478873, "drop": -0.019295774647887343}}, "iid_acc": 0.73, "selectivity": 0.06499999999999995} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5722543352601156, "drop": -0.01936971987550018, "rotation": {"subspace_principal_angle": 1.4113114437701857, "mean_class_cosine": 0.18732090600899934}}}, "iid_acc": 0.5528846153846154, "selectivity": 0.1586538461538462, "iid_split_rotation": {"subspace_principal_angle": 1.4113877250233597, "mean_class_cosine": 0.1900533891006572}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.41346153846153844, "drop": 0.0}, "paraphrase": {"acc": 0.44508670520231214, "drop": -0.0316251667407737, "rotation": {"subspace_principal_angle": 1.4614748229703205, "mean_class_cosine": 0.41743847989205585}}}, "iid_acc": 0.41346153846153844, "selectivity": 0.019230769230769218, "iid_split_rotation": {"subspace_principal_angle": 1.1472146976181432, "mean_class_cosine": 0.7289310712969407}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.5336538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": -0.044380835927078754}}, "iid_acc": 0.5336538461538461, "selectivity": 0.13942307692307693} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8848821081830791, "drop": -0.009882108183079108, "rotation": {"subspace_principal_angle": 1.2416497520853838, "mean_class_cosine": 0.32323553402514815}}}, "iid_acc": 0.875, "selectivity": 0.11124999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1798700863789469, "mean_class_cosine": 0.3810449400671245}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5423023578363384, "drop": 0.022697642163661547, "rotation": {"subspace_principal_angle": 1.0521118048612201, "mean_class_cosine": 0.46745139889316134}}}, "iid_acc": 0.565, "selectivity": -0.1987500000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8706941393205756, "mean_class_cosine": 0.8476985688696614}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.84875, "drop": 0.0}, "paraphrase": {"acc": 0.855755894590846, "drop": -0.007005894590845996}}, "iid_acc": 0.84875, "selectivity": 0.08499999999999996} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.81375, "drop": 0.0}, "paraphrase": {"acc": 0.7830985915492957, "drop": 0.030651408450704243, "rotation": {"subspace_principal_angle": 1.2474741652442436, "mean_class_cosine": 0.3177183323280257}}, "domain": {"acc": 0.7875, "drop": 0.026249999999999996, "rotation": {"subspace_principal_angle": 1.1724916667604595, "mean_class_cosine": 0.38785627191420236}}}, "iid_acc": 0.81375, "selectivity": 0.27125, "iid_split_rotation": {"subspace_principal_angle": 1.2460595658666687, "mean_class_cosine": 0.31905931609907595}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.52625, "drop": 0.0}, "paraphrase": {"acc": 0.5450704225352113, "drop": -0.018820422535211323, "rotation": {"subspace_principal_angle": 1.217800297410841, "mean_class_cosine": 0.9938492827921086}}, "domain": {"acc": 0.4875, "drop": 0.03875000000000001, "rotation": {"subspace_principal_angle": 1.4265907543600744, "mean_class_cosine": 0.524717137763018}}}, "iid_acc": 0.52625, "selectivity": -0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.0792601449175692, "mean_class_cosine": 0.9975233270077284}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.8225, "drop": 0.0}, "paraphrase": {"acc": 0.7887323943661971, "drop": 0.03376760563380288}, "domain": {"acc": 0.78875, "drop": 0.03375000000000006}}, "iid_acc": 0.8225, "selectivity": 0.28} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8269961977186312, "drop": 0.02675380228136881, "rotation": {"subspace_principal_angle": 1.0856620278217086, "mean_class_cosine": 0.46632716618126924}}, "domain": {"acc": 0.59125, "drop": 0.26249999999999996, "rotation": {"subspace_principal_angle": 1.2546499825836788, "mean_class_cosine": 0.3109062073482294}}, "length": {"acc": 0.8508371385083714, "drop": 0.0029128614916286155, "rotation": {"subspace_principal_angle": 1.0562705758313071, "mean_class_cosine": 0.492122058178101}}}, "iid_acc": 0.85375, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.04571866168401, "mean_class_cosine": 0.5012802086419837}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.6844106463878327, "drop": -0.03316064638783267, "rotation": {"subspace_principal_angle": 1.2624327988110398, "mean_class_cosine": 0.9443739768139503}}, "domain": {"acc": 0.55625, "drop": 0.09499999999999997, "rotation": {"subspace_principal_angle": 1.2485837708508047, "mean_class_cosine": 0.5140204613064764}}, "length": {"acc": 0.6727549467275494, "drop": -0.021504946727549434, "rotation": {"subspace_principal_angle": 1.3927565265280373, "mean_class_cosine": 0.891203465902359}}}, "iid_acc": 0.65125, "selectivity": 0.17375000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4689242704183265, "mean_class_cosine": 0.9409811928992509}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8136882129277566, "drop": 0.023811787072243407}, "domain": {"acc": 0.6275, "drop": 0.21000000000000008}, "length": {"acc": 0.8340943683409436, "drop": 0.003405631659056385}}, "iid_acc": 0.8375, "selectivity": 0.36000000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.86, "drop": 0.0}, "paraphrase": {"acc": 0.8533541341653667, "drop": 0.0066458658346333355, "rotation": {"subspace_principal_angle": 1.3730733013635024, "mean_class_cosine": 0.2667006488076803}}}, "iid_acc": 0.86, "selectivity": 0.59125, "iid_split_rotation": {"subspace_principal_angle": 1.4307610404327311, "mean_class_cosine": 0.277445881216753}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": -0.006445007800311986, "rotation": {"subspace_principal_angle": 1.4820823089446629, "mean_class_cosine": 0.9745838181982655}}}, "iid_acc": 0.86875, "selectivity": 0.6000000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5380315983443016, "mean_class_cosine": 0.9691609768158005}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.40875, "drop": 0.0}, "paraphrase": {"acc": 0.38221528861154447, "drop": 0.026534711388455534}}, "iid_acc": 0.40875, "selectivity": 0.14} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.9475, "drop": 0.0}, "paraphrase": {"acc": 0.935251798561151, "drop": 0.012248201438848971, "rotation": {"subspace_principal_angle": 1.1046415070501514, "mean_class_cosine": 0.6973108185521933}}}, "iid_acc": 0.9475, "selectivity": 0.87375, "iid_split_rotation": {"subspace_principal_angle": 1.0670730680867, "mean_class_cosine": 0.747139744826814}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.4628297362110312, "drop": 0.03967026378896876, "rotation": {"subspace_principal_angle": 1.5435774572835343, "mean_class_cosine": 0.8716859158620075}}}, "iid_acc": 0.5025, "selectivity": 0.42874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.389055684030504, "mean_class_cosine": 0.7772996017727697}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.92625, "drop": 0.0}, "paraphrase": {"acc": 0.9160671462829736, "drop": 0.01018285371702643}}, "iid_acc": 0.92625, "selectivity": 0.8525} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5176470588235295, "drop": 0.018602941176470544, "rotation": {"subspace_principal_angle": 1.5270884311030197, "mean_class_cosine": 0.04369398057174439}}}, "iid_acc": 0.53625, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.476543368793087, "mean_class_cosine": 0.094113468729472}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.5045751633986928, "drop": 0.011674836601307192, "rotation": {"subspace_principal_angle": 1.5294742379185886, "mean_class_cosine": 0.18570440239474292}}}, "iid_acc": 0.51625, "selectivity": 0.04249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.444417852756359, "mean_class_cosine": 0.20646522396999623}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 2, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5333333333333333, "drop": 0.0029166666666666785}}, "iid_acc": 0.53625, "selectivity": 0.0625} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.5431266846361186, "drop": 0.0518733153638814, "rotation": {"subspace_principal_angle": 1.467102949540591, "mean_class_cosine": 0.29891759428990633}}}, "iid_acc": 0.595, "selectivity": 0.31375, "iid_split_rotation": {"subspace_principal_angle": 1.5242732907156675, "mean_class_cosine": 0.27641104971513886}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.1925, "drop": 0.0}, "paraphrase": {"acc": 0.20754716981132076, "drop": -0.01504716981132076, "rotation": {"subspace_principal_angle": 1.3006810428081945, "mean_class_cosine": 0.5938365988587205}}}, "iid_acc": 0.1925, "selectivity": -0.08875, "iid_split_rotation": {"subspace_principal_angle": 1.5008582164851936, "mean_class_cosine": 0.6730802808549522}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5309973045822103, "drop": 0.017752695417789677}}, "iid_acc": 0.54875, "selectivity": 0.26749999999999996} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7429048414023373, "drop": -0.002904841402337266, "rotation": {"subspace_principal_angle": 1.321041649041896, "mean_class_cosine": 0.24716625603971074}}}, "iid_acc": 0.74, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3832981091452556, "mean_class_cosine": 0.18640154565031142}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5976627712854758, "drop": -0.03766277128547579, "rotation": {"subspace_principal_angle": 1.3172594721297197, "mean_class_cosine": 0.9641971900851679}}}, "iid_acc": 0.56, "selectivity": 0.0525000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.1295344939397884, "mean_class_cosine": 0.9754037717574164}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.72875, "drop": 0.0}, "paraphrase": {"acc": 0.7378964941569283, "drop": -0.009146494156928253}}, "iid_acc": 0.72875, "selectivity": 0.22125000000000006} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.6551155115511551, "drop": 0.01988448844884494, "rotation": {"subspace_principal_angle": 1.4103931207701927, "mean_class_cosine": 0.1597162496412831}}}, "iid_acc": 0.675, "selectivity": 0.18500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.451975006440924, "mean_class_cosine": 0.1185419210711762}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5396039603960396, "drop": 0.0016460396039603697, "rotation": {"subspace_principal_angle": 1.5446898330353456, "mean_class_cosine": 0.9934586731916171}}}, "iid_acc": 0.54125, "selectivity": 0.05125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3541190216750216, "mean_class_cosine": 0.9952103287013051}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.6975, "drop": 0.0}, "paraphrase": {"acc": 0.6617161716171617, "drop": 0.035783828382838334}}, "iid_acc": 0.6975, "selectivity": 0.20750000000000002} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7523510971786834, "drop": -0.00735109717868343, "rotation": {"subspace_principal_angle": 1.3838768695601789, "mean_class_cosine": 0.18583289760856309}}}, "iid_acc": 0.745, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.411496556459018, "mean_class_cosine": 0.1586268817831123}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.485, "drop": 0.0}, "paraphrase": {"acc": 0.5219435736677116, "drop": -0.036943573667711616, "rotation": {"subspace_principal_angle": 1.4598677944473912, "mean_class_cosine": 0.7189138058825407}}}, "iid_acc": 0.485, "selectivity": -0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.257226644633361, "mean_class_cosine": 0.8721290560249453}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7507836990595611, "drop": 0.005466300940438873}}, "iid_acc": 0.75625, "selectivity": 0.17374999999999996} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9121338912133892, "drop": 0.010366108786610817, "rotation": {"subspace_principal_angle": 1.190236906942168, "mean_class_cosine": 0.3714399247796692}}}, "iid_acc": 0.9225, "selectivity": 0.38749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1180814498000364, "mean_class_cosine": 0.43740853126331475}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.8019525801952581, "drop": 0.0642974198047419, "rotation": {"subspace_principal_angle": 1.316257361185483, "mean_class_cosine": 0.9428379882047717}}}, "iid_acc": 0.86625, "selectivity": 0.33124999999999993, "iid_split_rotation": {"subspace_principal_angle": 0.8984797392010688, "mean_class_cosine": 0.9658337076254448}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.9107391910739191, "drop": 0.009260808926080943}}, "iid_acc": 0.92, "selectivity": 0.385} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9950900163666121, "drop": 0.0011599836333878732, "rotation": {"subspace_principal_angle": 0.8576615569303944, "mean_class_cosine": 0.6542078543716676}}}, "iid_acc": 0.99625, "selectivity": 0.15000000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9012913552860777, "mean_class_cosine": 0.6205978969084885}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.68125, "drop": 0.0}, "paraphrase": {"acc": 0.5548281505728314, "drop": 0.1264218494271686, "rotation": {"subspace_principal_angle": 0.6861186089833221, "mean_class_cosine": 0.9976598565566228}}}, "iid_acc": 0.68125, "selectivity": -0.16499999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.6195594346223908, "mean_class_cosine": 0.9998339526246938}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": -0.0009533551554827824}}, "iid_acc": 0.9925, "selectivity": 0.1462500000000001} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6915492957746479, "drop": -0.004049295774647854, "rotation": {"subspace_principal_angle": 1.4821835383859001, "mean_class_cosine": 0.08849686598665059}}}, "iid_acc": 0.6875, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5422652050607666, "mean_class_cosine": 0.0285272510510537}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.55375, "drop": 0.0}, "paraphrase": {"acc": 0.5140845070422535, "drop": 0.03966549295774646, "rotation": {"subspace_principal_angle": 1.411758063021396, "mean_class_cosine": -0.05543427344626703}}}, "iid_acc": 0.55375, "selectivity": -0.08500000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.1826722547195974, "mean_class_cosine": -0.8190873052106897}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7450704225352113, "drop": -0.01757042253521124}}, "iid_acc": 0.7275, "selectivity": 0.08875} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5673076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.5549132947976878, "drop": 0.012394397510004485, "rotation": {"subspace_principal_angle": 1.435539235192927, "mean_class_cosine": 0.18347039301699317}}}, "iid_acc": 0.5673076923076923, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.3743463171910144, "mean_class_cosine": 0.24089153525080556}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.4326923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.41040462427745666, "drop": 0.022287683414851045, "rotation": {"subspace_principal_angle": 1.3988658583948614, "mean_class_cosine": -0.27577383548106055}}}, "iid_acc": 0.4326923076923077, "selectivity": 0.04326923076923078, "iid_split_rotation": {"subspace_principal_angle": 1.0687827569428319, "mean_class_cosine": 0.8344718355538371}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5528846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.5953757225433526, "drop": -0.04249110715873716}}, "iid_acc": 0.5528846153846154, "selectivity": 0.1634615384615385} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.89625, "drop": 0.0}, "paraphrase": {"acc": 0.9029126213592233, "drop": -0.006662621359223353, "rotation": {"subspace_principal_angle": 1.20453886521433, "mean_class_cosine": 0.3581236367027224}}}, "iid_acc": 0.89625, "selectivity": 0.17374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.2309341256728386, "mean_class_cosine": 0.3333571784399262}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5298196948682385, "drop": 0.022680305131761447, "rotation": {"subspace_principal_angle": 1.310916058135534, "mean_class_cosine": -0.07981176620814859}}}, "iid_acc": 0.5525, "selectivity": -0.17000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1896244566676593, "mean_class_cosine": 0.9512750131760109}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.8575, "drop": 0.0}, "paraphrase": {"acc": 0.8640776699029126, "drop": -0.006577669902912531}}, "iid_acc": 0.8575, "selectivity": 0.135} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8056338028169014, "drop": 0.048116197183098564, "rotation": {"subspace_principal_angle": 1.2927332146049364, "mean_class_cosine": 0.2744936747860658}}, "domain": {"acc": 0.82375, "drop": 0.030000000000000027, "rotation": {"subspace_principal_angle": 1.2846999556990648, "mean_class_cosine": 0.28220942740249133}}}, "iid_acc": 0.85375, "selectivity": 0.32125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.220104644031074, "mean_class_cosine": 0.3435474732924775}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.52, "drop": 0.0}, "paraphrase": {"acc": 0.532394366197183, "drop": -0.012394366197183038, "rotation": {"subspace_principal_angle": 0.365026570963859, "mean_class_cosine": 0.9987360435705452}}, "domain": {"acc": 0.4875, "drop": 0.03250000000000003, "rotation": {"subspace_principal_angle": 1.5449752153763945, "mean_class_cosine": 0.4557918147259487}}}, "iid_acc": 0.52, "selectivity": -0.012499999999999956, "iid_split_rotation": {"subspace_principal_angle": 0.34709256477040135, "mean_class_cosine": 0.9995894414380515}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.8225352112676056, "drop": 0.022464788732394325}, "domain": {"acc": 0.8325, "drop": 0.012499999999999956}}, "iid_acc": 0.845, "selectivity": 0.3125} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8688212927756654, "drop": 0.016178707224334565, "rotation": {"subspace_principal_angle": 1.2184200366683278, "mean_class_cosine": 0.34512905932789906}}, "domain": {"acc": 0.7325, "drop": 0.15249999999999997, "rotation": {"subspace_principal_angle": 1.3710819230300715, "mean_class_cosine": 0.1983894193717613}}, "length": {"acc": 0.8919330289193302, "drop": -0.006933028919330231, "rotation": {"subspace_principal_angle": 1.1635770628342312, "mean_class_cosine": 0.3960575192257396}}}, "iid_acc": 0.885, "selectivity": 0.40625, "iid_split_rotation": {"subspace_principal_angle": 1.178058815472859, "mean_class_cosine": 0.38271893642492194}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.7585551330798479, "drop": -0.02605513307984786, "rotation": {"subspace_principal_angle": 1.0808603554768208, "mean_class_cosine": 0.9645911983303304}}, "domain": {"acc": 0.5575, "drop": 0.17500000000000004, "rotation": {"subspace_principal_angle": 1.4332455458492723, "mean_class_cosine": 0.4266662959732691}}, "length": {"acc": 0.7595129375951294, "drop": -0.02701293759512935, "rotation": {"subspace_principal_angle": 1.1028359956060956, "mean_class_cosine": 0.9361757660470457}}}, "iid_acc": 0.7325, "selectivity": 0.25375000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.340774604922017, "mean_class_cosine": 0.9544915504856384}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.876425855513308, "drop": 0.01732414448669206}, "domain": {"acc": 0.57875, "drop": 0.31500000000000006}, "length": {"acc": 0.898021308980213, "drop": -0.0042713089802129955}}, "iid_acc": 0.89375, "selectivity": 0.41500000000000004} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": 0.0023049921996879386, "rotation": {"subspace_principal_angle": 1.2939750508328909, "mean_class_cosine": 0.3697587563743038}}}, "iid_acc": 0.8775, "selectivity": 0.60625, "iid_split_rotation": {"subspace_principal_angle": 1.3239667655285696, "mean_class_cosine": 0.3832240775461434}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.57375, "drop": 0.0}, "paraphrase": {"acc": 0.5772230889235569, "drop": -0.0034730889235569107, "rotation": {"subspace_principal_angle": 1.4461389808941905, "mean_class_cosine": 0.8942227001502406}}}, "iid_acc": 0.57375, "selectivity": 0.3025, "iid_split_rotation": {"subspace_principal_angle": 1.5670258105099684, "mean_class_cosine": 0.9555720866412488}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.87625, "drop": 0.0}, "paraphrase": {"acc": 0.875195007800312, "drop": 0.0010549921996879652}}, "iid_acc": 0.87625, "selectivity": 0.605} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.9625, "drop": 0.0}, "paraphrase": {"acc": 0.9568345323741008, "drop": 0.0056654676258992565, "rotation": {"subspace_principal_angle": 1.0889403294719116, "mean_class_cosine": 0.7097651658028382}}}, "iid_acc": 0.9625, "selectivity": 0.90625, "iid_split_rotation": {"subspace_principal_angle": 1.0742515810195217, "mean_class_cosine": 0.7520191017637144}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.475, "drop": 0.0}, "paraphrase": {"acc": 0.45083932853717024, "drop": 0.02416067146282974, "rotation": {"subspace_principal_angle": 1.1366627849078705, "mean_class_cosine": 0.861011502467378}}}, "iid_acc": 0.475, "selectivity": 0.41874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5498782331536536, "mean_class_cosine": 0.7592574844685773}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9400479616306955, "drop": 0.014952038369304477}}, "iid_acc": 0.955, "selectivity": 0.8987499999999999} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5830065359477125, "drop": 0.01699346405228752, "rotation": {"subspace_principal_angle": 1.4156934488627833, "mean_class_cosine": 0.1544817430571851}}}, "iid_acc": 0.6, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.3875062308714112, "mean_class_cosine": 0.18226553873635268}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.5375, "drop": 0.0}, "paraphrase": {"acc": 0.5254901960784314, "drop": 0.012009803921568585, "rotation": {"subspace_principal_angle": 0.8046612760829073, "mean_class_cosine": 0.9020520136987834}}}, "iid_acc": 0.5375, "selectivity": 0.03249999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.5910745370085485, "mean_class_cosine": -0.921862888991628}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.5825, "drop": 0.0}, "paraphrase": {"acc": 0.550326797385621, "drop": 0.03217320261437906}}, "iid_acc": 0.5825, "selectivity": 0.07750000000000001} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.5876010781671159, "drop": 0.011148921832884073, "rotation": {"subspace_principal_angle": 1.4649394534970999, "mean_class_cosine": 0.27348013078523753}}}, "iid_acc": 0.59875, "selectivity": 0.3075, "iid_split_rotation": {"subspace_principal_angle": 1.5210318775010727, "mean_class_cosine": 0.26014522379767274}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.18375, "drop": 0.0}, "paraphrase": {"acc": 0.1725067385444744, "drop": 0.011243261455525594, "rotation": {"subspace_principal_angle": 1.2577695007550802, "mean_class_cosine": 0.6357274521759256}}}, "iid_acc": 0.18375, "selectivity": -0.10750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4122075752855408, "mean_class_cosine": 0.6982530139834328}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.59375, "drop": 0.0}, "paraphrase": {"acc": 0.5592991913746631, "drop": 0.03445080862533689}}, "iid_acc": 0.59375, "selectivity": 0.3025} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7161936560934892, "drop": 0.0025563439065108273, "rotation": {"subspace_principal_angle": 1.4142647213981934, "mean_class_cosine": 0.1558931614185028}}}, "iid_acc": 0.71875, "selectivity": 0.23375, "iid_split_rotation": {"subspace_principal_angle": 1.4440239039597567, "mean_class_cosine": 0.12643313043512422}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5442404006677797, "drop": -0.014240400667779651, "rotation": {"subspace_principal_angle": 1.438598695547982, "mean_class_cosine": 0.9982834430354565}}}, "iid_acc": 0.53, "selectivity": 0.04500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.312217021230763, "mean_class_cosine": 0.9973052099136428}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7712854757929883, "drop": -0.0075354757929883}}, "iid_acc": 0.76375, "selectivity": 0.27875000000000005} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6567656765676567, "drop": 0.02323432343234333, "rotation": {"subspace_principal_angle": 1.4737611346239565, "mean_class_cosine": 0.09688298605763825}}}, "iid_acc": 0.68, "selectivity": 0.18875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4500535476739749, "mean_class_cosine": 0.12044961172303534}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5379537953795379, "drop": 0.002046204620462122, "rotation": {"subspace_principal_angle": 0.9674520996825667, "mean_class_cosine": 0.9975306978126972}}}, "iid_acc": 0.54, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 0.6210793518048389, "mean_class_cosine": 0.9983833459358811}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.6914191419141914, "drop": 0.041080858085808636}}, "iid_acc": 0.7325, "selectivity": 0.24125000000000002} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.71375, "drop": 0.0}, "paraphrase": {"acc": 0.7304075235109718, "drop": -0.016657523510971828, "rotation": {"subspace_principal_angle": 1.4688894465039684, "mean_class_cosine": 0.10173058782242872}}}, "iid_acc": 0.71375, "selectivity": 0.1775, "iid_split_rotation": {"subspace_principal_angle": 1.4124632331913725, "mean_class_cosine": 0.15767237056399389}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.49625, "drop": 0.0}, "paraphrase": {"acc": 0.5266457680250783, "drop": -0.030395768025078318, "rotation": {"subspace_principal_angle": 0.7632925636553622, "mean_class_cosine": 0.9245850252469596}}}, "iid_acc": 0.49625, "selectivity": -0.03999999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.8486480460237737, "mean_class_cosine": 0.7937846466717617}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.75125, "drop": 0.0}, "paraphrase": {"acc": 0.7570532915360502, "drop": -0.0058032915360501924}}, "iid_acc": 0.75125, "selectivity": 0.21499999999999997} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9316596931659693, "drop": 0.0070903068340306685, "rotation": {"subspace_principal_angle": 1.1500093845203234, "mean_class_cosine": 0.40847887501442204}}}, "iid_acc": 0.93875, "selectivity": 0.4149999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.0038823721844266, "mean_class_cosine": 0.5370313385967125}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.62375, "drop": 0.0}, "paraphrase": {"acc": 0.6192468619246861, "drop": 0.004503138075313884, "rotation": {"subspace_principal_angle": 1.486642282369967, "mean_class_cosine": 0.9620795603326425}}}, "iid_acc": 0.62375, "selectivity": 0.09999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4105065066746125, "mean_class_cosine": 0.9378064089839906}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9386331938633193, "drop": 0.00011680613668063611}}, "iid_acc": 0.93875, "selectivity": 0.4149999999999999} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9819967266775778, "drop": 0.009253273322422206, "rotation": {"subspace_principal_angle": 0.7667082993155465, "mean_class_cosine": 0.7201982449314277}}}, "iid_acc": 0.99125, "selectivity": 0.13374999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.7510140573282249, "mean_class_cosine": 0.7309972720101955}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.939443535188216, "drop": -0.006943535188215999, "rotation": {"subspace_principal_angle": 0.996635177095526, "mean_class_cosine": 0.953156648540542}}}, "iid_acc": 0.9325, "selectivity": 0.07499999999999996, "iid_split_rotation": {"subspace_principal_angle": 0.9887430183481719, "mean_class_cosine": 0.9831308468715932}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9819967266775778, "drop": 0.00675327332242226}}, "iid_acc": 0.98875, "selectivity": 0.13124999999999998} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.6873239436619718, "drop": 0.02267605633802816, "rotation": {"subspace_principal_angle": 1.4852509416429518, "mean_class_cosine": 0.08544108628330754}}}, "iid_acc": 0.71, "selectivity": 0.12124999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4922855379226647, "mean_class_cosine": 0.0784301577092764}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5154929577464789, "drop": 0.03325704225352111, "rotation": {"subspace_principal_angle": 0.5600755724824341, "mean_class_cosine": -0.860265927958924}}}, "iid_acc": 0.54875, "selectivity": -0.040000000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.2092432879483297, "mean_class_cosine": 0.6330568261731295}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7464788732394366, "drop": -0.006478873239436633}}, "iid_acc": 0.74, "selectivity": 0.15125} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.6057692307692307, "drop": 0.0}, "paraphrase": {"acc": 0.6184971098265896, "drop": -0.012727879057358837, "rotation": {"subspace_principal_angle": 1.4119061291333692, "mean_class_cosine": 0.1704373905622454}}}, "iid_acc": 0.6057692307692307, "selectivity": 0.24999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4297364412104765, "mean_class_cosine": 0.20248169581710895}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.4567307692307692, "drop": 0.0}, "paraphrase": {"acc": 0.4508670520231214, "drop": 0.005863717207647834, "rotation": {"subspace_principal_angle": 1.3366057576399157, "mean_class_cosine": 0.7129666544008023}}}, "iid_acc": 0.4567307692307692, "selectivity": 0.10096153846153844, "iid_split_rotation": {"subspace_principal_angle": 1.417385145003071, "mean_class_cosine": 0.826148295807795}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.5480769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.5433526011560693, "drop": 0.004724321920853813}}, "iid_acc": 0.5480769230769231, "selectivity": 0.19230769230769235} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8932038834951457, "drop": 0.0017961165048543393, "rotation": {"subspace_principal_angle": 1.311372754664119, "mean_class_cosine": 0.2565234550971107}}}, "iid_acc": 0.895, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.2262739349985579, "mean_class_cosine": 0.337747173335497}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5436893203883495, "drop": 0.022560679611650536, "rotation": {"subspace_principal_angle": 1.2898034193621133, "mean_class_cosine": -0.3258881311565935}}}, "iid_acc": 0.56625, "selectivity": -0.1825, "iid_split_rotation": {"subspace_principal_angle": 1.2264589211138428, "mean_class_cosine": 0.9728987695098255}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.90375, "drop": 0.0}, "paraphrase": {"acc": 0.9042995839112344, "drop": -0.0005495839112343859}}, "iid_acc": 0.90375, "selectivity": 0.15500000000000003} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.8323943661971831, "drop": 0.02135563380281691, "rotation": {"subspace_principal_angle": 1.2914523895644616, "mean_class_cosine": 0.27572507644251076}}, "domain": {"acc": 0.87125, "drop": -0.01749999999999996, "rotation": {"subspace_principal_angle": 1.4039511655723063, "mean_class_cosine": 0.16607215124792452}}}, "iid_acc": 0.85375, "selectivity": 0.33125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.250730833076561, "mean_class_cosine": 0.3146287288985242}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5352112676056338, "drop": -0.011461267605633707, "rotation": {"subspace_principal_angle": 0.9170707486812013, "mean_class_cosine": 0.9954078358133513}}, "domain": {"acc": 0.4875, "drop": 0.03625000000000006, "rotation": {"subspace_principal_angle": 1.5523343990205853, "mean_class_cosine": 0.22749359327828106}}}, "iid_acc": 0.52375, "selectivity": 0.0012500000000000844, "iid_split_rotation": {"subspace_principal_angle": 1.4480914990441032, "mean_class_cosine": 0.9983051052181077}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8436619718309859, "drop": 0.033838028169014045}, "domain": {"acc": 0.82, "drop": 0.057499999999999996}}, "iid_acc": 0.8775, "selectivity": 0.355} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8840304182509505, "drop": 0.014719581749049526, "rotation": {"subspace_principal_angle": 1.3732084918619123, "mean_class_cosine": 0.19630467274275237}}, "domain": {"acc": 0.655, "drop": 0.24375000000000002, "rotation": {"subspace_principal_angle": 1.4452895556036265, "mean_class_cosine": 0.1251775345482053}}, "length": {"acc": 0.8995433789954338, "drop": -0.0007933789954337467, "rotation": {"subspace_principal_angle": 1.335992218473073, "mean_class_cosine": 0.2326524735797551}}}, "iid_acc": 0.89875, "selectivity": 0.39875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3151033264444465, "mean_class_cosine": 0.2529159390133714}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.8365019011406845, "drop": 0.032248098859315566, "rotation": {"subspace_principal_angle": 1.0533503571330138, "mean_class_cosine": 0.9658970994045344}}, "domain": {"acc": 0.5575, "drop": 0.31125, "rotation": {"subspace_principal_angle": 1.3501859541358534, "mean_class_cosine": 0.2666305404277387}}, "length": {"acc": 0.8751902587519026, "drop": -0.006440258751902572, "rotation": {"subspace_principal_angle": 1.4766030164776944, "mean_class_cosine": 0.9618498962901108}}}, "iid_acc": 0.86875, "selectivity": 0.36875, "iid_split_rotation": {"subspace_principal_angle": 0.8666123632485124, "mean_class_cosine": 0.9567597495837794}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.91875, "drop": 0.0}, "paraphrase": {"acc": 0.8954372623574145, "drop": 0.023312737642585457}, "domain": {"acc": 0.83125, "drop": 0.08749999999999991}, "length": {"acc": 0.923896499238965, "drop": -0.005146499238965041}}, "iid_acc": 0.91875, "selectivity": 0.41874999999999996} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.8829953198127926, "drop": -0.0004953198127926051, "rotation": {"subspace_principal_angle": 1.300584248001103, "mean_class_cosine": 0.33037107823356726}}}, "iid_acc": 0.8825, "selectivity": 0.5887499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3544227368327597, "mean_class_cosine": 0.32442597332270096}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7410296411856474, "drop": -0.013529641185647368, "rotation": {"subspace_principal_angle": 1.0031403826052068, "mean_class_cosine": 0.9483063492551265}}}, "iid_acc": 0.7275, "selectivity": 0.43375, "iid_split_rotation": {"subspace_principal_angle": 0.835980512632536, "mean_class_cosine": 0.9567132219152211}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.8975, "drop": 0.0}, "paraphrase": {"acc": 0.8939157566302652, "drop": 0.00358424336973473}}, "iid_acc": 0.8975, "selectivity": 0.60375} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.97375, "drop": 0.0}, "paraphrase": {"acc": 0.9592326139088729, "drop": 0.014517386091127071, "rotation": {"subspace_principal_angle": 1.0461394619384272, "mean_class_cosine": 0.7060277954956268}}}, "iid_acc": 0.97375, "selectivity": 0.90375, "iid_split_rotation": {"subspace_principal_angle": 1.0572533041916146, "mean_class_cosine": 0.7322189822775439}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6546762589928058, "drop": 0.005323741007194238, "rotation": {"subspace_principal_angle": 1.3147848288233726, "mean_class_cosine": 0.8625758379093228}}}, "iid_acc": 0.66, "selectivity": 0.5900000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.2749765898173022, "mean_class_cosine": 0.8458260421792687}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.97125, "drop": 0.0}, "paraphrase": {"acc": 0.9616306954436451, "drop": 0.009619304556354846}}, "iid_acc": 0.97125, "selectivity": 0.9012499999999999} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.62875, "drop": 0.0}, "paraphrase": {"acc": 0.6065359477124183, "drop": 0.022214052287581776, "rotation": {"subspace_principal_angle": 1.4471225946917146, "mean_class_cosine": 0.12335870420709219}}}, "iid_acc": 0.62875, "selectivity": 0.14375000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.396757975212198, "mean_class_cosine": 0.1731610965171537}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.54, "drop": 0.0}, "paraphrase": {"acc": 0.5202614379084968, "drop": 0.019738562091503264, "rotation": {"subspace_principal_angle": 1.017105332000735, "mean_class_cosine": 0.8150921314497699}}}, "iid_acc": 0.54, "selectivity": 0.05500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.3938330496938112, "mean_class_cosine": -0.5401181987718008}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.63125, "drop": 0.0}, "paraphrase": {"acc": 0.6143790849673203, "drop": 0.01687091503267968}}, "iid_acc": 0.63125, "selectivity": 0.14625} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.655, "drop": 0.0}, "paraphrase": {"acc": 0.6266846361185984, "drop": 0.028315363881401656, "rotation": {"subspace_principal_angle": 1.4639056921866045, "mean_class_cosine": 0.2607364800995969}}}, "iid_acc": 0.655, "selectivity": 0.38875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5392183501899661, "mean_class_cosine": 0.23109273289109802}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.1925, "drop": 0.0}, "paraphrase": {"acc": 0.18867924528301888, "drop": 0.003820754716981123, "rotation": {"subspace_principal_angle": 1.3025058812372758, "mean_class_cosine": 0.609760025746434}}}, "iid_acc": 0.1925, "selectivity": -0.07374999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.380643945283782, "mean_class_cosine": 0.7278668025302554}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.64875, "drop": 0.0}, "paraphrase": {"acc": 0.6145552560646901, "drop": 0.03419474393530997}}, "iid_acc": 0.64875, "selectivity": 0.38250000000000006} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.775, "drop": 0.0}, "paraphrase": {"acc": 0.7696160267111853, "drop": 0.005383973288814681, "rotation": {"subspace_principal_angle": 1.4245590947156432, "mean_class_cosine": 0.145716567257158}}}, "iid_acc": 0.775, "selectivity": 0.28875, "iid_split_rotation": {"subspace_principal_angle": 1.471068816739365, "mean_class_cosine": 0.09956228431408685}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5425709515859767, "drop": -0.017570951585976657, "rotation": {"subspace_principal_angle": 1.4049876290028978, "mean_class_cosine": 0.8302182537519393}}}, "iid_acc": 0.525, "selectivity": 0.03875000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.628667819566516, "mean_class_cosine": 0.9886859497074707}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.7875, "drop": 0.0}, "paraphrase": {"acc": 0.7863105175292153, "drop": 0.001189482470784653}}, "iid_acc": 0.7875, "selectivity": 0.30124999999999996} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6666666666666666, "drop": 0.00708333333333333, "rotation": {"subspace_principal_angle": 1.4652126246929733, "mean_class_cosine": 0.10538763834081566}}}, "iid_acc": 0.67375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.453567266755897, "mean_class_cosine": 0.1169607381052836}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5412541254125413, "drop": 0.00874587458745879, "rotation": {"subspace_principal_angle": 1.435416177004221, "mean_class_cosine": -0.8400641400418787}}}, "iid_acc": 0.55, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.353882512217704, "mean_class_cosine": 0.9873913376016163}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 18, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.6897689768976898, "drop": 0.026481023102310264}}, "iid_acc": 0.71625, "selectivity": 0.22250000000000003} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7476489028213166, "drop": -0.02014890282131654, "rotation": {"subspace_principal_angle": 1.5031766373236868, "mean_class_cosine": 0.06756817028714371}}}, "iid_acc": 0.7275, "selectivity": 0.21375, "iid_split_rotation": {"subspace_principal_angle": 1.4985421261579637, "mean_class_cosine": 0.07219134782882808}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5297805642633229, "drop": -0.016030564263322833, "rotation": {"subspace_principal_angle": 1.4269512459642844, "mean_class_cosine": 0.6191468418063922}}}, "iid_acc": 0.51375, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 0.6941036459644455, "mean_class_cosine": 0.6890861724335152}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7586206896551724, "drop": -0.008620689655172376}}, "iid_acc": 0.75, "selectivity": 0.23624999999999996} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9400278940027894, "drop": 0.01122210599721063, "rotation": {"subspace_principal_angle": 1.212406540346009, "mean_class_cosine": 0.3507667834923325}}}, "iid_acc": 0.95125, "selectivity": 0.41875000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.1491505155466, "mean_class_cosine": 0.40926267218064927}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.785, "drop": 0.0}, "paraphrase": {"acc": 0.796373779637378, "drop": -0.011373779637377956, "rotation": {"subspace_principal_angle": 0.5776320485859313, "mean_class_cosine": 0.9642980239049406}}}, "iid_acc": 0.785, "selectivity": 0.25250000000000006, "iid_split_rotation": {"subspace_principal_angle": 0.8609996534230742, "mean_class_cosine": 0.9712015306482873}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 17, "dists": {"iid": {"acc": 0.95125, "drop": 0.0}, "paraphrase": {"acc": 0.9372384937238494, "drop": 0.014011506276150665}}, "iid_acc": 0.95125, "selectivity": 0.41875000000000007} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.9975, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.004046644844517222, "rotation": {"subspace_principal_angle": 0.8761164033837602, "mean_class_cosine": 0.6401395707229799}}}, "iid_acc": 0.9975, "selectivity": 0.1825000000000001, "iid_split_rotation": {"subspace_principal_angle": 0.8963642783199154, "mean_class_cosine": 0.6244538122618211}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.4075286415711948, "drop": 0.3112213584288052, "rotation": {"subspace_principal_angle": 1.3896444565022879, "mean_class_cosine": 0.8607515326377927}}}, "iid_acc": 0.71875, "selectivity": -0.09624999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.1374176741267867, "mean_class_cosine": 0.9949908284390802}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.99625, "drop": 0.0}, "paraphrase": {"acc": 0.9918166939443536, "drop": 0.004433306055646402}}, "iid_acc": 0.99625, "selectivity": 0.18125000000000002} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7112676056338029, "drop": 0.02498239436619709, "rotation": {"subspace_principal_angle": 1.512652494056396, "mean_class_cosine": 0.05811107708207547}}}, "iid_acc": 0.73625, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.5022867064091796, "mean_class_cosine": 0.06845604053333303}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5338028169014084, "drop": 0.03119718309859154, "rotation": {"subspace_principal_angle": 1.2680709408982833, "mean_class_cosine": -0.35755644118795693}}}, "iid_acc": 0.565, "selectivity": -0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.547048981725239, "mean_class_cosine": 0.09058745031393456}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.78125, "drop": 0.0}, "paraphrase": {"acc": 0.7507042253521127, "drop": 0.030545774647887325}}, "iid_acc": 0.78125, "selectivity": 0.19125000000000003} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5961538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5953757225433526, "drop": 0.000778123610493564, "rotation": {"subspace_principal_angle": 1.4398635945392528, "mean_class_cosine": 0.13929059880991962}}}, "iid_acc": 0.5961538461538461, "selectivity": 0.22115384615384615, "iid_split_rotation": {"subspace_principal_angle": 1.4596191174582307, "mean_class_cosine": 0.18093919697525562}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5192307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.5491329479768786, "drop": -0.029902178746109342, "rotation": {"subspace_principal_angle": 1.0835579282901093, "mean_class_cosine": 0.6041012733975983}}}, "iid_acc": 0.5192307692307693, "selectivity": 0.14423076923076927, "iid_split_rotation": {"subspace_principal_angle": 1.0998389268494577, "mean_class_cosine": 0.6208614359967286}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5721153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": -0.0059192974655403185}}, "iid_acc": 0.5721153846153846, "selectivity": 0.19711538461538458} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.9056865464632455, "drop": -0.0006865464632455076, "rotation": {"subspace_principal_angle": 1.3934785192501797, "mean_class_cosine": 0.17639007447692165}}}, "iid_acc": 0.905, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3193210248676257, "mean_class_cosine": 0.24883312785983533}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.57, "drop": 0.0}, "paraphrase": {"acc": 0.5533980582524272, "drop": 0.01660194174757279, "rotation": {"subspace_principal_angle": 1.5354546576355417, "mean_class_cosine": 0.1563657422847014}}}, "iid_acc": 0.57, "selectivity": -0.14250000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.445383158458667, "mean_class_cosine": 0.9556115381599443}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9084604715672677, "drop": 0.006539528432732311}}, "iid_acc": 0.915, "selectivity": 0.2025} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.776056338028169, "drop": 0.033943661971831074, "rotation": {"subspace_principal_angle": 1.3558532015708853, "mean_class_cosine": 0.2132918626421837}}, "domain": {"acc": 0.8, "drop": 0.010000000000000009, "rotation": {"subspace_principal_angle": 1.3213395199914533, "mean_class_cosine": 0.24687761614550957}}}, "iid_acc": 0.81, "selectivity": 0.30000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2949820598326707, "mean_class_cosine": 0.27233051788958473}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5309859154929577, "drop": -0.002235915492957652, "rotation": {"subspace_principal_angle": 0.949671245424435, "mean_class_cosine": 0.9981407854560767}}, "domain": {"acc": 0.4875, "drop": 0.041250000000000064, "rotation": {"subspace_principal_angle": 1.5209675895490495, "mean_class_cosine": 0.5347820929835523}}}, "iid_acc": 0.52875, "selectivity": 0.018750000000000044, "iid_split_rotation": {"subspace_principal_angle": 0.5977677192699846, "mean_class_cosine": 0.9995599104407848}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.825, "drop": 0.0}, "paraphrase": {"acc": 0.7295774647887324, "drop": 0.09542253521126753}, "domain": {"acc": 0.73125, "drop": 0.09375}}, "iid_acc": 0.825, "selectivity": 0.31499999999999995} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.825, "drop": 0.0}, "paraphrase": {"acc": 0.8022813688212928, "drop": 0.022718631178707205, "rotation": {"subspace_principal_angle": 1.3701838535641038, "mean_class_cosine": 0.19926955803809265}}, "domain": {"acc": 0.56, "drop": 0.2649999999999999, "rotation": {"subspace_principal_angle": 1.3671422264532063, "mean_class_cosine": 0.2022492580826582}}, "length": {"acc": 0.8356164383561644, "drop": -0.010616438356164437, "rotation": {"subspace_principal_angle": 1.3466907976398639, "mean_class_cosine": 0.22223434808814393}}}, "iid_acc": 0.825, "selectivity": 0.32499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3129495009920322, "mean_class_cosine": 0.25499915160174697}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.595, "drop": 0.0}, "paraphrase": {"acc": 0.6406844106463878, "drop": -0.04568441064638784, "rotation": {"subspace_principal_angle": 1.5652743111758125, "mean_class_cosine": 0.9561201033378162}}, "domain": {"acc": 0.5575, "drop": 0.03749999999999998, "rotation": {"subspace_principal_angle": 1.307138927137268, "mean_class_cosine": 0.5860658652565125}}, "length": {"acc": 0.6164383561643836, "drop": -0.02143835616438361, "rotation": {"subspace_principal_angle": 0.9884986790030585, "mean_class_cosine": 0.9027214333318725}}}, "iid_acc": 0.595, "selectivity": 0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.2061930031896504, "mean_class_cosine": 0.9278347514353975}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.85375, "drop": 0.0}, "paraphrase": {"acc": 0.844106463878327, "drop": 0.009643536121673013}, "domain": {"acc": 0.66875, "drop": 0.18500000000000005}, "length": {"acc": 0.8493150684931506, "drop": 0.004434931506849371}}, "iid_acc": 0.85375, "selectivity": 0.35375} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8642745709828393, "drop": 0.005725429017160666, "rotation": {"subspace_principal_angle": 1.3957840280918499, "mean_class_cosine": 0.27977067989217674}}}, "iid_acc": 0.87, "selectivity": 0.5974999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3533239187236508, "mean_class_cosine": 0.32998309013733307}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.5425, "drop": 0.0}, "paraphrase": {"acc": 0.5460218408736349, "drop": -0.0035218408736349405, "rotation": {"subspace_principal_angle": 1.5216614620859656, "mean_class_cosine": 0.8734402521973395}}}, "iid_acc": 0.5425, "selectivity": 0.26999999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.0674047358789585, "mean_class_cosine": 0.9542985146974496}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 4, "dists": {"iid": {"acc": 0.8825, "drop": 0.0}, "paraphrase": {"acc": 0.890795631825273, "drop": -0.008295631825273042}}, "iid_acc": 0.8825, "selectivity": 0.6099999999999999} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.96125, "drop": 0.0}, "paraphrase": {"acc": 0.9376498800959233, "drop": 0.023600119904076733, "rotation": {"subspace_principal_angle": 1.1479338257378555, "mean_class_cosine": 0.6070660513833379}}}, "iid_acc": 0.96125, "selectivity": 0.89375, "iid_split_rotation": {"subspace_principal_angle": 1.1171811389695125, "mean_class_cosine": 0.6559427626785893}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6139088729016786, "drop": 0.008591127098321438, "rotation": {"subspace_principal_angle": 1.2623715388420953, "mean_class_cosine": 0.8819688751478222}}}, "iid_acc": 0.6225, "selectivity": 0.555, "iid_split_rotation": {"subspace_principal_angle": 1.2562343632402415, "mean_class_cosine": 0.8445383158129988}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.95375, "drop": 0.0}, "paraphrase": {"acc": 0.9328537170263789, "drop": 0.020896282973621116}}, "iid_acc": 0.95375, "selectivity": 0.88625} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": 0.011323529411764732, "rotation": {"subspace_principal_angle": 1.4587999072142643, "mean_class_cosine": 0.11176243416412668}}}, "iid_acc": 0.5525, "selectivity": 0.02749999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4400549078317155, "mean_class_cosine": 0.13036926971343207}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.51625, "drop": 0.0}, "paraphrase": {"acc": 0.538562091503268, "drop": -0.02231209150326796, "rotation": {"subspace_principal_angle": 0.9958229238018077, "mean_class_cosine": 0.9168202372116476}}}, "iid_acc": 0.51625, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.4327159822922633, "mean_class_cosine": -0.5177790256388579}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.49673202614379086, "drop": 0.054517973856209156}}, "iid_acc": 0.55125, "selectivity": 0.026249999999999996} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5619946091644205, "drop": 0.039255390835579496, "rotation": {"subspace_principal_angle": 1.5685155167026745, "mean_class_cosine": 0.2634805112669454}}}, "iid_acc": 0.60125, "selectivity": 0.3212499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.498476982651726, "mean_class_cosine": 0.251518325093681}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.26125, "drop": 0.0}, "paraphrase": {"acc": 0.261455525606469, "drop": -0.00020552560646902585, "rotation": {"subspace_principal_angle": 1.3916019392156598, "mean_class_cosine": 0.6531664347702235}}}, "iid_acc": 0.26125, "selectivity": -0.018750000000000044, "iid_split_rotation": {"subspace_principal_angle": 1.3668566895557186, "mean_class_cosine": 0.6689391992705035}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.58625, "drop": 0.0}, "paraphrase": {"acc": 0.5700808625336927, "drop": 0.016169137466307326}}, "iid_acc": 0.58625, "selectivity": 0.30625} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.7195325542570952, "drop": -0.009532554257095205, "rotation": {"subspace_principal_angle": 1.4001021400462081, "mean_class_cosine": 0.16986648813286045}}}, "iid_acc": 0.71, "selectivity": 0.21249999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.397132197602315, "mean_class_cosine": 0.17279251519584937}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.57625, "drop": 0.0}, "paraphrase": {"acc": 0.5859766277128547, "drop": -0.009726627712854707, "rotation": {"subspace_principal_angle": 1.328491446497459, "mean_class_cosine": 0.9377244571178811}}}, "iid_acc": 0.57625, "selectivity": 0.07875000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4036636361795571, "mean_class_cosine": 0.938290713035285}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7479131886477463, "drop": -0.00416318864774623}}, "iid_acc": 0.74375, "selectivity": 0.24625000000000002} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.6375, "drop": 0.0}, "paraphrase": {"acc": 0.6105610561056105, "drop": 0.026938943894389444, "rotation": {"subspace_principal_angle": 1.490662860294879, "mean_class_cosine": 0.08004773289195696}}}, "iid_acc": 0.6375, "selectivity": 0.14749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.516170603318211, "mean_class_cosine": 0.054598560612693414}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5412541254125413, "drop": 0.0049958745874587596, "rotation": {"subspace_principal_angle": 1.018522615405077, "mean_class_cosine": 0.998295366669647}}}, "iid_acc": 0.54625, "selectivity": 0.05625000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.799728898586944, "mean_class_cosine": 0.9985173926775679}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 5, "dists": {"iid": {"acc": 0.68875, "drop": 0.0}, "paraphrase": {"acc": 0.6831683168316832, "drop": 0.005581683168316753}}, "iid_acc": 0.68875, "selectivity": 0.19874999999999998} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.68125, "drop": 0.0}, "paraphrase": {"acc": 0.677115987460815, "drop": 0.004134012539184995, "rotation": {"subspace_principal_angle": 1.513447961089972, "mean_class_cosine": 0.05731693598787185}}}, "iid_acc": 0.68125, "selectivity": 0.10875000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.5579856772962786, "mean_class_cosine": 0.012810299103026977}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.50125, "drop": 0.0}, "paraphrase": {"acc": 0.5250783699059561, "drop": -0.02382836990595616, "rotation": {"subspace_principal_angle": 1.5103655352700989, "mean_class_cosine": 0.7751093816111925}}}, "iid_acc": 0.50125, "selectivity": -0.07125000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9404374749778482, "mean_class_cosine": 0.48400831659040194}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7570532915360502, "drop": -0.027053291536050184}}, "iid_acc": 0.73, "selectivity": 0.15749999999999997} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.91625, "drop": 0.0}, "paraphrase": {"acc": 0.9135285913528591, "drop": 0.002721408647140877, "rotation": {"subspace_principal_angle": 1.2733184253579535, "mean_class_cosine": 0.29310981653055773}}}, "iid_acc": 0.91625, "selectivity": 0.37250000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1920176049103017, "mean_class_cosine": 0.36978603528128035}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.83, "drop": 0.0}, "paraphrase": {"acc": 0.8145048814504882, "drop": 0.0154951185495118, "rotation": {"subspace_principal_angle": 1.4437573872568787, "mean_class_cosine": 0.9676721148007581}}}, "iid_acc": 0.83, "selectivity": 0.28625, "iid_split_rotation": {"subspace_principal_angle": 1.4314363898953524, "mean_class_cosine": 0.9613334451017563}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.91625, "drop": 0.0}, "paraphrase": {"acc": 0.9135285913528591, "drop": 0.002721408647140877}}, "iid_acc": 0.91625, "selectivity": 0.37250000000000005} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.0015466448445171643, "rotation": {"subspace_principal_angle": 1.0288570623244413, "mean_class_cosine": 0.5157983478150161}}}, "iid_acc": 0.995, "selectivity": 0.22999999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.069772811988795, "mean_class_cosine": 0.4803235060741538}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.734860883797054, "drop": -0.03611088379705407, "rotation": {"subspace_principal_angle": 1.4159558052810821, "mean_class_cosine": 0.998661624282861}}}, "iid_acc": 0.69875, "selectivity": -0.06625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.1246427235862093, "mean_class_cosine": 0.9995471474354687}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 10, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9918166939443536, "drop": -0.0018166939443535757}}, "iid_acc": 0.99, "selectivity": 0.22499999999999998} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6845070422535211, "drop": 0.009242957746478875, "rotation": {"subspace_principal_angle": 1.5353921160260495, "mean_class_cosine": 0.035396814949682484}}}, "iid_acc": 0.69375, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4561316678570049, "mean_class_cosine": 0.11441355588600202}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5295774647887324, "drop": -0.012077464788732395, "rotation": {"subspace_principal_angle": 1.1952282463550592, "mean_class_cosine": 0.34100773822159186}}}, "iid_acc": 0.5175, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.5260408088060127, "mean_class_cosine": 0.9004767260950717}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7535211267605634, "drop": -0.017271126760563416}}, "iid_acc": 0.73625, "selectivity": 0.09124999999999994} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.6127167630057804, "drop": -0.050216763005780374, "rotation": {"subspace_principal_angle": 1.4402506816259675, "mean_class_cosine": 0.1541271484023787}}}, "iid_acc": 0.5625, "selectivity": 0.17788461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.373294097014231, "mean_class_cosine": 0.19890553489034377}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.49710982658959535, "drop": -0.035571365051133785, "rotation": {"subspace_principal_angle": 1.1768674031248585, "mean_class_cosine": 0.7104337840281887}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.07692307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.1855767421726293, "mean_class_cosine": 0.8092177583847361}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5336538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.5549132947976878, "drop": -0.02125944864384166}}, "iid_acc": 0.5336538461538461, "selectivity": 0.1490384615384615} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.9001386962552012, "drop": -0.01013869625520114, "rotation": {"subspace_principal_angle": 1.3883765772077856, "mean_class_cosine": 0.18140970234637815}}}, "iid_acc": 0.89, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.3442390049803465, "mean_class_cosine": 0.22462415910132738}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5436893203883495, "drop": 0.018810679611650505, "rotation": {"subspace_principal_angle": 0.9241498843813925, "mean_class_cosine": -0.17909041195198278}}}, "iid_acc": 0.5625, "selectivity": -0.135, "iid_split_rotation": {"subspace_principal_angle": 0.7140290968434547, "mean_class_cosine": 0.9686729997957932}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.87875, "drop": 0.0}, "paraphrase": {"acc": 0.8890429958391124, "drop": -0.010292995839112362}}, "iid_acc": 0.87875, "selectivity": 0.18125000000000002} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8325, "drop": 0.0}, "paraphrase": {"acc": 0.7746478873239436, "drop": 0.05785211267605639, "rotation": {"subspace_principal_angle": 1.2990940486645208, "mean_class_cosine": 0.26837165555553494}}, "domain": {"acc": 0.7975, "drop": 0.03500000000000003, "rotation": {"subspace_principal_angle": 1.3921776610825947, "mean_class_cosine": 0.17767038602968485}}}, "iid_acc": 0.8325, "selectivity": 0.32625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2834752018104276, "mean_class_cosine": 0.28338418654918335}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.52875, "drop": 0.0}, "paraphrase": {"acc": 0.5309859154929577, "drop": -0.002235915492957652, "rotation": {"subspace_principal_angle": 1.0586485465355164, "mean_class_cosine": 0.9956582503304628}}, "domain": {"acc": 0.4875, "drop": 0.041250000000000064, "rotation": {"subspace_principal_angle": 1.361095756118237, "mean_class_cosine": 0.3175504588818687}}}, "iid_acc": 0.52875, "selectivity": 0.022500000000000075, "iid_split_rotation": {"subspace_principal_angle": 0.8302811510741681, "mean_class_cosine": 0.9988764885217147}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.819718309859155, "drop": 0.04653169014084502}, "domain": {"acc": 0.8425, "drop": 0.023749999999999938}}, "iid_acc": 0.86625, "selectivity": 0.36} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8829953198127926, "drop": -0.0029953198127925518, "rotation": {"subspace_principal_angle": 1.3716186408201203, "mean_class_cosine": 0.26184135314471124}}}, "iid_acc": 0.88, "selectivity": 0.59875, "iid_split_rotation": {"subspace_principal_angle": 1.3986787504321423, "mean_class_cosine": 0.27403222616028045}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.80625, "drop": 0.0}, "paraphrase": {"acc": 0.7862714508580343, "drop": 0.01997854914196573, "rotation": {"subspace_principal_angle": 1.1213682836223715, "mean_class_cosine": 0.9645304775406084}}}, "iid_acc": 0.80625, "selectivity": 0.525, "iid_split_rotation": {"subspace_principal_angle": 1.190218420873915, "mean_class_cosine": 0.9648000749883475}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 21, "dists": {"iid": {"acc": 0.8925, "drop": 0.0}, "paraphrase": {"acc": 0.8923556942277691, "drop": 0.00014430577223090157}}, "iid_acc": 0.8925, "selectivity": 0.61125} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.54125, "drop": 0.0}, "paraphrase": {"acc": 0.5437908496732026, "drop": -0.002540849673202561, "rotation": {"subspace_principal_angle": 1.4224273033339936, "mean_class_cosine": 0.14782527199893394}}}, "iid_acc": 0.54125, "selectivity": 0.048750000000000016, "iid_split_rotation": {"subspace_principal_angle": 1.4416538946189268, "mean_class_cosine": 0.1287837634804017}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.538562091503268, "drop": -0.021062091503267988, "rotation": {"subspace_principal_angle": 1.2749369531499732, "mean_class_cosine": 0.8687066652493326}}}, "iid_acc": 0.5175, "selectivity": 0.024999999999999967, "iid_split_rotation": {"subspace_principal_angle": 1.539011739741241, "mean_class_cosine": -0.5381342265082592}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.54625, "drop": 0.0}, "paraphrase": {"acc": 0.5294117647058824, "drop": 0.016838235294117654}}, "iid_acc": 0.54625, "selectivity": 0.05375000000000002} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.5916442048517521, "drop": 0.023355795148247926, "rotation": {"subspace_principal_angle": 1.3790979385283946, "mean_class_cosine": 0.3195081442711161}}}, "iid_acc": 0.615, "selectivity": 0.36375, "iid_split_rotation": {"subspace_principal_angle": 1.4199621331697765, "mean_class_cosine": 0.29428965873721996}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.2875, "drop": 0.0}, "paraphrase": {"acc": 0.2803234501347709, "drop": 0.007176549865229087, "rotation": {"subspace_principal_angle": 1.1732397784339854, "mean_class_cosine": 0.6776335920045801}}}, "iid_acc": 0.2875, "selectivity": 0.036250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3510532914197495, "mean_class_cosine": 0.6935310313683204}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.63, "drop": 0.0}, "paraphrase": {"acc": 0.6226415094339622, "drop": 0.0073584905660377675}}, "iid_acc": 0.63, "selectivity": 0.37875000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.70375, "drop": 0.0}, "paraphrase": {"acc": 0.6828046744574291, "drop": 0.020945325542570892, "rotation": {"subspace_principal_angle": 1.4506727333643843, "mean_class_cosine": 0.1198349109994324}}}, "iid_acc": 0.70375, "selectivity": 0.20375, "iid_split_rotation": {"subspace_principal_angle": 1.4560638773991705, "mean_class_cosine": 0.1144809009150342}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.5225, "drop": 0.0}, "paraphrase": {"acc": 0.5442404006677797, "drop": -0.021740400667779713, "rotation": {"subspace_principal_angle": 1.3673823057780032, "mean_class_cosine": 0.9981711848367865}}}, "iid_acc": 0.5225, "selectivity": 0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 1.1930110372511171, "mean_class_cosine": 0.9981828476111254}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7362270450751253, "drop": 0.013772954924874736}}, "iid_acc": 0.75, "selectivity": 0.25} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.7084639498432602, "drop": -0.02096394984326022, "rotation": {"subspace_principal_angle": 1.5051962181158443, "mean_class_cosine": 0.06555306849859195}}}, "iid_acc": 0.6875, "selectivity": 0.14125, "iid_split_rotation": {"subspace_principal_angle": 1.5172834822024313, "mean_class_cosine": 0.053487308133283416}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5658307210031348, "drop": -0.010830721003134758, "rotation": {"subspace_principal_angle": 1.1030041242538011, "mean_class_cosine": 0.6806746541237996}}}, "iid_acc": 0.555, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.0411245457901148, "mean_class_cosine": 0.7191236806689514}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 22, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.7492163009404389, "drop": -0.0029663009404389262}}, "iid_acc": 0.74625, "selectivity": 0.19999999999999996} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.94125, "drop": 0.0}, "paraphrase": {"acc": 0.9288702928870293, "drop": 0.012379707112970761, "rotation": {"subspace_principal_angle": 1.2301430406581242, "mean_class_cosine": 0.3341029094871022}}}, "iid_acc": 0.94125, "selectivity": 0.41500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.153997330636452, "mean_class_cosine": 0.4048355678035365}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.75875, "drop": 0.0}, "paraphrase": {"acc": 0.7642956764295676, "drop": -0.005545676429567603, "rotation": {"subspace_principal_angle": 1.5114298942376507, "mean_class_cosine": 0.967424461326506}}}, "iid_acc": 0.75875, "selectivity": 0.23250000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9983998428171448, "mean_class_cosine": 0.9658650662106624}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 16, "dists": {"iid": {"acc": 0.94875, "drop": 0.0}, "paraphrase": {"acc": 0.9428172942817294, "drop": 0.005932705718270537}}, "iid_acc": 0.94875, "selectivity": 0.4225} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.70375, "drop": 0.0}, "paraphrase": {"acc": 0.6732394366197183, "drop": 0.030510563380281686, "rotation": {"subspace_principal_angle": 1.5542729001951894, "mean_class_cosine": 0.01652267472899558}}}, "iid_acc": 0.70375, "selectivity": 0.13124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4874758177302438, "mean_class_cosine": 0.08322413642773308}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.5725, "drop": 0.0}, "paraphrase": {"acc": 0.5464788732394367, "drop": 0.02602112676056334, "rotation": {"subspace_principal_angle": 1.5278946651673369, "mean_class_cosine": -0.5372610043917487}}}, "iid_acc": 0.5725, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.4388173212776232, "mean_class_cosine": 0.7758844286751576}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 9, "dists": {"iid": {"acc": 0.7525, "drop": 0.0}, "paraphrase": {"acc": 0.7183098591549296, "drop": 0.03419014084507033}}, "iid_acc": 0.7525, "selectivity": 0.17999999999999994} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5606936416184971, "drop": 0.0018063583815028927, "rotation": {"subspace_principal_angle": 1.4713914583801808, "mean_class_cosine": 0.1086073697472787}}}, "iid_acc": 0.5625, "selectivity": 0.20192307692307693, "iid_split_rotation": {"subspace_principal_angle": 1.4025179262636047, "mean_class_cosine": 0.18370954160961964}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.46153846153846156, "drop": 0.0}, "paraphrase": {"acc": 0.47398843930635837, "drop": -0.012449977767896803, "rotation": {"subspace_principal_angle": 1.1101897330656618, "mean_class_cosine": 0.7275819529837535}}}, "iid_acc": 0.46153846153846156, "selectivity": 0.10096153846153849, "iid_split_rotation": {"subspace_principal_angle": 1.5497284875073214, "mean_class_cosine": 0.7930942268672401}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.6009615384615384, "drop": 0.0}, "paraphrase": {"acc": 0.5780346820809249, "drop": 0.022926856380613536}}, "iid_acc": 0.6009615384615384, "selectivity": 0.24038461538461536} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8987517337031901, "drop": -1.733703190009983e-06, "rotation": {"subspace_principal_angle": 1.403522863323226, "mean_class_cosine": 0.16649449067720934}}}, "iid_acc": 0.89875, "selectivity": 0.21750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.420138412202417, "mean_class_cosine": 0.15008862700519784}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.57125, "drop": 0.0}, "paraphrase": {"acc": 0.5617198335644937, "drop": 0.009530166435506304, "rotation": {"subspace_principal_angle": 0.94423285674327, "mean_class_cosine": 0.24684962591703655}}}, "iid_acc": 0.57125, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.5221499477378648, "mean_class_cosine": 0.9517198734637033}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 12, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9098474341192788, "drop": 0.012652565880721167}}, "iid_acc": 0.9225, "selectivity": 0.24124999999999996} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.85625, "drop": 0.0}, "paraphrase": {"acc": 0.8323943661971831, "drop": 0.023855633802816856, "rotation": {"subspace_principal_angle": 1.2594449210343017, "mean_class_cosine": 0.3063453465601831}}, "domain": {"acc": 0.8675, "drop": -0.011250000000000093, "rotation": {"subspace_principal_angle": 1.2223752440939522, "mean_class_cosine": 0.3414141888265658}}}, "iid_acc": 0.85625, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 1.1803322078545402, "mean_class_cosine": 0.38061764197303105}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5366197183098591, "drop": -0.004119718309859133, "rotation": {"subspace_principal_angle": 1.0843410916453302, "mean_class_cosine": 0.9995443885101234}}, "domain": {"acc": 0.4875, "drop": 0.044999999999999984, "rotation": {"subspace_principal_angle": 1.5626024602699073, "mean_class_cosine": 0.4830036010251207}}}, "iid_acc": 0.5325, "selectivity": 0.030000000000000027, "iid_split_rotation": {"subspace_principal_angle": 1.1107445344198659, "mean_class_cosine": 0.9998835461376956}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8028169014084507, "drop": 0.03468309859154928}, "domain": {"acc": 0.8225, "drop": 0.015000000000000013}}, "iid_acc": 0.8375, "selectivity": 0.3350000000000001} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.9125, "drop": 0.0}, "paraphrase": {"acc": 0.8954372623574145, "drop": 0.01706273764258548, "rotation": {"subspace_principal_angle": 1.0306140943276123, "mean_class_cosine": 0.5142922854845775}}, "domain": {"acc": 0.77875, "drop": 0.13374999999999992, "rotation": {"subspace_principal_angle": 1.2652458098581651, "mean_class_cosine": 0.30081823858041606}}, "length": {"acc": 0.9193302891933028, "drop": -0.006830289193302863, "rotation": {"subspace_principal_angle": 0.994782867288978, "mean_class_cosine": 0.5446849986678566}}}, "iid_acc": 0.9125, "selectivity": 0.4125, "iid_split_rotation": {"subspace_principal_angle": 1.0394992159651213, "mean_class_cosine": 0.5066520720072085}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.58125, "drop": 0.0}, "paraphrase": {"acc": 0.6254752851711026, "drop": -0.04422528517110258, "rotation": {"subspace_principal_angle": 0.9837366616608404, "mean_class_cosine": 0.9776968730440565}}, "domain": {"acc": 0.5575, "drop": 0.02375000000000005, "rotation": {"subspace_principal_angle": 1.3537717360413388, "mean_class_cosine": 0.5378898485059446}}, "length": {"acc": 0.6012176560121766, "drop": -0.019967656012176538, "rotation": {"subspace_principal_angle": 1.0664869055781216, "mean_class_cosine": 0.8881910220135161}}}, "iid_acc": 0.58125, "selectivity": 0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5132821504454912, "mean_class_cosine": 0.9660599604667153}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 3, "layer": 14, "dists": {"iid": {"acc": 0.91, "drop": 0.0}, "paraphrase": {"acc": 0.9049429657794676, "drop": 0.005057034220532386}, "domain": {"acc": 0.755, "drop": 0.15500000000000003}, "length": {"acc": 0.9178082191780822, "drop": -0.007808219178082165}}, "iid_acc": 0.91, "selectivity": 0.41000000000000003} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.88375, "drop": 0.0}, "paraphrase": {"acc": 0.8783151326053042, "drop": 0.005434867394695786, "rotation": {"subspace_principal_angle": 1.4018526090844314, "mean_class_cosine": 0.21584227023789393}}}, "iid_acc": 0.88375, "selectivity": 0.6412500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3568370650579906, "mean_class_cosine": 0.277933277471287}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.88125, "drop": 0.0}, "paraphrase": {"acc": 0.8798751950078003, "drop": 0.001374804992199663, "rotation": {"subspace_principal_angle": 1.5437428796709174, "mean_class_cosine": 0.9838278021789155}}}, "iid_acc": 0.88125, "selectivity": 0.6387499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.5098184352195878, "mean_class_cosine": 0.9768614686576095}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 3, "layer": 23, "dists": {"iid": {"acc": 0.9, "drop": 0.0}, "paraphrase": {"acc": 0.8923556942277691, "drop": 0.007644305772230964}}, "iid_acc": 0.9, "selectivity": 0.6575} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.965, "drop": 0.0}, "paraphrase": {"acc": 0.9496402877697842, "drop": 0.015359712230215816, "rotation": {"subspace_principal_angle": 0.9402737529008706, "mean_class_cosine": 0.7722383597549551}}}, "iid_acc": 0.965, "selectivity": 0.9075, "iid_split_rotation": {"subspace_principal_angle": 0.8971810536010522, "mean_class_cosine": 0.8007298180881067}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.13875, "drop": 0.0}, "paraphrase": {"acc": 0.14628297362110312, "drop": -0.007532973621103112, "rotation": {"subspace_principal_angle": 1.1315401402990097, "mean_class_cosine": 0.6941035532694599}}}, "iid_acc": 0.13875, "selectivity": 0.08125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.385883009922417, "mean_class_cosine": 0.6014609571780535}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.93875, "drop": 0.0}, "paraphrase": {"acc": 0.9136690647482014, "drop": 0.025080935251798553}}, "iid_acc": 0.93875, "selectivity": 0.88125} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.6300653594771242, "drop": 0.008684640522875875, "rotation": {"subspace_principal_angle": 1.2787080353422497, "mean_class_cosine": 0.28795269263212087}}}, "iid_acc": 0.63875, "selectivity": 0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.306568966622189, "mean_class_cosine": 0.2611635211442275}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5411764705882353, "drop": -0.028676470588235303, "rotation": {"subspace_principal_angle": 1.4646247494771925, "mean_class_cosine": 0.7483798246130362}}}, "iid_acc": 0.5125, "selectivity": -0.02750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.5194053435641155, "mean_class_cosine": -0.9436278317396236}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5333333333333333, "drop": 0.0004166666666666208}}, "iid_acc": 0.53375, "selectivity": -0.006250000000000089} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.62, "drop": 0.0}, "paraphrase": {"acc": 0.5876010781671159, "drop": 0.032398921832884064, "rotation": {"subspace_principal_angle": 1.3946988954670643, "mean_class_cosine": 0.4588645470889185}}}, "iid_acc": 0.62, "selectivity": 0.31625, "iid_split_rotation": {"subspace_principal_angle": 1.4814328069291833, "mean_class_cosine": 0.3908719413079447}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.1825, "drop": 0.0}, "paraphrase": {"acc": 0.1765498652291105, "drop": 0.005950134770889487, "rotation": {"subspace_principal_angle": 1.428737416987744, "mean_class_cosine": 0.6603620114391797}}}, "iid_acc": 0.1825, "selectivity": -0.12125000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.3464001687915095, "mean_class_cosine": 0.917084029092293}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 3, "layer": 8, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.5175202156334232, "drop": 0.012479784366576818}}, "iid_acc": 0.53, "selectivity": 0.22625} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7612687813021702, "drop": -0.0037687813021702876, "rotation": {"subspace_principal_angle": 1.0841355564151438, "mean_class_cosine": 0.46767695778639146}}}, "iid_acc": 0.7575, "selectivity": 0.24624999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.1201130263537538, "mean_class_cosine": 0.4355807084230253}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6828046744574291, "drop": -0.012804674457429055, "rotation": {"subspace_principal_angle": 1.559246280818732, "mean_class_cosine": 0.8724262003165955}}}, "iid_acc": 0.67, "selectivity": 0.15875000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.208049879221657, "mean_class_cosine": 0.8740158426804394}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 3, "layer": 1, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7395659432387313, "drop": 0.01418405676126877}}, "iid_acc": 0.75375, "selectivity": 0.24250000000000005} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.68375, "drop": 0.0}, "paraphrase": {"acc": 0.6287128712871287, "drop": 0.05503712871287125, "rotation": {"subspace_principal_angle": 1.4304093704815346, "mean_class_cosine": 0.1399262745187435}}}, "iid_acc": 0.68375, "selectivity": 0.17999999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4264870806603527, "mean_class_cosine": 0.14380889026388544}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.5525, "drop": 0.0}, "paraphrase": {"acc": 0.5610561056105611, "drop": -0.008556105610561082, "rotation": {"subspace_principal_angle": 1.4505633244788272, "mean_class_cosine": 0.9993707117635033}}}, "iid_acc": 0.5525, "selectivity": 0.04874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.562613516437802, "mean_class_cosine": 0.9992208351646927}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 3, "layer": 3, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.6171617161716172, "drop": 0.04283828382838284}}, "iid_acc": 0.66, "selectivity": 0.15625} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7429467084639498, "drop": -0.011696708463949879, "rotation": {"subspace_principal_angle": 1.4343600588327907, "mean_class_cosine": 0.1360133715328124}}}, "iid_acc": 0.73125, "selectivity": 0.12874999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.3984660781436913, "mean_class_cosine": 0.17147854522739675}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.50875, "drop": 0.0}, "paraphrase": {"acc": 0.5266457680250783, "drop": -0.017895768025078307, "rotation": {"subspace_principal_angle": 1.5522402276263103, "mean_class_cosine": 0.9681835964660888}}}, "iid_acc": 0.50875, "selectivity": -0.09375, "iid_split_rotation": {"subspace_principal_angle": 1.278589503289651, "mean_class_cosine": 0.36941996233066643}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 3, "layer": 6, "dists": {"iid": {"acc": 0.73375, "drop": 0.0}, "paraphrase": {"acc": 0.7429467084639498, "drop": -0.009196708463949821}}, "iid_acc": 0.73375, "selectivity": 0.13124999999999998} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.9375, "drop": 0.0}, "paraphrase": {"acc": 0.9163179916317992, "drop": 0.021182008368200833, "rotation": {"subspace_principal_angle": 1.190401875207584, "mean_class_cosine": 0.371286753786929}}}, "iid_acc": 0.9375, "selectivity": 0.39625, "iid_split_rotation": {"subspace_principal_angle": 1.0454137579769693, "mean_class_cosine": 0.5015440139447017}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.55125, "drop": 0.0}, "paraphrase": {"acc": 0.5425383542538355, "drop": 0.008711645746164565, "rotation": {"subspace_principal_angle": 0.70122334453455, "mean_class_cosine": 0.9845649431106505}}}, "iid_acc": 0.55125, "selectivity": 0.010000000000000009, "iid_split_rotation": {"subspace_principal_angle": 0.5014314350811042, "mean_class_cosine": 0.9308041543855714}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.945, "drop": 0.0}, "paraphrase": {"acc": 0.9456066945606695, "drop": -0.0006066945606695295}}, "iid_acc": 0.945, "selectivity": 0.40374999999999994} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.995, "drop": 0.0}, "paraphrase": {"acc": 0.9950900163666121, "drop": -9.001636661210011e-05, "rotation": {"subspace_principal_angle": 0.9545129430981294, "mean_class_cosine": 0.578006280576806}}}, "iid_acc": 0.995, "selectivity": 0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 0.9634855001163742, "mean_class_cosine": 0.5706612157863881}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.69875, "drop": 0.0}, "paraphrase": {"acc": 0.723404255319149, "drop": -0.024654255319149, "rotation": {"subspace_principal_angle": 0.9334539669903851, "mean_class_cosine": 0.9999542616831976}}}, "iid_acc": 0.69875, "selectivity": -0.12875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4894565210879196, "mean_class_cosine": 0.9999767868513006}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 3, "layer": 11, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9934533551554828, "drop": 0.000296644844517191}}, "iid_acc": 0.99375, "selectivity": 0.16625} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.74375, "drop": 0.0}, "paraphrase": {"acc": 0.7183098591549296, "drop": 0.025440140845070403, "rotation": {"subspace_principal_angle": 1.4257021861565011, "mean_class_cosine": 0.14458558178778447}}}, "iid_acc": 0.74375, "selectivity": 0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4515121208658293, "mean_class_cosine": 0.11900153014654812}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.5625, "drop": 0.0}, "paraphrase": {"acc": 0.5183098591549296, "drop": 0.04419014084507045, "rotation": {"subspace_principal_angle": 1.5128863516688837, "mean_class_cosine": -0.06917035187866413}}}, "iid_acc": 0.5625, "selectivity": -0.03374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3373195084183762, "mean_class_cosine": 0.802090774256028}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 3, "layer": 15, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.752112676056338, "drop": -0.01711267605633804}}, "iid_acc": 0.735, "selectivity": 0.13875000000000004} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.625, "drop": 0.0}, "paraphrase": {"acc": 0.630057803468208, "drop": -0.005057803468208055, "rotation": {"subspace_principal_angle": 1.3578016617138275, "mean_class_cosine": 0.22714259169758785}}}, "iid_acc": 0.625, "selectivity": 0.24038461538461536, "iid_split_rotation": {"subspace_principal_angle": 1.34596465147534, "mean_class_cosine": 0.24687761314743237}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.4519230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.45664739884393063, "drop": -0.004724321920853702, "rotation": {"subspace_principal_angle": 1.4589762703207936, "mean_class_cosine": 0.9288307050828664}}}, "iid_acc": 0.4519230769230769, "selectivity": 0.06730769230769229, "iid_split_rotation": {"subspace_principal_angle": 1.2273866113889293, "mean_class_cosine": 0.9582100809299986}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 3, "layer": 7, "dists": {"iid": {"acc": 0.5865384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.5838150289017341, "drop": 0.002723432636727474}}, "iid_acc": 0.5865384615384616, "selectivity": 0.20192307692307693} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.895, "drop": 0.0}, "paraphrase": {"acc": 0.8904299583911235, "drop": 0.0045700416088765294, "rotation": {"subspace_principal_angle": 1.3187059663512182, "mean_class_cosine": 0.24942879345042712}}}, "iid_acc": 0.895, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2441873711885174, "mean_class_cosine": 0.3208330998449743}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.55, "drop": 0.0}, "paraphrase": {"acc": 0.5325936199722607, "drop": 0.01740638002773931, "rotation": {"subspace_principal_angle": 1.2286707925235416, "mean_class_cosine": -0.644580232070691}}}, "iid_acc": 0.55, "selectivity": -0.20875, "iid_split_rotation": {"subspace_principal_angle": 0.7593618744036684, "mean_class_cosine": 0.9814176976119293}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 3, "layer": 13, "dists": {"iid": {"acc": 0.89875, "drop": 0.0}, "paraphrase": {"acc": 0.8945908460471568, "drop": 0.004159153952843275}}, "iid_acc": 0.89875, "selectivity": 0.14} +{"model": "pythia-70m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7098150782361309, "drop": 0.04768492176386907, "rotation": {"subspace_principal_angle": 1.2799135817505813, "mean_class_cosine": 0.28679799861480304}}, "domain": {"acc": 0.73375, "drop": 0.023749999999999938, "rotation": {"subspace_principal_angle": 1.206450803740472, "mean_class_cosine": 0.3563378556067195}}}, "iid_acc": 0.7575, "selectivity": 0.2587499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2152898580705969, "mean_class_cosine": 0.34806520925621826}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.55625, "drop": 0.0}, "paraphrase": {"acc": 0.5889046941678521, "drop": -0.03265469416785205, "rotation": {"subspace_principal_angle": 1.51956074817224, "mean_class_cosine": 0.8695877050620235}}, "domain": {"acc": 0.54125, "drop": 0.015000000000000013, "rotation": {"subspace_principal_angle": 1.5651562278626985, "mean_class_cosine": 0.5820640074758123}}}, "iid_acc": 0.55625, "selectivity": 0.057499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.548228069141704, "mean_class_cosine": 0.8592588697201281}} +{"model": "pythia-70m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.7475, "drop": 0.0}, "paraphrase": {"acc": 0.7226173541963016, "drop": 0.024882645803698478}, "domain": {"acc": 0.7575, "drop": -0.009999999999999898}}, "iid_acc": 0.7475, "selectivity": 0.24875000000000003} +{"model": "pythia-70m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.807843137254902, "drop": -0.005343137254901986, "rotation": {"subspace_principal_angle": 0.9349235883028191, "mean_class_cosine": 0.5938799054215346}}, "domain": {"acc": 0.66, "drop": 0.14249999999999996, "rotation": {"subspace_principal_angle": 1.2644579436182715, "mean_class_cosine": 0.30156951858956016}}, "length": {"acc": 0.802962962962963, "drop": -0.0004629629629629983, "rotation": {"subspace_principal_angle": 0.8439392303466156, "mean_class_cosine": 0.6645243339696073}}}, "iid_acc": 0.8025, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 0.8932244951414783, "mean_class_cosine": 0.626903094729581}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.69625, "drop": 0.0}, "paraphrase": {"acc": 0.7137254901960784, "drop": -0.017475490196078414, "rotation": {"subspace_principal_angle": 1.1712305596628225, "mean_class_cosine": 0.9467024508176543}}, "domain": {"acc": 0.57875, "drop": 0.11750000000000005, "rotation": {"subspace_principal_angle": 1.3012077657528451, "mean_class_cosine": 0.356404654810238}}, "length": {"acc": 0.7051851851851851, "drop": -0.008935185185185102, "rotation": {"subspace_principal_angle": 1.1652413786828095, "mean_class_cosine": 0.9420615872580638}}}, "iid_acc": 0.69625, "selectivity": 0.22250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4630243777096839, "mean_class_cosine": 0.8983235883657518}} +{"model": "pythia-70m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.78375, "drop": 0.0}, "paraphrase": {"acc": 0.7862745098039216, "drop": -0.002524509803921604}, "domain": {"acc": 0.61125, "drop": 0.1725}, "length": {"acc": 0.7881481481481482, "drop": -0.0043981481481482065}}, "iid_acc": 0.78375, "selectivity": 0.30999999999999994} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.8475, "drop": 0.0}, "paraphrase": {"acc": 0.8348484848484848, "drop": 0.012651515151515191, "rotation": {"subspace_principal_angle": 1.4031997490020873, "mean_class_cosine": 0.28396714753242047}}}, "iid_acc": 0.8475, "selectivity": 0.5975, "iid_split_rotation": {"subspace_principal_angle": 1.3806377410186017, "mean_class_cosine": 0.32568678636943665}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8181818181818182, "drop": 0.008068181818181808, "rotation": {"subspace_principal_angle": 1.4758819738130697, "mean_class_cosine": 0.9741084999680931}}}, "iid_acc": 0.82625, "selectivity": 0.57625, "iid_split_rotation": {"subspace_principal_angle": 1.1310624243298266, "mean_class_cosine": 0.9619284246782831}} +{"model": "pythia-70m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.865, "drop": 0.0}, "paraphrase": {"acc": 0.8545454545454545, "drop": 0.010454545454545494}}, "iid_acc": 0.865, "selectivity": 0.615} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9343065693430657, "drop": 0.011943430656934373, "rotation": {"subspace_principal_angle": 1.1865093507054412, "mean_class_cosine": 0.7073522515216414}}}, "iid_acc": 0.94625, "selectivity": 0.865, "iid_split_rotation": {"subspace_principal_angle": 1.127510316714073, "mean_class_cosine": 0.7428054255764255}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5571776155717761, "drop": 0.00907238442822389, "rotation": {"subspace_principal_angle": 1.3920610393607629, "mean_class_cosine": 0.7826885130462339}}}, "iid_acc": 0.56625, "selectivity": 0.48500000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.3330395256421586, "mean_class_cosine": 0.8909496426331701}} +{"model": "pythia-70m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.87125, "drop": 0.0}, "paraphrase": {"acc": 0.8491484184914841, "drop": 0.022101581508515822}}, "iid_acc": 0.87125, "selectivity": 0.7899999999999999} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.4967490247074122, "drop": 0.027000975292587825, "rotation": {"subspace_principal_angle": 1.56369432512863, "mean_class_cosine": 0.0071019419641176074}}}, "iid_acc": 0.52375, "selectivity": -0.011249999999999982, "iid_split_rotation": {"subspace_principal_angle": 1.5297750902165248, "mean_class_cosine": 0.04100973285438006}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5025, "drop": 0.0}, "paraphrase": {"acc": 0.48764629388816644, "drop": 0.014853706111833509, "rotation": {"subspace_principal_angle": 1.4431237989109693, "mean_class_cosine": -0.12706524234551198}}}, "iid_acc": 0.5025, "selectivity": -0.032500000000000084, "iid_split_rotation": {"subspace_principal_angle": 1.5637689056526396, "mean_class_cosine": -0.006306566848071626}} +{"model": "pythia-70m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5175, "drop": 0.0}, "paraphrase": {"acc": 0.5006501950585176, "drop": 0.01684980494148236}}, "iid_acc": 0.5175, "selectivity": -0.01750000000000007} +{"model": "pythia-70m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.6, "drop": 0.0}, "paraphrase": {"acc": 0.5309973045822103, "drop": 0.0690026954177897, "rotation": {"subspace_principal_angle": 1.38978118980255, "mean_class_cosine": 0.3392687986531371}}}, "iid_acc": 0.6, "selectivity": 0.33875, "iid_split_rotation": {"subspace_principal_angle": 1.4386478796411275, "mean_class_cosine": 0.3036790886331675}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.1375, "drop": 0.0}, "paraphrase": {"acc": 0.1792452830188679, "drop": -0.0417452830188679, "rotation": {"subspace_principal_angle": 1.4906453725067592, "mean_class_cosine": 0.7942802395915041}}}, "iid_acc": 0.1375, "selectivity": -0.12374999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.4328823629930647, "mean_class_cosine": 0.6275373859348595}} +{"model": "pythia-70m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.5075, "drop": 0.0}, "paraphrase": {"acc": 0.49326145552560646, "drop": 0.014238544474393489}}, "iid_acc": 0.5075, "selectivity": 0.24624999999999997} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.7275, "drop": 0.0}, "paraphrase": {"acc": 0.7259136212624585, "drop": 0.0015863787375415273, "rotation": {"subspace_principal_angle": 1.3191750526863224, "mean_class_cosine": 0.24897450603487703}}}, "iid_acc": 0.7275, "selectivity": 0.21750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3632129148081515, "mean_class_cosine": 0.2060957956694738}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.54875, "drop": 0.0}, "paraphrase": {"acc": 0.5647840531561462, "drop": -0.016034053156146255, "rotation": {"subspace_principal_angle": 1.2764535135475243, "mean_class_cosine": 0.9760866896549567}}}, "iid_acc": 0.54875, "selectivity": 0.03874999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4318977466650444, "mean_class_cosine": 0.980247865923659}} +{"model": "pythia-70m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7325581395348837, "drop": 0.012441860465116306}}, "iid_acc": 0.745, "selectivity": 0.235} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.65125, "drop": 0.0}, "paraphrase": {"acc": 0.603648424543947, "drop": 0.047601575456053036, "rotation": {"subspace_principal_angle": 1.439031942391082, "mean_class_cosine": 0.13138343627378768}}}, "iid_acc": 0.65125, "selectivity": 0.12875000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.281323973512102, "mean_class_cosine": 0.2854465710869707}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.615257048092869, "drop": -0.00025704809286897223, "rotation": {"subspace_principal_angle": 1.3313954355137612, "mean_class_cosine": 0.9237404825006307}}}, "iid_acc": 0.615, "selectivity": 0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.5707267568617127, "mean_class_cosine": 0.8603555241845086}} +{"model": "pythia-70m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.66875, "drop": 0.0}, "paraphrase": {"acc": 0.6119402985074627, "drop": 0.05680970149253728}}, "iid_acc": 0.66875, "selectivity": 0.14625} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.7037643207855974, "drop": 0.013735679214402663, "rotation": {"subspace_principal_angle": 1.2968057041198295, "mean_class_cosine": 0.27057534861786225}}}, "iid_acc": 0.7175, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.2964308527675759, "mean_class_cosine": 0.2709361985468884}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.485, "drop": 0.0}, "paraphrase": {"acc": 0.5073649754500819, "drop": -0.02236497545008187, "rotation": {"subspace_principal_angle": 1.0405608352231381, "mean_class_cosine": 0.7548288630584341}}}, "iid_acc": 0.485, "selectivity": -0.12750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.4829069517857016, "mean_class_cosine": 0.8524232187677423}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7325, "drop": 0.0}, "paraphrase": {"acc": 0.6988543371522095, "drop": 0.03364566284779058}}, "iid_acc": 0.7325, "selectivity": 0.12} +{"model": "pythia-70m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8884353741496599, "drop": 0.017814625850340104, "rotation": {"subspace_principal_angle": 1.0443295050595705, "mean_class_cosine": 0.5024817409879795}}}, "iid_acc": 0.90625, "selectivity": 0.415, "iid_split_rotation": {"subspace_principal_angle": 1.0612122822978838, "mean_class_cosine": 0.4878141817807545}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.845, "drop": 0.0}, "paraphrase": {"acc": 0.7877551020408163, "drop": 0.05724489795918364, "rotation": {"subspace_principal_angle": 1.3338632968935769, "mean_class_cosine": 0.8769564447689145}}}, "iid_acc": 0.845, "selectivity": 0.35374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3069371602368167, "mean_class_cosine": 0.9412077325237846}} +{"model": "pythia-70m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.905, "drop": 0.0}, "paraphrase": {"acc": 0.8789115646258503, "drop": 0.026088435374149732}}, "iid_acc": 0.905, "selectivity": 0.41375} +{"model": "pythia-70m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.0027821252059307966, "rotation": {"subspace_principal_angle": 0.6856933607020452, "mean_class_cosine": 0.7739801903730489}}}, "iid_acc": 0.99125, "selectivity": 0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6339194714608114, "mean_class_cosine": 0.8057121716043742}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.85875, "drop": 0.0}, "paraphrase": {"acc": 0.8517298187808896, "drop": 0.007020181219110411, "rotation": {"subspace_principal_angle": 1.4167093648556626, "mean_class_cosine": 0.9633020957225444}}}, "iid_acc": 0.85875, "selectivity": -0.022499999999999964, "iid_split_rotation": {"subspace_principal_angle": 1.5217003295605112, "mean_class_cosine": 0.9881251646589356}} +{"model": "pythia-70m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.985, "drop": 0.0}, "paraphrase": {"acc": 0.9835255354200988, "drop": 0.0014744645799011913}}, "iid_acc": 0.985, "selectivity": 0.10375000000000001} +{"model": "pythia-70m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.65875, "drop": 0.0}, "paraphrase": {"acc": 0.6511299435028248, "drop": 0.007620056497175121, "rotation": {"subspace_principal_angle": 1.5394490513572545, "mean_class_cosine": 0.031342141781080114}}}, "iid_acc": 0.65875, "selectivity": 0.00374999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.529266372043682, "mean_class_cosine": 0.04151801773809659}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5268361581920904, "drop": -0.011836158192090385, "rotation": {"subspace_principal_angle": 1.5178073225930298, "mean_class_cosine": 0.21529478327501697}}}, "iid_acc": 0.515, "selectivity": -0.14, "iid_split_rotation": {"subspace_principal_angle": 1.0749998948016042, "mean_class_cosine": -0.653032824583841}} +{"model": "pythia-70m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.7231638418079096, "drop": -0.0031638418079096287}}, "iid_acc": 0.72, "selectivity": 0.06499999999999995} +{"model": "pythia-70m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.6105769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": -0.004807692307692291, "rotation": {"subspace_principal_angle": 1.399178613094325, "mean_class_cosine": 0.18957379279214578}}}, "iid_acc": 0.6105769230769231, "selectivity": 0.13942307692307698, "iid_split_rotation": {"subspace_principal_angle": 1.4319729985001466, "mean_class_cosine": 0.17742658711691725}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.4951923076923077, "drop": 0.0}, "paraphrase": {"acc": 0.5443786982248521, "drop": -0.049186390532544366, "rotation": {"subspace_principal_angle": 1.3438575364804408, "mean_class_cosine": 0.641655124351597}}}, "iid_acc": 0.4951923076923077, "selectivity": 0.024038461538461564, "iid_split_rotation": {"subspace_principal_angle": 1.2506869110680288, "mean_class_cosine": 0.4854077602261109}} +{"model": "pythia-70m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.5817307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": -0.033653846153846145}}, "iid_acc": 0.5817307692307693, "selectivity": 0.11057692307692313} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8706896551724138, "drop": 0.009310344827586192, "rotation": {"subspace_principal_angle": 1.2345926633445847, "mean_class_cosine": 0.3299056836099217}}}, "iid_acc": 0.88, "selectivity": 0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.219582897698587, "mean_class_cosine": 0.3440374169146968}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.735632183908046, "drop": 0.010617816091954002, "rotation": {"subspace_principal_angle": 1.5495300864536512, "mean_class_cosine": 0.7777397530079471}}}, "iid_acc": 0.74625, "selectivity": -0.028750000000000053, "iid_split_rotation": {"subspace_principal_angle": 1.4860573861271493, "mean_class_cosine": 0.7695588747682932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.85, "drop": 0.0}, "paraphrase": {"acc": 0.8520114942528736, "drop": -0.0020114942528736135}}, "iid_acc": 0.85, "selectivity": 0.07499999999999996} +{"model": "pythia-160m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.78, "drop": 0.0}, "paraphrase": {"acc": 0.732574679943101, "drop": 0.047425320056899034, "rotation": {"subspace_principal_angle": 1.2608414774243961, "mean_class_cosine": 0.30501563767234113}}, "domain": {"acc": 0.73, "drop": 0.050000000000000044, "rotation": {"subspace_principal_angle": 1.2628486007462063, "mean_class_cosine": 0.30310354600989187}}}, "iid_acc": 0.78, "selectivity": 0.275, "iid_split_rotation": {"subspace_principal_angle": 1.2752896290640992, "mean_class_cosine": 0.29122462225458356}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.51375, "drop": 0.0}, "paraphrase": {"acc": 0.5419630156472262, "drop": -0.028213015647226114, "rotation": {"subspace_principal_angle": 1.2025920048727983, "mean_class_cosine": 0.9965846798687883}}, "domain": {"acc": 0.54125, "drop": -0.02749999999999997, "rotation": {"subspace_principal_angle": 1.4516540873608994, "mean_class_cosine": 0.4073452915520843}}}, "iid_acc": 0.51375, "selectivity": 0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 0.9220355133014713, "mean_class_cosine": 0.9960516328052383}} +{"model": "pythia-160m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.7695590327169275, "drop": 0.04044096728307256}, "domain": {"acc": 0.81, "drop": 0.0}}, "iid_acc": 0.81, "selectivity": 0.30500000000000005} +{"model": "pythia-160m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8475, "drop": 0.0}, "paraphrase": {"acc": 0.8176470588235294, "drop": 0.029852941176470638, "rotation": {"subspace_principal_angle": 1.2030769237259111, "mean_class_cosine": 0.35948823052238327}}, "domain": {"acc": 0.52875, "drop": 0.31875, "rotation": {"subspace_principal_angle": 1.3485990887438482, "mean_class_cosine": 0.2203733735747998}}, "length": {"acc": 0.845925925925926, "drop": 0.001574074074074061, "rotation": {"subspace_principal_angle": 1.1802807331888672, "mean_class_cosine": 0.3806652417654258}}}, "iid_acc": 0.8475, "selectivity": 0.32125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.1501816602046397, "mean_class_cosine": 0.40832162126076316}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.6941176470588235, "drop": 0.017132352941176543, "rotation": {"subspace_principal_angle": 1.2416590458376622, "mean_class_cosine": 0.9484259293470975}}, "domain": {"acc": 0.555, "drop": 0.15625, "rotation": {"subspace_principal_angle": 1.321419755161037, "mean_class_cosine": 0.2893218657278887}}, "length": {"acc": 0.72, "drop": -0.008749999999999925, "rotation": {"subspace_principal_angle": 1.35683655045995, "mean_class_cosine": 0.9567589393114461}}}, "iid_acc": 0.71125, "selectivity": 0.18500000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5633547700573491, "mean_class_cosine": 0.9200623797757711}} +{"model": "pythia-160m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.8117647058823529, "drop": 0.014485294117647096}, "domain": {"acc": 0.58125, "drop": 0.245}, "length": {"acc": 0.8266666666666667, "drop": -0.0004166666666666208}}, "iid_acc": 0.82625, "selectivity": 0.30000000000000004} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.8575, "drop": 0.0}, "paraphrase": {"acc": 0.8515151515151516, "drop": 0.005984848484848482, "rotation": {"subspace_principal_angle": 1.4452099258783577, "mean_class_cosine": 0.24580852369730838}}}, "iid_acc": 0.8575, "selectivity": 0.6312500000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3943169835208806, "mean_class_cosine": 0.29648597603430116}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.83625, "drop": 0.0}, "paraphrase": {"acc": 0.8181818181818182, "drop": 0.018068181818181817, "rotation": {"subspace_principal_angle": 1.4625849696271982, "mean_class_cosine": 0.976912859464512}}}, "iid_acc": 0.83625, "selectivity": 0.6100000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3260532260143207, "mean_class_cosine": 0.9681319283557925}} +{"model": "pythia-160m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.89, "drop": 0.0}, "paraphrase": {"acc": 0.8727272727272727, "drop": 0.01727272727272733}}, "iid_acc": 0.89, "selectivity": 0.6637500000000001} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.955, "drop": 0.0}, "paraphrase": {"acc": 0.9343065693430657, "drop": 0.020693430656934297, "rotation": {"subspace_principal_angle": 1.197650762223889, "mean_class_cosine": 0.684597725874047}}}, "iid_acc": 0.955, "selectivity": 0.8875, "iid_split_rotation": {"subspace_principal_angle": 1.55650137949423, "mean_class_cosine": 0.7066198619660665}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5328467153284672, "drop": -0.00909671532846712, "rotation": {"subspace_principal_angle": 1.3016766284172454, "mean_class_cosine": 0.7910526489249412}}}, "iid_acc": 0.52375, "selectivity": 0.45625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2176842386779647, "mean_class_cosine": 0.8907647870321528}} +{"model": "pythia-160m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.92, "drop": 0.0}, "paraphrase": {"acc": 0.8953771289537713, "drop": 0.024622871046228734}}, "iid_acc": 0.92, "selectivity": 0.8525} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.53625, "drop": 0.0}, "paraphrase": {"acc": 0.5110533159947984, "drop": 0.025196684005201586, "rotation": {"subspace_principal_angle": 1.5229524083126806, "mean_class_cosine": 0.04782566779303716}}}, "iid_acc": 0.53625, "selectivity": 0.07124999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.447992810822192, "mean_class_cosine": 0.12249508806050173}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.48875, "drop": 0.0}, "paraphrase": {"acc": 0.4967490247074122, "drop": -0.007999024707412206, "rotation": {"subspace_principal_angle": 1.0259677749536884, "mean_class_cosine": -0.6292438791105933}}}, "iid_acc": 0.48875, "selectivity": 0.023749999999999993, "iid_split_rotation": {"subspace_principal_angle": 1.512376005309043, "mean_class_cosine": 0.4076926470016709}} +{"model": "pythia-160m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.4975, "drop": 0.0}, "paraphrase": {"acc": 0.4759427828348505, "drop": 0.021557217165149523}}, "iid_acc": 0.4975, "selectivity": 0.03249999999999997} +{"model": "pythia-160m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.64, "drop": 0.0}, "paraphrase": {"acc": 0.6078167115902965, "drop": 0.032183288409703525, "rotation": {"subspace_principal_angle": 1.5433122888881867, "mean_class_cosine": 0.42874421872110796}}}, "iid_acc": 0.64, "selectivity": 0.34750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.5081587697047814, "mean_class_cosine": 0.4198126394888509}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.3575, "drop": 0.0}, "paraphrase": {"acc": 0.31266846361185985, "drop": 0.044831536388140136, "rotation": {"subspace_principal_angle": 1.530582416292941, "mean_class_cosine": 0.6444087690823997}}}, "iid_acc": 0.3575, "selectivity": 0.065, "iid_split_rotation": {"subspace_principal_angle": 1.459558734848993, "mean_class_cosine": 0.588656959531061}} +{"model": "pythia-160m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.5053908355795148, "drop": -0.0003908355795148033}}, "iid_acc": 0.505, "selectivity": 0.21250000000000002} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.7408637873754153, "drop": -0.023363787375415224, "rotation": {"subspace_principal_angle": 1.3639308415465492, "mean_class_cosine": 0.20539322845895566}}}, "iid_acc": 0.7175, "selectivity": 0.20500000000000007, "iid_split_rotation": {"subspace_principal_angle": 1.3710931416929857, "mean_class_cosine": 0.19837842368661718}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.615, "drop": 0.0}, "paraphrase": {"acc": 0.6212624584717608, "drop": -0.006262458471760768, "rotation": {"subspace_principal_angle": 1.5406708151005613, "mean_class_cosine": 0.8985506716369838}}}, "iid_acc": 0.615, "selectivity": 0.10250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5633757536581958, "mean_class_cosine": 0.9281917050542527}} +{"model": "pythia-160m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.7558139534883721, "drop": -0.0020639534883720723}}, "iid_acc": 0.75375, "selectivity": 0.24125000000000008} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6086235489220564, "drop": 0.02637645107794362, "rotation": {"subspace_principal_angle": 1.490593398738083, "mean_class_cosine": 0.08011697135563739}}}, "iid_acc": 0.635, "selectivity": 0.12, "iid_split_rotation": {"subspace_principal_angle": 1.4638090458782234, "mean_class_cosine": 0.10678329666310886}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.59, "drop": 0.0}, "paraphrase": {"acc": 0.593698175787728, "drop": -0.0036981757877280197, "rotation": {"subspace_principal_angle": 1.5381806288996867, "mean_class_cosine": 0.9646917936431184}}}, "iid_acc": 0.59, "selectivity": 0.07499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.47230491449444, "mean_class_cosine": 0.9335513939805022}} +{"model": "pythia-160m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6650082918739635, "drop": 0.014991708126036563}}, "iid_acc": 0.68, "selectivity": 0.16500000000000004} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.6939443535188216, "drop": 0.02605564648117842, "rotation": {"subspace_principal_angle": 1.3786060400308688, "mean_class_cosine": 0.19100931112364286}}}, "iid_acc": 0.72, "selectivity": 0.07750000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.264300653737959, "mean_class_cosine": 0.3017194819812067}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.4828150572831424, "drop": 0.01718494271685761, "rotation": {"subspace_principal_angle": 1.5623310465881193, "mean_class_cosine": 0.5999071653626515}}}, "iid_acc": 0.5, "selectivity": -0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4335616759166658, "mean_class_cosine": 0.882837690979637}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.73625, "drop": 0.0}, "paraphrase": {"acc": 0.7217675941080196, "drop": 0.014482405891980354}}, "iid_acc": 0.73625, "selectivity": 0.09375} +{"model": "pythia-160m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.92875, "drop": 0.0}, "paraphrase": {"acc": 0.9224489795918367, "drop": 0.0063010204081632715, "rotation": {"subspace_principal_angle": 1.0136784498875773, "mean_class_cosine": 0.5287421015976792}}}, "iid_acc": 0.92875, "selectivity": 0.46124999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.0680431153367775, "mean_class_cosine": 0.4818398888533543}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.66, "drop": 0.0}, "paraphrase": {"acc": 0.5931972789115646, "drop": 0.06680272108843543, "rotation": {"subspace_principal_angle": 1.4285742922049103, "mean_class_cosine": 0.8425489186964704}}}, "iid_acc": 0.66, "selectivity": 0.1925, "iid_split_rotation": {"subspace_principal_angle": 1.0889790851620134, "mean_class_cosine": 0.7904713230276075}} +{"model": "pythia-160m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.92125, "drop": 0.0}, "paraphrase": {"acc": 0.9414965986394558, "drop": -0.02024659863945577}}, "iid_acc": 0.92125, "selectivity": 0.45375} +{"model": "pythia-160m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9835255354200988, "drop": 0.008974464579901253, "rotation": {"subspace_principal_angle": 1.0095289992802334, "mean_class_cosine": 0.5322595207735084}}}, "iid_acc": 0.9925, "selectivity": 0.13375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.9575577798603117, "mean_class_cosine": 0.5755189198683394}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.68875, "drop": 0.0}, "paraphrase": {"acc": 0.5535420098846787, "drop": 0.13520799011532125, "rotation": {"subspace_principal_angle": 0.9140922175459604, "mean_class_cosine": 0.9973997191189731}}}, "iid_acc": 0.68875, "selectivity": -0.17000000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2996635368722853, "mean_class_cosine": 0.9996246655982532}} +{"model": "pythia-160m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.98875, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.001929571663920937}}, "iid_acc": 0.98875, "selectivity": 0.13} +{"model": "pythia-160m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.67375, "drop": 0.0}, "paraphrase": {"acc": 0.6511299435028248, "drop": 0.022620056497175134, "rotation": {"subspace_principal_angle": 1.4877740078390536, "mean_class_cosine": 0.08292697708829477}}}, "iid_acc": 0.67375, "selectivity": 0.039999999999999925, "iid_split_rotation": {"subspace_principal_angle": 1.4794201635002593, "mean_class_cosine": 0.09124905725263827}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.4971751412429379, "drop": 0.03032485875706209, "rotation": {"subspace_principal_angle": 1.4466641276425085, "mean_class_cosine": 0.1695136595431389}}}, "iid_acc": 0.5275, "selectivity": -0.10625000000000007, "iid_split_rotation": {"subspace_principal_angle": 0.9893876475048883, "mean_class_cosine": -0.7941063384626064}} +{"model": "pythia-160m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.72375, "drop": 0.0}, "paraphrase": {"acc": 0.730225988700565, "drop": -0.006475988700564983}}, "iid_acc": 0.72375, "selectivity": 0.08999999999999997} +{"model": "pythia-160m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.6346153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.6449704142011834, "drop": -0.010355029585798814, "rotation": {"subspace_principal_angle": 1.401130641487477, "mean_class_cosine": 0.20190956693568948}}}, "iid_acc": 0.6346153846153846, "selectivity": 0.14423076923076922, "iid_split_rotation": {"subspace_principal_angle": 1.5640320536219627, "mean_class_cosine": 0.22919831492919082}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.5480769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.5739644970414202, "drop": -0.025887573964497035, "rotation": {"subspace_principal_angle": 1.3084230282064335, "mean_class_cosine": 0.6381937638179063}}}, "iid_acc": 0.5480769230769231, "selectivity": 0.057692307692307765, "iid_split_rotation": {"subspace_principal_angle": 1.1829697589881627, "mean_class_cosine": 0.46077624553615687}} +{"model": "pythia-160m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 2, "dists": {"iid": {"acc": 0.5913461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.621301775147929, "drop": -0.029955621301775093}}, "iid_acc": 0.5913461538461539, "selectivity": 0.10096153846153849} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.9025, "drop": 0.0}, "paraphrase": {"acc": 0.896551724137931, "drop": 0.005948275862068919, "rotation": {"subspace_principal_angle": 1.1447057554079878, "mean_class_cosine": 0.41331408912500195}}}, "iid_acc": 0.9025, "selectivity": 0.11624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.074604861072015, "mean_class_cosine": 0.4760797664026806}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.7775, "drop": 0.0}, "paraphrase": {"acc": 0.7873563218390804, "drop": -0.009856321839080473, "rotation": {"subspace_principal_angle": 1.4653362403173142, "mean_class_cosine": 0.5927810736166856}}}, "iid_acc": 0.7775, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.5560355491414277, "mean_class_cosine": 0.4882192393355668}} +{"model": "pythia-160m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.81875, "drop": 0.0}, "paraphrase": {"acc": 0.8117816091954023, "drop": 0.006968390804597657}}, "iid_acc": 0.81875, "selectivity": 0.03249999999999997} +{"model": "pythia-410m", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7937411095305832, "drop": 0.010008890469416776, "rotation": {"subspace_principal_angle": 1.3185889844917067, "mean_class_cosine": 0.2495420761739433}}, "domain": {"acc": 0.8825, "drop": -0.07874999999999999, "rotation": {"subspace_principal_angle": 1.3325721583114072, "mean_class_cosine": 0.23597731993808926}}}, "iid_acc": 0.80375, "selectivity": 0.29874999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.269745108593276, "mean_class_cosine": 0.2965243102986965}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5391180654338549, "drop": -0.02661806543385492, "rotation": {"subspace_principal_angle": 0.6582429807151361, "mean_class_cosine": 0.99926299379126}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.5312218618656819, "mean_class_cosine": 0.3356468163658598}}}, "iid_acc": 0.5125, "selectivity": 0.007499999999999951, "iid_split_rotation": {"subspace_principal_angle": 0.8629499226938586, "mean_class_cosine": 0.9991531829071234}} +{"model": "pythia-410m", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.802275960170697, "drop": 0.008974039829303004}, "domain": {"acc": 0.84625, "drop": -0.03499999999999992}}, "iid_acc": 0.81125, "selectivity": 0.30625} +{"model": "pythia-410m", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8666666666666667, "drop": 0.0008333333333333526, "rotation": {"subspace_principal_angle": 1.1792412886091128, "mean_class_cosine": 0.38162622377493116}}, "domain": {"acc": 0.54125, "drop": 0.32625000000000004, "rotation": {"subspace_principal_angle": 1.3149401585776963, "mean_class_cosine": 0.25307379864091606}}, "length": {"acc": 0.8725925925925926, "drop": -0.0050925925925925375, "rotation": {"subspace_principal_angle": 1.129874880545573, "mean_class_cosine": 0.42677296414992083}}}, "iid_acc": 0.8675, "selectivity": 0.37625000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.1059112988008373, "mean_class_cosine": 0.44832003402375775}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.80375, "drop": 0.0}, "paraphrase": {"acc": 0.7941176470588235, "drop": 0.00963235294117648, "rotation": {"subspace_principal_angle": 1.2208647478151893, "mean_class_cosine": 0.9486786112928252}}, "domain": {"acc": 0.555, "drop": 0.24874999999999992, "rotation": {"subspace_principal_angle": 1.4628719512711352, "mean_class_cosine": 0.16922760359562494}}, "length": {"acc": 0.8103703703703704, "drop": -0.006620370370370443, "rotation": {"subspace_principal_angle": 1.4518778845432394, "mean_class_cosine": 0.9386719672471594}}}, "iid_acc": 0.80375, "selectivity": 0.31249999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.1012982204245179, "mean_class_cosine": 0.8734916313973369}} +{"model": "pythia-410m", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8666666666666667, "drop": 0.01083333333333325}, "domain": {"acc": 0.525, "drop": 0.3524999999999999}, "length": {"acc": 0.8740740740740741, "drop": 0.0034259259259258323}}, "iid_acc": 0.8775, "selectivity": 0.3862499999999999} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.86625, "drop": 0.0}, "paraphrase": {"acc": 0.8681818181818182, "drop": -0.0019318181818182012, "rotation": {"subspace_principal_angle": 1.348541899340855, "mean_class_cosine": 0.33742946528630835}}}, "iid_acc": 0.86625, "selectivity": 0.6087499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3243203014914424, "mean_class_cosine": 0.3890416789316658}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.58875, "drop": 0.0}, "paraphrase": {"acc": 0.5772727272727273, "drop": 0.011477272727272725, "rotation": {"subspace_principal_angle": 1.1652446863372317, "mean_class_cosine": 0.943376790057927}}}, "iid_acc": 0.58875, "selectivity": 0.33125, "iid_split_rotation": {"subspace_principal_angle": 1.0212787219956532, "mean_class_cosine": 0.9231286978282006}} +{"model": "pythia-410m", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.875, "drop": 0.0}, "paraphrase": {"acc": 0.8727272727272727, "drop": 0.002272727272727315}}, "iid_acc": 0.875, "selectivity": 0.6174999999999999} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.97, "drop": 0.0}, "paraphrase": {"acc": 0.9586374695863747, "drop": 0.011362530413625227, "rotation": {"subspace_principal_angle": 1.0788515585794918, "mean_class_cosine": 0.680525963213341}}}, "iid_acc": 0.97, "selectivity": 0.91625, "iid_split_rotation": {"subspace_principal_angle": 1.0595584522895658, "mean_class_cosine": 0.7030153073160818}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.732360097323601, "drop": 0.0026399026763990197, "rotation": {"subspace_principal_angle": 1.5116341861027407, "mean_class_cosine": 0.873531191022212}}}, "iid_acc": 0.735, "selectivity": 0.68125, "iid_split_rotation": {"subspace_principal_angle": 1.4944238666561842, "mean_class_cosine": 0.9319808959810626}} +{"model": "pythia-410m", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 22, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9513381995133819, "drop": 0.004911800486618101}}, "iid_acc": 0.95625, "selectivity": 0.9025000000000001} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.56125, "drop": 0.0}, "paraphrase": {"acc": 0.5565669700910273, "drop": 0.004683029908972736, "rotation": {"subspace_principal_angle": 1.4553250675447744, "mean_class_cosine": 0.11521482211569181}}}, "iid_acc": 0.56125, "selectivity": 0.0625, "iid_split_rotation": {"subspace_principal_angle": 1.4999753911814653, "mean_class_cosine": 0.0707617488193961}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5097529258777633, "drop": -0.01975292587776334, "rotation": {"subspace_principal_angle": 1.5663024506899672, "mean_class_cosine": 0.00763914984376122}}}, "iid_acc": 0.49, "selectivity": -0.008750000000000036, "iid_split_rotation": {"subspace_principal_angle": 1.2427849178014092, "mean_class_cosine": 0.3406471787964678}} +{"model": "pythia-410m", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.48, "drop": 0.0}, "paraphrase": {"acc": 0.49154746423927176, "drop": -0.011547464239271776}}, "iid_acc": 0.48, "selectivity": -0.018750000000000044} +{"model": "pythia-410m", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.610512129380054, "drop": 0.028237870619946093, "rotation": {"subspace_principal_angle": 1.430640605755257, "mean_class_cosine": 0.2631769683858712}}}, "iid_acc": 0.63875, "selectivity": 0.39625000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.378148191691318, "mean_class_cosine": 0.297004025063675}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.12, "drop": 0.0}, "paraphrase": {"acc": 0.1280323450134771, "drop": -0.008032345013477105, "rotation": {"subspace_principal_angle": 1.543620776429251, "mean_class_cosine": 0.9288205176328912}}}, "iid_acc": 0.12, "selectivity": -0.1225, "iid_split_rotation": {"subspace_principal_angle": 1.43244607202274, "mean_class_cosine": 0.6582331054084928}} +{"model": "pythia-410m", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.59875, "drop": 0.0}, "paraphrase": {"acc": 0.6185983827493261, "drop": -0.0198483827493261}}, "iid_acc": 0.59875, "selectivity": 0.35625} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.74125, "drop": 0.0}, "paraphrase": {"acc": 0.7657807308970099, "drop": -0.024530730897009967, "rotation": {"subspace_principal_angle": 1.3660163460568273, "mean_class_cosine": 0.20335174270412884}}}, "iid_acc": 0.74125, "selectivity": 0.18999999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.413083036491105, "mean_class_cosine": 0.1570602898398632}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5475, "drop": 0.0}, "paraphrase": {"acc": 0.5647840531561462, "drop": -0.01728405315614623, "rotation": {"subspace_principal_angle": 0.8355530661327796, "mean_class_cosine": 0.9953186629963564}}}, "iid_acc": 0.5475, "selectivity": -0.003750000000000031, "iid_split_rotation": {"subspace_principal_angle": 1.4453486412673606, "mean_class_cosine": 0.996980243254297}} +{"model": "pythia-410m", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.7725, "drop": 0.0}, "paraphrase": {"acc": 0.7774086378737541, "drop": -0.004908637873754174}}, "iid_acc": 0.7725, "selectivity": 0.22124999999999995} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.67, "drop": 0.0}, "paraphrase": {"acc": 0.6252072968490879, "drop": 0.044792703150912105, "rotation": {"subspace_principal_angle": 1.4695161231652571, "mean_class_cosine": 0.10110714243323128}}}, "iid_acc": 0.67, "selectivity": 0.17250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4419616097046257, "mean_class_cosine": 0.12847860473843392}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5174129353233831, "drop": 0.010087064676616908, "rotation": {"subspace_principal_angle": 0.9292634922449805, "mean_class_cosine": 0.9948800762828822}}}, "iid_acc": 0.5275, "selectivity": 0.02999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.044667492526286, "mean_class_cosine": 0.9982719754571383}} +{"model": "pythia-410m", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.69375, "drop": 0.0}, "paraphrase": {"acc": 0.6882255389718076, "drop": 0.005524461028192373}}, "iid_acc": 0.69375, "selectivity": 0.19624999999999998} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.6890343698854338, "drop": 0.022215630114566287, "rotation": {"subspace_principal_angle": 1.44866296261917, "mean_class_cosine": 0.12182995563984489}}}, "iid_acc": 0.71125, "selectivity": 0.17875000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.4353636321031356, "mean_class_cosine": 0.13501905614072562}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.45625, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.018381751227495913, "rotation": {"subspace_principal_angle": 1.4711476398340722, "mean_class_cosine": -0.12720149911565562}}}, "iid_acc": 0.45625, "selectivity": -0.07624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3619831712096353, "mean_class_cosine": 0.9877340756601919}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.7575, "drop": 0.0}, "paraphrase": {"acc": 0.7495908346972177, "drop": 0.007909165302782295}}, "iid_acc": 0.7575, "selectivity": 0.22499999999999998} +{"model": "pythia-410m", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9333333333333333, "drop": 0.009166666666666656, "rotation": {"subspace_principal_angle": 1.1452162894508524, "mean_class_cosine": 0.41284914883949264}}}, "iid_acc": 0.9425, "selectivity": 0.45375, "iid_split_rotation": {"subspace_principal_angle": 1.1544129671241614, "mean_class_cosine": 0.4044554792046704}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.82125, "drop": 0.0}, "paraphrase": {"acc": 0.8108843537414966, "drop": 0.010365646258503447, "rotation": {"subspace_principal_angle": 1.4778841101146312, "mean_class_cosine": 0.9112587174612317}}}, "iid_acc": 0.82125, "selectivity": 0.3325, "iid_split_rotation": {"subspace_principal_angle": 1.2685804950112414, "mean_class_cosine": 0.7130834799426322}} +{"model": "pythia-410m", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9374149659863945, "drop": 0.005085034013605494}}, "iid_acc": 0.9425, "selectivity": 0.45375} +{"model": "pythia-410m", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9818780889621087, "drop": 0.011871911037891314, "rotation": {"subspace_principal_angle": 1.1022275614770123, "mean_class_cosine": 0.4516097785057144}}}, "iid_acc": 0.99375, "selectivity": 0.23875000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.0082315927465428, "mean_class_cosine": 0.5333574325360171}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9654036243822076, "drop": -0.00915362438220757, "rotation": {"subspace_principal_angle": 1.43644764939401, "mean_class_cosine": 0.9461248783860318}}}, "iid_acc": 0.95625, "selectivity": 0.20125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.5120745206596544, "mean_class_cosine": 0.9885336177553897}} +{"model": "pythia-410m", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.008577018121911029}}, "iid_acc": 0.99375, "selectivity": 0.23875000000000002} +{"model": "pythia-410m", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.71125, "drop": 0.0}, "paraphrase": {"acc": 0.693502824858757, "drop": 0.01774717514124302, "rotation": {"subspace_principal_angle": 1.4474830685331326, "mean_class_cosine": 0.12300097560505913}}}, "iid_acc": 0.71125, "selectivity": 0.09625000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3987832690834892, "mean_class_cosine": 0.1711660439449748}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5155367231638418, "drop": -0.015536723163841804, "rotation": {"subspace_principal_angle": 1.3401905016501925, "mean_class_cosine": 0.29339747507161273}}}, "iid_acc": 0.5, "selectivity": -0.11499999999999999, "iid_split_rotation": {"subspace_principal_angle": 0.6179518034185482, "mean_class_cosine": -0.9377606995696546}} +{"model": "pythia-410m", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.71625, "drop": 0.0}, "paraphrase": {"acc": 0.7259887005649718, "drop": -0.009738700564971725}}, "iid_acc": 0.71625, "selectivity": 0.10125000000000006} +{"model": "pythia-410m", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.6586538461538461, "drop": 0.0}, "paraphrase": {"acc": 0.6568047337278107, "drop": 0.0018491124260354708, "rotation": {"subspace_principal_angle": 1.4172417633253025, "mean_class_cosine": 0.17412301577924724}}}, "iid_acc": 0.6586538461538461, "selectivity": 0.21634615384615385, "iid_split_rotation": {"subspace_principal_angle": 1.3955447682800144, "mean_class_cosine": 0.20505793596312397}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.49038461538461536, "drop": 0.0}, "paraphrase": {"acc": 0.4319526627218935, "drop": 0.05843195266272189, "rotation": {"subspace_principal_angle": 1.0501848723626588, "mean_class_cosine": 0.8199704533724792}}}, "iid_acc": 0.49038461538461536, "selectivity": 0.04807692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.1738154155015759, "mean_class_cosine": 0.567012811542814}} +{"model": "pythia-410m", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.6778846153846154, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.0625}}, "iid_acc": 0.6778846153846154, "selectivity": 0.23557692307692313} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.91125, "drop": 0.0}, "paraphrase": {"acc": 0.8951149425287356, "drop": 0.016135057471264425, "rotation": {"subspace_principal_angle": 1.3118705655633425, "mean_class_cosine": 0.2560422701778298}}}, "iid_acc": 0.91125, "selectivity": 0.16249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3182833995213499, "mean_class_cosine": 0.24983798198888385}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.60125, "drop": 0.0}, "paraphrase": {"acc": 0.5804597701149425, "drop": 0.02079022988505741, "rotation": {"subspace_principal_angle": 1.0152273026792804, "mean_class_cosine": 0.8731250404929112}}}, "iid_acc": 0.60125, "selectivity": -0.14750000000000008, "iid_split_rotation": {"subspace_principal_angle": 0.9386885075423576, "mean_class_cosine": 0.5900924520237005}} +{"model": "pythia-410m", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.90875, "drop": 0.0}, "paraphrase": {"acc": 0.9037356321839081, "drop": 0.00501436781609188}}, "iid_acc": 0.90875, "selectivity": 0.15999999999999992} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.86875, "drop": 0.0}, "paraphrase": {"acc": 0.8421052631578947, "drop": 0.026644736842105332, "rotation": {"subspace_principal_angle": 1.3051008608954664, "mean_class_cosine": 0.262580393680383}}, "domain": {"acc": 0.865, "drop": 0.003750000000000031, "rotation": {"subspace_principal_angle": 1.4287631781026644, "mean_class_cosine": 0.1415560812007921}}}, "iid_acc": 0.86875, "selectivity": 0.35750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.2792444334922417, "mean_class_cosine": 0.28743897242092714}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.51125, "drop": 0.0}, "paraphrase": {"acc": 0.5391180654338549, "drop": -0.027868065433854894, "rotation": {"subspace_principal_angle": 1.5601359902088487, "mean_class_cosine": 0.9979053013441271}}, "domain": {"acc": 0.54125, "drop": -0.030000000000000027, "rotation": {"subspace_principal_angle": 1.4395920974393557, "mean_class_cosine": 0.23107881587582263}}}, "iid_acc": 0.51125, "selectivity": 0.0, "iid_split_rotation": {"subspace_principal_angle": 1.1884475439272368, "mean_class_cosine": 0.9977411853469901}} +{"model": "pythia-1.4b", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.87, "drop": 0.0}, "paraphrase": {"acc": 0.8492176386913229, "drop": 0.020782361308677055}, "domain": {"acc": 0.83375, "drop": 0.036250000000000004}}, "iid_acc": 0.87, "selectivity": 0.35875} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8647058823529412, "drop": 0.02904411764705883, "rotation": {"subspace_principal_angle": 1.3891107177208437, "mean_class_cosine": 0.1806876941601903}}, "domain": {"acc": 0.49, "drop": 0.40375000000000005, "rotation": {"subspace_principal_angle": 1.470272439761965, "mean_class_cosine": 0.10035467270286502}}, "length": {"acc": 0.8933333333333333, "drop": 0.0004166666666667318, "rotation": {"subspace_principal_angle": 1.336894896267935, "mean_class_cosine": 0.23177447065599882}}}, "iid_acc": 0.89375, "selectivity": 0.40125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.274478152252649, "mean_class_cosine": 0.2920008293919314}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.83125, "drop": 0.0}, "paraphrase": {"acc": 0.8117647058823529, "drop": 0.0194852941176471, "rotation": {"subspace_principal_angle": 1.3046911642334313, "mean_class_cosine": 0.9572532608121695}}, "domain": {"acc": 0.48125, "drop": 0.35000000000000003, "rotation": {"subspace_principal_angle": 1.2118756513387043, "mean_class_cosine": -0.011293529250558751}}, "length": {"acc": 0.8355555555555556, "drop": -0.0043055555555555625, "rotation": {"subspace_principal_angle": 1.2439536531709152, "mean_class_cosine": 0.9646290213720963}}}, "iid_acc": 0.83125, "selectivity": 0.33875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.5181831450402061, "mean_class_cosine": 0.9548748861745605}} +{"model": "pythia-1.4b", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8862745098039215, "drop": 0.007475490196078516}, "domain": {"acc": 0.7925, "drop": 0.10125000000000006}, "length": {"acc": 0.8962962962962963, "drop": -0.0025462962962962132}}, "iid_acc": 0.89375, "selectivity": 0.40125000000000005} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.89125, "drop": 0.0}, "paraphrase": {"acc": 0.8818181818181818, "drop": 0.009431818181818152, "rotation": {"subspace_principal_angle": 1.4369187993598345, "mean_class_cosine": 0.22041085502998395}}}, "iid_acc": 0.89125, "selectivity": 0.675, "iid_split_rotation": {"subspace_principal_angle": 1.4078654977043972, "mean_class_cosine": 0.2586407992914274}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.81, "drop": 0.0}, "paraphrase": {"acc": 0.7727272727272727, "drop": 0.037272727272727346, "rotation": {"subspace_principal_angle": 1.0175502832205285, "mean_class_cosine": 0.972353587476467}}}, "iid_acc": 0.81, "selectivity": 0.59375, "iid_split_rotation": {"subspace_principal_angle": 0.7420502510291406, "mean_class_cosine": 0.9634510604904396}} +{"model": "pythia-1.4b", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8875, "drop": 0.0}, "paraphrase": {"acc": 0.8833333333333333, "drop": 0.004166666666666652}}, "iid_acc": 0.8875, "selectivity": 0.6712499999999999} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9683698296836983, "drop": 0.007880170316301616, "rotation": {"subspace_principal_angle": 1.0992412190119936, "mean_class_cosine": 0.6867400981100495}}}, "iid_acc": 0.97625, "selectivity": 0.91875, "iid_split_rotation": {"subspace_principal_angle": 1.0034069622469282, "mean_class_cosine": 0.7160948548382569}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.80125, "drop": 0.0}, "paraphrase": {"acc": 0.8175182481751825, "drop": -0.016268248175182465, "rotation": {"subspace_principal_angle": 1.379334954656771, "mean_class_cosine": 0.8678086437576017}}}, "iid_acc": 0.80125, "selectivity": 0.74375, "iid_split_rotation": {"subspace_principal_angle": 0.8811167846778503, "mean_class_cosine": 0.9382525612852962}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 17, "dists": {"iid": {"acc": 0.97625, "drop": 0.0}, "paraphrase": {"acc": 0.9635036496350365, "drop": 0.01274635036496341}}, "iid_acc": 0.97625, "selectivity": 0.91875} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.5925, "drop": 0.0}, "paraphrase": {"acc": 0.599479843953186, "drop": -0.00697984395318596, "rotation": {"subspace_principal_angle": 1.4685167248622502, "mean_class_cosine": 0.10210136870418976}}}, "iid_acc": 0.5925, "selectivity": 0.09625, "iid_split_rotation": {"subspace_principal_angle": 1.492714330722181, "mean_class_cosine": 0.07800267856136348}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.49, "drop": 0.0}, "paraphrase": {"acc": 0.5097529258777633, "drop": -0.01975292587776334, "rotation": {"subspace_principal_angle": 1.3905538624434135, "mean_class_cosine": 0.43937650291700403}}}, "iid_acc": 0.49, "selectivity": -0.006250000000000033, "iid_split_rotation": {"subspace_principal_angle": 1.4581021828043095, "mean_class_cosine": -0.1678392152718658}} +{"model": "pythia-1.4b", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.577373211963589, "drop": -0.017373211963588986}}, "iid_acc": 0.56, "selectivity": 0.06375000000000003} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.65625, "drop": 0.0}, "paraphrase": {"acc": 0.6320754716981132, "drop": 0.024174528301886822, "rotation": {"subspace_principal_angle": 1.4488303719308924, "mean_class_cosine": 0.2608030814489897}}}, "iid_acc": 0.65625, "selectivity": 0.39, "iid_split_rotation": {"subspace_principal_angle": 1.4365260084834874, "mean_class_cosine": 0.27287586977099626}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.12, "drop": 0.0}, "paraphrase": {"acc": 0.12533692722371967, "drop": -0.005336927223719673, "rotation": {"subspace_principal_angle": 1.4252705910030379, "mean_class_cosine": 0.9453360722802503}}}, "iid_acc": 0.12, "selectivity": -0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.444135686833852, "mean_class_cosine": 0.6596132535425966}} +{"model": "pythia-1.4b", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.68, "drop": 0.0}, "paraphrase": {"acc": 0.6792452830188679, "drop": 0.0007547169811321641}}, "iid_acc": 0.68, "selectivity": 0.41375000000000006} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.75375, "drop": 0.0}, "paraphrase": {"acc": 0.760797342192691, "drop": -0.007047342192690986, "rotation": {"subspace_principal_angle": 1.4082028248717875, "mean_class_cosine": 0.16187804379501777}}}, "iid_acc": 0.75375, "selectivity": 0.21875, "iid_split_rotation": {"subspace_principal_angle": 1.4629230727487192, "mean_class_cosine": 0.10766416205491924}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5813953488372093, "drop": -0.021395348837209283, "rotation": {"subspace_principal_angle": 0.7137539003362412, "mean_class_cosine": 0.9509360173967503}}}, "iid_acc": 0.56, "selectivity": 0.025000000000000022, "iid_split_rotation": {"subspace_principal_angle": 1.0125517193691644, "mean_class_cosine": 0.9691372563411869}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.77375, "drop": 0.0}, "paraphrase": {"acc": 0.782392026578073, "drop": -0.008642026578073003}}, "iid_acc": 0.77375, "selectivity": 0.23875000000000002} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6451077943615257, "drop": 0.032392205638474336, "rotation": {"subspace_principal_angle": 1.4944467452146757, "mean_class_cosine": 0.0762754262877726}}}, "iid_acc": 0.6775, "selectivity": 0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.510090306300586, "mean_class_cosine": 0.06066874151454521}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.525, "drop": 0.0}, "paraphrase": {"acc": 0.5190713101160862, "drop": 0.005928689883913818, "rotation": {"subspace_principal_angle": 0.9851631097477104, "mean_class_cosine": 0.9870998388876457}}}, "iid_acc": 0.525, "selectivity": 0.01375000000000004, "iid_split_rotation": {"subspace_principal_angle": 0.8525820765093011, "mean_class_cosine": 0.9943272408460271}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.71, "drop": 0.0}, "paraphrase": {"acc": 0.681592039800995, "drop": 0.028407960199004934}}, "iid_acc": 0.71, "selectivity": 0.19874999999999998} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7184942716857611, "drop": 0.01650572831423891, "rotation": {"subspace_principal_angle": 1.4867843877013787, "mean_class_cosine": 0.08391314783631819}}}, "iid_acc": 0.735, "selectivity": 0.18499999999999994, "iid_split_rotation": {"subspace_principal_angle": 1.4727261303836074, "mean_class_cosine": 0.09791306933287178}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.4575, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.017131751227495884, "rotation": {"subspace_principal_angle": 1.4184867997055783, "mean_class_cosine": -0.12546411952087166}}}, "iid_acc": 0.4575, "selectivity": -0.09250000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.0826072069356356, "mean_class_cosine": 0.9756432630642227}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.74, "drop": 0.0}, "paraphrase": {"acc": 0.7446808510638298, "drop": -0.004680851063829761}}, "iid_acc": 0.74, "selectivity": 0.18999999999999995} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9401360544217687, "drop": 0.01861394557823126, "rotation": {"subspace_principal_angle": 1.1843775125331726, "mean_class_cosine": 0.3768737131337715}}}, "iid_acc": 0.95875, "selectivity": 0.5037499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.1630113814353649, "mean_class_cosine": 0.3965768790298291}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8025, "drop": 0.0}, "paraphrase": {"acc": 0.7755102040816326, "drop": 0.026989795918367365, "rotation": {"subspace_principal_angle": 0.9611960505376331, "mean_class_cosine": 0.9267288005728522}}}, "iid_acc": 0.8025, "selectivity": 0.3475, "iid_split_rotation": {"subspace_principal_angle": 0.6069137112776986, "mean_class_cosine": 0.8431794446206526}} +{"model": "pythia-1.4b", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.95625, "drop": 0.0}, "paraphrase": {"acc": 0.9523809523809523, "drop": 0.0038690476190477163}}, "iid_acc": 0.95625, "selectivity": 0.50125} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.005282125205930854, "rotation": {"subspace_principal_angle": 1.0067410745718517, "mean_class_cosine": 0.5346176539849523}}}, "iid_acc": 0.99375, "selectivity": 0.16125, "iid_split_rotation": {"subspace_principal_angle": 0.9448974604279329, "mean_class_cosine": 0.5858259839474748}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.66625, "drop": 0.0}, "paraphrase": {"acc": 0.40362438220757824, "drop": 0.2626256177924218, "rotation": {"subspace_principal_angle": 0.5567698167152252, "mean_class_cosine": 0.9809168363033169}}}, "iid_acc": 0.66625, "selectivity": -0.16625, "iid_split_rotation": {"subspace_principal_angle": 1.4341855078247439, "mean_class_cosine": 0.9999348952126837}} +{"model": "pythia-1.4b", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 5, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.006077018121910971}}, "iid_acc": 0.99125, "selectivity": 0.15874999999999995} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.6878531073446328, "drop": 0.030896892655367214, "rotation": {"subspace_principal_angle": 1.4853872573002862, "mean_class_cosine": 0.08530526830833424}}}, "iid_acc": 0.71875, "selectivity": 0.11499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3987163989876108, "mean_class_cosine": 0.1712319268023629}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.505, "drop": 0.0}, "paraphrase": {"acc": 0.5240112994350282, "drop": -0.019011299435028217, "rotation": {"subspace_principal_angle": 1.4845790668606034, "mean_class_cosine": 0.3310855041248773}}}, "iid_acc": 0.505, "selectivity": -0.09875, "iid_split_rotation": {"subspace_principal_angle": 1.4077945456675773, "mean_class_cosine": -0.8399837311477907}} +{"model": "pythia-1.4b", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.76375, "drop": 0.0}, "paraphrase": {"acc": 0.7372881355932204, "drop": 0.026461864406779667}}, "iid_acc": 0.76375, "selectivity": 0.16000000000000003} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6686390532544378, "drop": 0.018860946745562157, "rotation": {"subspace_principal_angle": 1.4431099310758673, "mean_class_cosine": 0.13799385145730425}}}, "iid_acc": 0.6875, "selectivity": 0.25961538461538464, "iid_split_rotation": {"subspace_principal_angle": 1.4632019943344636, "mean_class_cosine": 0.12184824535087391}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.5913461538461539, "drop": 0.0}, "paraphrase": {"acc": 0.5502958579881657, "drop": 0.04105029585798814, "rotation": {"subspace_principal_angle": 1.0140814812234873, "mean_class_cosine": 0.7438713263454885}}}, "iid_acc": 0.5913461538461539, "selectivity": 0.1634615384615385, "iid_split_rotation": {"subspace_principal_angle": 0.9988210743938787, "mean_class_cosine": 0.6629369773883759}} +{"model": "pythia-1.4b", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6745562130177515, "drop": 0.012943786982248517}}, "iid_acc": 0.6875, "selectivity": 0.25961538461538464} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.9075, "drop": 0.0}, "paraphrase": {"acc": 0.8936781609195402, "drop": 0.013821839080459752, "rotation": {"subspace_principal_angle": 1.3954839068042866, "mean_class_cosine": 0.17441577734760833}}}, "iid_acc": 0.9075, "selectivity": 0.16749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3326472336535684, "mean_class_cosine": 0.23590436416653315}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6135057471264368, "drop": 0.021494252873563213, "rotation": {"subspace_principal_angle": 0.9373099242351036, "mean_class_cosine": 0.8137785112536557}}}, "iid_acc": 0.635, "selectivity": -0.10499999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.3179859570982269, "mean_class_cosine": 0.6358483449489087}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 8, "dists": {"iid": {"acc": 0.915, "drop": 0.0}, "paraphrase": {"acc": 0.9051724137931034, "drop": 0.00982758620689661}}, "iid_acc": 0.915, "selectivity": 0.17500000000000004} +{"model": "gpt2", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.795, "drop": 0.0}, "paraphrase": {"acc": 0.786628733997155, "drop": 0.00837126600284499, "rotation": {"subspace_principal_angle": 1.310108733461078, "mean_class_cosine": 0.2577449743747703}}, "domain": {"acc": 0.79, "drop": 0.0050000000000000044, "rotation": {"subspace_principal_angle": 1.3864147968628893, "mean_class_cosine": 0.18333858173678275}}}, "iid_acc": 0.795, "selectivity": 0.28500000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.3577658636568546, "mean_class_cosine": 0.2114228247278906}} +{"model": "gpt2", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5362731152204836, "drop": -0.023773115220483643, "rotation": {"subspace_principal_angle": 1.4911474327942673, "mean_class_cosine": 0.9991258530703471}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.4889114257860925, "mean_class_cosine": 0.43982299953686566}}}, "iid_acc": 0.5125, "selectivity": 0.0024999999999999467, "iid_split_rotation": {"subspace_principal_angle": 1.5599854028146705, "mean_class_cosine": 0.9990812404452749}} +{"model": "gpt2", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 7, "dists": {"iid": {"acc": 0.82, "drop": 0.0}, "paraphrase": {"acc": 0.7923186344238976, "drop": 0.027681365576102346}, "domain": {"acc": 0.83125, "drop": -0.011250000000000093}}, "iid_acc": 0.82, "selectivity": 0.30999999999999994} +{"model": "gpt2", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.8375, "drop": 0.0}, "paraphrase": {"acc": 0.8058823529411765, "drop": 0.03161764705882353, "rotation": {"subspace_principal_angle": 1.3708466120257015, "mean_class_cosine": 0.19862004766983304}}, "domain": {"acc": 0.68125, "drop": 0.15625, "rotation": {"subspace_principal_angle": 1.384027843480055, "mean_class_cosine": 0.1856845513313584}}, "length": {"acc": 0.8488888888888889, "drop": -0.011388888888888893, "rotation": {"subspace_principal_angle": 1.3282357766524422, "mean_class_cosine": 0.24018900406085658}}}, "iid_acc": 0.8375, "selectivity": 0.32875, "iid_split_rotation": {"subspace_principal_angle": 1.3949678570258965, "mean_class_cosine": 0.17492389391787147}} +{"model": "gpt2", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.77625, "drop": 0.0}, "paraphrase": {"acc": 0.7823529411764706, "drop": -0.006102941176470589, "rotation": {"subspace_principal_angle": 0.8158694259835028, "mean_class_cosine": 0.9437326197394826}}, "domain": {"acc": 0.555, "drop": 0.22124999999999995, "rotation": {"subspace_principal_angle": 1.4680655032048706, "mean_class_cosine": 0.22680383044754415}}, "length": {"acc": 0.7866666666666666, "drop": -0.01041666666666663, "rotation": {"subspace_principal_angle": 1.075088915720483, "mean_class_cosine": 0.9383123548837031}}}, "iid_acc": 0.77625, "selectivity": 0.26749999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3496789801104079, "mean_class_cosine": 0.92006570726007}} +{"model": "gpt2", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.85125, "drop": 0.0}, "paraphrase": {"acc": 0.8431372549019608, "drop": 0.008112745098039165}, "domain": {"acc": 0.55, "drop": 0.3012499999999999}, "length": {"acc": 0.8548148148148148, "drop": -0.003564814814814854}}, "iid_acc": 0.85125, "selectivity": 0.3424999999999999} +{"model": "gpt2", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8363636363636363, "drop": 0.0023863636363636864, "rotation": {"subspace_principal_angle": 1.3907909651765882, "mean_class_cosine": 0.27160223854429794}}}, "iid_acc": 0.83875, "selectivity": 0.58375, "iid_split_rotation": {"subspace_principal_angle": 1.374984073519207, "mean_class_cosine": 0.3084279719861868}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.53125, "drop": 0.0}, "paraphrase": {"acc": 0.5393939393939394, "drop": -0.008143939393939426, "rotation": {"subspace_principal_angle": 1.3095900045768096, "mean_class_cosine": 0.9315569392551215}}}, "iid_acc": 0.53125, "selectivity": 0.27625, "iid_split_rotation": {"subspace_principal_angle": 1.3512999965296404, "mean_class_cosine": 0.9302928191576891}} +{"model": "gpt2", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8712121212121212, "drop": 0.002537878787878811}}, "iid_acc": 0.87375, "selectivity": 0.61875} +{"model": "gpt2", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9416058394160584, "drop": 0.017144160583941637, "rotation": {"subspace_principal_angle": 1.169967417171358, "mean_class_cosine": 0.6021107227776437}}}, "iid_acc": 0.95875, "selectivity": 0.88, "iid_split_rotation": {"subspace_principal_angle": 1.0879671022275426, "mean_class_cosine": 0.6463882048277961}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.56625, "drop": 0.0}, "paraphrase": {"acc": 0.5717761557177615, "drop": -0.005526155717761494, "rotation": {"subspace_principal_angle": 1.5297708862576174, "mean_class_cosine": 0.8076540626245426}}}, "iid_acc": 0.56625, "selectivity": 0.48750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.31250485165038, "mean_class_cosine": 0.8967378658864777}} +{"model": "gpt2", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.95875, "drop": 0.0}, "paraphrase": {"acc": 0.9245742092457421, "drop": 0.03417579075425792}}, "iid_acc": 0.95875, "selectivity": 0.88} +{"model": "gpt2", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5675, "drop": 0.0}, "paraphrase": {"acc": 0.5370611183355006, "drop": 0.030438881664499373, "rotation": {"subspace_principal_angle": 1.4333280858176263, "mean_class_cosine": 0.13703568181783926}}}, "iid_acc": 0.5675, "selectivity": 0.08000000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.5372359367014987, "mean_class_cosine": 0.03355409060481619}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5019505851755527, "drop": -0.0019505851755526882, "rotation": {"subspace_principal_angle": 1.5166639134271087, "mean_class_cosine": 0.7284040860032104}}}, "iid_acc": 0.5, "selectivity": 0.012500000000000011, "iid_split_rotation": {"subspace_principal_angle": 1.5345005286615787, "mean_class_cosine": -0.9244877025726085}} +{"model": "gpt2", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5136540962288687, "drop": -0.0011540962288687506}}, "iid_acc": 0.5125, "selectivity": 0.024999999999999967} +{"model": "gpt2", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.6475, "drop": 0.0}, "paraphrase": {"acc": 0.578167115902965, "drop": 0.06933288409703497, "rotation": {"subspace_principal_angle": 1.4396284306360443, "mean_class_cosine": 0.2858703332393371}}}, "iid_acc": 0.6475, "selectivity": 0.39249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4241221420081531, "mean_class_cosine": 0.3414796507506154}} +{"model": "gpt2", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.2, "drop": 0.0}, "paraphrase": {"acc": 0.2169811320754717, "drop": -0.016981132075471694, "rotation": {"subspace_principal_angle": 1.478985421617415, "mean_class_cosine": 0.7260803181295756}}}, "iid_acc": 0.2, "selectivity": -0.05499999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.3657973968851722, "mean_class_cosine": 0.6865447650543287}} +{"model": "gpt2", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.62375, "drop": 0.0}, "paraphrase": {"acc": 0.5956873315363881, "drop": 0.02806266846361194}}, "iid_acc": 0.62375, "selectivity": 0.36875} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.75, "drop": 0.0}, "paraphrase": {"acc": 0.7541528239202658, "drop": -0.004152823920265836, "rotation": {"subspace_principal_angle": 1.3055947110598365, "mean_class_cosine": 0.26210384065900993}}}, "iid_acc": 0.75, "selectivity": 0.19625000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4135255349018365, "mean_class_cosine": 0.15662326790731992}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6225, "drop": 0.0}, "paraphrase": {"acc": 0.6295681063122923, "drop": -0.007068106312292266, "rotation": {"subspace_principal_angle": 1.412398245668166, "mean_class_cosine": 0.9872471083520573}}}, "iid_acc": 0.6225, "selectivity": 0.06875000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.0308446538703597, "mean_class_cosine": 0.9830930319085469}} +{"model": "gpt2", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.74625, "drop": 0.0}, "paraphrase": {"acc": 0.7458471760797342, "drop": 0.0004028239202658046}}, "iid_acc": 0.74625, "selectivity": 0.1925} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.64625, "drop": 0.0}, "paraphrase": {"acc": 0.593698175787728, "drop": 0.052551824212272, "rotation": {"subspace_principal_angle": 1.4774953080617963, "mean_class_cosine": 0.09316571216572649}}}, "iid_acc": 0.64625, "selectivity": 0.14249999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.4132962154086606, "mean_class_cosine": 0.15684975311512517}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.5275, "drop": 0.0}, "paraphrase": {"acc": 0.5174129353233831, "drop": 0.010087064676616908, "rotation": {"subspace_principal_angle": 1.4727791381339745, "mean_class_cosine": 0.9962811296518983}}}, "iid_acc": 0.5275, "selectivity": 0.023749999999999938, "iid_split_rotation": {"subspace_principal_angle": 1.324949369047214, "mean_class_cosine": 0.9986827286127624}} +{"model": "gpt2", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.70125, "drop": 0.0}, "paraphrase": {"acc": 0.6633499170812603, "drop": 0.0379000829187397}}, "iid_acc": 0.70125, "selectivity": 0.1975} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.705, "drop": 0.0}, "paraphrase": {"acc": 0.6972176759410802, "drop": 0.007782324058919765, "rotation": {"subspace_principal_angle": 1.5034674219806008, "mean_class_cosine": 0.06727804731954892}}}, "iid_acc": 0.705, "selectivity": 0.13624999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4178710295321864, "mean_class_cosine": 0.1523299382782468}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.46125, "drop": 0.0}, "paraphrase": {"acc": 0.469721767594108, "drop": -0.008471767594108004, "rotation": {"subspace_principal_angle": 1.2701697435580983, "mean_class_cosine": -0.4177592757825601}}}, "iid_acc": 0.46125, "selectivity": -0.10749999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.4728680034154755, "mean_class_cosine": 0.9927211138662637}} +{"model": "gpt2", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 3, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.723404255319149, "drop": 0.0015957446808509967}}, "iid_acc": 0.725, "selectivity": 0.15625} +{"model": "gpt2", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90625, "drop": 0.0}, "paraphrase": {"acc": 0.8993197278911564, "drop": 0.00693027210884356, "rotation": {"subspace_principal_angle": 1.2974026238207301, "mean_class_cosine": 0.2700006465670901}}}, "iid_acc": 0.90625, "selectivity": 0.44125, "iid_split_rotation": {"subspace_principal_angle": 1.2752165891310814, "mean_class_cosine": 0.29129449547307984}} +{"model": "gpt2", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.87375, "drop": 0.0}, "paraphrase": {"acc": 0.8761904761904762, "drop": -0.0024404761904761596, "rotation": {"subspace_principal_angle": 1.0942486888882677, "mean_class_cosine": 0.942913260405956}}}, "iid_acc": 0.87375, "selectivity": 0.40875, "iid_split_rotation": {"subspace_principal_angle": 0.9061762984312886, "mean_class_cosine": 0.8552558967615458}} +{"model": "gpt2", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9306122448979591, "drop": -0.008112244897959142}}, "iid_acc": 0.9225, "selectivity": 0.45749999999999996} +{"model": "gpt2", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.985172981878089, "drop": 0.007327018121911055, "rotation": {"subspace_principal_angle": 1.072930156314306, "mean_class_cosine": 0.4775518369824139}}}, "iid_acc": 0.9925, "selectivity": 0.22375, "iid_split_rotation": {"subspace_principal_angle": 1.0267208542859751, "mean_class_cosine": 0.5176272803692485}} +{"model": "gpt2", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.7100494233937397, "drop": -0.022549423393739665, "rotation": {"subspace_principal_angle": 1.2433728178446086, "mean_class_cosine": 0.997937410347562}}}, "iid_acc": 0.6875, "selectivity": -0.08125000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.4234682182286518, "mean_class_cosine": 0.9995278492579525}} +{"model": "gpt2", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9925, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.005679571663920968}}, "iid_acc": 0.9925, "selectivity": 0.22375} +{"model": "gpt2", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.6675, "drop": 0.0}, "paraphrase": {"acc": 0.6652542372881356, "drop": 0.0022457627118643853, "rotation": {"subspace_principal_angle": 1.5101964181742233, "mean_class_cosine": 0.0605628247623603}}}, "iid_acc": 0.6675, "selectivity": 0.06499999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4116411659990913, "mean_class_cosine": 0.1584841015440742}} +{"model": "gpt2", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.495, "drop": 0.0}, "paraphrase": {"acc": 0.5112994350282486, "drop": -0.0162994350282486, "rotation": {"subspace_principal_angle": 1.2440118776947484, "mean_class_cosine": 0.4397585570740069}}}, "iid_acc": 0.495, "selectivity": -0.10750000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.0189797005862058, "mean_class_cosine": -0.9017093672978498}} +{"model": "gpt2", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 9, "dists": {"iid": {"acc": 0.72625, "drop": 0.0}, "paraphrase": {"acc": 0.7090395480225988, "drop": 0.01721045197740112}}, "iid_acc": 0.72625, "selectivity": 0.12374999999999992} +{"model": "gpt2", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.5976331360946746, "drop": 0.04178994082840226, "rotation": {"subspace_principal_angle": 1.3670769632091069, "mean_class_cosine": 0.221961109804074}}}, "iid_acc": 0.6394230769230769, "selectivity": 0.13942307692307687, "iid_split_rotation": {"subspace_principal_angle": 1.3869516790493233, "mean_class_cosine": 0.2430729393469231}} +{"model": "gpt2", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.3269230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.35502958579881655, "drop": -0.028106508875739622, "rotation": {"subspace_principal_angle": 1.4355728624955333, "mean_class_cosine": 0.7267142033085312}}}, "iid_acc": 0.3269230769230769, "selectivity": -0.17307692307692307, "iid_split_rotation": {"subspace_principal_angle": 1.1849716820736076, "mean_class_cosine": 0.7548318595334315}} +{"model": "gpt2", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.6490384615384616, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.033653846153846145}}, "iid_acc": 0.6490384615384616, "selectivity": 0.14903846153846156} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8936781609195402, "drop": 0.007571839080459775, "rotation": {"subspace_principal_angle": 1.404231999409768, "mean_class_cosine": 0.1657952106350371}}}, "iid_acc": 0.90125, "selectivity": 0.19125000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.415744792030871, "mean_class_cosine": 0.1544310160272101}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.60375, "drop": 0.0}, "paraphrase": {"acc": 0.5919540229885057, "drop": 0.011795977011494263, "rotation": {"subspace_principal_angle": 0.829164366422975, "mean_class_cosine": 0.8319128806446898}}}, "iid_acc": 0.60375, "selectivity": -0.10624999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.1149004124569097, "mean_class_cosine": 0.34312396741171547}} +{"model": "gpt2", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.9175, "drop": 0.0}, "paraphrase": {"acc": 0.9123563218390804, "drop": 0.00514367816091954}}, "iid_acc": 0.9175, "selectivity": 0.20750000000000002} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.82625, "drop": 0.0}, "paraphrase": {"acc": 0.7837837837837838, "drop": 0.04246621621621627, "rotation": {"subspace_principal_angle": 1.3737563564873747, "mean_class_cosine": 0.1957674385129709}}, "domain": {"acc": 0.80125, "drop": 0.025000000000000022, "rotation": {"subspace_principal_angle": 1.376738696045229, "mean_class_cosine": 0.19284193990423557}}}, "iid_acc": 0.82625, "selectivity": 0.32000000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.3152150432578507, "mean_class_cosine": 0.2528078527302202}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5362731152204836, "drop": -0.023773115220483643, "rotation": {"subspace_principal_angle": 0.8284726401777219, "mean_class_cosine": 0.9987588658556924}}, "domain": {"acc": 0.54125, "drop": -0.028750000000000053, "rotation": {"subspace_principal_angle": 1.559015217957762, "mean_class_cosine": 0.37022421141777156}}}, "iid_acc": 0.5125, "selectivity": 0.006249999999999978, "iid_split_rotation": {"subspace_principal_angle": 1.4590601880015512, "mean_class_cosine": 0.9985735288523132}} +{"model": "gpt2-medium", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.84125, "drop": 0.0}, "paraphrase": {"acc": 0.7809388335704125, "drop": 0.06031116642958756}, "domain": {"acc": 0.84125, "drop": 0.0}}, "iid_acc": 0.84125, "selectivity": 0.3350000000000001} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.88, "drop": 0.0}, "paraphrase": {"acc": 0.8606060606060606, "drop": 0.019393939393939408, "rotation": {"subspace_principal_angle": 1.3448125803660131, "mean_class_cosine": 0.38575539047723395}}}, "iid_acc": 0.88, "selectivity": 0.67125, "iid_split_rotation": {"subspace_principal_angle": 1.2768731344200692, "mean_class_cosine": 0.4190057988176132}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.355, "drop": 0.0}, "paraphrase": {"acc": 0.3106060606060606, "drop": 0.044393939393939374, "rotation": {"subspace_principal_angle": 1.1355626344445542, "mean_class_cosine": 0.905045264153325}}}, "iid_acc": 0.355, "selectivity": 0.14625, "iid_split_rotation": {"subspace_principal_angle": 1.5495618026124416, "mean_class_cosine": 0.9314215323578507}} +{"model": "gpt2-medium", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.885, "drop": 0.0}, "paraphrase": {"acc": 0.8833333333333333, "drop": 0.0016666666666667052}}, "iid_acc": 0.885, "selectivity": 0.67625} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.56375, "drop": 0.0}, "paraphrase": {"acc": 0.540962288686606, "drop": 0.022787711313393966, "rotation": {"subspace_principal_angle": 1.4370366006567117, "mean_class_cosine": 0.13336121878297752}}}, "iid_acc": 0.56375, "selectivity": 0.039999999999999925, "iid_split_rotation": {"subspace_principal_angle": 1.49957415048444, "mean_class_cosine": 0.07116197799766466}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5, "drop": 0.0}, "paraphrase": {"acc": 0.5019505851755527, "drop": -0.0019505851755526882, "rotation": {"subspace_principal_angle": 1.5354758515901374, "mean_class_cosine": 0.6588408805283696}}}, "iid_acc": 0.5, "selectivity": -0.02375000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.1055065787936926, "mean_class_cosine": -0.9020263448542752}} +{"model": "gpt2-medium", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.53, "drop": 0.0}, "paraphrase": {"acc": 0.4837451235370611, "drop": 0.04625487646293891}}, "iid_acc": 0.53, "selectivity": 0.006249999999999978} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6226415094339622, "drop": 0.054858490566037754, "rotation": {"subspace_principal_angle": 1.4132260830128582, "mean_class_cosine": 0.315857679610795}}}, "iid_acc": 0.6775, "selectivity": 0.4175, "iid_split_rotation": {"subspace_principal_angle": 1.41912790604836, "mean_class_cosine": 0.32533634812236173}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.245, "drop": 0.0}, "paraphrase": {"acc": 0.2628032345013477, "drop": -0.017803234501347687, "rotation": {"subspace_principal_angle": 1.2435931646346163, "mean_class_cosine": 0.704084265739651}}}, "iid_acc": 0.245, "selectivity": -0.015000000000000013, "iid_split_rotation": {"subspace_principal_angle": 1.2112395431952154, "mean_class_cosine": 0.6851956806446887}} +{"model": "gpt2-medium", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 1, "dists": {"iid": {"acc": 0.63875, "drop": 0.0}, "paraphrase": {"acc": 0.6185983827493261, "drop": 0.020151617250673937}}, "iid_acc": 0.63875, "selectivity": 0.37875000000000003} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.73, "drop": 0.0}, "paraphrase": {"acc": 0.7475083056478405, "drop": -0.01750830564784056, "rotation": {"subspace_principal_angle": 1.4592105747301425, "mean_class_cosine": 0.11135433008645393}}}, "iid_acc": 0.73, "selectivity": 0.23249999999999998, "iid_split_rotation": {"subspace_principal_angle": 1.5311242177574582, "mean_class_cosine": 0.03966170335806373}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.565, "drop": 0.0}, "paraphrase": {"acc": 0.5830564784053156, "drop": -0.018056478405315657, "rotation": {"subspace_principal_angle": 1.3688794358413638, "mean_class_cosine": 0.9651761136342984}}}, "iid_acc": 0.565, "selectivity": 0.06749999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.3920800768445374, "mean_class_cosine": 0.9821284088039223}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.7625, "drop": 0.0}, "paraphrase": {"acc": 0.7790697674418605, "drop": -0.01656976744186056}}, "iid_acc": 0.7625, "selectivity": 0.26499999999999996} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6775, "drop": 0.0}, "paraphrase": {"acc": 0.6677577741407529, "drop": 0.009742225859247111, "rotation": {"subspace_principal_angle": 1.5176621403016026, "mean_class_cosine": 0.05310918824653445}}}, "iid_acc": 0.6775, "selectivity": 0.11250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.457199891551373, "mean_class_cosine": 0.11335228191779778}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.46125, "drop": 0.0}, "paraphrase": {"acc": 0.4746317512274959, "drop": -0.013381751227495908, "rotation": {"subspace_principal_angle": 1.5461043074662548, "mean_class_cosine": -0.13586943898113613}}}, "iid_acc": 0.46125, "selectivity": -0.10374999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.5005876536533236, "mean_class_cosine": 0.9790471285771307}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.745, "drop": 0.0}, "paraphrase": {"acc": 0.7364975450081833, "drop": 0.008502454991816677}}, "iid_acc": 0.745, "selectivity": 0.18000000000000005} +{"model": "gpt2-medium", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.935, "drop": 0.0}, "paraphrase": {"acc": 0.9333333333333333, "drop": 0.0016666666666667052, "rotation": {"subspace_principal_angle": 1.2444931970939492, "mean_class_cosine": 0.32054342619861337}}}, "iid_acc": 0.935, "selectivity": 0.43750000000000006, "iid_split_rotation": {"subspace_principal_angle": 1.2405200440265627, "mean_class_cosine": 0.32430439122172083}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.85, "drop": 0.0}, "paraphrase": {"acc": 0.8448979591836735, "drop": 0.005102040816326481, "rotation": {"subspace_principal_angle": 0.8606756929177377, "mean_class_cosine": 0.946865026478741}}}, "iid_acc": 0.85, "selectivity": 0.3525, "iid_split_rotation": {"subspace_principal_angle": 1.1743193451678584, "mean_class_cosine": 0.8825257604902363}} +{"model": "gpt2-medium", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.94625, "drop": 0.0}, "paraphrase": {"acc": 0.9442176870748299, "drop": 0.0020323129251701433}}, "iid_acc": 0.94625, "selectivity": 0.44875000000000004} +{"model": "gpt2-medium", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": 0.0011346787479407094, "rotation": {"subspace_principal_angle": 1.075164913899876, "mean_class_cosine": 0.4755871801312971}}}, "iid_acc": 0.99125, "selectivity": 0.20875, "iid_split_rotation": {"subspace_principal_angle": 0.9579353285296144, "mean_class_cosine": 0.575210123910989}} +{"model": "gpt2-medium", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.6725, "drop": 0.0}, "paraphrase": {"acc": 0.6935749588138386, "drop": -0.021074958813838585, "rotation": {"subspace_principal_angle": 1.2610093649983862, "mean_class_cosine": 0.9991933388093968}}}, "iid_acc": 0.6725, "selectivity": -0.10999999999999999, "iid_split_rotation": {"subspace_principal_angle": 1.2635441752944665, "mean_class_cosine": 0.9997669069494608}} +{"model": "gpt2-medium", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99125, "drop": 0.0}, "paraphrase": {"acc": 0.9868204283360791, "drop": 0.004429571663920884}}, "iid_acc": 0.99125, "selectivity": 0.20875} +{"model": "gpt2-medium", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.68625, "drop": 0.0}, "paraphrase": {"acc": 0.655367231638418, "drop": 0.03088276836158199, "rotation": {"subspace_principal_angle": 1.5026236950761325, "mean_class_cosine": 0.06811983851611561}}}, "iid_acc": 0.68625, "selectivity": 0.09750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.4122166058344536, "mean_class_cosine": 0.15791590818006312}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.49875, "drop": 0.0}, "paraphrase": {"acc": 0.5098870056497176, "drop": -0.011137005649717535, "rotation": {"subspace_principal_angle": 1.2304337374701209, "mean_class_cosine": 0.42959183939934587}}}, "iid_acc": 0.49875, "selectivity": -0.08999999999999997, "iid_split_rotation": {"subspace_principal_angle": 1.504970426427495, "mean_class_cosine": -0.8759318900080222}} +{"model": "gpt2-medium", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 13, "dists": {"iid": {"acc": 0.72, "drop": 0.0}, "paraphrase": {"acc": 0.7189265536723164, "drop": 0.0010734463276835804}}, "iid_acc": 0.72, "selectivity": 0.13124999999999998} +{"model": "gpt2-medium", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.6442307692307693, "drop": 0.0}, "paraphrase": {"acc": 0.6449704142011834, "drop": -0.0007396449704141217, "rotation": {"subspace_principal_angle": 1.4713950529119475, "mean_class_cosine": 0.14817274811366463}}}, "iid_acc": 0.6442307692307693, "selectivity": 0.2211538461538462, "iid_split_rotation": {"subspace_principal_angle": 1.5076060847864945, "mean_class_cosine": 0.10760657740009509}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5048076923076923, "drop": 0.0}, "paraphrase": {"acc": 0.4319526627218935, "drop": 0.07285502958579881, "rotation": {"subspace_principal_angle": 1.3335837272128281, "mean_class_cosine": 0.760631306748815}}}, "iid_acc": 0.5048076923076923, "selectivity": 0.08173076923076922, "iid_split_rotation": {"subspace_principal_angle": 1.2437180651878326, "mean_class_cosine": 0.6436823456850219}} +{"model": "gpt2-medium", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.6875, "drop": 0.0}, "paraphrase": {"acc": 0.6568047337278107, "drop": 0.030695266272189325}}, "iid_acc": 0.6875, "selectivity": 0.2644230769230769} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8836206896551724, "drop": 0.010129310344827669, "rotation": {"subspace_principal_angle": 1.3913906147486017, "mean_class_cosine": 0.17844485539062008}}}, "iid_acc": 0.89375, "selectivity": 0.1725000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.4139643022148158, "mean_class_cosine": 0.15618990061442609}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.6120689655172413, "drop": 0.014181034482758648, "rotation": {"subspace_principal_angle": 0.7525157984620534, "mean_class_cosine": 0.7829464834163522}}}, "iid_acc": 0.62625, "selectivity": -0.09499999999999997, "iid_split_rotation": {"subspace_principal_angle": 0.6467108614757874, "mean_class_cosine": 0.4684027921683622}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.9225, "drop": 0.0}, "paraphrase": {"acc": 0.9051724137931034, "drop": 0.01732758620689656}}, "iid_acc": 0.9225, "selectivity": 0.20125000000000004} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "logreg", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.8525, "drop": 0.0}, "paraphrase": {"acc": 0.8264580369843528, "drop": 0.026041963015647207, "rotation": {"subspace_principal_angle": 1.205280976722454, "mean_class_cosine": 0.3574306478032766}}, "domain": {"acc": 0.85, "drop": 0.0025000000000000577, "rotation": {"subspace_principal_angle": 1.1833422855356077, "mean_class_cosine": 0.37783240518966293}}}, "iid_acc": 0.8525, "selectivity": 0.35375, "iid_split_rotation": {"subspace_principal_angle": 1.1308070932804455, "mean_class_cosine": 0.4259297240456055}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mass_mean", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.515, "drop": 0.0}, "paraphrase": {"acc": 0.5405405405405406, "drop": -0.025540540540540557, "rotation": {"subspace_principal_angle": 1.3060835605730696, "mean_class_cosine": 0.9997239873200385}}, "domain": {"acc": 0.54125, "drop": -0.026249999999999996, "rotation": {"subspace_principal_angle": 1.4432212340779689, "mean_class_cosine": 0.4463453977855254}}}, "iid_acc": 0.515, "selectivity": 0.016249999999999987, "iid_split_rotation": {"subspace_principal_angle": 1.4900771909518655, "mean_class_cosine": 0.9997738232242801}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "probe": "mlp", "seed": 4, "layer": 12, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8122332859174964, "drop": 0.026516714082503556}, "domain": {"acc": 0.835, "drop": 0.003750000000000031}}, "iid_acc": 0.83875, "selectivity": 0.33999999999999997} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "logreg", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8675, "drop": 0.0}, "paraphrase": {"acc": 0.8333333333333334, "drop": 0.03416666666666668, "rotation": {"subspace_principal_angle": 1.3661248872725396, "mean_class_cosine": 0.20324546817811073}}, "domain": {"acc": 0.56875, "drop": 0.29875000000000007, "rotation": {"subspace_principal_angle": 1.424143642480157, "mean_class_cosine": 0.14612757252528197}}, "length": {"acc": 0.8725925925925926, "drop": -0.0050925925925925375, "rotation": {"subspace_principal_angle": 1.304086875035765, "mean_class_cosine": 0.26355866374052267}}}, "iid_acc": 0.8675, "selectivity": 0.34750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.2174020916768917, "mean_class_cosine": 0.34608427795190094}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mass_mean", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.81125, "drop": 0.0}, "paraphrase": {"acc": 0.7941176470588235, "drop": 0.017132352941176543, "rotation": {"subspace_principal_angle": 1.0543437038158343, "mean_class_cosine": 0.9698412280291595}}, "domain": {"acc": 0.5875, "drop": 0.22375, "rotation": {"subspace_principal_angle": 1.5050071211756375, "mean_class_cosine": 0.07636329311882721}}, "length": {"acc": 0.8177777777777778, "drop": -0.006527777777777799, "rotation": {"subspace_principal_angle": 1.0067642663923582, "mean_class_cosine": 0.9690199508303958}}}, "iid_acc": 0.81125, "selectivity": 0.29125, "iid_split_rotation": {"subspace_principal_angle": 1.3667918849667973, "mean_class_cosine": 0.9283366058714894}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "probe": "mlp", "seed": 4, "layer": 20, "dists": {"iid": {"acc": 0.8775, "drop": 0.0}, "paraphrase": {"acc": 0.8725490196078431, "drop": 0.004950980392156801}, "domain": {"acc": 0.53125, "drop": 0.34624999999999995}, "length": {"acc": 0.8785185185185185, "drop": -0.0010185185185185297}}, "iid_acc": 0.8775, "selectivity": 0.35749999999999993} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "logreg", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.89375, "drop": 0.0}, "paraphrase": {"acc": 0.8803030303030303, "drop": 0.01344696969696979, "rotation": {"subspace_principal_angle": 1.3932379682070741, "mean_class_cosine": 0.2875718296573594}}}, "iid_acc": 0.89375, "selectivity": 0.6375000000000001, "iid_split_rotation": {"subspace_principal_angle": 1.3409039719806688, "mean_class_cosine": 0.32901256562331466}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mass_mean", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.5975, "drop": 0.0}, "paraphrase": {"acc": 0.5787878787878787, "drop": 0.01871212121212129, "rotation": {"subspace_principal_angle": 0.6350926877279145, "mean_class_cosine": 0.9549271920995692}}}, "iid_acc": 0.5975, "selectivity": 0.34125000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.562730533540774, "mean_class_cosine": 0.945283431787009}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "probe": "mlp", "seed": 4, "layer": 16, "dists": {"iid": {"acc": 0.88875, "drop": 0.0}, "paraphrase": {"acc": 0.8712121212121212, "drop": 0.017537878787878824}}, "iid_acc": 0.88875, "selectivity": 0.6325000000000001} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "logreg", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.96625, "drop": 0.0}, "paraphrase": {"acc": 0.9586374695863747, "drop": 0.007612530413625307, "rotation": {"subspace_principal_angle": 1.1642922551102963, "mean_class_cosine": 0.6176664792575828}}}, "iid_acc": 0.96625, "selectivity": 0.89625, "iid_split_rotation": {"subspace_principal_angle": 1.0828409682713176, "mean_class_cosine": 0.6471064246781972}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mass_mean", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.83875, "drop": 0.0}, "paraphrase": {"acc": 0.8321167883211679, "drop": 0.006633211678832129, "rotation": {"subspace_principal_angle": 1.5226979113961525, "mean_class_cosine": 0.9230064437182347}}}, "iid_acc": 0.83875, "selectivity": 0.76875, "iid_split_rotation": {"subspace_principal_angle": 1.319128362171639, "mean_class_cosine": 0.9669627199357128}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "probe": "mlp", "seed": 4, "layer": 24, "dists": {"iid": {"acc": 0.9625, "drop": 0.0}, "paraphrase": {"acc": 0.9464720194647201, "drop": 0.016027980535279873}}, "iid_acc": 0.9625, "selectivity": 0.8925000000000001} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.62625, "drop": 0.0}, "paraphrase": {"acc": 0.635890767230169, "drop": -0.009640767230169045, "rotation": {"subspace_principal_angle": 1.2265731114344571, "mean_class_cosine": 0.3374655623320725}}}, "iid_acc": 0.62625, "selectivity": 0.14125, "iid_split_rotation": {"subspace_principal_angle": 1.3292992619939612, "mean_class_cosine": 0.239156515406049}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.48875, "drop": 0.0}, "paraphrase": {"acc": 0.5084525357607282, "drop": -0.019702535760728224, "rotation": {"subspace_principal_angle": 1.5180213673447036, "mean_class_cosine": 0.7337434200066699}}}, "iid_acc": 0.48875, "selectivity": 0.003750000000000031, "iid_split_rotation": {"subspace_principal_angle": 0.9576309376243967, "mean_class_cosine": -0.9826509562242909}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5125, "drop": 0.0}, "paraphrase": {"acc": 0.5149544863459038, "drop": -0.002454486345903839}}, "iid_acc": 0.5125, "selectivity": 0.02749999999999997} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "logreg", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.62125, "drop": 0.0}, "paraphrase": {"acc": 0.5700808625336927, "drop": 0.051169137466307246, "rotation": {"subspace_principal_angle": 1.3744117706906038, "mean_class_cosine": 0.4523473567422777}}}, "iid_acc": 0.62125, "selectivity": 0.33499999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.3212695884395511, "mean_class_cosine": 0.4413545946757409}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mass_mean", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.11875, "drop": 0.0}, "paraphrase": {"acc": 0.11455525606469003, "drop": 0.004194743935309969, "rotation": {"subspace_principal_angle": 1.3057209282680173, "mean_class_cosine": 0.9958332765626513}}}, "iid_acc": 0.11875, "selectivity": -0.1675, "iid_split_rotation": {"subspace_principal_angle": 1.5288056782240633, "mean_class_cosine": 0.6645736917680091}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "probe": "mlp", "seed": 4, "layer": 6, "dists": {"iid": {"acc": 0.555, "drop": 0.0}, "paraphrase": {"acc": 0.5619946091644205, "drop": -0.006994609164420407}}, "iid_acc": 0.555, "selectivity": 0.26875000000000004} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "logreg", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.735, "drop": 0.0}, "paraphrase": {"acc": 0.7275747508305648, "drop": 0.007425249169435211, "rotation": {"subspace_principal_angle": 1.4274592750130581, "mean_class_cosine": 0.14284673359650063}}}, "iid_acc": 0.735, "selectivity": 0.21250000000000002, "iid_split_rotation": {"subspace_principal_angle": 1.4743386316433527, "mean_class_cosine": 0.09630818958862122}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mass_mean", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.56, "drop": 0.0}, "paraphrase": {"acc": 0.5631229235880398, "drop": -0.003122923588039783, "rotation": {"subspace_principal_angle": 1.2083317440135473, "mean_class_cosine": 0.995078207599845}}}, "iid_acc": 0.56, "selectivity": 0.03750000000000009, "iid_split_rotation": {"subspace_principal_angle": 1.2327106774088032, "mean_class_cosine": 0.9974410369073173}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "probe": "mlp", "seed": 4, "layer": 19, "dists": {"iid": {"acc": 0.75625, "drop": 0.0}, "paraphrase": {"acc": 0.7574750830564784, "drop": -0.0012250830564783932}}, "iid_acc": 0.75625, "selectivity": 0.23375} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "logreg", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.6925, "drop": 0.0}, "paraphrase": {"acc": 0.648424543946932, "drop": 0.04407545605306795, "rotation": {"subspace_principal_angle": 1.4115718181974544, "mean_class_cosine": 0.15855257251441335}}}, "iid_acc": 0.6925, "selectivity": 0.20625, "iid_split_rotation": {"subspace_principal_angle": 1.3640799568494573, "mean_class_cosine": 0.20524729008148926}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mass_mean", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.52375, "drop": 0.0}, "paraphrase": {"acc": 0.5223880597014925, "drop": 0.001361940298507558, "rotation": {"subspace_principal_angle": 1.3460720235249697, "mean_class_cosine": 0.9994305298973165}}}, "iid_acc": 0.52375, "selectivity": 0.03750000000000003, "iid_split_rotation": {"subspace_principal_angle": 1.238382906553322, "mean_class_cosine": 0.999840093834897}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "probe": "mlp", "seed": 4, "layer": 4, "dists": {"iid": {"acc": 0.635, "drop": 0.0}, "paraphrase": {"acc": 0.6401326699834162, "drop": -0.005132669983416216}}, "iid_acc": 0.635, "selectivity": 0.14875} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "logreg", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.7175, "drop": 0.0}, "paraphrase": {"acc": 0.6988543371522095, "drop": 0.018645662847790567, "rotation": {"subspace_principal_angle": 1.4345047521354637, "mean_class_cosine": 0.13587002144149035}}}, "iid_acc": 0.7175, "selectivity": 0.14875000000000005, "iid_split_rotation": {"subspace_principal_angle": 1.448693729965042, "mean_class_cosine": 0.12179941742267139}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mass_mean", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.4675, "drop": 0.0}, "paraphrase": {"acc": 0.4779050736497545, "drop": -0.01040507364975446, "rotation": {"subspace_principal_angle": 1.4886165492494845, "mean_class_cosine": -0.5750441624077374}}}, "iid_acc": 0.4675, "selectivity": -0.10124999999999995, "iid_split_rotation": {"subspace_principal_angle": 1.4864385355642995, "mean_class_cosine": 0.995645299950886}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "probe": "mlp", "seed": 4, "layer": 14, "dists": {"iid": {"acc": 0.725, "drop": 0.0}, "paraphrase": {"acc": 0.7217675941080196, "drop": 0.003232405891980372}}, "iid_acc": 0.725, "selectivity": 0.15625} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.9425, "drop": 0.0}, "paraphrase": {"acc": 0.9414965986394558, "drop": 0.0010034013605442205, "rotation": {"subspace_principal_angle": 1.0554320916574222, "mean_class_cosine": 0.49285180754683094}}}, "iid_acc": 0.9425, "selectivity": 0.50875, "iid_split_rotation": {"subspace_principal_angle": 1.0718759182874942, "mean_class_cosine": 0.4784778287309053}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.53375, "drop": 0.0}, "paraphrase": {"acc": 0.5360544217687074, "drop": -0.0023044217687074875, "rotation": {"subspace_principal_angle": 0.6230061876035811, "mean_class_cosine": 0.8605846805322723}}}, "iid_acc": 0.53375, "selectivity": 0.09999999999999992, "iid_split_rotation": {"subspace_principal_angle": 0.7433840996009756, "mean_class_cosine": 0.3607772763702933}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.9325, "drop": 0.0}, "paraphrase": {"acc": 0.9401360544217687, "drop": -0.007636054421768734}}, "iid_acc": 0.9325, "selectivity": 0.49874999999999997} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "logreg", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99375, "drop": 0.0}, "paraphrase": {"acc": 0.9901153212520593, "drop": 0.003634678747940767, "rotation": {"subspace_principal_angle": 1.0513045046339702, "mean_class_cosine": 0.49643906722894926}}}, "iid_acc": 0.99375, "selectivity": 0.14500000000000002, "iid_split_rotation": {"subspace_principal_angle": 0.9685923811624354, "mean_class_cosine": 0.5664600954062968}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mass_mean", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.675, "drop": 0.0}, "paraphrase": {"acc": 0.685337726523888, "drop": -0.01033772652388798, "rotation": {"subspace_principal_angle": 1.3240904740257078, "mean_class_cosine": 0.999937048647147}}}, "iid_acc": 0.675, "selectivity": -0.17374999999999996, "iid_split_rotation": {"subspace_principal_angle": 1.5089394180470712, "mean_class_cosine": 0.9999798272733277}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "probe": "mlp", "seed": 4, "layer": 11, "dists": {"iid": {"acc": 0.99, "drop": 0.0}, "paraphrase": {"acc": 0.9884678747940692, "drop": 0.0015321252059308232}}, "iid_acc": 0.99, "selectivity": 0.14125} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "logreg", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.73125, "drop": 0.0}, "paraphrase": {"acc": 0.7259887005649718, "drop": 0.005261299435028177, "rotation": {"subspace_principal_angle": 1.48146441422371, "mean_class_cosine": 0.08921314568865987}}}, "iid_acc": 0.73125, "selectivity": 0.1087499999999999, "iid_split_rotation": {"subspace_principal_angle": 1.4204367412204741, "mean_class_cosine": 0.1497936706206605}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mass_mean", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.5325, "drop": 0.0}, "paraphrase": {"acc": 0.5141242937853108, "drop": 0.018375706214689203, "rotation": {"subspace_principal_angle": 1.4237156435330802, "mean_class_cosine": -0.529348589575781}}}, "iid_acc": 0.5325, "selectivity": -0.09000000000000008, "iid_split_rotation": {"subspace_principal_angle": 1.0016401557930925, "mean_class_cosine": -0.9687547350705455}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "probe": "mlp", "seed": 4, "layer": 15, "dists": {"iid": {"acc": 0.71875, "drop": 0.0}, "paraphrase": {"acc": 0.7245762711864406, "drop": -0.005826271186440635}}, "iid_acc": 0.71875, "selectivity": 0.09624999999999995} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "logreg", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6730769230769231, "drop": 0.0}, "paraphrase": {"acc": 0.6627218934911243, "drop": 0.010355029585798814, "rotation": {"subspace_principal_angle": 1.3926151506687268, "mean_class_cosine": 0.16905691896718814}}}, "iid_acc": 0.6730769230769231, "selectivity": 0.21153846153846156, "iid_split_rotation": {"subspace_principal_angle": 1.455424707460169, "mean_class_cosine": 0.13141358197406708}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mass_mean", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.5096153846153846, "drop": 0.0}, "paraphrase": {"acc": 0.4378698224852071, "drop": 0.07174556213017746, "rotation": {"subspace_principal_angle": 1.2579697526624707, "mean_class_cosine": 0.8469173188143175}}}, "iid_acc": 0.5096153846153846, "selectivity": 0.04807692307692302, "iid_split_rotation": {"subspace_principal_angle": 1.545165675478419, "mean_class_cosine": 0.5995767861930784}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "probe": "mlp", "seed": 4, "layer": 21, "dists": {"iid": {"acc": 0.6394230769230769, "drop": 0.0}, "paraphrase": {"acc": 0.6153846153846154, "drop": 0.024038461538461453}}, "iid_acc": 0.6394230769230769, "selectivity": 0.1778846153846153} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "logreg", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8922413793103449, "drop": 0.009008620689655134, "rotation": {"subspace_principal_angle": 1.2161263841678638, "mean_class_cosine": 0.34728086917165485}}}, "iid_acc": 0.90125, "selectivity": 0.11250000000000004, "iid_split_rotation": {"subspace_principal_angle": 1.242802856120801, "mean_class_cosine": 0.3221441156385047}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mass_mean", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.5975, "drop": 0.0}, "paraphrase": {"acc": 0.5747126436781609, "drop": 0.022787356321839147, "rotation": {"subspace_principal_angle": 1.0414849280010454, "mean_class_cosine": 0.8882768003569734}}}, "iid_acc": 0.5975, "selectivity": -0.19124999999999992, "iid_split_rotation": {"subspace_principal_angle": 1.075377185201994, "mean_class_cosine": 0.00826524737963924}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "probe": "mlp", "seed": 4, "layer": 10, "dists": {"iid": {"acc": 0.90125, "drop": 0.0}, "paraphrase": {"acc": 0.8836206896551724, "drop": 0.01762931034482762}}, "iid_acc": 0.90125, "selectivity": 0.11250000000000004} diff --git a/repro_bundle/results_tier2/results/predictors.jsonl b/repro_bundle/results_tier2/results/predictors.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..1932dcd50c5b0912b9208511e29762246bcc7990 --- /dev/null +++ b/repro_bundle/results_tier2/results/predictors.jsonl @@ -0,0 +1,490 @@ +{"model": "pythia-70m", "dataset": "sst2", "seed": 0, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921994833, "raptor_stability": 0.8375464081764221, "fragility": 0.25, "augmentation_robustness": 0.5702075197703796, "xie_feature_dispersion": 0.8227753483650888, "pac": 0.7038769639734008, "whitened_cosine_id": 0.8294597268104553}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 0, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998517, "raptor_stability": 0.9239972829818726, "fragility": 1.0, "augmentation_robustness": 0.8889069939502979, "xie_feature_dispersion": 1.0993059236404399, "pac": 0.9064521384660853, "whitened_cosine_id": 0.8396446704864502}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 0, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6360429269373853, "raptor_stability": 0.8923418521881104, "fragility": 1.5, "augmentation_robustness": 0.8610546015390961, "xie_feature_dispersion": 3.4534575785421744, "pac": 0.8766982268636032, "whitened_cosine_id": 0.9029156565666199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 0, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14189668585488155, "raptor_stability": 0.9457845091819763, "fragility": 1.5, "augmentation_robustness": 0.8719582541089957, "xie_feature_dispersion": 5.2630010355712145, "pac": 0.9088713816454861, "whitened_cosine_id": 0.9317190051078796}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 0, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974287, "raptor_stability": 0.7641952633857727, "fragility": 1.0, "augmentation_robustness": 0.6648445683071406, "xie_feature_dispersion": 0.4779918901319739, "pac": 0.7145199158464566, "whitened_cosine_id": 0.6775354743003845}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 0, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5953587445214507, "raptor_stability": 0.8714625239372253, "fragility": 0.5, "augmentation_robustness": 0.7563053805185295, "xie_feature_dispersion": 1.5333196894740788, "pac": 0.8138839522278773, "whitened_cosine_id": 0.8480013012886047}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 0, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994247, "raptor_stability": 0.8230767250061035, "fragility": 0.5, "augmentation_robustness": 0.6879123320435471, "xie_feature_dispersion": 1.779944270255231, "pac": 0.7554945285248253, "whitened_cosine_id": 0.8223811388015747}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 0, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.711632992199055, "raptor_stability": 0.8190855383872986, "fragility": 0.25, "augmentation_robustness": 0.6171877949935642, "xie_feature_dispersion": 2.512801472259043, "pac": 0.7181366666904314, "whitened_cosine_id": 0.7729045748710632}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989692, "raptor_stability": 0.8473518490791321, "fragility": 1.0, "augmentation_robustness": 0.7986513028646275, "xie_feature_dispersion": 1.163949469847951, "pac": 0.8230015759718798, "whitened_cosine_id": 0.7645858526229858}} +{"model": "pythia-70m", "dataset": "subj", "seed": 0, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002171, "raptor_stability": 0.9112874269485474, "fragility": 1.5, "augmentation_robustness": 0.808805287561379, "xie_feature_dispersion": 2.3756762827181857, "pac": 0.8600463572549633, "whitened_cosine_id": 0.9034857153892517}} +{"model": "pythia-70m", "dataset": "spam", "seed": 0, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004674, "raptor_stability": 0.9211491942405701, "fragility": 1.5, "augmentation_robustness": 0.7947530573404484, "xie_feature_dispersion": 10.408354824969607, "pac": 0.8579511257905093, "whitened_cosine_id": 0.9505216479301453}} +{"model": "pythia-70m", "dataset": "cola", "seed": 0, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921980065, "raptor_stability": 0.7842410206794739, "fragility": 1.0, "augmentation_robustness": 0.7059047928449348, "xie_feature_dispersion": 0.5820240387291664, "pac": 0.7450729067622044, "whitened_cosine_id": 0.7026472687721252}} +{"model": "pythia-70m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.7516991834926783, "raptor_stability": 0.8351565003395081, "fragility": 1.0, "augmentation_robustness": 0.783567124817086, "xie_feature_dispersion": 1.9650544929284912, "pac": 0.8093618125782971, "whitened_cosine_id": 0.8273500800132751}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 0, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998, "raptor_stability": 0.8865668177604675, "fragility": 1.5, "augmentation_robustness": 0.7726527290870765, "xie_feature_dispersion": 1.822712974246315, "pac": 0.829609773423772, "whitened_cosine_id": 0.8655645847320557}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350087, "raptor_stability": 0.8526354432106018, "fragility": 0.25, "augmentation_robustness": 0.5730135750873497, "xie_feature_dispersion": 1.4101369914642357, "pac": 0.7128245091489758, "whitened_cosine_id": 0.8668140769004822}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 0, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350252, "raptor_stability": 0.8803290128707886, "fragility": 1.0, "augmentation_robustness": 0.7816552796335788, "xie_feature_dispersion": 1.3391749548314755, "pac": 0.8309921462521836, "whitened_cosine_id": 0.8713250756263733}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3953680487548481, "raptor_stability": 0.8455236554145813, "fragility": 0.25, "augmentation_robustness": 0.7613131386356026, "xie_feature_dispersion": 0.9318287616318928, "pac": 0.8034183970250919, "whitened_cosine_id": 0.8953917026519775}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 0, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10457984207175601, "raptor_stability": 0.9457212686538696, "fragility": 1.5, "augmentation_robustness": 0.8741609981641726, "xie_feature_dispersion": 5.199477497190765, "pac": 0.909941133409021, "whitened_cosine_id": 0.9181342720985413}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 0, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342227, "raptor_stability": 0.7698441743850708, "fragility": 0.25, "augmentation_robustness": 0.6134415015890253, "xie_feature_dispersion": 0.25779876141354974, "pac": 0.6916428379870481, "whitened_cosine_id": 0.7203242182731628}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.574379658862361, "raptor_stability": 0.8597204685211182, "fragility": 0.25, "augmentation_robustness": 0.7115808236459255, "xie_feature_dispersion": 1.6192479136149953, "pac": 0.7856506460835218, "whitened_cosine_id": 0.8496574759483337}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348304, "raptor_stability": 0.8552483320236206, "fragility": 1.0, "augmentation_robustness": 0.735566226119687, "xie_feature_dispersion": 2.2631913339622987, "pac": 0.7954072790716538, "whitened_cosine_id": 0.8250994682312012}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 0, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934627, "raptor_stability": 0.8244509100914001, "fragility": 0.5, "augmentation_robustness": 0.6308774870695307, "xie_feature_dispersion": 1.9587240977086187, "pac": 0.7276641985804655, "whitened_cosine_id": 0.7910383939743042}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934696, "raptor_stability": 0.8365420699119568, "fragility": 1.0, "augmentation_robustness": 0.750756907821624, "xie_feature_dispersion": 1.5342572755652866, "pac": 0.7936494888667904, "whitened_cosine_id": 0.8055051565170288}} +{"model": "pythia-160m", "dataset": "subj", "seed": 0, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351844, "raptor_stability": 0.8689209222793579, "fragility": 1.0, "augmentation_robustness": 0.6997853402308795, "xie_feature_dispersion": 1.77134285145721, "pac": 0.7843531312551186, "whitened_cosine_id": 0.9039409160614014}} +{"model": "pythia-160m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352976, "raptor_stability": 0.943769633769989, "fragility": 3.0, "augmentation_robustness": 0.847956265457393, "xie_feature_dispersion": 6.963799146343127, "pac": 0.895862949613691, "whitened_cosine_id": 0.9455703496932983}} +{"model": "pythia-160m", "dataset": "cola", "seed": 0, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344465, "raptor_stability": 0.7988646626472473, "fragility": 0.25, "augmentation_robustness": 0.5594474756256655, "xie_feature_dispersion": 0.7758972276639047, "pac": 0.6791560691364564, "whitened_cosine_id": 0.7851360440254211}} +{"model": "pythia-160m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.4494122497402506, "raptor_stability": 0.8332012295722961, "fragility": 1.5, "augmentation_robustness": 0.7827926863582254, "xie_feature_dispersion": 2.5953837186456754, "pac": 0.8079969579652608, "whitened_cosine_id": 0.8253653049468994}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 0, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935011, "raptor_stability": 0.866767942905426, "fragility": 0.5, "augmentation_robustness": 0.7083306947816466, "xie_feature_dispersion": 1.5213101554825748, "pac": 0.7875493188435363, "whitened_cosine_id": 0.8781046867370605}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 0, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881098, "raptor_stability": 0.8615295886993408, "fragility": 0.25, "augmentation_robustness": 0.5615804178671275, "xie_feature_dispersion": 1.6884023301086555, "pac": 0.7115550032832342, "whitened_cosine_id": 0.8637628555297852}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881813, "raptor_stability": 0.8620725274085999, "fragility": 1.0, "augmentation_robustness": 0.7514519723873301, "xie_feature_dispersion": 1.923545301649778, "pac": 0.806762249897965, "whitened_cosine_id": 0.8767724633216858}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.32280286192962354, "raptor_stability": 0.8491001725196838, "fragility": 1.0, "augmentation_robustness": 0.7680864251666882, "xie_feature_dispersion": 6.130394237589204, "pac": 0.8085932988431861, "whitened_cosine_id": 0.8844512701034546}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.1116614706117857, "raptor_stability": 0.945551872253418, "fragility": 2.0, "augmentation_robustness": 0.8767712545738563, "xie_feature_dispersion": 7.687914979274747, "pac": 0.9111615634136372, "whitened_cosine_id": 0.8968205451965332}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 0, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879029, "raptor_stability": 0.787240743637085, "fragility": 0.25, "augmentation_robustness": 0.6335049509446696, "xie_feature_dispersion": 0.12206039095754939, "pac": 0.7103728472908772, "whitened_cosine_id": 0.7804538607597351}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 0, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.4882081459032719, "raptor_stability": 0.8632135391235352, "fragility": 0.25, "augmentation_robustness": 0.7301808505878356, "xie_feature_dispersion": 2.173352441480384, "pac": 0.7966971948556854, "whitened_cosine_id": 0.8471425175666809}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 0, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880458, "raptor_stability": 0.8126577138900757, "fragility": 0.25, "augmentation_robustness": 0.6377721250064905, "xie_feature_dispersion": 4.485569514407045, "pac": 0.7252149194482831, "whitened_cosine_id": 0.827558696269989}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 0, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879928, "raptor_stability": 0.7987161874771118, "fragility": 0.25, "augmentation_robustness": 0.59189319563508, "xie_feature_dispersion": 2.5329861784126293, "pac": 0.6953046915560959, "whitened_cosine_id": 0.810127854347229}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 0, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880347, "raptor_stability": 0.814401388168335, "fragility": 0.25, "augmentation_robustness": 0.6407410712281117, "xie_feature_dispersion": 0.7524282237703519, "pac": 0.7275712296982233, "whitened_cosine_id": 0.8191361427307129}} +{"model": "pythia-410m", "dataset": "subj", "seed": 0, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882566, "raptor_stability": 0.8862898945808411, "fragility": 0.5, "augmentation_robustness": 0.7428187785144447, "xie_feature_dispersion": 1.428972872571398, "pac": 0.8145543365476429, "whitened_cosine_id": 0.8959206342697144}} +{"model": "pythia-410m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882899, "raptor_stability": 0.9429970979690552, "fragility": 3.0, "augmentation_robustness": 0.8404945670527885, "xie_feature_dispersion": 6.72893787243834, "pac": 0.8917458325109218, "whitened_cosine_id": 0.933451235294342}} +{"model": "pythia-410m", "dataset": "cola", "seed": 0, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879473, "raptor_stability": 0.7842450737953186, "fragility": 0.25, "augmentation_robustness": 0.4686022827832364, "xie_feature_dispersion": 1.1149557488538653, "pac": 0.6264236782892775, "whitened_cosine_id": 0.8007057905197144}} +{"model": "pythia-410m", "dataset": "stance", "seed": 0, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.5795818589574081, "raptor_stability": 0.8285464644432068, "fragility": 1.0, "augmentation_robustness": 0.7555137866104972, "xie_feature_dispersion": 1.8361604596405292, "pac": 0.792030125526852, "whitened_cosine_id": 0.8221842646598816}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 0, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881513, "raptor_stability": 0.8521958589553833, "fragility": 0.5, "augmentation_robustness": 0.6802066641488503, "xie_feature_dispersion": 1.5630966219328584, "pac": 0.7662012615521168, "whitened_cosine_id": 0.8685176968574524}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.8599012494087219, "fragility": 0.25, "augmentation_robustness": 0.5241726726161993, "xie_feature_dispersion": 2.122662038410848, "pac": 0.6920369610124606, "whitened_cosine_id": 0.8232980370521545}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.825421154499054, "fragility": 1.0, "augmentation_robustness": 0.6705289577483847, "xie_feature_dispersion": 2.576726583434504, "pac": 0.7479750561237193, "whitened_cosine_id": 0.8465724587440491}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 0, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22041139949027433, "raptor_stability": 0.8484893441200256, "fragility": 1.0, "augmentation_robustness": 0.7715857915506817, "xie_feature_dispersion": 5.917035974988243, "pac": 0.8100375678353537, "whitened_cosine_id": 0.8215498328208923}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06228035557873579, "raptor_stability": 0.9404226541519165, "fragility": 3.0, "augmentation_robustness": 0.8674278851786935, "xie_feature_dispersion": 9.493553367505593, "pac": 0.9039252696653051, "whitened_cosine_id": 0.8300829529762268}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 0, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.7848396301269531, "fragility": 0.25, "augmentation_robustness": 0.6320998218394226, "xie_feature_dispersion": 0.2653617366201949, "pac": 0.7084697259831878, "whitened_cosine_id": 0.817557156085968}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.019767617442798265, "raptor_stability": 0.8671453595161438, "fragility": 0.25, "augmentation_robustness": 0.6900615345666754, "xie_feature_dispersion": 2.8877683311134636, "pac": 0.7786034470414096, "whitened_cosine_id": 0.830573558807373}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 0, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010617, "raptor_stability": 0.8155287504196167, "fragility": 0.25, "augmentation_robustness": 0.5947305676343537, "xie_feature_dispersion": 5.342310320027777, "pac": 0.7051296590269852, "whitened_cosine_id": 0.8247697353363037}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 0, "layer": 22, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.7958077192306519, "fragility": 0.25, "augmentation_robustness": 0.5722304723886765, "xie_feature_dispersion": 2.217429056075072, "pac": 0.6840190958096641, "whitened_cosine_id": 0.8194364309310913}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 0, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8089179396629333, "fragility": 0.5, "augmentation_robustness": 0.6530652465185355, "xie_feature_dispersion": 1.5482508686542642, "pac": 0.7309915930907345, "whitened_cosine_id": 0.8221121430397034}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 0, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010872, "raptor_stability": 0.8672937750816345, "fragility": 1.0, "augmentation_robustness": 0.694108081388879, "xie_feature_dispersion": 2.5476390335776844, "pac": 0.7807009282352568, "whitened_cosine_id": 0.8334078788757324}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 0, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010434, "raptor_stability": 0.9216615557670593, "fragility": 1.0, "augmentation_robustness": 0.8105170199587214, "xie_feature_dispersion": 23.432722674772293, "pac": 0.8660892878628903, "whitened_cosine_id": 0.8588989973068237}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 0, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010818, "raptor_stability": 0.8071193099021912, "fragility": 0.25, "augmentation_robustness": 0.4910409126593769, "xie_feature_dispersion": 1.549092757263006, "pac": 0.649080111280784, "whitened_cosine_id": 0.8179035186767578}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 0, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.36672114356076757, "raptor_stability": 0.8304018974304199, "fragility": 1.0, "augmentation_robustness": 0.7571886661476842, "xie_feature_dispersion": 3.593233328751099, "pac": 0.7937952817890521, "whitened_cosine_id": 0.8171507716178894}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8476687073707581, "fragility": 0.5, "augmentation_robustness": 0.6670456540942392, "xie_feature_dispersion": 2.045523026748139, "pac": 0.7573571807324986, "whitened_cosine_id": 0.8353421092033386}} +{"model": "gpt2", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349679, "raptor_stability": 0.8233915567398071, "fragility": 0.25, "augmentation_robustness": 0.4566662273595645, "xie_feature_dispersion": 1.6815976407755642, "pac": 0.6400288920496858, "whitened_cosine_id": 0.857738196849823}} +{"model": "gpt2", "dataset": "imdb", "seed": 0, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350731, "raptor_stability": 0.7829911708831787, "fragility": 0.25, "augmentation_robustness": 0.530870321802199, "xie_feature_dispersion": 1.6579993440971876, "pac": 0.6569307463426888, "whitened_cosine_id": 0.8754125237464905}} +{"model": "gpt2", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.34635368740834577, "raptor_stability": 0.8845524787902832, "fragility": 0.25, "augmentation_robustness": 0.8737520489801199, "xie_feature_dispersion": 0.8678892394256513, "pac": 0.8791522638852016, "whitened_cosine_id": 0.910406231880188}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 0, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.12115672881499416, "raptor_stability": 0.9302389025688171, "fragility": 1.5, "augmentation_robustness": 0.8117314786216752, "xie_feature_dispersion": 6.121815304004312, "pac": 0.8709851905952462, "whitened_cosine_id": 0.9043624997138977}} +{"model": "gpt2", "dataset": "counterfact", "seed": 0, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859340033, "raptor_stability": 0.7690413594245911, "fragility": 0.25, "augmentation_robustness": 0.7125524136549481, "xie_feature_dispersion": 0.1095429604074235, "pac": 0.7407968865397696, "whitened_cosine_id": 0.7072374820709229}} +{"model": "gpt2", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.7366388390202255, "raptor_stability": 0.8725014328956604, "fragility": 0.25, "augmentation_robustness": 0.7768639767211045, "xie_feature_dispersion": 1.77009486868241, "pac": 0.8246827048083825, "whitened_cosine_id": 0.8444538116455078}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347128, "raptor_stability": 0.7693442106246948, "fragility": 0.25, "augmentation_robustness": 0.4900059843963644, "xie_feature_dispersion": 3.694561283935865, "pac": 0.6296750975105296, "whitened_cosine_id": 0.8187122344970703}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 0, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346073, "raptor_stability": 0.7867448925971985, "fragility": 0.25, "augmentation_robustness": 0.6153990422956858, "xie_feature_dispersion": 2.309585917543921, "pac": 0.7010719674464421, "whitened_cosine_id": 0.8086420297622681}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 0, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346306, "raptor_stability": 0.7805789113044739, "fragility": 0.25, "augmentation_robustness": 0.636728510469453, "xie_feature_dispersion": 0.25866709333672444, "pac": 0.7086537108869635, "whitened_cosine_id": 0.803403377532959}} +{"model": "gpt2", "dataset": "subj", "seed": 0, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352106, "raptor_stability": 0.8574187755584717, "fragility": 0.25, "augmentation_robustness": 0.6556398588385755, "xie_feature_dispersion": 0.951143558340801, "pac": 0.7565293171985236, "whitened_cosine_id": 0.9090107679367065}} +{"model": "gpt2", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593527, "raptor_stability": 0.9161890149116516, "fragility": 2.0, "augmentation_robustness": 0.8054413549470808, "xie_feature_dispersion": 6.497801646724056, "pac": 0.8608151849293662, "whitened_cosine_id": 0.9386647939682007}} +{"model": "gpt2", "dataset": "cola", "seed": 0, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859346528, "raptor_stability": 0.7938734889030457, "fragility": 0.25, "augmentation_robustness": 0.46439785386376703, "xie_feature_dispersion": 0.9649847502240455, "pac": 0.6291356713834063, "whitened_cosine_id": 0.807263970375061}} +{"model": "gpt2", "dataset": "stance", "seed": 0, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.490010700608385, "raptor_stability": 0.8106656670570374, "fragility": 0.25, "augmentation_robustness": 0.7245922829104868, "xie_feature_dispersion": 2.664520632491477, "pac": 0.7676289749837621, "whitened_cosine_id": 0.8153484463691711}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935004, "raptor_stability": 0.8306007981300354, "fragility": 0.25, "augmentation_robustness": 0.5872621905723052, "xie_feature_dispersion": 1.1399162300669312, "pac": 0.7089314943511703, "whitened_cosine_id": 0.8771829605102539}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 0, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688191, "raptor_stability": 0.8413395881652832, "fragility": 0.25, "augmentation_robustness": 0.5136767036968108, "xie_feature_dispersion": 1.7719582872902115, "pac": 0.677508145931047, "whitened_cosine_id": 0.8595471382141113}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 0, "layer": 23, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881988, "raptor_stability": 0.9016227722167969, "fragility": 0.25, "augmentation_robustness": 0.7456883958977025, "xie_feature_dispersion": 1.107946359548132, "pac": 0.8236555840572497, "whitened_cosine_id": 0.8640339970588684}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.26407762597782747, "raptor_stability": 0.8502908945083618, "fragility": 1.0, "augmentation_robustness": 0.7617860523249808, "xie_feature_dispersion": 4.268927113799556, "pac": 0.8060384734166712, "whitened_cosine_id": 0.8772361874580383}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.097419516062648, "raptor_stability": 0.928425133228302, "fragility": 1.5, "augmentation_robustness": 0.8139603155889766, "xie_feature_dispersion": 6.700705765199243, "pac": 0.8711927244086393, "whitened_cosine_id": 0.8741976618766785}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 0, "layer": 3, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878758, "raptor_stability": 0.7550429105758667, "fragility": 0.25, "augmentation_robustness": 0.7384688990847975, "xie_feature_dispersion": 0.2304978644612739, "pac": 0.746755904830332, "whitened_cosine_id": 0.7729927897453308}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 0, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6820984754002586, "raptor_stability": 0.8745980262756348, "fragility": 0.25, "augmentation_robustness": 0.713824744543151, "xie_feature_dispersion": 0.8621445356293869, "pac": 0.7942113854093928, "whitened_cosine_id": 0.8492467999458313}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880176, "raptor_stability": 0.8115680813789368, "fragility": 0.25, "augmentation_robustness": 0.5889832766438082, "xie_feature_dispersion": 1.137989115163849, "pac": 0.7002756790113724, "whitened_cosine_id": 0.8243570923805237}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 0, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878862, "raptor_stability": 0.777171790599823, "fragility": 0.25, "augmentation_robustness": 0.5409047648889594, "xie_feature_dispersion": 2.1506698277748155, "pac": 0.6590382777443913, "whitened_cosine_id": 0.8085216283798218}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 0, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880017, "raptor_stability": 0.7901369333267212, "fragility": 0.25, "augmentation_robustness": 0.5844201102203882, "xie_feature_dispersion": 0.462184243542827, "pac": 0.6872785217735546, "whitened_cosine_id": 0.8134031891822815}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 0, "layer": 24, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688252, "raptor_stability": 0.8832855224609375, "fragility": 0.25, "augmentation_robustness": 0.7291403967687851, "xie_feature_dispersion": 0.4906864607581503, "pac": 0.8062129596148613, "whitened_cosine_id": 0.8952683210372925}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882066, "raptor_stability": 0.914933979511261, "fragility": 1.0, "augmentation_robustness": 0.8072328905323266, "xie_feature_dispersion": 4.701874448406257, "pac": 0.8610834350217937, "whitened_cosine_id": 0.915591835975647}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 0, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880159, "raptor_stability": 0.7916848063468933, "fragility": 0.25, "augmentation_robustness": 0.4476010668333125, "xie_feature_dispersion": 0.79552886696272, "pac": 0.6196429365901029, "whitened_cosine_id": 0.8178503513336182}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 0, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.42249695846338925, "raptor_stability": 0.8128862977027893, "fragility": 0.5, "augmentation_robustness": 0.7398084898037611, "xie_feature_dispersion": 2.2610696467251814, "pac": 0.7763473937532752, "whitened_cosine_id": 0.7861542701721191}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 0, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881549, "raptor_stability": 0.8164692521095276, "fragility": 0.25, "augmentation_robustness": 0.5951602317050955, "xie_feature_dispersion": 1.1005465582305016, "pac": 0.7058147419073115, "whitened_cosine_id": 0.8596013784408569}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648414, "raptor_stability": 0.8818655014038086, "fragility": 0.25, "augmentation_robustness": 0.6289366997740752, "xie_feature_dispersion": 1.8204152059660144, "pac": 0.7554011005889418, "whitened_cosine_id": 0.879472017288208}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 0, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.29387292376502, "raptor_stability": 0.8775492906570435, "fragility": 1.0, "augmentation_robustness": 0.7836625681170646, "xie_feature_dispersion": 2.218615090585379, "pac": 0.830605929387054, "whitened_cosine_id": 0.8882995247840881}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.33450754332538074, "raptor_stability": 0.8566421270370483, "fragility": 1.0, "augmentation_robustness": 0.7926578157046991, "xie_feature_dispersion": 3.7561628051141023, "pac": 0.8246499713708737, "whitened_cosine_id": 0.9043422937393188}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07399960414320991, "raptor_stability": 0.9483939409255981, "fragility": 1.5, "augmentation_robustness": 0.881505837440225, "xie_feature_dispersion": 7.327642739764713, "pac": 0.9149498891829115, "whitened_cosine_id": 0.9209960699081421}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 0, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237645932, "raptor_stability": 0.8066194653511047, "fragility": 0.25, "augmentation_robustness": 0.7512795380738843, "xie_feature_dispersion": 0.06081582657176685, "pac": 0.7789495017124946, "whitened_cosine_id": 0.7630186080932617}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6131208141710327, "raptor_stability": 0.8868826031684875, "fragility": 0.25, "augmentation_robustness": 0.848475526665586, "xie_feature_dispersion": 1.992273238599221, "pac": 0.8676790649170367, "whitened_cosine_id": 0.8511517643928528}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645996, "raptor_stability": 0.8185003995895386, "fragility": 0.25, "augmentation_robustness": 0.6978679311233282, "xie_feature_dispersion": 4.466765724760196, "pac": 0.7581841653564334, "whitened_cosine_id": 0.8152397871017456}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 0, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647149, "raptor_stability": 0.811665415763855, "fragility": 0.25, "augmentation_robustness": 0.6512162425704373, "xie_feature_dispersion": 2.8843177774025395, "pac": 0.7314408291671461, "whitened_cosine_id": 0.8048377633094788}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 0, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647015, "raptor_stability": 0.8171485662460327, "fragility": 0.25, "augmentation_robustness": 0.7362926278030426, "xie_feature_dispersion": 0.15381271196675575, "pac": 0.7767205970245377, "whitened_cosine_id": 0.8043486475944519}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 0, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651123, "raptor_stability": 0.8921712040901184, "fragility": 0.5, "augmentation_robustness": 0.7485023298249472, "xie_feature_dispersion": 0.9955751494789675, "pac": 0.8203367669575328, "whitened_cosine_id": 0.9154570698738098}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 0, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923765166, "raptor_stability": 0.9529117941856384, "fragility": 3.0, "augmentation_robustness": 0.87503706844258, "xie_feature_dispersion": 9.854132791030557, "pac": 0.9139744313141092, "whitened_cosine_id": 0.9523190855979919}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 0, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764728, "raptor_stability": 0.8035885691642761, "fragility": 0.25, "augmentation_robustness": 0.5005855255082153, "xie_feature_dispersion": 1.0666210514085848, "pac": 0.6520870473362457, "whitened_cosine_id": 0.818631649017334}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 0, "layer": 20, "concept": "stance", "predictors": {"sip_eigengap": 0.4157006657150525, "raptor_stability": 0.8240557312965393, "fragility": 0.5, "augmentation_robustness": 0.7331066076432546, "xie_feature_dispersion": 3.234317390572246, "pac": 0.778581169469897, "whitened_cosine_id": 0.8179960250854492}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 0, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649436, "raptor_stability": 0.8530629873275757, "fragility": 0.25, "augmentation_robustness": 0.7025335558592776, "xie_feature_dispersion": 1.1742203784808831, "pac": 0.7777982715934266, "whitened_cosine_id": 0.8717560768127441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 1, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921994895, "raptor_stability": 0.8314626216888428, "fragility": 0.25, "augmentation_robustness": 0.5808293731010495, "xie_feature_dispersion": 1.6689369031429147, "pac": 0.7061459973949462, "whitened_cosine_id": 0.824091911315918}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998223, "raptor_stability": 0.8874242901802063, "fragility": 1.0, "augmentation_robustness": 0.8228104146559615, "xie_feature_dispersion": 1.271550857614664, "pac": 0.855117352418084, "whitened_cosine_id": 0.8205482363700867}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6806422452118264, "raptor_stability": 0.8989484906196594, "fragility": 1.5, "augmentation_robustness": 0.8787513256973658, "xie_feature_dispersion": 3.4157347611498774, "pac": 0.8888499081585126, "whitened_cosine_id": 0.9129272699356079}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 1, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.15525918027531763, "raptor_stability": 0.9445532560348511, "fragility": 1.5, "augmentation_robustness": 0.8628763882540946, "xie_feature_dispersion": 5.228730991204411, "pac": 0.9037148221444729, "whitened_cosine_id": 0.9321261644363403}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 1, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974072, "raptor_stability": 0.7612232565879822, "fragility": 0.25, "augmentation_robustness": 0.6405111879943068, "xie_feature_dispersion": 0.7015570516775577, "pac": 0.7008672222911445, "whitened_cosine_id": 0.6881313323974609}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 1, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5841926037551581, "raptor_stability": 0.8477641940116882, "fragility": 0.5, "augmentation_robustness": 0.6474137169579742, "xie_feature_dispersion": 1.484633799742414, "pac": 0.7475889554848312, "whitened_cosine_id": 0.8390121459960938}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 1, "layer": 3, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994224, "raptor_stability": 0.8421536087989807, "fragility": 0.5, "augmentation_robustness": 0.7239927602573658, "xie_feature_dispersion": 2.981639036084935, "pac": 0.7830731845281733, "whitened_cosine_id": 0.8215951323509216}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 1, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987687, "raptor_stability": 0.819472074508667, "fragility": 1.0, "augmentation_robustness": 0.6380691012431728, "xie_feature_dispersion": 2.073055712645641, "pac": 0.7287705878759199, "whitened_cosine_id": 0.7341217398643494}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 1, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989095, "raptor_stability": 0.843525767326355, "fragility": 1.0, "augmentation_robustness": 0.8007193546782, "xie_feature_dispersion": 1.5447695946615063, "pac": 0.8221225610022775, "whitened_cosine_id": 0.7792574167251587}} +{"model": "pythia-70m", "dataset": "subj", "seed": 1, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002709, "raptor_stability": 0.8913278579711914, "fragility": 1.0, "augmentation_robustness": 0.7332091297611866, "xie_feature_dispersion": 1.980053369512756, "pac": 0.812268493866189, "whitened_cosine_id": 0.9182124137878418}} +{"model": "pythia-70m", "dataset": "spam", "seed": 1, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922003908, "raptor_stability": 0.9203832745552063, "fragility": 1.5, "augmentation_robustness": 0.7922849095394744, "xie_feature_dispersion": 10.611895676124647, "pac": 0.8563340920473403, "whitened_cosine_id": 0.950851321220398}} +{"model": "pythia-70m", "dataset": "cola", "seed": 1, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.7986350655555725, "fragility": 1.0, "augmentation_robustness": 0.7043344717903556, "xie_feature_dispersion": 0.7104026437376014, "pac": 0.751484768672964, "whitened_cosine_id": 0.7400056719779968}} +{"model": "pythia-70m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6754527167162487, "raptor_stability": 0.8382856249809265, "fragility": 1.0, "augmentation_robustness": 0.8019897778297428, "xie_feature_dispersion": 1.8257872743628552, "pac": 0.8201377014053346, "whitened_cosine_id": 0.8341732621192932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 1, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997764, "raptor_stability": 0.8679760098457336, "fragility": 0.5, "augmentation_robustness": 0.7205293551358967, "xie_feature_dispersion": 0.9711058139858004, "pac": 0.7942526824908152, "whitened_cosine_id": 0.8618689775466919}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 1, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349757, "raptor_stability": 0.8496716618537903, "fragility": 0.25, "augmentation_robustness": 0.5842367312683088, "xie_feature_dispersion": 1.8689219377096642, "pac": 0.7169541965610495, "whitened_cosine_id": 0.8571608662605286}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.39754248593502, "raptor_stability": 0.8464771509170532, "fragility": 0.5, "augmentation_robustness": 0.6850682519540424, "xie_feature_dispersion": 1.7078889299200177, "pac": 0.7657727014355478, "whitened_cosine_id": 0.8704335689544678}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 1, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.3976585069526139, "raptor_stability": 0.8461608290672302, "fragility": 1.0, "augmentation_robustness": 0.7765248733340492, "xie_feature_dispersion": 2.8951389779509134, "pac": 0.8113428512006398, "whitened_cosine_id": 0.9028950929641724}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.12810819210433996, "raptor_stability": 0.9377708435058594, "fragility": 3.0, "augmentation_robustness": 0.833317612225739, "xie_feature_dispersion": 10.50628418134277, "pac": 0.8855442278657992, "whitened_cosine_id": 0.9117757678031921}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 1, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.397542485934175, "raptor_stability": 0.7833874821662903, "fragility": 0.25, "augmentation_robustness": 0.6108469106562627, "xie_feature_dispersion": 0.3790863394835801, "pac": 0.6971171964112766, "whitened_cosine_id": 0.7308042049407959}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 1, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.5499358318667639, "raptor_stability": 0.8611683249473572, "fragility": 0.25, "augmentation_robustness": 0.7033012402006117, "xie_feature_dispersion": 1.9883915877863603, "pac": 0.7822347825739844, "whitened_cosine_id": 0.8481037616729736}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 1, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347079, "raptor_stability": 0.8096680641174316, "fragility": 0.25, "augmentation_robustness": 0.6360886444392633, "xie_feature_dispersion": 2.3558677799388796, "pac": 0.7228783542783475, "whitened_cosine_id": 0.8148756623268127}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 1, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345162, "raptor_stability": 0.7992299795150757, "fragility": 0.25, "augmentation_robustness": 0.5915203833323763, "xie_feature_dispersion": 3.2725825993057094, "pac": 0.695375181423726, "whitened_cosine_id": 0.7864978909492493}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 1, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345495, "raptor_stability": 0.8322464823722839, "fragility": 1.0, "augmentation_robustness": 0.768225719172241, "xie_feature_dispersion": 1.8790072833719011, "pac": 0.8002361007722625, "whitened_cosine_id": 0.7750239968299866}} +{"model": "pythia-160m", "dataset": "subj", "seed": 1, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352137, "raptor_stability": 0.8875827193260193, "fragility": 1.0, "augmentation_robustness": 0.7169304038534543, "xie_feature_dispersion": 1.9523142841304229, "pac": 0.8022565615897368, "whitened_cosine_id": 0.9101552963256836}} +{"model": "pythia-160m", "dataset": "spam", "seed": 1, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859347847, "raptor_stability": 0.913774847984314, "fragility": 1.0, "augmentation_robustness": 0.8208256677071334, "xie_feature_dispersion": 14.489404204949587, "pac": 0.8673002578457236, "whitened_cosine_id": 0.93967604637146}} +{"model": "pythia-160m", "dataset": "cola", "seed": 1, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859342416, "raptor_stability": 0.7926198244094849, "fragility": 1.0, "augmentation_robustness": 0.6447861391769918, "xie_feature_dispersion": 0.9165363040019758, "pac": 0.7187029817932383, "whitened_cosine_id": 0.7564545273780823}} +{"model": "pythia-160m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6971690127369661, "raptor_stability": 0.8396480679512024, "fragility": 1.0, "augmentation_robustness": 0.7875725989177652, "xie_feature_dispersion": 2.4375720598203166, "pac": 0.8136103334344837, "whitened_cosine_id": 0.8327767848968506}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 1, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349908, "raptor_stability": 0.8594450950622559, "fragility": 0.5, "augmentation_robustness": 0.6977040298979517, "xie_feature_dispersion": 0.7745815425512376, "pac": 0.7785745624801037, "whitened_cosine_id": 0.8690269589424133}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 1, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688131, "raptor_stability": 0.8608813881874084, "fragility": 0.25, "augmentation_robustness": 0.5666047397490934, "xie_feature_dispersion": 2.4202065461787297, "pac": 0.7137430639682509, "whitened_cosine_id": 0.8642251491546631}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 1, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881893, "raptor_stability": 0.8481273055076599, "fragility": 1.0, "augmentation_robustness": 0.7130798265252153, "xie_feature_dispersion": 2.320769849966657, "pac": 0.7806035660164377, "whitened_cosine_id": 0.8769779801368713}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 1, "layer": 13, "concept": "topic", "predictors": {"sip_eigengap": 0.30704864737844834, "raptor_stability": 0.8611399531364441, "fragility": 1.0, "augmentation_robustness": 0.7903941411680551, "xie_feature_dispersion": 3.4011674827208287, "pac": 0.8257670471522496, "whitened_cosine_id": 0.8893525004386902}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 1, "layer": 19, "concept": "topic", "predictors": {"sip_eigengap": 0.08244206897104008, "raptor_stability": 0.9409602880477905, "fragility": 2.0, "augmentation_robustness": 0.8566799506256193, "xie_feature_dispersion": 8.808736850172968, "pac": 0.8988201193367049, "whitened_cosine_id": 0.891150951385498}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 1, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878758, "raptor_stability": 0.7924683690071106, "fragility": 0.25, "augmentation_robustness": 0.6334062400656503, "xie_feature_dispersion": 0.6634691181873394, "pac": 0.7129373045363805, "whitened_cosine_id": 0.7721397280693054}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 1, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.519816512768835, "raptor_stability": 0.8667904734611511, "fragility": 0.25, "augmentation_robustness": 0.7478177367365705, "xie_feature_dispersion": 2.8557298314061703, "pac": 0.8073041050988607, "whitened_cosine_id": 0.8457608222961426}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 1, "layer": 10, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956878145, "raptor_stability": 0.8067510724067688, "fragility": 0.25, "augmentation_robustness": 0.6562650597345253, "xie_feature_dispersion": 5.470678455409298, "pac": 0.731508066070647, "whitened_cosine_id": 0.8200975656509399}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 1, "layer": 21, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879326, "raptor_stability": 0.7820147275924683, "fragility": 0.25, "augmentation_robustness": 0.587986767048995, "xie_feature_dispersion": 2.5969424874908773, "pac": 0.6850007473207316, "whitened_cosine_id": 0.7967674136161804}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 1, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879308, "raptor_stability": 0.7950331568717957, "fragility": 0.25, "augmentation_robustness": 0.6646175791835158, "xie_feature_dispersion": 2.059673087772778, "pac": 0.7298253680276557, "whitened_cosine_id": 0.8044969439506531}} +{"model": "pythia-410m", "dataset": "subj", "seed": 1, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882595, "raptor_stability": 0.8856076598167419, "fragility": 0.5, "augmentation_robustness": 0.7139391967127188, "xie_feature_dispersion": 1.5476544469738573, "pac": 0.7997734282647304, "whitened_cosine_id": 0.8976815342903137}} +{"model": "pythia-410m", "dataset": "spam", "seed": 1, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956874568, "raptor_stability": 0.9077131152153015, "fragility": 0.5, "augmentation_robustness": 0.8020763670174436, "xie_feature_dispersion": 17.967343252945344, "pac": 0.8548947411163725, "whitened_cosine_id": 0.9222725629806519}} +{"model": "pythia-410m", "dataset": "cola", "seed": 1, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879513, "raptor_stability": 0.7959128022193909, "fragility": 0.25, "augmentation_robustness": 0.45896954560148306, "xie_feature_dispersion": 0.6838657969146573, "pac": 0.627441173910437, "whitened_cosine_id": 0.8098257780075073}} +{"model": "pythia-410m", "dataset": "stance", "seed": 1, "layer": 19, "concept": "stance", "predictors": {"sip_eigengap": 0.520575334062942, "raptor_stability": 0.8262770771980286, "fragility": 0.5, "augmentation_robustness": 0.7565572182499646, "xie_feature_dispersion": 2.363460584912377, "pac": 0.7914171477239966, "whitened_cosine_id": 0.8227141499519348}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 1, "layer": 15, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881407, "raptor_stability": 0.8368763327598572, "fragility": 0.5, "augmentation_robustness": 0.6726761343337394, "xie_feature_dispersion": 0.8861988690029199, "pac": 0.7547762335467982, "whitened_cosine_id": 0.8615521788597107}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010796, "raptor_stability": 0.8518760800361633, "fragility": 0.25, "augmentation_robustness": 0.5469758605057219, "xie_feature_dispersion": 3.1855570820407504, "pac": 0.6994259702709427, "whitened_cosine_id": 0.8287058472633362}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8483022451400757, "fragility": 1.0, "augmentation_robustness": 0.6553477126801958, "xie_feature_dispersion": 2.6155304376742743, "pac": 0.7518249789101357, "whitened_cosine_id": 0.8468784689903259}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22290771506820922, "raptor_stability": 0.8515928983688354, "fragility": 1.5, "augmentation_robustness": 0.7746267660688861, "xie_feature_dispersion": 5.927804802925069, "pac": 0.8131098322188608, "whitened_cosine_id": 0.8236908912658691}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.058822548518337014, "raptor_stability": 0.9422677159309387, "fragility": 2.0, "augmentation_robustness": 0.8595274869237916, "xie_feature_dispersion": 8.534984669917208, "pac": 0.9008976014273651, "whitened_cosine_id": 0.8359106779098511}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.7920423746109009, "fragility": 0.25, "augmentation_robustness": 0.6413142308147114, "xie_feature_dispersion": 0.860340632017618, "pac": 0.7166783027128061, "whitened_cosine_id": 0.8168047666549683}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 1, "layer": 9, "concept": "emotion", "predictors": {"sip_eigengap": 0.4313770796442383, "raptor_stability": 0.8590365052223206, "fragility": 0.25, "augmentation_robustness": 0.694955833506802, "xie_feature_dispersion": 3.6526929613427925, "pac": 0.7769961693645613, "whitened_cosine_id": 0.8269678950309753}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 1, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 0.855816496101069, "raptor_stability": 0.810732364654541, "fragility": 0.25, "augmentation_robustness": 0.6270267466287328, "xie_feature_dispersion": 6.292509564537482, "pac": 0.718879555641637, "whitened_cosine_id": 0.8283994793891907}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 1, "layer": 13, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010526, "raptor_stability": 0.8113015294075012, "fragility": 0.25, "augmentation_robustness": 0.6088162548597921, "xie_feature_dispersion": 3.2324734107222866, "pac": 0.7100588921336466, "whitened_cosine_id": 0.8244034051895142}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 1, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010755, "raptor_stability": 0.791317880153656, "fragility": 0.25, "augmentation_robustness": 0.6080442076224152, "xie_feature_dispersion": 2.2409906655814193, "pac": 0.6996810438880356, "whitened_cosine_id": 0.818917989730835}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 1, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010871, "raptor_stability": 0.8893104195594788, "fragility": 1.0, "augmentation_robustness": 0.696504331799587, "xie_feature_dispersion": 2.547582289724496, "pac": 0.7929073756795328, "whitened_cosine_id": 0.832260251045227}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 1, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010578, "raptor_stability": 0.9184004664421082, "fragility": 1.0, "augmentation_robustness": 0.7929347080961948, "xie_feature_dispersion": 23.49076580939176, "pac": 0.8556675872691515, "whitened_cosine_id": 0.8580988049507141}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 1, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961009699, "raptor_stability": 0.8054239749908447, "fragility": 0.25, "augmentation_robustness": 0.46022627880181477, "xie_feature_dispersion": 0.7167139402414644, "pac": 0.6328251268963297, "whitened_cosine_id": 0.8232492804527283}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 1, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.4012148911144739, "raptor_stability": 0.8381373882293701, "fragility": 1.0, "augmentation_robustness": 0.7598258349622733, "xie_feature_dispersion": 2.6386343651247124, "pac": 0.7989816115958217, "whitened_cosine_id": 0.8214442133903503}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 1, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010833, "raptor_stability": 0.8418669700622559, "fragility": 1.0, "augmentation_robustness": 0.6524348508596854, "xie_feature_dispersion": 1.6761850609669111, "pac": 0.7471509104609706, "whitened_cosine_id": 0.8340412378311157}} +{"model": "gpt2", "dataset": "sst2", "seed": 1, "layer": 8, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485935016, "raptor_stability": 0.8106276392936707, "fragility": 0.25, "augmentation_robustness": 0.49014044266582973, "xie_feature_dispersion": 1.6638828788327318, "pac": 0.6503840409797502, "whitened_cosine_id": 0.8658217191696167}} +{"model": "gpt2", "dataset": "imdb", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350674, "raptor_stability": 0.8155258893966675, "fragility": 0.25, "augmentation_robustness": 0.582233966857331, "xie_feature_dispersion": 1.679121568516715, "pac": 0.6988799281269993, "whitened_cosine_id": 0.8795225024223328}} +{"model": "gpt2", "dataset": "ag_news", "seed": 1, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.31952355566470164, "raptor_stability": 0.8801834583282471, "fragility": 0.25, "augmentation_robustness": 0.8715217181573922, "xie_feature_dispersion": 0.9325824128470682, "pac": 0.8758525882428196, "whitened_cosine_id": 0.9088584184646606}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.13768748508218773, "raptor_stability": 0.9254063367843628, "fragility": 1.5, "augmentation_robustness": 0.7871717482295453, "xie_feature_dispersion": 5.392898240224899, "pac": 0.856289042506954, "whitened_cosine_id": 0.9057890176773071}} +{"model": "gpt2", "dataset": "counterfact", "seed": 1, "layer": 7, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343133, "raptor_stability": 0.7852461338043213, "fragility": 0.25, "augmentation_robustness": 0.6622392322460525, "xie_feature_dispersion": 0.5647271084995943, "pac": 0.723742683025187, "whitened_cosine_id": 0.7448322176933289}} +{"model": "gpt2", "dataset": "emotion", "seed": 1, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5436635722625862, "raptor_stability": 0.8717889785766602, "fragility": 0.25, "augmentation_robustness": 0.7592733911032505, "xie_feature_dispersion": 2.524501993095808, "pac": 0.8155311848399553, "whitened_cosine_id": 0.8443379998207092}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348176, "raptor_stability": 0.8326132297515869, "fragility": 0.25, "augmentation_robustness": 0.7215291300706207, "xie_feature_dispersion": 0.8694719793739331, "pac": 0.7770711799111039, "whitened_cosine_id": 0.8273347020149231}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 1, "layer": 10, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346122, "raptor_stability": 0.7642949819564819, "fragility": 0.25, "augmentation_robustness": 0.5635955618554486, "xie_feature_dispersion": 2.167418535028266, "pac": 0.6639452719059653, "whitened_cosine_id": 0.7877233624458313}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 1, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346217, "raptor_stability": 0.7664420008659363, "fragility": 0.25, "augmentation_robustness": 0.4934067076281954, "xie_feature_dispersion": 1.3905365467341602, "pac": 0.6299243542470658, "whitened_cosine_id": 0.7959058284759521}} +{"model": "gpt2", "dataset": "subj", "seed": 1, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352203, "raptor_stability": 0.853984534740448, "fragility": 0.5, "augmentation_robustness": 0.6319009208667878, "xie_feature_dispersion": 1.4255917684674573, "pac": 0.7429427278036179, "whitened_cosine_id": 0.9104035496711731}} +{"model": "gpt2", "dataset": "spam", "seed": 1, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353107, "raptor_stability": 0.9109200239181519, "fragility": 1.0, "augmentation_robustness": 0.7966288421897234, "xie_feature_dispersion": 14.38720114784055, "pac": 0.8537744330539376, "whitened_cosine_id": 0.9343566298484802}} +{"model": "gpt2", "dataset": "cola", "seed": 1, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859345384, "raptor_stability": 0.7919601202011108, "fragility": 0.25, "augmentation_robustness": 0.5130744044422636, "xie_feature_dispersion": 0.34394405556746344, "pac": 0.6525172623216873, "whitened_cosine_id": 0.7920956015586853}} +{"model": "gpt2", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.5721762371971069, "raptor_stability": 0.8337757587432861, "fragility": 0.5, "augmentation_robustness": 0.7456673050417738, "xie_feature_dispersion": 1.1954515513989168, "pac": 0.7897215318925299, "whitened_cosine_id": 0.8244547843933105}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 1, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349934, "raptor_stability": 0.8276537656784058, "fragility": 0.25, "augmentation_robustness": 0.6092711286506372, "xie_feature_dispersion": 0.655742490317081, "pac": 0.7184624471645215, "whitened_cosine_id": 0.8719987273216248}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.21030729568816, "raptor_stability": 0.8340166807174683, "fragility": 0.25, "augmentation_robustness": 0.5218252189937355, "xie_feature_dispersion": 1.7679503238709116, "pac": 0.6779209498556018, "whitened_cosine_id": 0.8651541471481323}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 1, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882157, "raptor_stability": 0.8258817195892334, "fragility": 0.25, "augmentation_robustness": 0.5752874121442174, "xie_feature_dispersion": 1.3475744664628955, "pac": 0.7005845658667254, "whitened_cosine_id": 0.8705449104309082}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 1, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.26782873777331245, "raptor_stability": 0.841077983379364, "fragility": 1.0, "augmentation_robustness": 0.7470633913167877, "xie_feature_dispersion": 4.180838494364298, "pac": 0.7940706873480758, "whitened_cosine_id": 0.8797165155410767}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.10399119928589491, "raptor_stability": 0.9248352646827698, "fragility": 1.5, "augmentation_robustness": 0.7945295897496373, "xie_feature_dispersion": 5.61182173262905, "pac": 0.8596824272162036, "whitened_cosine_id": 0.8758760094642639}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 1, "layer": 14, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878629, "raptor_stability": 0.7786945700645447, "fragility": 0.25, "augmentation_robustness": 0.6144804383339777, "xie_feature_dispersion": 0.5748324075475697, "pac": 0.6965875041992612, "whitened_cosine_id": 0.7664059400558472}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5592522550238864, "raptor_stability": 0.8738617300987244, "fragility": 0.25, "augmentation_robustness": 0.720919459890958, "xie_feature_dispersion": 0.9112899816215909, "pac": 0.7973905949948412, "whitened_cosine_id": 0.8513531684875488}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 1, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880974, "raptor_stability": 0.8126772046089172, "fragility": 0.25, "augmentation_robustness": 0.7050632297874023, "xie_feature_dispersion": 1.322075514358211, "pac": 0.7588702171981598, "whitened_cosine_id": 0.8305314779281616}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 1, "layer": 13, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879664, "raptor_stability": 0.7759368419647217, "fragility": 0.25, "augmentation_robustness": 0.5622012822384077, "xie_feature_dispersion": 2.3188187063045684, "pac": 0.6690690621015647, "whitened_cosine_id": 0.8039388656616211}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 1, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880128, "raptor_stability": 0.7699597477912903, "fragility": 0.25, "augmentation_robustness": 0.6476643493242761, "xie_feature_dispersion": 0.7754531474850158, "pac": 0.7088120485577831, "whitened_cosine_id": 0.8113746047019958}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 1, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882637, "raptor_stability": 0.8628992438316345, "fragility": 1.0, "augmentation_robustness": 0.6397751392051434, "xie_feature_dispersion": 2.0316047842765594, "pac": 0.751337191518389, "whitened_cosine_id": 0.8833309412002563}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 1, "layer": 16, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882794, "raptor_stability": 0.9118919968605042, "fragility": 1.5, "augmentation_robustness": 0.7830820986182032, "xie_feature_dispersion": 13.506990469815198, "pac": 0.8474870477393537, "whitened_cosine_id": 0.908318042755127}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 1, "layer": 16, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880065, "raptor_stability": 0.8000022768974304, "fragility": 0.25, "augmentation_robustness": 0.4811244638624909, "xie_feature_dispersion": 0.37533178694301084, "pac": 0.6405633703799607, "whitened_cosine_id": 0.8167566061019897}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 1, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.4372306810871382, "raptor_stability": 0.8121774792671204, "fragility": 0.5, "augmentation_robustness": 0.7197757150389671, "xie_feature_dispersion": 1.6812559297796246, "pac": 0.7659765971530437, "whitened_cosine_id": 0.7655757069587708}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 1, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881318, "raptor_stability": 0.8076351881027222, "fragility": 0.25, "augmentation_robustness": 0.6171652138408624, "xie_feature_dispersion": 0.9086262406903829, "pac": 0.7124002009717922, "whitened_cosine_id": 0.8581160306930542}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648992, "raptor_stability": 0.8649455904960632, "fragility": 0.25, "augmentation_robustness": 0.6131424180860278, "xie_feature_dispersion": 2.15799302448076, "pac": 0.7390440042910456, "whitened_cosine_id": 0.8817294239997864}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 1, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650475, "raptor_stability": 0.8865694999694824, "fragility": 1.0, "augmentation_robustness": 0.7989302272108586, "xie_feature_dispersion": 2.4686831296081597, "pac": 0.8427498635901705, "whitened_cosine_id": 0.8952975869178772}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 1, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.32396935001382504, "raptor_stability": 0.8518233299255371, "fragility": 1.0, "augmentation_robustness": 0.7712947063545464, "xie_feature_dispersion": 4.51435777726234, "pac": 0.8115590181400418, "whitened_cosine_id": 0.902996301651001}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.0634212938889773, "raptor_stability": 0.9348612427711487, "fragility": 3.0, "augmentation_robustness": 0.8289800859783713, "xie_feature_dispersion": 10.742096589376187, "pac": 0.8819206643747599, "whitened_cosine_id": 0.8965702056884766}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646287, "raptor_stability": 0.8297770023345947, "fragility": 0.25, "augmentation_robustness": 0.7628796605466598, "xie_feature_dispersion": 0.5968347455995571, "pac": 0.7963283314406273, "whitened_cosine_id": 0.7946618795394897}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 1, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.5646054321397148, "raptor_stability": 0.877227246761322, "fragility": 0.25, "augmentation_robustness": 0.8188598337200217, "xie_feature_dispersion": 2.8243621674803006, "pac": 0.8480435402406719, "whitened_cosine_id": 0.849128246307373}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 1, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237648017, "raptor_stability": 0.7998038530349731, "fragility": 0.25, "augmentation_robustness": 0.6147531583223901, "xie_feature_dispersion": 5.29862303981088, "pac": 0.7072785056786817, "whitened_cosine_id": 0.8292998671531677}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 1, "layer": 10, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237642246, "raptor_stability": 0.8075653314590454, "fragility": 0.25, "augmentation_robustness": 0.6484486881494425, "xie_feature_dispersion": 3.06991914353266, "pac": 0.7280070098042439, "whitened_cosine_id": 0.801740288734436}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 1, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237646085, "raptor_stability": 0.8232854008674622, "fragility": 0.25, "augmentation_robustness": 0.7361198321668313, "xie_feature_dispersion": 2.0361420955712815, "pac": 0.7797026165171468, "whitened_cosine_id": 0.8059210181236267}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 1, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237650999, "raptor_stability": 0.9056962132453918, "fragility": 0.5, "augmentation_robustness": 0.7644724917188301, "xie_feature_dispersion": 1.206422317161318, "pac": 0.835084352482111, "whitened_cosine_id": 0.9149098992347717}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 1, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237633957, "raptor_stability": 0.9107893109321594, "fragility": 0.5, "augmentation_robustness": 0.8112238006915025, "xie_feature_dispersion": 18.322389911250944, "pac": 0.8610065558118309, "whitened_cosine_id": 0.9440927505493164}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 1, "layer": 18, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764712, "raptor_stability": 0.8045408129692078, "fragility": 0.25, "augmentation_robustness": 0.48280717514059407, "xie_feature_dispersion": 0.5417228937870184, "pac": 0.6436739940549009, "whitened_cosine_id": 0.8147454857826233}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 1, "layer": 20, "concept": "stance", "predictors": {"sip_eigengap": 0.4812620447442306, "raptor_stability": 0.8222463130950928, "fragility": 0.5, "augmentation_robustness": 0.7353172970670624, "xie_feature_dispersion": 2.449239368783959, "pac": 0.7787818050810775, "whitened_cosine_id": 0.8223068118095398}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 1, "layer": 17, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764944, "raptor_stability": 0.8373395800590515, "fragility": 0.25, "augmentation_robustness": 0.6654539817611203, "xie_feature_dispersion": 0.6952088262981461, "pac": 0.7513967809100859, "whitened_cosine_id": 0.8643338680267334}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921995475, "raptor_stability": 0.826153576374054, "fragility": 0.25, "augmentation_robustness": 0.5550600140566192, "xie_feature_dispersion": 1.2685202535875624, "pac": 0.6906067952153365, "whitened_cosine_id": 0.8190510272979736}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998612, "raptor_stability": 0.8727290630340576, "fragility": 1.0, "augmentation_robustness": 0.8023565468625505, "xie_feature_dispersion": 1.1304006682636438, "pac": 0.837542804948304, "whitened_cosine_id": 0.8432037830352783}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 2, "layer": 1, "concept": "topic", "predictors": {"sip_eigengap": 0.6563267159870434, "raptor_stability": 0.9060066938400269, "fragility": 1.5, "augmentation_robustness": 0.8994135427836584, "xie_feature_dispersion": 3.2940888567741142, "pac": 0.9027101183118427, "whitened_cosine_id": 0.9018473625183105}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 2, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.1409559124523405, "raptor_stability": 0.9160086512565613, "fragility": 0.5, "augmentation_robustness": 0.8052056873288177, "xie_feature_dispersion": 1.129241758791748, "pac": 0.8606071692926895, "whitened_cosine_id": 0.9244332909584045}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 2, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921978088, "raptor_stability": 0.7689433693885803, "fragility": 0.5, "augmentation_robustness": 0.664331639648433, "xie_feature_dispersion": 0.4906246722259355, "pac": 0.7166375045185067, "whitened_cosine_id": 0.6844558715820312}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6049173382330815, "raptor_stability": 0.8735766410827637, "fragility": 1.0, "augmentation_robustness": 0.7882025082836114, "xie_feature_dispersion": 1.5346222774203366, "pac": 0.8308895746831875, "whitened_cosine_id": 0.8501214385032654}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 2, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921991698, "raptor_stability": 0.8122056126594543, "fragility": 0.5, "augmentation_robustness": 0.6586846606652739, "xie_feature_dispersion": 1.9447361037026398, "pac": 0.7354451366623641, "whitened_cosine_id": 0.7998017072677612}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987974, "raptor_stability": 0.8192676901817322, "fragility": 1.0, "augmentation_robustness": 0.6847445561526596, "xie_feature_dispersion": 1.857559017759172, "pac": 0.752006123167196, "whitened_cosine_id": 0.7365460991859436}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 2, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989806, "raptor_stability": 0.7996385097503662, "fragility": 0.5, "augmentation_robustness": 0.673138203990154, "xie_feature_dispersion": 1.2956329589634799, "pac": 0.73638835687026, "whitened_cosine_id": 0.7707058787345886}} +{"model": "pythia-70m", "dataset": "subj", "seed": 2, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.711632992200211, "raptor_stability": 0.8931885957717896, "fragility": 1.5, "augmentation_robustness": 0.7723343164195375, "xie_feature_dispersion": 2.385960236411005, "pac": 0.8327614560956635, "whitened_cosine_id": 0.9023992419242859}} +{"model": "pythia-70m", "dataset": "spam", "seed": 2, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004394, "raptor_stability": 0.9427627921104431, "fragility": 1.5, "augmentation_robustness": 0.8262282460478666, "xie_feature_dispersion": 10.434188689895425, "pac": 0.8844955190791548, "whitened_cosine_id": 0.9575247764587402}} +{"model": "pythia-70m", "dataset": "cola", "seed": 2, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921978055, "raptor_stability": 0.7819207310676575, "fragility": 1.0, "augmentation_robustness": 0.6892368500077494, "xie_feature_dispersion": 0.5849615167254145, "pac": 0.7355787905377034, "whitened_cosine_id": 0.7173330187797546}} +{"model": "pythia-70m", "dataset": "stance", "seed": 2, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.467708916223557, "raptor_stability": 0.823031485080719, "fragility": 1.0, "augmentation_robustness": 0.7788880873984344, "xie_feature_dispersion": 1.6869566033029149, "pac": 0.8009597862395768, "whitened_cosine_id": 0.8191723823547363}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 2, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998568, "raptor_stability": 0.8905608654022217, "fragility": 1.5, "augmentation_robustness": 0.7918130449642642, "xie_feature_dispersion": 1.7648002764851518, "pac": 0.8411869551832429, "whitened_cosine_id": 0.8727987408638}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 2, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349053, "raptor_stability": 0.8420531153678894, "fragility": 0.25, "augmentation_robustness": 0.5578391048029369, "xie_feature_dispersion": 1.9830546413363874, "pac": 0.6999461100854132, "whitened_cosine_id": 0.8557217121124268}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 2, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350019, "raptor_stability": 0.8570092916488647, "fragility": 1.0, "augmentation_robustness": 0.7611505060754636, "xie_feature_dispersion": 1.332977638352146, "pac": 0.8090798988621641, "whitened_cosine_id": 0.8642057180404663}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 2, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.5427793527803491, "raptor_stability": 0.8923298716545105, "fragility": 1.0, "augmentation_robustness": 0.8591541728510862, "xie_feature_dispersion": 3.279140555736114, "pac": 0.8757420222527983, "whitened_cosine_id": 0.9014958739280701}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.10936523074513556, "raptor_stability": 0.938740074634552, "fragility": 2.0, "augmentation_robustness": 0.8369704413137592, "xie_feature_dispersion": 8.126249426404561, "pac": 0.8878552579741557, "whitened_cosine_id": 0.9102428555488586}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 2, "layer": 5, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342813, "raptor_stability": 0.7747091054916382, "fragility": 0.25, "augmentation_robustness": 0.6417974566818402, "xie_feature_dispersion": 0.2710043288534134, "pac": 0.7082532810867392, "whitened_cosine_id": 0.7429371476173401}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 2, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.536126639231569, "raptor_stability": 0.8551614880561829, "fragility": 0.25, "augmentation_robustness": 0.7039385026715385, "xie_feature_dispersion": 2.1343450512794986, "pac": 0.7795499953638607, "whitened_cosine_id": 0.8513872623443604}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 2, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859346717, "raptor_stability": 0.8405966758728027, "fragility": 1.0, "augmentation_robustness": 0.7310447997932741, "xie_feature_dispersion": 2.1154557972849912, "pac": 0.7858207378330384, "whitened_cosine_id": 0.7988827228546143}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934474, "raptor_stability": 0.809672474861145, "fragility": 0.25, "augmentation_robustness": 0.6381545155767033, "xie_feature_dispersion": 3.2921137785738233, "pac": 0.7239134952189241, "whitened_cosine_id": 0.7898976802825928}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 2, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345826, "raptor_stability": 0.8115760087966919, "fragility": 0.25, "augmentation_robustness": 0.6743335197367628, "xie_feature_dispersion": 2.0708787972712512, "pac": 0.7429547642667274, "whitened_cosine_id": 0.7857599854469299}} +{"model": "pythia-160m", "dataset": "subj", "seed": 2, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352095, "raptor_stability": 0.8895910382270813, "fragility": 1.0, "augmentation_robustness": 0.7135129309774878, "xie_feature_dispersion": 1.9799099583102993, "pac": 0.8015519846022845, "whitened_cosine_id": 0.9134976267814636}} +{"model": "pythia-160m", "dataset": "spam", "seed": 2, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353311, "raptor_stability": 0.923969566822052, "fragility": 3.0, "augmentation_robustness": 0.7712028495795479, "xie_feature_dispersion": 10.474610541637501, "pac": 0.8475862082007999, "whitened_cosine_id": 0.9371472001075745}} +{"model": "pythia-160m", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934449, "raptor_stability": 0.7818026542663574, "fragility": 0.25, "augmentation_robustness": 0.5589858962588444, "xie_feature_dispersion": 0.36909274612641374, "pac": 0.6703942752626009, "whitened_cosine_id": 0.7660055756568909}} +{"model": "pythia-160m", "dataset": "stance", "seed": 2, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.550133319968111, "raptor_stability": 0.8319461345672607, "fragility": 1.5, "augmentation_robustness": 0.7869835159029682, "xie_feature_dispersion": 2.567878497868356, "pac": 0.8094648252351144, "whitened_cosine_id": 0.8248243927955627}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350152, "raptor_stability": 0.8500860929489136, "fragility": 0.5, "augmentation_robustness": 0.7049551414527425, "xie_feature_dispersion": 1.015287990938809, "pac": 0.777520617200828, "whitened_cosine_id": 0.8699454069137573}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880925, "raptor_stability": 0.8507713079452515, "fragility": 0.25, "augmentation_robustness": 0.56023098369132, "xie_feature_dispersion": 2.5382673428676514, "pac": 0.7055011458182857, "whitened_cosine_id": 0.8639923334121704}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881815, "raptor_stability": 0.8507294654846191, "fragility": 1.0, "augmentation_robustness": 0.6945310589611032, "xie_feature_dispersion": 2.1999878871216194, "pac": 0.7726302622228611, "whitened_cosine_id": 0.876013994216919}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 2, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.26365195418554105, "raptor_stability": 0.8327406644821167, "fragility": 1.0, "augmentation_robustness": 0.7542952777812159, "xie_feature_dispersion": 6.585161526633954, "pac": 0.7935179711316662, "whitened_cosine_id": 0.8844253420829773}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 2, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.09434498750832387, "raptor_stability": 0.9433990716934204, "fragility": 3.0, "augmentation_robustness": 0.8439835278167953, "xie_feature_dispersion": 13.538073364650904, "pac": 0.8936912997551079, "whitened_cosine_id": 0.882512092590332}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 2, "layer": 21, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878156, "raptor_stability": 0.7554473280906677, "fragility": 0.25, "augmentation_robustness": 0.5795980822750073, "xie_feature_dispersion": 0.25302338379103195, "pac": 0.6675227051828375, "whitened_cosine_id": 0.7533318400382996}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 2, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.4286496963727474, "raptor_stability": 0.8515519499778748, "fragility": 0.25, "augmentation_robustness": 0.7217973601571539, "xie_feature_dispersion": 2.4117533415561345, "pac": 0.7866746550675143, "whitened_cosine_id": 0.8424076437950134}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 2, "layer": 10, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880234, "raptor_stability": 0.8253244757652283, "fragility": 0.25, "augmentation_robustness": 0.6525579225683095, "xie_feature_dispersion": 3.286387642880714, "pac": 0.7389411991667689, "whitened_cosine_id": 0.8348305821418762}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295687971, "raptor_stability": 0.7986416220664978, "fragility": 0.5, "augmentation_robustness": 0.6102543792217037, "xie_feature_dispersion": 1.8402433518812902, "pac": 0.7044480006441007, "whitened_cosine_id": 0.7969480156898499}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 2, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687956, "raptor_stability": 0.8041142225265503, "fragility": 0.25, "augmentation_robustness": 0.6348348918631781, "xie_feature_dispersion": 2.8843251781307306, "pac": 0.7194745571948642, "whitened_cosine_id": 0.8083853721618652}} +{"model": "pythia-410m", "dataset": "subj", "seed": 2, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882677, "raptor_stability": 0.8861598372459412, "fragility": 1.0, "augmentation_robustness": 0.7199515086420826, "xie_feature_dispersion": 1.5647630291131758, "pac": 0.8030556729440119, "whitened_cosine_id": 0.9003214240074158}} +{"model": "pythia-410m", "dataset": "spam", "seed": 2, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882948, "raptor_stability": 0.9486513733863831, "fragility": 3.0, "augmentation_robustness": 0.8147735901930376, "xie_feature_dispersion": 8.118654593687403, "pac": 0.8817124817897104, "whitened_cosine_id": 0.933245062828064}} +{"model": "pythia-410m", "dataset": "cola", "seed": 2, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879588, "raptor_stability": 0.801864504814148, "fragility": 0.25, "augmentation_robustness": 0.5568185846503796, "xie_feature_dispersion": 0.47412534702494263, "pac": 0.6793415447322637, "whitened_cosine_id": 0.8032734394073486}} +{"model": "pythia-410m", "dataset": "stance", "seed": 2, "layer": 10, "concept": "stance", "predictors": {"sip_eigengap": 0.47549258334344463, "raptor_stability": 0.8262801170349121, "fragility": 0.5, "augmentation_robustness": 0.7669381088312476, "xie_feature_dispersion": 3.2823490583418935, "pac": 0.7966091129330799, "whitened_cosine_id": 0.8195136189460754}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 2, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881449, "raptor_stability": 0.8327568769454956, "fragility": 0.5, "augmentation_robustness": 0.6987375826417322, "xie_feature_dispersion": 1.0417000979829214, "pac": 0.7657472297936139, "whitened_cosine_id": 0.864019513130188}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 0.855816496101061, "raptor_stability": 0.8567962050437927, "fragility": 0.25, "augmentation_robustness": 0.5671640624141537, "xie_feature_dispersion": 2.9934403270968586, "pac": 0.7119801337289733, "whitened_cosine_id": 0.8268300294876099}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010875, "raptor_stability": 0.8453766107559204, "fragility": 1.0, "augmentation_robustness": 0.6545753307454034, "xie_feature_dispersion": 2.1868653519527914, "pac": 0.7499759707506619, "whitened_cosine_id": 0.8455048203468323}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 2, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.25959190862405174, "raptor_stability": 0.8485456705093384, "fragility": 1.5, "augmentation_robustness": 0.7696896788348585, "xie_feature_dispersion": 6.11360094217724, "pac": 0.8091176746720985, "whitened_cosine_id": 0.82154780626297}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 2, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.054741897344356236, "raptor_stability": 0.9400308728218079, "fragility": 3.0, "augmentation_robustness": 0.8422214207799361, "xie_feature_dispersion": 25.830217797796784, "pac": 0.891126146800872, "whitened_cosine_id": 0.8212482333183289}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.7967193126678467, "fragility": 0.25, "augmentation_robustness": 0.6376161223358435, "xie_feature_dispersion": 0.32762786809964206, "pac": 0.7171677175018452, "whitened_cosine_id": 0.8186400532722473}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 2, "layer": 14, "concept": "emotion", "predictors": {"sip_eigengap": 0.3939989620104987, "raptor_stability": 0.8490006923675537, "fragility": 0.5, "augmentation_robustness": 0.6994232263413324, "xie_feature_dispersion": 2.8039065456029375, "pac": 0.774211959354443, "whitened_cosine_id": 0.8225671648979187}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 2, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010765, "raptor_stability": 0.8104034066200256, "fragility": 0.25, "augmentation_robustness": 0.6280345727884112, "xie_feature_dispersion": 3.5385168087713166, "pac": 0.7192189897042185, "whitened_cosine_id": 0.8256248831748962}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010356, "raptor_stability": 0.8042046427726746, "fragility": 0.25, "augmentation_robustness": 0.5798721253653247, "xie_feature_dispersion": 4.885251661924462, "pac": 0.6920383840689996, "whitened_cosine_id": 0.8246341943740845}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 2, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010591, "raptor_stability": 0.7971631288528442, "fragility": 0.25, "augmentation_robustness": 0.6109212963398408, "xie_feature_dispersion": 3.0958681822178455, "pac": 0.7040422125963426, "whitened_cosine_id": 0.825272798538208}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 2, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010877, "raptor_stability": 0.8769533038139343, "fragility": 1.0, "augmentation_robustness": 0.686316534527354, "xie_feature_dispersion": 2.781727808686068, "pac": 0.7816349191706442, "whitened_cosine_id": 0.8311170339584351}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 2, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010558, "raptor_stability": 0.9453728795051575, "fragility": 1.0, "augmentation_robustness": 0.820275837399458, "xie_feature_dispersion": 23.518955158553926, "pac": 0.8828243584523077, "whitened_cosine_id": 0.8724843859672546}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 2, "layer": 13, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010875, "raptor_stability": 0.8096633553504944, "fragility": 0.25, "augmentation_robustness": 0.4820399124790271, "xie_feature_dispersion": 0.6272978464384222, "pac": 0.6458516339147607, "whitened_cosine_id": 0.8209949731826782}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 2, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.33865901445727387, "raptor_stability": 0.8298296332359314, "fragility": 1.0, "augmentation_robustness": 0.7626508775947324, "xie_feature_dispersion": 3.230859359898638, "pac": 0.7962402554153318, "whitened_cosine_id": 0.8170763850212097}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 2, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010808, "raptor_stability": 0.8269712924957275, "fragility": 1.0, "augmentation_robustness": 0.6705085135623837, "xie_feature_dispersion": 1.7905194345731703, "pac": 0.7487399030290556, "whitened_cosine_id": 0.8267220854759216}} +{"model": "gpt2", "dataset": "sst2", "seed": 2, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349719, "raptor_stability": 0.8054147958755493, "fragility": 0.25, "augmentation_robustness": 0.452677434814117, "xie_feature_dispersion": 1.9950046355062208, "pac": 0.6290461153448331, "whitened_cosine_id": 0.8579275012016296}} +{"model": "gpt2", "dataset": "imdb", "seed": 2, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350525, "raptor_stability": 0.8147286772727966, "fragility": 0.25, "augmentation_robustness": 0.5420364534350757, "xie_feature_dispersion": 1.5795679944440126, "pac": 0.6783825653539362, "whitened_cosine_id": 0.8795053362846375}} +{"model": "gpt2", "dataset": "ag_news", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.37606835826916657, "raptor_stability": 0.831476092338562, "fragility": 1.0, "augmentation_robustness": 0.7160835130077355, "xie_feature_dispersion": 3.9948278981935426, "pac": 0.7737798026731488, "whitened_cosine_id": 0.8991539478302002}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 2, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.1066838897374052, "raptor_stability": 0.926025390625, "fragility": 1.5, "augmentation_robustness": 0.7998744401134587, "xie_feature_dispersion": 5.30333228141077, "pac": 0.8629499153692293, "whitened_cosine_id": 0.9026139974594116}} +{"model": "gpt2", "dataset": "counterfact", "seed": 2, "layer": 5, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342753, "raptor_stability": 0.7606139779090881, "fragility": 0.25, "augmentation_robustness": 0.6853486887621179, "xie_feature_dispersion": 0.33002840995944466, "pac": 0.722981333335603, "whitened_cosine_id": 0.73723304271698}} +{"model": "gpt2", "dataset": "emotion", "seed": 2, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5291604831105917, "raptor_stability": 0.8539204597473145, "fragility": 0.25, "augmentation_robustness": 0.6936184867435239, "xie_feature_dispersion": 2.010880210817857, "pac": 0.7737694732454192, "whitened_cosine_id": 0.852493941783905}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 2, "layer": 3, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859346874, "raptor_stability": 0.7885839939117432, "fragility": 0.25, "augmentation_robustness": 0.6401645091390288, "xie_feature_dispersion": 2.7009497954700477, "pac": 0.7143742515253859, "whitened_cosine_id": 0.8214506506919861}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 2, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859344792, "raptor_stability": 0.7899907827377319, "fragility": 0.25, "augmentation_robustness": 0.6988157056142623, "xie_feature_dispersion": 1.1126097391994054, "pac": 0.7444032441759971, "whitened_cosine_id": 0.7636452317237854}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 2, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346346, "raptor_stability": 0.8108178973197937, "fragility": 0.25, "augmentation_robustness": 0.7288817265620731, "xie_feature_dispersion": 0.9421790599719136, "pac": 0.7698498119409334, "whitened_cosine_id": 0.7841159701347351}} +{"model": "gpt2", "dataset": "subj", "seed": 2, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935236, "raptor_stability": 0.8722711205482483, "fragility": 0.5, "augmentation_robustness": 0.671483073623656, "xie_feature_dispersion": 1.095645923787708, "pac": 0.7718770970859521, "whitened_cosine_id": 0.9157454967498779}} +{"model": "gpt2", "dataset": "spam", "seed": 2, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935328, "raptor_stability": 0.9371232390403748, "fragility": 1.0, "augmentation_robustness": 0.796134201470319, "xie_feature_dispersion": 13.868407438115556, "pac": 0.8666287202553469, "whitened_cosine_id": 0.9425339698791504}} +{"model": "gpt2", "dataset": "cola", "seed": 2, "layer": 8, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344947, "raptor_stability": 0.7808783650398254, "fragility": 0.25, "augmentation_robustness": 0.4942423131561597, "xie_feature_dispersion": 0.38598825222094824, "pac": 0.6375603390979926, "whitened_cosine_id": 0.7848895788192749}} +{"model": "gpt2", "dataset": "stance", "seed": 2, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.5282860197072011, "raptor_stability": 0.8207774758338928, "fragility": 0.25, "augmentation_robustness": 0.7408748851031105, "xie_feature_dispersion": 2.2792793583875905, "pac": 0.7808261804685017, "whitened_cosine_id": 0.8131434321403503}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 2, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350183, "raptor_stability": 0.8160192370414734, "fragility": 0.25, "augmentation_robustness": 0.5989172462486304, "xie_feature_dispersion": 0.7330282513183846, "pac": 0.707468241645052, "whitened_cosine_id": 0.8756778240203857}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881522, "raptor_stability": 0.8280349969863892, "fragility": 0.25, "augmentation_robustness": 0.5057969177918578, "xie_feature_dispersion": 2.142674805730929, "pac": 0.6669159573891235, "whitened_cosine_id": 0.8551852107048035}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 2, "layer": 17, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882153, "raptor_stability": 0.8116974234580994, "fragility": 0.25, "augmentation_robustness": 0.5686588039011631, "xie_feature_dispersion": 1.607819045435618, "pac": 0.6901781136796312, "whitened_cosine_id": 0.8731050491333008}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 2, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.31591694292321826, "raptor_stability": 0.8379892706871033, "fragility": 1.0, "augmentation_robustness": 0.742913021052252, "xie_feature_dispersion": 4.204689835168313, "pac": 0.7904511458696777, "whitened_cosine_id": 0.8770087361335754}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 2, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.09406374121141926, "raptor_stability": 0.9260669350624084, "fragility": 1.5, "augmentation_robustness": 0.7954546313237745, "xie_feature_dispersion": 6.374717972535422, "pac": 0.8607607831930915, "whitened_cosine_id": 0.8749038577079773}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 2, "layer": 15, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878642, "raptor_stability": 0.7624801993370056, "fragility": 0.25, "augmentation_robustness": 0.6618500550845569, "xie_feature_dispersion": 0.3458533220193969, "pac": 0.7121651272107812, "whitened_cosine_id": 0.7618997693061829}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5517110732147434, "raptor_stability": 0.8649501204490662, "fragility": 0.25, "augmentation_robustness": 0.7080522584774469, "xie_feature_dispersion": 0.9242099717833849, "pac": 0.7865011894632565, "whitened_cosine_id": 0.8501837849617004}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 2, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879795, "raptor_stability": 0.8065077662467957, "fragility": 0.25, "augmentation_robustness": 0.6075362981901802, "xie_feature_dispersion": 2.637665840049803, "pac": 0.7070220322184879, "whitened_cosine_id": 0.8130406141281128}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 2, "layer": 24, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879593, "raptor_stability": 0.7929912805557251, "fragility": 0.25, "augmentation_robustness": 0.687746633761177, "xie_feature_dispersion": 1.7178207547218787, "pac": 0.7403689571584511, "whitened_cosine_id": 0.8002586364746094}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 2, "layer": 10, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879999, "raptor_stability": 0.7896746397018433, "fragility": 0.25, "augmentation_robustness": 0.6055377963859948, "xie_feature_dispersion": 2.184985172419946, "pac": 0.6976062180439191, "whitened_cosine_id": 0.8094580769538879}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 2, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882808, "raptor_stability": 0.8819090127944946, "fragility": 0.5, "augmentation_robustness": 0.671921924564452, "xie_feature_dispersion": 1.4175250970378874, "pac": 0.7769154686794733, "whitened_cosine_id": 0.8918017148971558}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 2, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956883163, "raptor_stability": 0.9394904971122742, "fragility": 1.0, "augmentation_robustness": 0.789939038367017, "xie_feature_dispersion": 13.48331423128814, "pac": 0.8647147677396456, "whitened_cosine_id": 0.917801022529602}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 2, "layer": 20, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880179, "raptor_stability": 0.7839298248291016, "fragility": 0.25, "augmentation_robustness": 0.5005989571125445, "xie_feature_dispersion": 0.4452074705477342, "pac": 0.642264390970823, "whitened_cosine_id": 0.8129088282585144}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 2, "layer": 19, "concept": "stance", "predictors": {"sip_eigengap": 0.42059041720080603, "raptor_stability": 0.8081520199775696, "fragility": 0.5, "augmentation_robustness": 0.7271794999740994, "xie_feature_dispersion": 2.1040215143183647, "pac": 0.7676657599758345, "whitened_cosine_id": 0.7815761566162109}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 2, "layer": 11, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881302, "raptor_stability": 0.800855815410614, "fragility": 0.25, "augmentation_robustness": 0.6448667243445377, "xie_feature_dispersion": 0.9884816369621833, "pac": 0.7228612698775758, "whitened_cosine_id": 0.8472304344177246}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237649451, "raptor_stability": 0.848286509513855, "fragility": 0.25, "augmentation_robustness": 0.5634090093483565, "xie_feature_dispersion": 2.429164692187356, "pac": 0.7058477594311057, "whitened_cosine_id": 0.8725934624671936}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923765047, "raptor_stability": 0.8775961995124817, "fragility": 1.0, "augmentation_robustness": 0.774986382310033, "xie_feature_dispersion": 2.073190764464318, "pac": 0.8262912909112574, "whitened_cosine_id": 0.8976234793663025}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 2, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.39588586732534015, "raptor_stability": 0.8559584021568298, "fragility": 1.0, "augmentation_robustness": 0.7820646549848398, "xie_feature_dispersion": 3.6437178512980712, "pac": 0.8190115285708348, "whitened_cosine_id": 0.9048267006874084}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 2, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.058756522874283214, "raptor_stability": 0.9501736760139465, "fragility": 1.5, "augmentation_robustness": 0.8683291628685149, "xie_feature_dispersion": 6.332238643735282, "pac": 0.9092514194412307, "whitened_cosine_id": 0.9210717082023621}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 2, "layer": 14, "concept": "truth", "predictors": {"sip_eigengap": 1.29387292376465, "raptor_stability": 0.8335102796554565, "fragility": 0.25, "augmentation_robustness": 0.7324644634343902, "xie_feature_dispersion": 0.2367801772803424, "pac": 0.7829873715449234, "whitened_cosine_id": 0.7776349186897278}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 2, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.5021817224915929, "raptor_stability": 0.8779242038726807, "fragility": 0.25, "augmentation_robustness": 0.8391776121680871, "xie_feature_dispersion": 2.253764943264769, "pac": 0.8585509080203839, "whitened_cosine_id": 0.848435640335083}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 2, "layer": 7, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237646702, "raptor_stability": 0.8149078488349915, "fragility": 0.25, "augmentation_robustness": 0.6656494198846684, "xie_feature_dispersion": 3.241751193008559, "pac": 0.74027863435983, "whitened_cosine_id": 0.8149601221084595}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 2, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237645712, "raptor_stability": 0.8159747123718262, "fragility": 0.25, "augmentation_robustness": 0.6734582597657116, "xie_feature_dispersion": 3.9245245325964895, "pac": 0.744716486068769, "whitened_cosine_id": 0.7945672273635864}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647373, "raptor_stability": 0.8471121788024902, "fragility": 1.5, "augmentation_robustness": 0.7818095671541662, "xie_feature_dispersion": 2.234140033744398, "pac": 0.8144608729783283, "whitened_cosine_id": 0.8059022426605225}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 2, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651356, "raptor_stability": 0.8970006108283997, "fragility": 0.5, "augmentation_robustness": 0.7567696692336531, "xie_feature_dispersion": 1.2621909015930055, "pac": 0.8268851400310264, "whitened_cosine_id": 0.9196778535842896}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 2, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237647655, "raptor_stability": 0.9248191714286804, "fragility": 1.0, "augmentation_robustness": 0.82350992677322, "xie_feature_dispersion": 18.125591348203603, "pac": 0.8741645491009502, "whitened_cosine_id": 0.9478391408920288}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 2, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647155, "raptor_stability": 0.7994158864021301, "fragility": 0.25, "augmentation_robustness": 0.5431886973103349, "xie_feature_dispersion": 0.28432650457876973, "pac": 0.6713022918562326, "whitened_cosine_id": 0.8063071370124817}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 2, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.47687488644054804, "raptor_stability": 0.8371041417121887, "fragility": 0.25, "augmentation_robustness": 0.797300656029293, "xie_feature_dispersion": 2.6985372907022853, "pac": 0.8172023988707409, "whitened_cosine_id": 0.8233371376991272}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 2, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764932, "raptor_stability": 0.8735290765762329, "fragility": 0.25, "augmentation_robustness": 0.7483144226433699, "xie_feature_dispersion": 0.37743185068024515, "pac": 0.8109217496098013, "whitened_cosine_id": 0.8706146478652954}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996696, "raptor_stability": 0.8488397598266602, "fragility": 0.25, "augmentation_robustness": 0.5179823378479093, "xie_feature_dispersion": 2.9434633106744124, "pac": 0.6834110488372847, "whitened_cosine_id": 0.8510316610336304}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 3, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921997957, "raptor_stability": 0.9214267730712891, "fragility": 1.0, "augmentation_robustness": 0.8748364094526995, "xie_feature_dispersion": 1.0370053025918007, "pac": 0.8981315912619943, "whitened_cosine_id": 0.8452136516571045}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 3, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5774925151839428, "raptor_stability": 0.8593819737434387, "fragility": 1.5, "augmentation_robustness": 0.7534819327566528, "xie_feature_dispersion": 4.6426103470172615, "pac": 0.8064319532500457, "whitened_cosine_id": 0.914084255695343}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 3, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.13231466369730585, "raptor_stability": 0.9453434348106384, "fragility": 2.0, "augmentation_robustness": 0.8580053912222981, "xie_feature_dispersion": 7.46273693031987, "pac": 0.9016744130164682, "whitened_cosine_id": 0.930860161781311}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 3, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921978637, "raptor_stability": 0.7810012698173523, "fragility": 1.0, "augmentation_robustness": 0.7041587287479656, "xie_feature_dispersion": 0.5352386761634802, "pac": 0.7425799992826589, "whitened_cosine_id": 0.6924300193786621}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.7155300362438014, "raptor_stability": 0.8798012733459473, "fragility": 1.0, "augmentation_robustness": 0.7953262771468999, "xie_feature_dispersion": 1.4738735759860853, "pac": 0.8375637752464236, "whitened_cosine_id": 0.8543421626091003}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.711632992199115, "raptor_stability": 0.8559250831604004, "fragility": 1.0, "augmentation_robustness": 0.7877299458146467, "xie_feature_dispersion": 1.5138918067092333, "pac": 0.8218275144875236, "whitened_cosine_id": 0.7800295948982239}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 3, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921988689, "raptor_stability": 0.8289263248443604, "fragility": 1.0, "augmentation_robustness": 0.6909281594330826, "xie_feature_dispersion": 1.8985337702191092, "pac": 0.7599272421387215, "whitened_cosine_id": 0.7646318078041077}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 3, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.711632992198979, "raptor_stability": 0.8319928050041199, "fragility": 1.0, "augmentation_robustness": 0.7511852191768399, "xie_feature_dispersion": 1.1045209668465537, "pac": 0.7915890120904798, "whitened_cosine_id": 0.7830063700675964}} +{"model": "pythia-70m", "dataset": "subj", "seed": 3, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002202, "raptor_stability": 0.9258168935775757, "fragility": 1.5, "augmentation_robustness": 0.8378077411249061, "xie_feature_dispersion": 2.402632179496979, "pac": 0.8818123173512409, "whitened_cosine_id": 0.9089578986167908}} +{"model": "pythia-70m", "dataset": "spam", "seed": 3, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004663, "raptor_stability": 0.9411624073982239, "fragility": 3.0, "augmentation_robustness": 0.8491174045572158, "xie_feature_dispersion": 8.333983109527576, "pac": 0.8951399059777199, "whitened_cosine_id": 0.9507425427436829}} +{"model": "pythia-70m", "dataset": "cola", "seed": 3, "layer": 4, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921976148, "raptor_stability": 0.7598171830177307, "fragility": 0.25, "augmentation_robustness": 0.5844171391849297, "xie_feature_dispersion": 0.1566522503868137, "pac": 0.6721171611013301, "whitened_cosine_id": 0.6859370470046997}} +{"model": "pythia-70m", "dataset": "stance", "seed": 3, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.764248302629601, "raptor_stability": 0.8229038119316101, "fragility": 0.5, "augmentation_robustness": 0.7459009481197789, "xie_feature_dispersion": 2.080304032697393, "pac": 0.7844023800256945, "whitened_cosine_id": 0.8263651728630066}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 3, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.711632992199843, "raptor_stability": 0.8560130596160889, "fragility": 1.0, "augmentation_robustness": 0.6862800186116383, "xie_feature_dispersion": 1.5912468339045849, "pac": 0.7711465391138637, "whitened_cosine_id": 0.8750572800636292}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 3, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349443, "raptor_stability": 0.8484313488006592, "fragility": 0.25, "augmentation_robustness": 0.5622190281630771, "xie_feature_dispersion": 4.044131456746566, "pac": 0.7053251884818681, "whitened_cosine_id": 0.8518450260162354}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350076, "raptor_stability": 0.871891438961029, "fragility": 1.0, "augmentation_robustness": 0.7693997298355439, "xie_feature_dispersion": 1.4704300817182188, "pac": 0.8206455843982865, "whitened_cosine_id": 0.8666113018989563}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 3, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.36025022356685565, "raptor_stability": 0.8447644710540771, "fragility": 1.5, "augmentation_robustness": 0.7221689699035224, "xie_feature_dispersion": 6.085221682949123, "pac": 0.7834667204787997, "whitened_cosine_id": 0.9049122333526611}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 3, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10505729790354523, "raptor_stability": 0.9525120854377747, "fragility": 1.5, "augmentation_robustness": 0.8703692378009392, "xie_feature_dispersion": 5.186371589560047, "pac": 0.9114406616193569, "whitened_cosine_id": 0.9208746552467346}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342074, "raptor_stability": 0.78183913230896, "fragility": 0.5, "augmentation_robustness": 0.6746312909390834, "xie_feature_dispersion": 0.5886846655394216, "pac": 0.7282352116240216, "whitened_cosine_id": 0.7248976230621338}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 3, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.6992640027033808, "raptor_stability": 0.8696766495704651, "fragility": 0.5, "augmentation_robustness": 0.7326524720518076, "xie_feature_dispersion": 1.5546891203751845, "pac": 0.8011645608111364, "whitened_cosine_id": 0.8558102250099182}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 3, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347303, "raptor_stability": 0.8296602964401245, "fragility": 0.25, "augmentation_robustness": 0.6793747577236712, "xie_feature_dispersion": 2.1630123007277384, "pac": 0.7545175270818978, "whitened_cosine_id": 0.8202237486839294}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 3, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346095, "raptor_stability": 0.8123674988746643, "fragility": 0.5, "augmentation_robustness": 0.6624179901863742, "xie_feature_dispersion": 1.81846361791661, "pac": 0.7373927445305193, "whitened_cosine_id": 0.7944690585136414}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 3, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346586, "raptor_stability": 0.8003823757171631, "fragility": 0.25, "augmentation_robustness": 0.6676009627104235, "xie_feature_dispersion": 0.7182211892213498, "pac": 0.7339916692137933, "whitened_cosine_id": 0.7915064692497253}} +{"model": "pythia-160m", "dataset": "subj", "seed": 3, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352172, "raptor_stability": 0.8996919989585876, "fragility": 1.0, "augmentation_robustness": 0.7410086684825309, "xie_feature_dispersion": 1.9890978207832326, "pac": 0.8203503337205593, "whitened_cosine_id": 0.9151029586791992}} +{"model": "pythia-160m", "dataset": "spam", "seed": 3, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859350205, "raptor_stability": 0.9280425310134888, "fragility": 1.0, "augmentation_robustness": 0.8124903597056025, "xie_feature_dispersion": 13.64857545430944, "pac": 0.8702664453595457, "whitened_cosine_id": 0.9369125366210938}} +{"model": "pythia-160m", "dataset": "cola", "seed": 3, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343062, "raptor_stability": 0.7884514331817627, "fragility": 0.25, "augmentation_robustness": 0.5611738456088626, "xie_feature_dispersion": 0.17857729363649755, "pac": 0.6748126393953127, "whitened_cosine_id": 0.7649105191230774}} +{"model": "pythia-160m", "dataset": "stance", "seed": 3, "layer": 4, "concept": "stance", "predictors": {"sip_eigengap": 0.6506681129814839, "raptor_stability": 0.8327218890190125, "fragility": 0.5, "augmentation_robustness": 0.7643668163368509, "xie_feature_dispersion": 1.7873782154455453, "pac": 0.7985443526779317, "whitened_cosine_id": 0.8258171081542969}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 3, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350412, "raptor_stability": 0.8567770719528198, "fragility": 0.5, "augmentation_robustness": 0.690159632122668, "xie_feature_dispersion": 1.305808488628754, "pac": 0.7734683520377439, "whitened_cosine_id": 0.8785260319709778}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 3, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881342, "raptor_stability": 0.8563058376312256, "fragility": 0.25, "augmentation_robustness": 0.5454463736012634, "xie_feature_dispersion": 4.914029083398948, "pac": 0.7008761056162445, "whitened_cosine_id": 0.8632848858833313}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688189, "raptor_stability": 0.8555104732513428, "fragility": 1.0, "augmentation_robustness": 0.7200850357383629, "xie_feature_dispersion": 2.288214053878471, "pac": 0.7877977544948529, "whitened_cosine_id": 0.8782402873039246}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 3, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.3288136919930267, "raptor_stability": 0.8414013981819153, "fragility": 2.0, "augmentation_robustness": 0.7266437337043424, "xie_feature_dispersion": 12.434412392508191, "pac": 0.7840225659431288, "whitened_cosine_id": 0.8793582320213318}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 3, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.0776609124998989, "raptor_stability": 0.9545353055000305, "fragility": 2.0, "augmentation_robustness": 0.8723893129234274, "xie_feature_dispersion": 7.487139994618903, "pac": 0.9134623092117289, "whitened_cosine_id": 0.8982532620429993}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 3, "layer": 14, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879506, "raptor_stability": 0.8133618831634521, "fragility": 0.25, "augmentation_robustness": 0.6928116161487408, "xie_feature_dispersion": 0.25094480473639263, "pac": 0.7530867496560965, "whitened_cosine_id": 0.7940764427185059}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 3, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.647965606663094, "raptor_stability": 0.8681758046150208, "fragility": 0.5, "augmentation_robustness": 0.7102638120672203, "xie_feature_dispersion": 1.6253563675189826, "pac": 0.7892198083411206, "whitened_cosine_id": 0.8482173085212708}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 3, "layer": 2, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295687982, "raptor_stability": 0.8303574919700623, "fragility": 1.0, "augmentation_robustness": 0.6867166191679835, "xie_feature_dispersion": 1.7750280665756264, "pac": 0.758537055569023, "whitened_cosine_id": 0.8150797486305237}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 3, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879206, "raptor_stability": 0.8025309443473816, "fragility": 0.25, "augmentation_robustness": 0.6179124255783944, "xie_feature_dispersion": 4.2516203791362885, "pac": 0.710221684962888, "whitened_cosine_id": 0.8091989755630493}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 3, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879957, "raptor_stability": 0.8084143996238708, "fragility": 0.25, "augmentation_robustness": 0.6400424522618103, "xie_feature_dispersion": 0.98880123212057, "pac": 0.7242284259428406, "whitened_cosine_id": 0.8083922863006592}} +{"model": "pythia-410m", "dataset": "subj", "seed": 3, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882546, "raptor_stability": 0.9100797772407532, "fragility": 1.0, "augmentation_robustness": 0.7426358153217486, "xie_feature_dispersion": 1.4905914800590063, "pac": 0.826357796281251, "whitened_cosine_id": 0.8978877663612366}} +{"model": "pythia-410m", "dataset": "spam", "seed": 3, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882908, "raptor_stability": 0.9304795265197754, "fragility": 3.0, "augmentation_robustness": 0.8261601242245535, "xie_feature_dispersion": 8.165725328081457, "pac": 0.8783198253721645, "whitened_cosine_id": 0.9207106232643127}} +{"model": "pythia-410m", "dataset": "cola", "seed": 3, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.210307295687961, "raptor_stability": 0.7938510775566101, "fragility": 0.25, "augmentation_robustness": 0.5212292454913044, "xie_feature_dispersion": 0.90277616899984, "pac": 0.6575401615239573, "whitened_cosine_id": 0.8084408640861511}} +{"model": "pythia-410m", "dataset": "stance", "seed": 3, "layer": 17, "concept": "stance", "predictors": {"sip_eigengap": 0.4936540695216237, "raptor_stability": 0.8370916247367859, "fragility": 0.5, "augmentation_robustness": 0.7826824738040318, "xie_feature_dispersion": 2.485399050185102, "pac": 0.8098870492704089, "whitened_cosine_id": 0.8208258748054504}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 3, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881564, "raptor_stability": 0.8418619632720947, "fragility": 0.5, "augmentation_robustness": 0.6628138222662973, "xie_feature_dispersion": 1.7478308880725917, "pac": 0.7523378927691959, "whitened_cosine_id": 0.8692632913589478}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 3, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010637, "raptor_stability": 0.8572840690612793, "fragility": 0.25, "augmentation_robustness": 0.5195046452531485, "xie_feature_dispersion": 5.520447159983482, "pac": 0.688394357157214, "whitened_cosine_id": 0.8275399804115295}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 3, "layer": 20, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010867, "raptor_stability": 0.8348312973976135, "fragility": 0.5, "augmentation_robustness": 0.6270832143777221, "xie_feature_dispersion": 1.9157864290958875, "pac": 0.7309572558876678, "whitened_cosine_id": 0.8293268084526062}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 3, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.20484503848029767, "raptor_stability": 0.85955810546875, "fragility": 1.5, "augmentation_robustness": 0.7594005762230035, "xie_feature_dispersion": 5.718643787688656, "pac": 0.8094793408458767, "whitened_cosine_id": 0.828823447227478}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 3, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05273817789404076, "raptor_stability": 0.9474886655807495, "fragility": 3.0, "augmentation_robustness": 0.8532070565094912, "xie_feature_dispersion": 9.58314422780073, "pac": 0.9003478610451203, "whitened_cosine_id": 0.832189679145813}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 3, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010883, "raptor_stability": 0.8169371485710144, "fragility": 0.25, "augmentation_robustness": 0.6690915350986811, "xie_feature_dispersion": 0.42029681121277923, "pac": 0.7430143418348478, "whitened_cosine_id": 0.8196472525596619}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 3, "layer": 9, "concept": "emotion", "predictors": {"sip_eigengap": 0.473623796458844, "raptor_stability": 0.8568776249885559, "fragility": 0.25, "augmentation_robustness": 0.6824483342515307, "xie_feature_dispersion": 2.6608823301213778, "pac": 0.7696629796200434, "whitened_cosine_id": 0.83025723695755}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 3, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010806, "raptor_stability": 0.8051812052726746, "fragility": 0.25, "augmentation_robustness": 0.6006977265762596, "xie_feature_dispersion": 4.0725358408621375, "pac": 0.7029394659244671, "whitened_cosine_id": 0.8263804912567139}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 3, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010875, "raptor_stability": 0.8060323596000671, "fragility": 0.5, "augmentation_robustness": 0.6189902560303079, "xie_feature_dispersion": 2.2388020889990288, "pac": 0.7125113078151875, "whitened_cosine_id": 0.8258398771286011}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 3, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010296, "raptor_stability": 0.7851132154464722, "fragility": 0.25, "augmentation_robustness": 0.5954573958337406, "xie_feature_dispersion": 1.0772966237421127, "pac": 0.6902853056401064, "whitened_cosine_id": 0.8228062987327576}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 3, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.8939151167869568, "fragility": 1.0, "augmentation_robustness": 0.7047598539681612, "xie_feature_dispersion": 2.9162655683918848, "pac": 0.799337485377559, "whitened_cosine_id": 0.8292585611343384}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 3, "layer": 3, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010838, "raptor_stability": 0.925938069820404, "fragility": 3.0, "augmentation_robustness": 0.8210687839915044, "xie_feature_dispersion": 10.71642677493898, "pac": 0.8735034269059543, "whitened_cosine_id": 0.8716909289360046}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 3, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010886, "raptor_stability": 0.7957292199134827, "fragility": 0.25, "augmentation_robustness": 0.45052250604191046, "xie_feature_dispersion": 0.7106391233544392, "pac": 0.6231258629776966, "whitened_cosine_id": 0.8221802115440369}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 3, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.37531177284512046, "raptor_stability": 0.8395811915397644, "fragility": 1.0, "augmentation_robustness": 0.7703567449624081, "xie_feature_dispersion": 2.787120846289539, "pac": 0.8049689682510863, "whitened_cosine_id": 0.8196287155151367}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 3, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.855816496101079, "raptor_stability": 0.8419731855392456, "fragility": 1.0, "augmentation_robustness": 0.6515139260319379, "xie_feature_dispersion": 2.265789759921114, "pac": 0.7467435557855917, "whitened_cosine_id": 0.8332340121269226}} +{"model": "gpt2", "dataset": "sst2", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934796, "raptor_stability": 0.8351662158966064, "fragility": 0.25, "augmentation_robustness": 0.46886485771385517, "xie_feature_dispersion": 4.2622215814473865, "pac": 0.6520155368052308, "whitened_cosine_id": 0.8634011149406433}} +{"model": "gpt2", "dataset": "imdb", "seed": 3, "layer": 8, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350514, "raptor_stability": 0.8268616795539856, "fragility": 0.5, "augmentation_robustness": 0.6116287845851899, "xie_feature_dispersion": 1.47086614284479, "pac": 0.7192452320695877, "whitened_cosine_id": 0.876645565032959}} +{"model": "gpt2", "dataset": "ag_news", "seed": 3, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.37970761633998423, "raptor_stability": 0.8910131454467773, "fragility": 0.25, "augmentation_robustness": 0.8581536528187313, "xie_feature_dispersion": 0.8261918893870008, "pac": 0.8745833991327543, "whitened_cosine_id": 0.9110751152038574}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 3, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.10991110313414279, "raptor_stability": 0.9330034255981445, "fragility": 1.5, "augmentation_robustness": 0.8025370623665369, "xie_feature_dispersion": 5.468890231915903, "pac": 0.8677702439823407, "whitened_cosine_id": 0.9063879251480103}} +{"model": "gpt2", "dataset": "counterfact", "seed": 3, "layer": 7, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343042, "raptor_stability": 0.7817078232765198, "fragility": 0.25, "augmentation_robustness": 0.6861783480825473, "xie_feature_dispersion": 0.1934064316642644, "pac": 0.7339430856795335, "whitened_cosine_id": 0.7454338073730469}} +{"model": "gpt2", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6182210878004821, "raptor_stability": 0.8550314903259277, "fragility": 0.25, "augmentation_robustness": 0.6785777254477671, "xie_feature_dispersion": 1.1654170019303518, "pac": 0.7668046078868473, "whitened_cosine_id": 0.8550549149513245}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 3, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347265, "raptor_stability": 0.8183044791221619, "fragility": 0.25, "augmentation_robustness": 0.7349596865260767, "xie_feature_dispersion": 0.7005527274806919, "pac": 0.7766320828241193, "whitened_cosine_id": 0.8098105788230896}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 3, "layer": 7, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346741, "raptor_stability": 0.7654048204421997, "fragility": 0.25, "augmentation_robustness": 0.5347742087458891, "xie_feature_dispersion": 3.241040228671077, "pac": 0.6500895145940444, "whitened_cosine_id": 0.8011150360107422}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 3, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346404, "raptor_stability": 0.7987147569656372, "fragility": 0.25, "augmentation_robustness": 0.7470243212681252, "xie_feature_dispersion": 0.9553360233273932, "pac": 0.7728695391168812, "whitened_cosine_id": 0.7885229587554932}} +{"model": "gpt2", "dataset": "subj", "seed": 3, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352408, "raptor_stability": 0.8722814321517944, "fragility": 0.5, "augmentation_robustness": 0.645032578503623, "xie_feature_dispersion": 1.9236159050902961, "pac": 0.7586570053277087, "whitened_cosine_id": 0.9171640276908875}} +{"model": "gpt2", "dataset": "spam", "seed": 3, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593528, "raptor_stability": 0.9187144041061401, "fragility": 1.5, "augmentation_robustness": 0.812376554161553, "xie_feature_dispersion": 12.08932579677981, "pac": 0.8655454791338466, "whitened_cosine_id": 0.9336515665054321}} +{"model": "gpt2", "dataset": "cola", "seed": 3, "layer": 8, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344077, "raptor_stability": 0.7706639766693115, "fragility": 0.25, "augmentation_robustness": 0.5024593786840387, "xie_feature_dispersion": 0.689280134170416, "pac": 0.6365616776766752, "whitened_cosine_id": 0.782214343547821}} +{"model": "gpt2", "dataset": "stance", "seed": 3, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.6228451679003366, "raptor_stability": 0.8234939575195312, "fragility": 0.5, "augmentation_robustness": 0.7234382249579733, "xie_feature_dispersion": 1.7228146499923553, "pac": 0.7734660912387523, "whitened_cosine_id": 0.8058009147644043}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 3, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935006, "raptor_stability": 0.8277425765991211, "fragility": 0.25, "augmentation_robustness": 0.6110880945528435, "xie_feature_dispersion": 1.2604141020751956, "pac": 0.7194153355759823, "whitened_cosine_id": 0.8723833560943604}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 3, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881538, "raptor_stability": 0.8475278615951538, "fragility": 0.25, "augmentation_robustness": 0.5298442019954345, "xie_feature_dispersion": 4.500703580339019, "pac": 0.6886860317952941, "whitened_cosine_id": 0.8626024723052979}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 3, "layer": 24, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881826, "raptor_stability": 0.896838903427124, "fragility": 0.25, "augmentation_robustness": 0.6259364774466486, "xie_feature_dispersion": 0.14721586133397682, "pac": 0.7613876904368864, "whitened_cosine_id": 0.8744872808456421}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 3, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.27417661774965685, "raptor_stability": 0.8498548269271851, "fragility": 1.0, "augmentation_robustness": 0.7432784283933686, "xie_feature_dispersion": 2.8627715211803557, "pac": 0.7965666276602767, "whitened_cosine_id": 0.8807106018066406}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 3, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.07699019521608971, "raptor_stability": 0.9369117021560669, "fragility": 1.5, "augmentation_robustness": 0.8223812079136419, "xie_feature_dispersion": 6.380122368170902, "pac": 0.8796464550348544, "whitened_cosine_id": 0.877031683921814}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 3, "layer": 20, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878205, "raptor_stability": 0.765117883682251, "fragility": 0.25, "augmentation_robustness": 0.6346673610552594, "xie_feature_dispersion": 0.2291144061643789, "pac": 0.6998926223687552, "whitened_cosine_id": 0.7576854825019836}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6234881751702148, "raptor_stability": 0.8669646382331848, "fragility": 0.25, "augmentation_robustness": 0.7006085301347155, "xie_feature_dispersion": 0.8652954065662566, "pac": 0.7837865841839502, "whitened_cosine_id": 0.8504673838615417}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 3, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880274, "raptor_stability": 0.813629686832428, "fragility": 0.25, "augmentation_robustness": 0.7004445795324097, "xie_feature_dispersion": 1.0577105967646572, "pac": 0.7570371331824188, "whitened_cosine_id": 0.8194169998168945}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 3, "layer": 16, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956880063, "raptor_stability": 0.7668350338935852, "fragility": 0.25, "augmentation_robustness": 0.5387038041591353, "xie_feature_dispersion": 3.0633639663829166, "pac": 0.6527694190263602, "whitened_cosine_id": 0.8101463317871094}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 3, "layer": 24, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687966, "raptor_stability": 0.7879273891448975, "fragility": 0.25, "augmentation_robustness": 0.7205035666208062, "xie_feature_dispersion": 0.933971243574555, "pac": 0.7542154778828518, "whitened_cosine_id": 0.7968817353248596}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 3, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882648, "raptor_stability": 0.8908157348632812, "fragility": 0.5, "augmentation_robustness": 0.6905186912348897, "xie_feature_dispersion": 1.3649873458587798, "pac": 0.7906672130490855, "whitened_cosine_id": 0.8880553841590881}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 3, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688045, "raptor_stability": 0.9237306714057922, "fragility": 1.0, "augmentation_robustness": 0.8008774808792517, "xie_feature_dispersion": 14.196302533817459, "pac": 0.862304076142522, "whitened_cosine_id": 0.9053286910057068}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 3, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879595, "raptor_stability": 0.786989688873291, "fragility": 0.25, "augmentation_robustness": 0.4592027295956805, "xie_feature_dispersion": 0.6920920489375771, "pac": 0.6230962092344857, "whitened_cosine_id": 0.8074944615364075}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 3, "layer": 16, "concept": "stance", "predictors": {"sip_eigengap": 0.4985627032597986, "raptor_stability": 0.8289973139762878, "fragility": 0.5, "augmentation_robustness": 0.7594746024395039, "xie_feature_dispersion": 1.849522827224973, "pac": 0.7942359582078959, "whitened_cosine_id": 0.7846319079399109}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 3, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881336, "raptor_stability": 0.8215765357017517, "fragility": 0.25, "augmentation_robustness": 0.5954299844747202, "xie_feature_dispersion": 1.2573193221997883, "pac": 0.7085032600882359, "whitened_cosine_id": 0.8571715950965881}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 3, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237647006, "raptor_stability": 0.8808207511901855, "fragility": 0.25, "augmentation_robustness": 0.6288068642959735, "xie_feature_dispersion": 5.001977030703906, "pac": 0.7548138077430795, "whitened_cosine_id": 0.8865893483161926}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650275, "raptor_stability": 0.8786996006965637, "fragility": 1.0, "augmentation_robustness": 0.7573141076406408, "xie_feature_dispersion": 2.4507641389974713, "pac": 0.8180068541686023, "whitened_cosine_id": 0.8870015740394592}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 3, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.27907133932759765, "raptor_stability": 0.8452746272087097, "fragility": 1.5, "augmentation_robustness": 0.7357258977124443, "xie_feature_dispersion": 8.01989836662422, "pac": 0.790500262460577, "whitened_cosine_id": 0.8956002593040466}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 3, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.07227979648704244, "raptor_stability": 0.9585252404212952, "fragility": 1.0, "augmentation_robustness": 0.8819010246269907, "xie_feature_dispersion": 6.547993580545896, "pac": 0.9202131325241429, "whitened_cosine_id": 0.9235095977783203}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 3, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646929, "raptor_stability": 0.8464949131011963, "fragility": 0.25, "augmentation_robustness": 0.7631963084432197, "xie_feature_dispersion": 0.07258280090844692, "pac": 0.804845610772208, "whitened_cosine_id": 0.8042153120040894}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 3, "layer": 11, "concept": "emotion", "predictors": {"sip_eigengap": 0.5364682673815797, "raptor_stability": 0.8717088103294373, "fragility": 0.25, "augmentation_robustness": 0.8095188795391789, "xie_feature_dispersion": 1.9155672487876931, "pac": 0.8406138449343081, "whitened_cosine_id": 0.8489202857017517}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 3, "layer": 6, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237646567, "raptor_stability": 0.8316985368728638, "fragility": 0.25, "augmentation_robustness": 0.6779121169500498, "xie_feature_dispersion": 4.322967748214793, "pac": 0.7548053269114567, "whitened_cosine_id": 0.8318672776222229}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 3, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237646929, "raptor_stability": 0.8263404369354248, "fragility": 1.0, "augmentation_robustness": 0.7174827260397666, "xie_feature_dispersion": 2.0942342832154073, "pac": 0.7719115814875956, "whitened_cosine_id": 0.791140615940094}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 3, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.293872923764742, "raptor_stability": 0.8533427119255066, "fragility": 2.0, "augmentation_robustness": 0.8330055291842721, "xie_feature_dispersion": 2.823288787076928, "pac": 0.8431741205548893, "whitened_cosine_id": 0.7997245192527771}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 3, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651214, "raptor_stability": 0.900323212146759, "fragility": 0.5, "augmentation_robustness": 0.7438287558717409, "xie_feature_dispersion": 1.6449955999096242, "pac": 0.82207598400925, "whitened_cosine_id": 0.9154244661331177}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 3, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237634952, "raptor_stability": 0.9340345859527588, "fragility": 0.5, "augmentation_robustness": 0.8432807256360871, "xie_feature_dispersion": 18.65723826433361, "pac": 0.888657655794423, "whitened_cosine_id": 0.9443875551223755}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 3, "layer": 14, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646847, "raptor_stability": 0.8058444857597351, "fragility": 0.25, "augmentation_robustness": 0.5270067886058506, "xie_feature_dispersion": 0.7691357194942805, "pac": 0.6664256371827928, "whitened_cosine_id": 0.8194195032119751}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 3, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.5814899203429595, "raptor_stability": 0.8490622639656067, "fragility": 0.25, "augmentation_robustness": 0.8046175979185222, "xie_feature_dispersion": 2.1435570572895988, "pac": 0.8268399309420644, "whitened_cosine_id": 0.8264880776405334}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 3, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764957, "raptor_stability": 0.8473092317581177, "fragility": 0.25, "augmentation_robustness": 0.6972751540596228, "xie_feature_dispersion": 1.2429462131243278, "pac": 0.7722921929088702, "whitened_cosine_id": 0.8770169615745544}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 4, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.711632992199527, "raptor_stability": 0.8324896097183228, "fragility": 0.25, "augmentation_robustness": 0.5885938862832164, "xie_feature_dispersion": 1.3807377703846688, "pac": 0.7105417480007696, "whitened_cosine_id": 0.8337145447731018}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 4, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.711632992199886, "raptor_stability": 0.9050891995429993, "fragility": 1.0, "augmentation_robustness": 0.8489347430216386, "xie_feature_dispersion": 1.2126650382704005, "pac": 0.877011971282319, "whitened_cosine_id": 0.850266695022583}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 4, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6455833206335958, "raptor_stability": 0.9030647277832031, "fragility": 1.5, "augmentation_robustness": 0.8698773067049781, "xie_feature_dispersion": 3.394769818942791, "pac": 0.8864710172440906, "whitened_cosine_id": 0.9047799110412598}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.1519243074719573, "raptor_stability": 0.9499274492263794, "fragility": 1.5, "augmentation_robustness": 0.874702657535151, "xie_feature_dispersion": 5.282104233636098, "pac": 0.9123150533807651, "whitened_cosine_id": 0.9344291687011719}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 4, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.711632992197464, "raptor_stability": 0.7690514922142029, "fragility": 0.5, "augmentation_robustness": 0.646014274798961, "xie_feature_dispersion": 0.46179379874179, "pac": 0.7075328835065819, "whitened_cosine_id": 0.6736164689064026}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6855318380406782, "raptor_stability": 0.8837265372276306, "fragility": 1.0, "augmentation_robustness": 0.7879341856049312, "xie_feature_dispersion": 1.5036821374561455, "pac": 0.835830361416281, "whitened_cosine_id": 0.8548471331596375}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 4, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921993338, "raptor_stability": 0.858695387840271, "fragility": 1.0, "augmentation_robustness": 0.7810593679700619, "xie_feature_dispersion": 1.841863426224599, "pac": 0.8198773779051665, "whitened_cosine_id": 0.7971394658088684}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 4, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921988385, "raptor_stability": 0.81805419921875, "fragility": 0.5, "augmentation_robustness": 0.6270972563452893, "xie_feature_dispersion": 2.0918980749205067, "pac": 0.7225757277820197, "whitened_cosine_id": 0.7637650966644287}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 4, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990021, "raptor_stability": 0.8543492555618286, "fragility": 1.0, "augmentation_robustness": 0.8033994268781987, "xie_feature_dispersion": 2.0592689851956822, "pac": 0.8288743412200137, "whitened_cosine_id": 0.7868671417236328}} +{"model": "pythia-70m", "dataset": "subj", "seed": 4, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002564, "raptor_stability": 0.8824465274810791, "fragility": 1.0, "augmentation_robustness": 0.7248629327605404, "xie_feature_dispersion": 1.925210310106005, "pac": 0.8036547301208097, "whitened_cosine_id": 0.9232692122459412}} +{"model": "pythia-70m", "dataset": "spam", "seed": 4, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004778, "raptor_stability": 0.9337431788444519, "fragility": 3.0, "augmentation_robustness": 0.8271455702691285, "xie_feature_dispersion": 8.1620184111796, "pac": 0.8804443745567903, "whitened_cosine_id": 0.9525339603424072}} +{"model": "pythia-70m", "dataset": "cola", "seed": 4, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921978768, "raptor_stability": 0.7940040826797485, "fragility": 1.0, "augmentation_robustness": 0.7165809555085785, "xie_feature_dispersion": 0.5974938164403546, "pac": 0.7552925190941635, "whitened_cosine_id": 0.7271562218666077}} +{"model": "pythia-70m", "dataset": "stance", "seed": 4, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.8148210225469336, "raptor_stability": 0.8456482291221619, "fragility": 1.0, "augmentation_robustness": 0.8025558898811386, "xie_feature_dispersion": 1.966541164919816, "pac": 0.8241020595016502, "whitened_cosine_id": 0.8281933665275574}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 4, "layer": 3, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997497, "raptor_stability": 0.8669647574424744, "fragility": 0.5, "augmentation_robustness": 0.7238996722036884, "xie_feature_dispersion": 1.0213916251568427, "pac": 0.7954322148230815, "whitened_cosine_id": 0.8732953667640686}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 4, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349261, "raptor_stability": 0.842431366443634, "fragility": 0.25, "augmentation_robustness": 0.6006899879923536, "xie_feature_dispersion": 2.8701281427080465, "pac": 0.7215606772179939, "whitened_cosine_id": 0.8559513688087463}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 4, "layer": 8, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485935049, "raptor_stability": 0.8715534806251526, "fragility": 0.5, "augmentation_robustness": 0.7578237900499349, "xie_feature_dispersion": 1.468960362896107, "pac": 0.8146886353375438, "whitened_cosine_id": 0.8794116377830505}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 4, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3108708237634092, "raptor_stability": 0.8495153784751892, "fragility": 0.25, "augmentation_robustness": 0.7581191296084264, "xie_feature_dispersion": 0.9310101861217775, "pac": 0.8038172540418078, "whitened_cosine_id": 0.8954182863235474}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 4, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.11711906515590792, "raptor_stability": 0.9491503834724426, "fragility": 1.5, "augmentation_robustness": 0.872645584642886, "xie_feature_dispersion": 5.1845819525161145, "pac": 0.9108979840576643, "whitened_cosine_id": 0.9199619293212891}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 4, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.397542485934307, "raptor_stability": 0.7986763715744019, "fragility": 0.25, "augmentation_robustness": 0.6679982833044269, "xie_feature_dispersion": 0.5584164688455334, "pac": 0.7333373274394144, "whitened_cosine_id": 0.7404189705848694}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.7058740799382045, "raptor_stability": 0.8810455799102783, "fragility": 1.0, "augmentation_robustness": 0.7601115407446962, "xie_feature_dispersion": 2.0356355439733544, "pac": 0.8205785603274873, "whitened_cosine_id": 0.8614587187767029}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 4, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.397542485934676, "raptor_stability": 0.8066017627716064, "fragility": 0.5, "augmentation_robustness": 0.6529396166694911, "xie_feature_dispersion": 2.9800122485144014, "pac": 0.7297706897205487, "whitened_cosine_id": 0.8008232712745667}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 4, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934582, "raptor_stability": 0.8202213644981384, "fragility": 0.25, "augmentation_robustness": 0.6057950002763687, "xie_feature_dispersion": 3.616328383472754, "pac": 0.7130081823872536, "whitened_cosine_id": 0.806983470916748}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 4, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346277, "raptor_stability": 0.8139558434486389, "fragility": 0.25, "augmentation_robustness": 0.6972874057878297, "xie_feature_dispersion": 1.5561310456535857, "pac": 0.7556216246182343, "whitened_cosine_id": 0.8032292723655701}} +{"model": "pythia-160m", "dataset": "subj", "seed": 4, "layer": 5, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351835, "raptor_stability": 0.8953569531440735, "fragility": 1.0, "augmentation_robustness": 0.7336658853694705, "xie_feature_dispersion": 1.8870415060935894, "pac": 0.814511419256772, "whitened_cosine_id": 0.9113974571228027}} +{"model": "pythia-160m", "dataset": "spam", "seed": 4, "layer": 12, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353225, "raptor_stability": 0.9130350351333618, "fragility": 0.5, "augmentation_robustness": 0.7635738024186676, "xie_feature_dispersion": 1.0575657286947788, "pac": 0.8383044187760147, "whitened_cosine_id": 0.9334843754768372}} +{"model": "pythia-160m", "dataset": "cola", "seed": 4, "layer": 2, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343628, "raptor_stability": 0.7938650250434875, "fragility": 0.5, "augmentation_robustness": 0.6480257075060164, "xie_feature_dispersion": 0.4811558212116989, "pac": 0.720945366274752, "whitened_cosine_id": 0.7700868844985962}} +{"model": "pythia-160m", "dataset": "stance", "seed": 4, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.591622106050008, "raptor_stability": 0.8417823314666748, "fragility": 1.0, "augmentation_robustness": 0.7769617698504535, "xie_feature_dispersion": 1.8092532859980797, "pac": 0.8093720506585642, "whitened_cosine_id": 0.8301091194152832}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 4, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935008, "raptor_stability": 0.8665009140968323, "fragility": 0.5, "augmentation_robustness": 0.7086095119691946, "xie_feature_dispersion": 0.8634855022918472, "pac": 0.7875552130330135, "whitened_cosine_id": 0.8738407492637634}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 4, "layer": 20, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688061, "raptor_stability": 0.8289727568626404, "fragility": 0.25, "augmentation_robustness": 0.5141338849042595, "xie_feature_dispersion": 3.3676014696218965, "pac": 0.6715533208834499, "whitened_cosine_id": 0.8497597575187683}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 4, "layer": 18, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882124, "raptor_stability": 0.8689995408058167, "fragility": 0.5, "augmentation_robustness": 0.7288625926699196, "xie_feature_dispersion": 1.9498286768695636, "pac": 0.7989310667378682, "whitened_cosine_id": 0.8895055055618286}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 4, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.2618240585223393, "raptor_stability": 0.8512345552444458, "fragility": 1.0, "augmentation_robustness": 0.7582979472125323, "xie_feature_dispersion": 5.093989432901596, "pac": 0.804766251228489, "whitened_cosine_id": 0.8913780450820923}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 4, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.10955555398923562, "raptor_stability": 0.9442237615585327, "fragility": 3.0, "augmentation_robustness": 0.8451578901315185, "xie_feature_dispersion": 10.914315920866802, "pac": 0.8946908258450257, "whitened_cosine_id": 0.887626588344574}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 4, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 1.210307295687852, "raptor_stability": 0.7869765162467957, "fragility": 0.25, "augmentation_robustness": 0.6309538054038855, "xie_feature_dispersion": 0.4977947174675206, "pac": 0.7089651608253406, "whitened_cosine_id": 0.7760900855064392}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 4, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.5396430100267917, "raptor_stability": 0.8690226674079895, "fragility": 0.25, "augmentation_robustness": 0.7275986700802664, "xie_feature_dispersion": 1.771013879173235, "pac": 0.7983106687441279, "whitened_cosine_id": 0.8532538414001465}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 4, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879974, "raptor_stability": 0.8093463182449341, "fragility": 0.25, "augmentation_robustness": 0.6522689919211225, "xie_feature_dispersion": 4.342720640505695, "pac": 0.7308076550830283, "whitened_cosine_id": 0.8254486918449402}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879697, "raptor_stability": 0.801374614238739, "fragility": 0.5, "augmentation_robustness": 0.5927650536655581, "xie_feature_dispersion": 1.97527441416868, "pac": 0.6970698339521486, "whitened_cosine_id": 0.8055334091186523}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 4, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687872, "raptor_stability": 0.8144592642784119, "fragility": 0.25, "augmentation_robustness": 0.6563598700902292, "xie_feature_dispersion": 3.7665662682483, "pac": 0.7354095671843206, "whitened_cosine_id": 0.8185945153236389}} +{"model": "pythia-410m", "dataset": "subj", "seed": 4, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882532, "raptor_stability": 0.8692127466201782, "fragility": 0.5, "augmentation_robustness": 0.7034189811361117, "xie_feature_dispersion": 1.4726101086077301, "pac": 0.7863158638781449, "whitened_cosine_id": 0.8998460173606873}} +{"model": "pythia-410m", "dataset": "spam", "seed": 4, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882872, "raptor_stability": 0.9380085468292236, "fragility": 3.0, "augmentation_robustness": 0.8306074892035639, "xie_feature_dispersion": 6.631280295102294, "pac": 0.8843080180163938, "whitened_cosine_id": 0.9329432845115662}} +{"model": "pythia-410m", "dataset": "cola", "seed": 4, "layer": 13, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879954, "raptor_stability": 0.8124500513076782, "fragility": 0.25, "augmentation_robustness": 0.6007360344180946, "xie_feature_dispersion": 0.12968683683090573, "pac": 0.7065930428628864, "whitened_cosine_id": 0.8110272288322449}} +{"model": "pythia-410m", "dataset": "stance", "seed": 4, "layer": 16, "concept": "stance", "predictors": {"sip_eigengap": 0.5699871301215584, "raptor_stability": 0.8456130027770996, "fragility": 0.5, "augmentation_robustness": 0.7784342554427507, "xie_feature_dispersion": 2.7721624801456297, "pac": 0.8120236291099252, "whitened_cosine_id": 0.8285333514213562}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 4, "layer": 17, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881238, "raptor_stability": 0.8406742215156555, "fragility": 0.5, "augmentation_robustness": 0.6639522820533309, "xie_feature_dispersion": 1.2414049521527235, "pac": 0.7523132517844933, "whitened_cosine_id": 0.8544057011604309}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 4, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010629, "raptor_stability": 0.8421841859817505, "fragility": 0.25, "augmentation_robustness": 0.5255393994669172, "xie_feature_dispersion": 3.4302386066080803, "pac": 0.6838617927243338, "whitened_cosine_id": 0.8283126950263977}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 4, "layer": 17, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8507398962974548, "fragility": 0.5, "augmentation_robustness": 0.6571983080736206, "xie_feature_dispersion": 2.186069124316738, "pac": 0.7539691021855377, "whitened_cosine_id": 0.8371143341064453}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 4, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.2265558176551069, "raptor_stability": 0.8569103479385376, "fragility": 1.5, "augmentation_robustness": 0.775218514152037, "xie_feature_dispersion": 5.3593709934723845, "pac": 0.8160644310452874, "whitened_cosine_id": 0.8336088061332703}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 4, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.07870194711620748, "raptor_stability": 0.9465585350990295, "fragility": 3.0, "augmentation_robustness": 0.8529859375455171, "xie_feature_dispersion": 9.52666970556469, "pac": 0.8997722363222733, "whitened_cosine_id": 0.8307517170906067}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010886, "raptor_stability": 0.8017974495887756, "fragility": 0.25, "augmentation_robustness": 0.6581476958880402, "xie_feature_dispersion": 0.5566355201054835, "pac": 0.7299725727384079, "whitened_cosine_id": 0.8239882588386536}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.02238370320048682, "raptor_stability": 0.861964762210846, "fragility": 0.25, "augmentation_robustness": 0.683838498205867, "xie_feature_dispersion": 2.404190193181982, "pac": 0.7729016302083564, "whitened_cosine_id": 0.8354349136352539}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 4, "layer": 13, "concept": "hate", "predictors": {"sip_eigengap": 0.85581649610108, "raptor_stability": 0.815185010433197, "fragility": 0.25, "augmentation_robustness": 0.6213317320340237, "xie_feature_dispersion": 4.537467582180153, "pac": 0.7182583712336104, "whitened_cosine_id": 0.8315821886062622}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 4, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010853, "raptor_stability": 0.8050772547721863, "fragility": 0.5, "augmentation_robustness": 0.6219013344721133, "xie_feature_dispersion": 2.2302886171133345, "pac": 0.7134892946221498, "whitened_cosine_id": 0.825539231300354}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 4, "layer": 21, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010789, "raptor_stability": 0.8045533895492554, "fragility": 0.25, "augmentation_robustness": 0.6038154742221061, "xie_feature_dispersion": 2.038988255260799, "pac": 0.7041844318856807, "whitened_cosine_id": 0.8270390629768372}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 4, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.8694700002670288, "fragility": 1.0, "augmentation_robustness": 0.7011428092065101, "xie_feature_dispersion": 2.50098235434014, "pac": 0.7853064047367695, "whitened_cosine_id": 0.8374378681182861}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 4, "layer": 8, "concept": "spam", "predictors": {"sip_eigengap": 0.85581649610102, "raptor_stability": 0.9178888201713562, "fragility": 1.0, "augmentation_robustness": 0.7928219231647947, "xie_feature_dispersion": 23.000550231187145, "pac": 0.8553553716680755, "whitened_cosine_id": 0.8662302494049072}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 4, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010873, "raptor_stability": 0.8157230615615845, "fragility": 0.25, "augmentation_robustness": 0.4993857440040359, "xie_feature_dispersion": 0.31676967447098947, "pac": 0.6575544027828102, "whitened_cosine_id": 0.8299345970153809}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 4, "layer": 17, "concept": "stance", "predictors": {"sip_eigengap": 0.33898643987040145, "raptor_stability": 0.828758716583252, "fragility": 0.5, "augmentation_robustness": 0.7561307052369252, "xie_feature_dispersion": 2.4747830566647955, "pac": 0.7924447109100885, "whitened_cosine_id": 0.825951099395752}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 4, "layer": 17, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010853, "raptor_stability": 0.8331989645957947, "fragility": 1.0, "augmentation_robustness": 0.627322928762955, "xie_feature_dispersion": 1.9407650108278485, "pac": 0.7302609466793748, "whitened_cosine_id": 0.8166413307189941}} +{"model": "gpt2", "dataset": "sst2", "seed": 4, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349996, "raptor_stability": 0.8296358585357666, "fragility": 0.25, "augmentation_robustness": 0.538035138872437, "xie_feature_dispersion": 3.0119285802495273, "pac": 0.6838354987041018, "whitened_cosine_id": 0.8716096878051758}} +{"model": "gpt2", "dataset": "imdb", "seed": 4, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350842, "raptor_stability": 0.8212791681289673, "fragility": 0.25, "augmentation_robustness": 0.5719860488360142, "xie_feature_dispersion": 1.6921644874690207, "pac": 0.6966326084824908, "whitened_cosine_id": 0.8852264881134033}} +{"model": "gpt2", "dataset": "ag_news", "seed": 4, "layer": 1, "concept": "topic", "predictors": {"sip_eigengap": 0.5333226182809293, "raptor_stability": 0.8636436462402344, "fragility": 0.5, "augmentation_robustness": 0.7962607413710602, "xie_feature_dispersion": 2.079954695672695, "pac": 0.8299521938056473, "whitened_cosine_id": 0.9045447707176208}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.1380779079157012, "raptor_stability": 0.932190477848053, "fragility": 1.5, "augmentation_robustness": 0.8059798202543121, "xie_feature_dispersion": 5.433633428902341, "pac": 0.8690851490511826, "whitened_cosine_id": 0.9079684615135193}} +{"model": "gpt2", "dataset": "counterfact", "seed": 4, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341972, "raptor_stability": 0.778740644454956, "fragility": 0.25, "augmentation_robustness": 0.689096563147253, "xie_feature_dispersion": 0.14506327750165868, "pac": 0.7339186038011045, "whitened_cosine_id": 0.7351542115211487}} +{"model": "gpt2", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5149693378247766, "raptor_stability": 0.8657316565513611, "fragility": 0.25, "augmentation_robustness": 0.6810631291311643, "xie_feature_dispersion": 1.1678519242893954, "pac": 0.7733973928412627, "whitened_cosine_id": 0.8591601848602295}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 4, "layer": 7, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347578, "raptor_stability": 0.7834518551826477, "fragility": 0.25, "augmentation_robustness": 0.5369999970113795, "xie_feature_dispersion": 3.433541094108048, "pac": 0.6602259260970136, "whitened_cosine_id": 0.8224130868911743}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 4, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345638, "raptor_stability": 0.7776834964752197, "fragility": 0.25, "augmentation_robustness": 0.5766816651217024, "xie_feature_dispersion": 2.7506262893948623, "pac": 0.677182580798461, "whitened_cosine_id": 0.790766179561615}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 4, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859344674, "raptor_stability": 0.779086172580719, "fragility": 0.25, "augmentation_robustness": 0.636466393301742, "xie_feature_dispersion": 2.815574398399473, "pac": 0.7077762829412305, "whitened_cosine_id": 0.8022896647453308}} +{"model": "gpt2", "dataset": "subj", "seed": 4, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352041, "raptor_stability": 0.8775368332862854, "fragility": 0.25, "augmentation_robustness": 0.7423765328860231, "xie_feature_dispersion": 0.37989537180206523, "pac": 0.8099566830861542, "whitened_cosine_id": 0.9062554836273193}} +{"model": "gpt2", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935321, "raptor_stability": 0.9084468483924866, "fragility": 1.5, "augmentation_robustness": 0.7197045600505599, "xie_feature_dispersion": 10.910132668764545, "pac": 0.8140757042215232, "whitened_cosine_id": 0.9353939294815063}} +{"model": "gpt2", "dataset": "cola", "seed": 4, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934498, "raptor_stability": 0.8069943189620972, "fragility": 0.25, "augmentation_robustness": 0.5145383818654423, "xie_feature_dispersion": 0.1311472028168222, "pac": 0.6607663504137697, "whitened_cosine_id": 0.8018531203269958}} +{"model": "gpt2", "dataset": "stance", "seed": 4, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5528580887739023, "raptor_stability": 0.8414139747619629, "fragility": 0.25, "augmentation_robustness": 0.7880240633832077, "xie_feature_dispersion": 0.9522759018719159, "pac": 0.8147190190725853, "whitened_cosine_id": 0.8301456570625305}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 4, "layer": 11, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934983, "raptor_stability": 0.8080868124961853, "fragility": 0.5, "augmentation_robustness": 0.5740397734709618, "xie_feature_dispersion": 1.3792581222344824, "pac": 0.6910632929835736, "whitened_cosine_id": 0.8652515411376953}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 4, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881142, "raptor_stability": 0.8341349363327026, "fragility": 0.25, "augmentation_robustness": 0.522748227401526, "xie_feature_dispersion": 3.1215463045771212, "pac": 0.6784415818671143, "whitened_cosine_id": 0.8599640727043152}} +{"model": "gpt2-medium", "dataset": "imdb", "seed": 4, "layer": 23, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882204, "raptor_stability": 0.9387363791465759, "fragility": 0.25, "augmentation_robustness": 0.8625580835416802, "xie_feature_dispersion": 2.3252826460499807, "pac": 0.900647231344128, "whitened_cosine_id": 0.8744732737541199}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 4, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.2490734387361061, "raptor_stability": 0.8433712124824524, "fragility": 1.0, "augmentation_robustness": 0.728063364939073, "xie_feature_dispersion": 3.704335352266776, "pac": 0.7857172887107626, "whitened_cosine_id": 0.881250262260437}} +{"model": "gpt2-medium", "dataset": "dbpedia", "seed": 4, "layer": 19, "concept": "topic", "predictors": {"sip_eigengap": 0.1068009553418093, "raptor_stability": 0.93490070104599, "fragility": 1.5, "augmentation_robustness": 0.8151994842437004, "xie_feature_dispersion": 6.637992417041358, "pac": 0.8750500926448452, "whitened_cosine_id": 0.8799629211425781}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 4, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878789, "raptor_stability": 0.7866268754005432, "fragility": 0.25, "augmentation_robustness": 0.6947319575923, "xie_feature_dispersion": 0.1466747323393826, "pac": 0.7406794164964217, "whitened_cosine_id": 0.7769638895988464}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5850775560898813, "raptor_stability": 0.8753568530082703, "fragility": 0.25, "augmentation_robustness": 0.7021984842516064, "xie_feature_dispersion": 0.8499569074700448, "pac": 0.7887776686299384, "whitened_cosine_id": 0.8553555011749268}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 4, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880465, "raptor_stability": 0.8168267607688904, "fragility": 0.25, "augmentation_robustness": 0.6873079329545261, "xie_feature_dispersion": 1.292586904966985, "pac": 0.7520673468617083, "whitened_cosine_id": 0.8249880075454712}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956880108, "raptor_stability": 0.7787988185882568, "fragility": 0.25, "augmentation_robustness": 0.535601105761814, "xie_feature_dispersion": 3.144268329198213, "pac": 0.6571999621750354, "whitened_cosine_id": 0.8181383609771729}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 4, "layer": 19, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880503, "raptor_stability": 0.7947708964347839, "fragility": 0.25, "augmentation_robustness": 0.5712740238571347, "xie_feature_dispersion": 2.3591332178783015, "pac": 0.6830224601459594, "whitened_cosine_id": 0.8295685052871704}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 4, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882628, "raptor_stability": 0.8655866980552673, "fragility": 0.5, "augmentation_robustness": 0.658993852125021, "xie_feature_dispersion": 1.294561616399908, "pac": 0.7622902750901441, "whitened_cosine_id": 0.8870449066162109}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 4, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882712, "raptor_stability": 0.9078684449195862, "fragility": 1.5, "augmentation_robustness": 0.7717328160030563, "xie_feature_dispersion": 4.942103977943902, "pac": 0.8398006304613213, "whitened_cosine_id": 0.9151633977890015}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 4, "layer": 16, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.210307295688009, "raptor_stability": 0.8008159399032593, "fragility": 0.25, "augmentation_robustness": 0.4862514101376035, "xie_feature_dispersion": 0.25648875129106014, "pac": 0.6435336750204314, "whitened_cosine_id": 0.8185151219367981}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 4, "layer": 15, "concept": "stance", "predictors": {"sip_eigengap": 0.4591650118237964, "raptor_stability": 0.818568766117096, "fragility": 0.5, "augmentation_robustness": 0.7404297489156003, "xie_feature_dispersion": 1.9692841257546414, "pac": 0.7794992575163482, "whitened_cosine_id": 0.7844472527503967}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 4, "layer": 23, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881433, "raptor_stability": 0.821359395980835, "fragility": 0.5, "augmentation_robustness": 0.5781830082719558, "xie_feature_dispersion": 1.435474963980124, "pac": 0.6997712021263953, "whitened_cosine_id": 0.8453944325447083}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 4, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237647728, "raptor_stability": 0.8522672653198242, "fragility": 0.25, "augmentation_robustness": 0.5718218717950756, "xie_feature_dispersion": 3.34994169851392, "pac": 0.7120445685574499, "whitened_cosine_id": 0.8843919038772583}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 4, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650468, "raptor_stability": 0.8942583203315735, "fragility": 1.0, "augmentation_robustness": 0.7825299636425157, "xie_feature_dispersion": 2.138162107858256, "pac": 0.8383941419870446, "whitened_cosine_id": 0.904426097869873}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 4, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.34415459384006464, "raptor_stability": 0.8814667463302612, "fragility": 0.5, "augmentation_robustness": 0.8472979677532054, "xie_feature_dispersion": 2.489404569197885, "pac": 0.8643823570417333, "whitened_cosine_id": 0.9101595878601074}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 4, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07232027374703373, "raptor_stability": 0.9524329900741577, "fragility": 1.5, "augmentation_robustness": 0.8775071051796265, "xie_feature_dispersion": 7.195039088518054, "pac": 0.914970047626892, "whitened_cosine_id": 0.9238596558570862}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 4, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646105, "raptor_stability": 0.8380523920059204, "fragility": 0.25, "augmentation_robustness": 0.7531953570003166, "xie_feature_dispersion": 0.1475593716528689, "pac": 0.7956238745031186, "whitened_cosine_id": 0.7818605303764343}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 4, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6524644536078231, "raptor_stability": 0.8897152543067932, "fragility": 0.25, "augmentation_robustness": 0.8390363852029633, "xie_feature_dispersion": 1.6422070698599567, "pac": 0.8643758197548783, "whitened_cosine_id": 0.854346513748169}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 4, "layer": 6, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647435, "raptor_stability": 0.8176067471504211, "fragility": 0.25, "augmentation_robustness": 0.7060153262170692, "xie_feature_dispersion": 4.1435032919274235, "pac": 0.7618110366837452, "whitened_cosine_id": 0.8246268033981323}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 4, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237636255, "raptor_stability": 0.8205522894859314, "fragility": 0.25, "augmentation_robustness": 0.6705621714892224, "xie_feature_dispersion": 3.9671135412282537, "pac": 0.7455572304875768, "whitened_cosine_id": 0.8058186173439026}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 4, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237646216, "raptor_stability": 0.8083575367927551, "fragility": 0.25, "augmentation_robustness": 0.685570372616926, "xie_feature_dispersion": 3.445721700576395, "pac": 0.7469639547048406, "whitened_cosine_id": 0.810322642326355}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 4, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651185, "raptor_stability": 0.886184573173523, "fragility": 0.5, "augmentation_robustness": 0.7504325815914692, "xie_feature_dispersion": 0.9492959086809997, "pac": 0.8183085773824961, "whitened_cosine_id": 0.919140100479126}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 4, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237647513, "raptor_stability": 0.9303396940231323, "fragility": 0.5, "augmentation_robustness": 0.8326478848343983, "xie_feature_dispersion": 18.28477413612892, "pac": 0.8814937894287653, "whitened_cosine_id": 0.9515261054039001}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 4, "layer": 14, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646718, "raptor_stability": 0.8116881251335144, "fragility": 0.25, "augmentation_robustness": 0.5360347240146708, "xie_feature_dispersion": 0.15135759875003865, "pac": 0.6738614245740926, "whitened_cosine_id": 0.8114071488380432}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.5271400318991657, "raptor_stability": 0.8213255405426025, "fragility": 0.5, "augmentation_robustness": 0.7172001724128508, "xie_feature_dispersion": 2.6328433172283368, "pac": 0.7692628564777266, "whitened_cosine_id": 0.826519787311554}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 4, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649207, "raptor_stability": 0.853298008441925, "fragility": 0.25, "augmentation_robustness": 0.6940637989225106, "xie_feature_dispersion": 0.7459643011965803, "pac": 0.7736809036822179, "whitened_cosine_id": 0.8680458068847656}} diff --git a/repro_bundle/results_tier2/results/predictors_1aug.jsonl b/repro_bundle/results_tier2/results/predictors_1aug.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..bc5227abf3302ef9d7c185585f63b50ee47870b3 --- /dev/null +++ b/repro_bundle/results_tier2/results/predictors_1aug.jsonl @@ -0,0 +1,472 @@ +{"model": "pythia-70m", "dataset": "sst2", "seed": 0, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921994833, "raptor_stability": 0.8375464081764221, "fragility": 0.25, "augmentation_robustness": 0.6545744286726296, "xie_feature_dispersion": 0.8227753483650888, "pac": 0.7460604184245259, "whitened_cosine_id": 0.8294597268104553}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 0, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998517, "raptor_stability": 0.9239972829818726, "fragility": 1.0, "augmentation_robustness": 0.9364739773993105, "xie_feature_dispersion": 1.0993059236404399, "pac": 0.9302356301905915, "whitened_cosine_id": 0.8396446704864502}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 0, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6360429269373853, "raptor_stability": 0.8923418521881104, "fragility": 1.5, "augmentation_robustness": 0.9079627357769793, "xie_feature_dispersion": 3.4534575785421744, "pac": 0.9001522939825448, "whitened_cosine_id": 0.9029156565666199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 0, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14189668585488155, "raptor_stability": 0.9457845091819763, "fragility": 1.5, "augmentation_robustness": 0.8839691673684479, "xie_feature_dispersion": 5.2630010355712145, "pac": 0.9148768382752122, "whitened_cosine_id": 0.9317190051078796}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 0, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921974287, "raptor_stability": 0.7641952633857727, "fragility": 1.0, "augmentation_robustness": 0.7320095613414648, "xie_feature_dispersion": 0.4779918901319739, "pac": 0.7481024123636187, "whitened_cosine_id": 0.6775354743003845}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 0, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5953587445214507, "raptor_stability": 0.8714625239372253, "fragility": 0.5, "augmentation_robustness": 0.8065598522498184, "xie_feature_dispersion": 1.5333196894740788, "pac": 0.8390111880935218, "whitened_cosine_id": 0.8480013012886047}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 0, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994247, "raptor_stability": 0.8230767250061035, "fragility": 0.5, "augmentation_robustness": 0.7221715232065848, "xie_feature_dispersion": 1.779944270255231, "pac": 0.7726241241063442, "whitened_cosine_id": 0.8223811388015747}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 0, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.711632992199055, "raptor_stability": 0.8190855383872986, "fragility": 0.25, "augmentation_robustness": 0.676819096553469, "xie_feature_dispersion": 2.512801472259043, "pac": 0.7479523174703837, "whitened_cosine_id": 0.7729045748710632}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989692, "raptor_stability": 0.8473518490791321, "fragility": 1.0, "augmentation_robustness": 0.8442697431706384, "xie_feature_dispersion": 1.163949469847951, "pac": 0.8458107961248853, "whitened_cosine_id": 0.7645858526229858}} +{"model": "pythia-70m", "dataset": "subj", "seed": 0, "layer": 1, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002171, "raptor_stability": 0.9112874269485474, "fragility": 1.5, "augmentation_robustness": 0.8646995541446441, "xie_feature_dispersion": 2.3756762827181857, "pac": 0.8879934905465957, "whitened_cosine_id": 0.9034857153892517}} +{"model": "pythia-70m", "dataset": "spam", "seed": 0, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004674, "raptor_stability": 0.9211491942405701, "fragility": 1.5, "augmentation_robustness": 0.884000762861815, "xie_feature_dispersion": 10.408354824969607, "pac": 0.9025749785511925, "whitened_cosine_id": 0.9505216479301453}} +{"model": "pythia-70m", "dataset": "cola", "seed": 0, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921980065, "raptor_stability": 0.7842410206794739, "fragility": 1.0, "augmentation_robustness": 0.751228525306698, "xie_feature_dispersion": 0.5820240387291664, "pac": 0.7677347729930859, "whitened_cosine_id": 0.7026472687721252}} +{"model": "pythia-70m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.7516991834926783, "raptor_stability": 0.8351565003395081, "fragility": 1.0, "augmentation_robustness": 0.8460384036839802, "xie_feature_dispersion": 1.9650544929284912, "pac": 0.8405974520117441, "whitened_cosine_id": 0.8273500800132751}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 0, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998, "raptor_stability": 0.8865668177604675, "fragility": 1.5, "augmentation_robustness": 0.8695133740159424, "xie_feature_dispersion": 1.822712974246315, "pac": 0.878040095888205, "whitened_cosine_id": 0.8655645847320557}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350087, "raptor_stability": 0.8526354432106018, "fragility": 0.25, "augmentation_robustness": 0.6188354080698831, "xie_feature_dispersion": 1.4101369914642357, "pac": 0.7357354256402424, "whitened_cosine_id": 0.8668140769004822}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 0, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350252, "raptor_stability": 0.8803290128707886, "fragility": 1.0, "augmentation_robustness": 0.8724963955930902, "xie_feature_dispersion": 1.3391749548314755, "pac": 0.8764127042319394, "whitened_cosine_id": 0.8713250756263733}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.3953680487548481, "raptor_stability": 0.8455236554145813, "fragility": 0.25, "augmentation_robustness": 0.8483093395593638, "xie_feature_dispersion": 0.9318287616318928, "pac": 0.8469164974869725, "whitened_cosine_id": 0.8953917026519775}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 0, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.10457984207175601, "raptor_stability": 0.9457212686538696, "fragility": 1.5, "augmentation_robustness": 0.8857764499429269, "xie_feature_dispersion": 5.199477497190765, "pac": 0.9157488592983982, "whitened_cosine_id": 0.9181342720985413}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 0, "layer": 4, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342227, "raptor_stability": 0.7698441743850708, "fragility": 0.25, "augmentation_robustness": 0.6816650768598687, "xie_feature_dispersion": 0.25779876141354974, "pac": 0.7257546256224697, "whitened_cosine_id": 0.7203242182731628}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.574379658862361, "raptor_stability": 0.8597204685211182, "fragility": 0.25, "augmentation_robustness": 0.7682780634518579, "xie_feature_dispersion": 1.6192479136149953, "pac": 0.8139992659864881, "whitened_cosine_id": 0.8496574759483337}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348304, "raptor_stability": 0.8552483320236206, "fragility": 1.0, "augmentation_robustness": 0.7851229050246351, "xie_feature_dispersion": 2.2631913339622987, "pac": 0.8201856185241279, "whitened_cosine_id": 0.8250994682312012}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 0, "layer": 2, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934627, "raptor_stability": 0.8244509100914001, "fragility": 0.5, "augmentation_robustness": 0.7136331067294028, "xie_feature_dispersion": 1.9587240977086187, "pac": 0.7690420084104015, "whitened_cosine_id": 0.7910383939743042}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 0, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934696, "raptor_stability": 0.8365420699119568, "fragility": 1.0, "augmentation_robustness": 0.7890852919390294, "xie_feature_dispersion": 1.5342572755652866, "pac": 0.8128136809254931, "whitened_cosine_id": 0.8055051565170288}} +{"model": "pythia-160m", "dataset": "subj", "seed": 0, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351844, "raptor_stability": 0.8689209222793579, "fragility": 1.0, "augmentation_robustness": 0.7746642653183333, "xie_feature_dispersion": 1.77134285145721, "pac": 0.8217925937988456, "whitened_cosine_id": 0.9039409160614014}} +{"model": "pythia-160m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352976, "raptor_stability": 0.943769633769989, "fragility": 3.0, "augmentation_robustness": 0.9172945300081299, "xie_feature_dispersion": 6.963799146343127, "pac": 0.9305320818890594, "whitened_cosine_id": 0.9455703496932983}} +{"model": "pythia-160m", "dataset": "cola", "seed": 0, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344465, "raptor_stability": 0.7988646626472473, "fragility": 0.25, "augmentation_robustness": 0.5972243911972416, "xie_feature_dispersion": 0.7758972276639047, "pac": 0.6980445269222444, "whitened_cosine_id": 0.7851360440254211}} +{"model": "pythia-160m", "dataset": "stance", "seed": 0, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.4494122497402506, "raptor_stability": 0.8332012295722961, "fragility": 1.5, "augmentation_robustness": 0.8344763697080788, "xie_feature_dispersion": 2.5953837186456754, "pac": 0.8338387996401875, "whitened_cosine_id": 0.8253653049468994}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 0, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935011, "raptor_stability": 0.866767942905426, "fragility": 0.5, "augmentation_robustness": 0.7943799335400991, "xie_feature_dispersion": 1.5213101554825748, "pac": 0.8305739382227626, "whitened_cosine_id": 0.8781046867370605}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 0, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881098, "raptor_stability": 0.8615295886993408, "fragility": 0.25, "augmentation_robustness": 0.64759763388926, "xie_feature_dispersion": 1.6884023301086555, "pac": 0.7545636112943004, "whitened_cosine_id": 0.8637628555297852}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881813, "raptor_stability": 0.8620725274085999, "fragility": 1.0, "augmentation_robustness": 0.8437283807712266, "xie_feature_dispersion": 1.923545301649778, "pac": 0.8529004540899132, "whitened_cosine_id": 0.8767724633216858}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 0, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.32280286192962354, "raptor_stability": 0.8491001725196838, "fragility": 1.0, "augmentation_robustness": 0.8717545371887874, "xie_feature_dispersion": 6.130394237589204, "pac": 0.8604273548542356, "whitened_cosine_id": 0.8844512701034546}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.1116614706117857, "raptor_stability": 0.945551872253418, "fragility": 2.0, "augmentation_robustness": 0.8856969264519378, "xie_feature_dispersion": 7.687914979274747, "pac": 0.9156243993526778, "whitened_cosine_id": 0.8968205451965332}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 0, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879029, "raptor_stability": 0.787240743637085, "fragility": 0.25, "augmentation_robustness": 0.7337481325439938, "xie_feature_dispersion": 0.12206039095754939, "pac": 0.7604944380905394, "whitened_cosine_id": 0.7804538607597351}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 0, "layer": 12, "concept": "emotion", "predictors": {"sip_eigengap": 0.4882081459032719, "raptor_stability": 0.8632135391235352, "fragility": 0.25, "augmentation_robustness": 0.7884647975902315, "xie_feature_dispersion": 2.173352441480384, "pac": 0.8258391683568833, "whitened_cosine_id": 0.8471425175666809}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 0, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880458, "raptor_stability": 0.8126577138900757, "fragility": 0.25, "augmentation_robustness": 0.7034424602081344, "xie_feature_dispersion": 4.485569514407045, "pac": 0.758050087049105, "whitened_cosine_id": 0.827558696269989}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 0, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879928, "raptor_stability": 0.7987161874771118, "fragility": 0.25, "augmentation_robustness": 0.6753132678040145, "xie_feature_dispersion": 2.5329861784126293, "pac": 0.7370147276405632, "whitened_cosine_id": 0.810127854347229}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 0, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880347, "raptor_stability": 0.814401388168335, "fragility": 0.25, "augmentation_robustness": 0.6923253341445008, "xie_feature_dispersion": 0.7524282237703519, "pac": 0.7533633611564179, "whitened_cosine_id": 0.8191361427307129}} +{"model": "pythia-410m", "dataset": "subj", "seed": 0, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882566, "raptor_stability": 0.8862898945808411, "fragility": 0.5, "augmentation_robustness": 0.7988711466010014, "xie_feature_dispersion": 1.428972872571398, "pac": 0.8425805205909213, "whitened_cosine_id": 0.8959206342697144}} +{"model": "pythia-410m", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882899, "raptor_stability": 0.9429970979690552, "fragility": 3.0, "augmentation_robustness": 0.9052157104985206, "xie_feature_dispersion": 6.72893787243834, "pac": 0.9241064042337879, "whitened_cosine_id": 0.933451235294342}} +{"model": "pythia-410m", "dataset": "cola", "seed": 0, "layer": 23, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879473, "raptor_stability": 0.7842450737953186, "fragility": 0.25, "augmentation_robustness": 0.5039294079390225, "xie_feature_dispersion": 1.1149557488538653, "pac": 0.6440872408671705, "whitened_cosine_id": 0.8007057905197144}} +{"model": "pythia-410m", "dataset": "stance", "seed": 0, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.5795818589574081, "raptor_stability": 0.8285464644432068, "fragility": 1.0, "augmentation_robustness": 0.8175707678986336, "xie_feature_dispersion": 1.8361604596405292, "pac": 0.8230586161709201, "whitened_cosine_id": 0.8221842646598816}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 0, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881513, "raptor_stability": 0.8521958589553833, "fragility": 0.5, "augmentation_robustness": 0.7625019501536046, "xie_feature_dispersion": 1.5630966219328584, "pac": 0.8073489045544939, "whitened_cosine_id": 0.8685176968574524}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 0, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.8599012494087219, "fragility": 0.25, "augmentation_robustness": 0.5848209179858852, "xie_feature_dispersion": 2.122662038410848, "pac": 0.7223610836973036, "whitened_cosine_id": 0.8232980370521545}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.825421154499054, "fragility": 1.0, "augmentation_robustness": 0.7804717428621477, "xie_feature_dispersion": 2.576726583434504, "pac": 0.8029464486806008, "whitened_cosine_id": 0.8465724587440491}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 0, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.22041139949027433, "raptor_stability": 0.8484893441200256, "fragility": 1.0, "augmentation_robustness": 0.8650683659984831, "xie_feature_dispersion": 5.917035974988243, "pac": 0.8567788550592543, "whitened_cosine_id": 0.8215498328208923}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 0, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06228035557873579, "raptor_stability": 0.9404226541519165, "fragility": 3.0, "augmentation_robustness": 0.8779983715906402, "xie_feature_dispersion": 9.493553367505593, "pac": 0.9092105128712784, "whitened_cosine_id": 0.8300829529762268}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 0, "layer": 11, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.7848396301269531, "fragility": 0.25, "augmentation_robustness": 0.7408315387590054, "xie_feature_dispersion": 0.2653617366201949, "pac": 0.7628355844429793, "whitened_cosine_id": 0.817557156085968}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.019767617442798265, "raptor_stability": 0.8671453595161438, "fragility": 0.25, "augmentation_robustness": 0.755285317721517, "xie_feature_dispersion": 2.8877683311134636, "pac": 0.8112153386188303, "whitened_cosine_id": 0.830573558807373}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 0, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010617, "raptor_stability": 0.8155287504196167, "fragility": 0.25, "augmentation_robustness": 0.6718229230312025, "xie_feature_dispersion": 5.342310320027777, "pac": 0.7436758367254096, "whitened_cosine_id": 0.8247697353363037}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 0, "layer": 22, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.7958077192306519, "fragility": 0.25, "augmentation_robustness": 0.6470101589682584, "xie_feature_dispersion": 2.217429056075072, "pac": 0.7214089390994551, "whitened_cosine_id": 0.8194364309310913}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 0, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8089179396629333, "fragility": 0.5, "augmentation_robustness": 0.7119613985713426, "xie_feature_dispersion": 1.5482508686542642, "pac": 0.760439669117138, "whitened_cosine_id": 0.8221121430397034}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 0, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010872, "raptor_stability": 0.8672937750816345, "fragility": 1.0, "augmentation_robustness": 0.7720944652439713, "xie_feature_dispersion": 2.5476390335776844, "pac": 0.819694120162803, "whitened_cosine_id": 0.8334078788757324}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 0, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010434, "raptor_stability": 0.9216615557670593, "fragility": 1.0, "augmentation_robustness": 0.895838461689926, "xie_feature_dispersion": 23.432722674772293, "pac": 0.9087500087284927, "whitened_cosine_id": 0.8588989973068237}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 0, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010818, "raptor_stability": 0.8071193099021912, "fragility": 0.25, "augmentation_robustness": 0.5249275775049415, "xie_feature_dispersion": 1.549092757263006, "pac": 0.6660234437035664, "whitened_cosine_id": 0.8179035186767578}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 0, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.36672114356076757, "raptor_stability": 0.8304018974304199, "fragility": 1.0, "augmentation_robustness": 0.8219455087263939, "xie_feature_dispersion": 3.593233328751099, "pac": 0.8261737030784069, "whitened_cosine_id": 0.8171507716178894}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8476687073707581, "fragility": 0.5, "augmentation_robustness": 0.7513937676225657, "xie_feature_dispersion": 2.045523026748139, "pac": 0.7995312374966619, "whitened_cosine_id": 0.8353421092033386}} +{"model": "gpt2", "dataset": "sst2", "seed": 0, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349679, "raptor_stability": 0.8233915567398071, "fragility": 0.25, "augmentation_robustness": 0.5123065291062753, "xie_feature_dispersion": 1.6815976407755642, "pac": 0.6678490429230413, "whitened_cosine_id": 0.857738196849823}} +{"model": "gpt2", "dataset": "imdb", "seed": 0, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350731, "raptor_stability": 0.7829911708831787, "fragility": 0.25, "augmentation_robustness": 0.6517344633023527, "xie_feature_dispersion": 1.6579993440971876, "pac": 0.7173628170927657, "whitened_cosine_id": 0.8754125237464905}} +{"model": "gpt2", "dataset": "ag_news", "seed": 0, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.34635368740834577, "raptor_stability": 0.8845524787902832, "fragility": 0.25, "augmentation_robustness": 0.9187097662639041, "xie_feature_dispersion": 0.8678892394256513, "pac": 0.9016311225270937, "whitened_cosine_id": 0.910406231880188}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 0, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.12115672881499416, "raptor_stability": 0.9302389025688171, "fragility": 1.5, "augmentation_robustness": 0.8321901130650252, "xie_feature_dispersion": 6.121815304004312, "pac": 0.8812145078169211, "whitened_cosine_id": 0.9043624997138977}} +{"model": "gpt2", "dataset": "counterfact", "seed": 0, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859340033, "raptor_stability": 0.7690413594245911, "fragility": 0.25, "augmentation_robustness": 0.809669912657523, "xie_feature_dispersion": 0.1095429604074235, "pac": 0.789355636041057, "whitened_cosine_id": 0.7072374820709229}} +{"model": "gpt2", "dataset": "emotion", "seed": 0, "layer": 4, "concept": "emotion", "predictors": {"sip_eigengap": 0.7366388390202255, "raptor_stability": 0.8725014328956604, "fragility": 0.25, "augmentation_robustness": 0.8145765018284364, "xie_feature_dispersion": 1.77009486868241, "pac": 0.8435389673620484, "whitened_cosine_id": 0.8444538116455078}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347128, "raptor_stability": 0.7693442106246948, "fragility": 0.25, "augmentation_robustness": 0.551602767544549, "xie_feature_dispersion": 3.694561283935865, "pac": 0.6604734890846219, "whitened_cosine_id": 0.8187122344970703}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 0, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346073, "raptor_stability": 0.7867448925971985, "fragility": 0.25, "augmentation_robustness": 0.6861444856119838, "xie_feature_dispersion": 2.309585917543921, "pac": 0.7364446891045912, "whitened_cosine_id": 0.8086420297622681}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 0, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346306, "raptor_stability": 0.7805789113044739, "fragility": 0.25, "augmentation_robustness": 0.657242871403201, "xie_feature_dispersion": 0.25866709333672444, "pac": 0.7189108913538375, "whitened_cosine_id": 0.803403377532959}} +{"model": "gpt2", "dataset": "subj", "seed": 0, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352106, "raptor_stability": 0.8574187755584717, "fragility": 0.25, "augmentation_robustness": 0.7295931094481585, "xie_feature_dispersion": 0.951143558340801, "pac": 0.7935059425033151, "whitened_cosine_id": 0.9090107679367065}} +{"model": "gpt2", "dataset": "spam", "seed": 0, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.39754248593527, "raptor_stability": 0.9161890149116516, "fragility": 2.0, "augmentation_robustness": 0.9103539019844458, "xie_feature_dispersion": 6.497801646724056, "pac": 0.9132714584480487, "whitened_cosine_id": 0.9386647939682007}} +{"model": "gpt2", "dataset": "cola", "seed": 0, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859346528, "raptor_stability": 0.7938734889030457, "fragility": 0.25, "augmentation_robustness": 0.49778314433790527, "xie_feature_dispersion": 0.9649847502240455, "pac": 0.6458283166204755, "whitened_cosine_id": 0.807263970375061}} +{"model": "gpt2", "dataset": "stance", "seed": 0, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.490010700608385, "raptor_stability": 0.8106656670570374, "fragility": 0.25, "augmentation_robustness": 0.7954562938419368, "xie_feature_dispersion": 2.664520632491477, "pac": 0.803060980449487, "whitened_cosine_id": 0.8153484463691711}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 0, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485935004, "raptor_stability": 0.8306007981300354, "fragility": 0.25, "augmentation_robustness": 0.6972854536891078, "xie_feature_dispersion": 1.1399162300669312, "pac": 0.7639431259095716, "whitened_cosine_id": 0.8771829605102539}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 0, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688191, "raptor_stability": 0.8413395881652832, "fragility": 0.25, "augmentation_robustness": 0.5985845188957462, "xie_feature_dispersion": 1.7719582872902115, "pac": 0.7199620535305147, "whitened_cosine_id": 0.8595471382141113}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 0, "layer": 3, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878758, "raptor_stability": 0.7550429105758667, "fragility": 0.25, "augmentation_robustness": 0.8450710203949588, "xie_feature_dispersion": 0.2304978644612739, "pac": 0.8000569654854128, "whitened_cosine_id": 0.7729927897453308}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 0, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6820984754002586, "raptor_stability": 0.8745980262756348, "fragility": 0.25, "augmentation_robustness": 0.7780721034137521, "xie_feature_dispersion": 0.8621445356293869, "pac": 0.8263350648446934, "whitened_cosine_id": 0.8492467999458313}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 0, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880176, "raptor_stability": 0.8115680813789368, "fragility": 0.25, "augmentation_robustness": 0.6729005392577236, "xie_feature_dispersion": 1.137989115163849, "pac": 0.7422343103183302, "whitened_cosine_id": 0.8243570923805237}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 0, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878862, "raptor_stability": 0.777171790599823, "fragility": 0.25, "augmentation_robustness": 0.6061778394389484, "xie_feature_dispersion": 2.1506698277748155, "pac": 0.6916748150193857, "whitened_cosine_id": 0.8085216283798218}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 0, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880017, "raptor_stability": 0.7901369333267212, "fragility": 0.25, "augmentation_robustness": 0.6555666038259415, "xie_feature_dispersion": 0.462184243542827, "pac": 0.7228517685763314, "whitened_cosine_id": 0.8134031891822815}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 0, "layer": 24, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688252, "raptor_stability": 0.8832855224609375, "fragility": 0.25, "augmentation_robustness": 0.7951101769713113, "xie_feature_dispersion": 0.4906864607581503, "pac": 0.8391978497161243, "whitened_cosine_id": 0.8952683210372925}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 0, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880159, "raptor_stability": 0.7916848063468933, "fragility": 0.25, "augmentation_robustness": 0.5224171661394365, "xie_feature_dispersion": 0.79552886696272, "pac": 0.6570509862431648, "whitened_cosine_id": 0.8178503513336182}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 0, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.42249695846338925, "raptor_stability": 0.8128862977027893, "fragility": 0.5, "augmentation_robustness": 0.8037561319272016, "xie_feature_dispersion": 2.2610696467251814, "pac": 0.8083212148149954, "whitened_cosine_id": 0.7861542701721191}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 0, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881549, "raptor_stability": 0.8164692521095276, "fragility": 0.25, "augmentation_robustness": 0.6904412835869388, "xie_feature_dispersion": 1.1005465582305016, "pac": 0.7534552678482331, "whitened_cosine_id": 0.8596013784408569}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 0, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648414, "raptor_stability": 0.8818655014038086, "fragility": 0.25, "augmentation_robustness": 0.67470015678154, "xie_feature_dispersion": 1.8204152059660144, "pac": 0.7782828290926743, "whitened_cosine_id": 0.879472017288208}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 0, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.29387292376502, "raptor_stability": 0.8775492906570435, "fragility": 1.0, "augmentation_robustness": 0.8720676344266011, "xie_feature_dispersion": 2.218615090585379, "pac": 0.8748084625418222, "whitened_cosine_id": 0.8882995247840881}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.33450754332538074, "raptor_stability": 0.8566421270370483, "fragility": 1.0, "augmentation_robustness": 0.8744251513008834, "xie_feature_dispersion": 3.7561628051141023, "pac": 0.8655336391689659, "whitened_cosine_id": 0.9043422937393188}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 0, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.07399960414320991, "raptor_stability": 0.9483939409255981, "fragility": 1.5, "augmentation_robustness": 0.891918719098521, "xie_feature_dispersion": 7.327642739764713, "pac": 0.9201563300120597, "whitened_cosine_id": 0.9209960699081421}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 0, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237645932, "raptor_stability": 0.8066194653511047, "fragility": 0.25, "augmentation_robustness": 0.8373520059306747, "xie_feature_dispersion": 0.06081582657176685, "pac": 0.8219857356408897, "whitened_cosine_id": 0.7630186080932617}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 0, "layer": 7, "concept": "emotion", "predictors": {"sip_eigengap": 0.6131208141710327, "raptor_stability": 0.8868826031684875, "fragility": 0.25, "augmentation_robustness": 0.8904501661922035, "xie_feature_dispersion": 1.992273238599221, "pac": 0.8886663846803455, "whitened_cosine_id": 0.8511517643928528}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 0, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645996, "raptor_stability": 0.8185003995895386, "fragility": 0.25, "augmentation_robustness": 0.7607385932812474, "xie_feature_dispersion": 4.466765724760196, "pac": 0.789619496435393, "whitened_cosine_id": 0.8152397871017456}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 0, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647149, "raptor_stability": 0.811665415763855, "fragility": 0.25, "augmentation_robustness": 0.7276199921424138, "xie_feature_dispersion": 2.8843177774025395, "pac": 0.7696427039531344, "whitened_cosine_id": 0.8048377633094788}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 0, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647015, "raptor_stability": 0.8171485662460327, "fragility": 0.25, "augmentation_robustness": 0.7939897357140582, "xie_feature_dispersion": 0.15381271196675575, "pac": 0.8055691509800454, "whitened_cosine_id": 0.8043486475944519}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 0, "layer": 13, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651123, "raptor_stability": 0.8921712040901184, "fragility": 0.5, "augmentation_robustness": 0.8243175866171568, "xie_feature_dispersion": 0.9955751494789675, "pac": 0.8582443953536376, "whitened_cosine_id": 0.9154570698738098}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 0, "layer": 2, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923765166, "raptor_stability": 0.9529117941856384, "fragility": 3.0, "augmentation_robustness": 0.94099314226979, "xie_feature_dispersion": 9.854132791030557, "pac": 0.9469524682277142, "whitened_cosine_id": 0.9523190855979919}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 0, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.293872923764728, "raptor_stability": 0.8035885691642761, "fragility": 0.25, "augmentation_robustness": 0.5589292088936479, "xie_feature_dispersion": 1.0666210514085848, "pac": 0.681258889028962, "whitened_cosine_id": 0.818631649017334}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 0, "layer": 20, "concept": "stance", "predictors": {"sip_eigengap": 0.4157006657150525, "raptor_stability": 0.8240557312965393, "fragility": 0.5, "augmentation_robustness": 0.8015463605814978, "xie_feature_dispersion": 3.234317390572246, "pac": 0.8128010459390185, "whitened_cosine_id": 0.8179960250854492}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 0, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649436, "raptor_stability": 0.8530629873275757, "fragility": 0.25, "augmentation_robustness": 0.7821972078583704, "xie_feature_dispersion": 1.1742203784808831, "pac": 0.8176300975929731, "whitened_cosine_id": 0.8717560768127441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921995341, "raptor_stability": 0.8336094617843628, "fragility": 0.25, "augmentation_robustness": 0.6377618570084167, "xie_feature_dispersion": 1.7265608037193576, "pac": 0.7356856593963897, "whitened_cosine_id": 0.8218536972999573}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 1, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998223, "raptor_stability": 0.8874242901802063, "fragility": 1.0, "augmentation_robustness": 0.8992101261321568, "xie_feature_dispersion": 1.271550857614664, "pac": 0.8933172081561815, "whitened_cosine_id": 0.8205482363700867}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.6806422452118264, "raptor_stability": 0.8989484906196594, "fragility": 1.5, "augmentation_robustness": 0.9168698848955824, "xie_feature_dispersion": 3.4157347611498774, "pac": 0.9079091877576209, "whitened_cosine_id": 0.9129272699356079}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 1, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.18807531413389164, "raptor_stability": 0.9559001326560974, "fragility": 2.0, "augmentation_robustness": 0.9200740132537277, "xie_feature_dispersion": 5.534601025030786, "pac": 0.9379870729549126, "whitened_cosine_id": 0.933467447757721}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 1, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.777370810508728, "fragility": 0.5, "augmentation_robustness": 0.7673906542908651, "xie_feature_dispersion": 0.6104756443427384, "pac": 0.7723807323997965, "whitened_cosine_id": 0.6849023103713989}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 1, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.5332073250620768, "raptor_stability": 0.8536012172698975, "fragility": 0.25, "augmentation_robustness": 0.7542190503987464, "xie_feature_dispersion": 1.7320048864765314, "pac": 0.803910133834322, "whitened_cosine_id": 0.8371078372001648}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 1, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921994555, "raptor_stability": 0.8290720582008362, "fragility": 0.5, "augmentation_robustness": 0.7840608903875346, "xie_feature_dispersion": 2.6910245205930696, "pac": 0.8065664742941854, "whitened_cosine_id": 0.8217921853065491}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 1, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987687, "raptor_stability": 0.819472074508667, "fragility": 1.0, "augmentation_robustness": 0.698958315186553, "xie_feature_dispersion": 2.073055712645641, "pac": 0.75921519484761, "whitened_cosine_id": 0.7341217398643494}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 1, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989486, "raptor_stability": 0.7951779961585999, "fragility": 0.5, "augmentation_robustness": 0.7589872979724266, "xie_feature_dispersion": 1.2637336763777984, "pac": 0.7770826470655132, "whitened_cosine_id": 0.7604337334632874}} +{"model": "pythia-70m", "dataset": "subj", "seed": 1, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922001947, "raptor_stability": 0.8965046405792236, "fragility": 1.5, "augmentation_robustness": 0.8297915129525963, "xie_feature_dispersion": 2.3654705725557443, "pac": 0.86314807676591, "whitened_cosine_id": 0.9005034565925598}} +{"model": "pythia-70m", "dataset": "spam", "seed": 1, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004856, "raptor_stability": 0.9503062963485718, "fragility": 3.0, "augmentation_robustness": 0.9222602495310279, "xie_feature_dispersion": 6.604643787479847, "pac": 0.9362832729397998, "whitened_cosine_id": 0.9495452046394348}} +{"model": "pythia-70m", "dataset": "cola", "seed": 1, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977986, "raptor_stability": 0.7986350655555725, "fragility": 1.0, "augmentation_robustness": 0.720600733310639, "xie_feature_dispersion": 0.7104026437376014, "pac": 0.7596178994331058, "whitened_cosine_id": 0.7400056719779968}} +{"model": "pythia-70m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6754527167162487, "raptor_stability": 0.8382856249809265, "fragility": 1.0, "augmentation_robustness": 0.8615565989812328, "xie_feature_dispersion": 1.8257872743628552, "pac": 0.8499211119810797, "whitened_cosine_id": 0.8341732621192932}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 1, "layer": 2, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997502, "raptor_stability": 0.8919016122817993, "fragility": 1.5, "augmentation_robustness": 0.8621425329457455, "xie_feature_dispersion": 1.7499668565574926, "pac": 0.8770220726137724, "whitened_cosine_id": 0.8624804019927979}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 1, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349208, "raptor_stability": 0.8506006002426147, "fragility": 0.25, "augmentation_robustness": 0.637601559557224, "xie_feature_dispersion": 1.9524296354090338, "pac": 0.7441010798999194, "whitened_cosine_id": 0.8579458594322205}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 1, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350347, "raptor_stability": 0.8709588050842285, "fragility": 1.0, "augmentation_robustness": 0.8681998579081558, "xie_feature_dispersion": 1.4529956718728516, "pac": 0.8695793314961922, "whitened_cosine_id": 0.8694334030151367}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 1, "layer": 7, "concept": "topic", "predictors": {"sip_eigengap": 0.45824187669700944, "raptor_stability": 0.8648454546928406, "fragility": 1.0, "augmentation_robustness": 0.883777230516235, "xie_feature_dispersion": 2.411760223810931, "pac": 0.8743113426045378, "whitened_cosine_id": 0.9022970199584961}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.12810819210433996, "raptor_stability": 0.9377708435058594, "fragility": 3.0, "augmentation_robustness": 0.8310133850459603, "xie_feature_dispersion": 10.50628418134277, "pac": 0.8843921142759099, "whitened_cosine_id": 0.9117757678031921}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 1, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341767, "raptor_stability": 0.7650079131126404, "fragility": 0.5, "augmentation_robustness": 0.7779032363265277, "xie_feature_dispersion": 0.6428540179632124, "pac": 0.771455574719584, "whitened_cosine_id": 0.7203051447868347}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 1, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.5407329041738655, "raptor_stability": 0.8731709122657776, "fragility": 0.5, "augmentation_robustness": 0.8006721595864167, "xie_feature_dispersion": 1.6646203889081308, "pac": 0.8369215359260971, "whitened_cosine_id": 0.854195773601532}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 1, "layer": 3, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348242, "raptor_stability": 0.8365353345870972, "fragility": 1.0, "augmentation_robustness": 0.7581229958494136, "xie_feature_dispersion": 1.9822842109390966, "pac": 0.7973291652182554, "whitened_cosine_id": 0.8330134749412537}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 1, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346446, "raptor_stability": 0.8064267039299011, "fragility": 0.25, "augmentation_robustness": 0.6783393674021946, "xie_feature_dispersion": 3.1602498907933048, "pac": 0.7423830356660479, "whitened_cosine_id": 0.7950520515441895}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 1, "layer": 8, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934659, "raptor_stability": 0.7950870990753174, "fragility": 0.25, "augmentation_robustness": 0.7522003888311254, "xie_feature_dispersion": 1.0110021271866068, "pac": 0.7736437439532213, "whitened_cosine_id": 0.7926974892616272}} +{"model": "pythia-160m", "dataset": "subj", "seed": 1, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351837, "raptor_stability": 0.8696022033691406, "fragility": 1.0, "augmentation_robustness": 0.7639011148909334, "xie_feature_dispersion": 2.2528502880398076, "pac": 0.816751659130037, "whitened_cosine_id": 0.9035648703575134}} +{"model": "pythia-160m", "dataset": "spam", "seed": 1, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352847, "raptor_stability": 0.9360153079032898, "fragility": 3.0, "augmentation_robustness": 0.9010665268870804, "xie_feature_dispersion": 7.136742674643747, "pac": 0.9185409173951851, "whitened_cosine_id": 0.9397127628326416}} +{"model": "pythia-160m", "dataset": "cola", "seed": 1, "layer": 8, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343877, "raptor_stability": 0.799544095993042, "fragility": 0.25, "augmentation_robustness": 0.6070554401031638, "xie_feature_dispersion": 0.34675513205319614, "pac": 0.7032997680481029, "whitened_cosine_id": 0.7765609622001648}} +{"model": "pythia-160m", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.6971690127369661, "raptor_stability": 0.8396480679512024, "fragility": 1.0, "augmentation_robustness": 0.8287048173260642, "xie_feature_dispersion": 2.4375720598203166, "pac": 0.8341764426386333, "whitened_cosine_id": 0.8327767848968506}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 1, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349912, "raptor_stability": 0.8510211110115051, "fragility": 0.5, "augmentation_robustness": 0.7777689965959116, "xie_feature_dispersion": 0.793449068148806, "pac": 0.8143950538037084, "whitened_cosine_id": 0.873329222202301}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 1, "layer": 10, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688131, "raptor_stability": 0.8608813881874084, "fragility": 0.25, "augmentation_robustness": 0.6344172991669839, "xie_feature_dispersion": 2.4202065461787297, "pac": 0.7476493436771962, "whitened_cosine_id": 0.8642251491546631}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 1, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881893, "raptor_stability": 0.8481273055076599, "fragility": 1.0, "augmentation_robustness": 0.8003220169247562, "xie_feature_dispersion": 2.320769849966657, "pac": 0.8242246612162081, "whitened_cosine_id": 0.8769779801368713}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.31678079404561216, "raptor_stability": 0.8443418741226196, "fragility": 1.5, "augmentation_robustness": 0.8372749145627374, "xie_feature_dispersion": 6.763574998208828, "pac": 0.8408083943426785, "whitened_cosine_id": 0.8846130967140198}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 1, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.08895215281835345, "raptor_stability": 0.9383633732795715, "fragility": 3.0, "augmentation_robustness": 0.8350381950983168, "xie_feature_dispersion": 10.874814823071377, "pac": 0.8867007841889442, "whitened_cosine_id": 0.8853277564048767}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 1, "layer": 16, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878747, "raptor_stability": 0.7895128726959229, "fragility": 0.25, "augmentation_robustness": 0.7721623022098749, "xie_feature_dispersion": 0.6522465207916613, "pac": 0.7808375874528989, "whitened_cosine_id": 0.7736839652061462}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 1, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5847740085808558, "raptor_stability": 0.8644301295280457, "fragility": 1.0, "augmentation_robustness": 0.7506737064210864, "xie_feature_dispersion": 2.0251662058673543, "pac": 0.807551917974566, "whitened_cosine_id": 0.8443336486816406}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879302, "raptor_stability": 0.8073345422744751, "fragility": 0.25, "augmentation_robustness": 0.7273212412855212, "xie_feature_dispersion": 5.356927418014672, "pac": 0.7673278917799982, "whitened_cosine_id": 0.8217640519142151}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 1, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956880057, "raptor_stability": 0.8053412437438965, "fragility": 0.25, "augmentation_robustness": 0.6599905871716598, "xie_feature_dispersion": 2.9578595858885146, "pac": 0.7326659154577781, "whitened_cosine_id": 0.8191479444503784}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 1, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687935, "raptor_stability": 0.7920674681663513, "fragility": 0.25, "augmentation_robustness": 0.7530199776918786, "xie_feature_dispersion": 2.0304992840234406, "pac": 0.772543722929115, "whitened_cosine_id": 0.8043911457061768}} +{"model": "pythia-410m", "dataset": "subj", "seed": 1, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882515, "raptor_stability": 0.8890897631645203, "fragility": 0.5, "augmentation_robustness": 0.774157157496381, "xie_feature_dispersion": 1.4231569822537715, "pac": 0.8316234603304506, "whitened_cosine_id": 0.8959898948669434}} +{"model": "pythia-410m", "dataset": "spam", "seed": 1, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956882854, "raptor_stability": 0.9181560277938843, "fragility": 3.0, "augmentation_robustness": 0.8613923232327275, "xie_feature_dispersion": 13.35429844074093, "pac": 0.889774175513306, "whitened_cosine_id": 0.9216764569282532}} +{"model": "pythia-410m", "dataset": "cola", "seed": 1, "layer": 22, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879537, "raptor_stability": 0.8038613796234131, "fragility": 0.25, "augmentation_robustness": 0.5332025416049512, "xie_feature_dispersion": 0.5611200077386973, "pac": 0.6685319606141822, "whitened_cosine_id": 0.8066012859344482}} +{"model": "pythia-410m", "dataset": "stance", "seed": 1, "layer": 14, "concept": "stance", "predictors": {"sip_eigengap": 0.5266347984206482, "raptor_stability": 0.8334753513336182, "fragility": 0.5, "augmentation_robustness": 0.8255680326919436, "xie_feature_dispersion": 2.548552169168803, "pac": 0.8295216920127808, "whitened_cosine_id": 0.82421875}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 1, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.210307295688135, "raptor_stability": 0.8675035238265991, "fragility": 1.0, "augmentation_robustness": 0.7796142351617439, "xie_feature_dispersion": 2.10629803849922, "pac": 0.8235588794941715, "whitened_cosine_id": 0.8668701648712158}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 1, "layer": 22, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8375368714332581, "fragility": 0.25, "augmentation_robustness": 0.5707958641951811, "xie_feature_dispersion": 2.2166913426405066, "pac": 0.7041663678142196, "whitened_cosine_id": 0.8229580521583557}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 1, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8483022451400757, "fragility": 1.0, "augmentation_robustness": 0.773674431811487, "xie_feature_dispersion": 2.6155304376742743, "pac": 0.8109883384757813, "whitened_cosine_id": 0.8468784689903259}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 1, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.34505361071401985, "raptor_stability": 0.8623528480529785, "fragility": 1.5, "augmentation_robustness": 0.866225226560754, "xie_feature_dispersion": 4.689887830271909, "pac": 0.8642890373068662, "whitened_cosine_id": 0.8534741997718811}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.058822548518337014, "raptor_stability": 0.9422677159309387, "fragility": 2.0, "augmentation_robustness": 0.8626061839556527, "xie_feature_dispersion": 8.534984669917208, "pac": 0.9024369499432957, "whitened_cosine_id": 0.8359106779098511}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 1, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010858, "raptor_stability": 0.7976512312889099, "fragility": 0.25, "augmentation_robustness": 0.743646934055148, "xie_feature_dispersion": 0.8601541662036642, "pac": 0.7706490826720289, "whitened_cosine_id": 0.8195480704307556}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 1, "layer": 19, "concept": "emotion", "predictors": {"sip_eigengap": 0.40445308965852644, "raptor_stability": 0.8435872197151184, "fragility": 0.5, "augmentation_robustness": 0.7387802241177431, "xie_feature_dispersion": 2.4410346204498894, "pac": 0.7911837219164308, "whitened_cosine_id": 0.8220205903053284}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 1, "layer": 23, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.7921804785728455, "fragility": 0.25, "augmentation_robustness": 0.6227259022979126, "xie_feature_dispersion": 3.0062653097311682, "pac": 0.707453190435379, "whitened_cosine_id": 0.8225616812705994}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 1, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010583, "raptor_stability": 0.8012144565582275, "fragility": 0.25, "augmentation_robustness": 0.6679021398605163, "xie_feature_dispersion": 3.9129339183725933, "pac": 0.7345582982093719, "whitened_cosine_id": 0.8250377774238586}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 1, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010749, "raptor_stability": 0.7956711649894714, "fragility": 0.25, "augmentation_robustness": 0.7021864744022904, "xie_feature_dispersion": 2.452541305670078, "pac": 0.748928819695881, "whitened_cosine_id": 0.8224895000457764}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 1, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010877, "raptor_stability": 0.8905167579650879, "fragility": 1.0, "augmentation_robustness": 0.7827455359108311, "xie_feature_dispersion": 2.6061836065846578, "pac": 0.8366311469379595, "whitened_cosine_id": 0.8333907723426819}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 1, "layer": 13, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010649, "raptor_stability": 0.9169431924819946, "fragility": 1.5, "augmentation_robustness": 0.8734181256563801, "xie_feature_dispersion": 22.457465915951406, "pac": 0.8951806590691873, "whitened_cosine_id": 0.8557054400444031}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 1, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.855816496100951, "raptor_stability": 0.8068552017211914, "fragility": 0.25, "augmentation_robustness": 0.5148135283708677, "xie_feature_dispersion": 0.6902828208408504, "pac": 0.6608343650460295, "whitened_cosine_id": 0.8203985691070557}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 1, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.4012383857529635, "raptor_stability": 0.8363339304924011, "fragility": 0.5, "augmentation_robustness": 0.8057210469674079, "xie_feature_dispersion": 3.057908787470783, "pac": 0.8210274887299045, "whitened_cosine_id": 0.8217589259147644}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 1, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010833, "raptor_stability": 0.8418669700622559, "fragility": 1.0, "augmentation_robustness": 0.7293225446950196, "xie_feature_dispersion": 1.6761850609669111, "pac": 0.7855947573786377, "whitened_cosine_id": 0.8340412378311157}} +{"model": "gpt2", "dataset": "sst2", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.39754248593502, "raptor_stability": 0.8069735169410706, "fragility": 0.25, "augmentation_robustness": 0.5092073410955438, "xie_feature_dispersion": 1.648414982559877, "pac": 0.6580904290183072, "whitened_cosine_id": 0.8618525862693787}} +{"model": "gpt2", "dataset": "imdb", "seed": 1, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350674, "raptor_stability": 0.8155258893966675, "fragility": 0.25, "augmentation_robustness": 0.6918368797960317, "xie_feature_dispersion": 1.679121568516715, "pac": 0.7536813845963496, "whitened_cosine_id": 0.8795225024223328}} +{"model": "gpt2", "dataset": "ag_news", "seed": 1, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.3994197076471763, "raptor_stability": 0.8592859506607056, "fragility": 0.5, "augmentation_robustness": 0.8643756652798957, "xie_feature_dispersion": 1.95723297240118, "pac": 0.8618308079703006, "whitened_cosine_id": 0.9026321768760681}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 1, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.13768748508218773, "raptor_stability": 0.9254063367843628, "fragility": 1.5, "augmentation_robustness": 0.7987703771800989, "xie_feature_dispersion": 5.392898240224899, "pac": 0.8620883569822309, "whitened_cosine_id": 0.9057890176773071}} +{"model": "gpt2", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342838, "raptor_stability": 0.7716324925422668, "fragility": 0.25, "augmentation_robustness": 0.7788427795247373, "xie_feature_dispersion": 0.5555909547173986, "pac": 0.775237636033502, "whitened_cosine_id": 0.7441529631614685}} +{"model": "gpt2", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5908930964643848, "raptor_stability": 0.8605509400367737, "fragility": 0.25, "augmentation_robustness": 0.7363054353676359, "xie_feature_dispersion": 1.3002246100849617, "pac": 0.7984281877022048, "whitened_cosine_id": 0.859701931476593}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 1, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348176, "raptor_stability": 0.8326132297515869, "fragility": 0.25, "augmentation_robustness": 0.7692316541993737, "xie_feature_dispersion": 0.8694719793739331, "pac": 0.8009224419754803, "whitened_cosine_id": 0.8273347020149231}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 1, "layer": 6, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934648, "raptor_stability": 0.775840163230896, "fragility": 0.25, "augmentation_robustness": 0.6870212631337806, "xie_feature_dispersion": 2.3737390065905095, "pac": 0.7314307131823383, "whitened_cosine_id": 0.7987164855003357}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 1, "layer": 12, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346262, "raptor_stability": 0.8207367062568665, "fragility": 0.25, "augmentation_robustness": 0.7990409525750108, "xie_feature_dispersion": 1.0977706807422258, "pac": 0.8098888294159386, "whitened_cosine_id": 0.7928319573402405}} +{"model": "gpt2", "dataset": "subj", "seed": 1, "layer": 7, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351993, "raptor_stability": 0.8593925833702087, "fragility": 0.5, "augmentation_robustness": 0.6926262488173032, "xie_feature_dispersion": 1.2545052058156476, "pac": 0.776009416093756, "whitened_cosine_id": 0.9041862487792969}} +{"model": "gpt2", "dataset": "spam", "seed": 1, "layer": 12, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353376, "raptor_stability": 0.9029030203819275, "fragility": 0.25, "augmentation_robustness": 0.835250541372843, "xie_feature_dispersion": 1.4550691568045606, "pac": 0.8690767808773853, "whitened_cosine_id": 0.9343510270118713}} +{"model": "gpt2", "dataset": "cola", "seed": 1, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344225, "raptor_stability": 0.804100751876831, "fragility": 0.25, "augmentation_robustness": 0.6480747813745288, "xie_feature_dispersion": 0.35759111190860965, "pac": 0.72608776662568, "whitened_cosine_id": 0.7744043469429016}} +{"model": "gpt2", "dataset": "stance", "seed": 1, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.5721762371971069, "raptor_stability": 0.8337757587432861, "fragility": 0.5, "augmentation_robustness": 0.7933240817840598, "xie_feature_dispersion": 1.1954515513989168, "pac": 0.8135499202636729, "whitened_cosine_id": 0.8244547843933105}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 1, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934976, "raptor_stability": 0.854945182800293, "fragility": 0.25, "augmentation_robustness": 0.8063951328046506, "xie_feature_dispersion": 0.46463961416788957, "pac": 0.8306701578024718, "whitened_cosine_id": 0.8690304160118103}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 1, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688203, "raptor_stability": 0.8271149396896362, "fragility": 0.25, "augmentation_robustness": 0.5702797141411624, "xie_feature_dispersion": 1.6809287391267713, "pac": 0.6986973269153993, "whitened_cosine_id": 0.8607844114303589}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 1, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 1.210307295687949, "raptor_stability": 0.791394829750061, "fragility": 0.25, "augmentation_robustness": 0.8160921898342837, "xie_feature_dispersion": 0.5899893741394086, "pac": 0.8037435097921724, "whitened_cosine_id": 0.795337975025177}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 1, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5592522550238864, "raptor_stability": 0.8738617300987244, "fragility": 0.25, "augmentation_robustness": 0.7757794731716269, "xie_feature_dispersion": 0.9112899816215909, "pac": 0.8248206016351756, "whitened_cosine_id": 0.8513531684875488}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 1, "layer": 24, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880974, "raptor_stability": 0.8126772046089172, "fragility": 0.25, "augmentation_robustness": 0.7615894417916816, "xie_feature_dispersion": 1.322075514358211, "pac": 0.7871333232002995, "whitened_cosine_id": 0.8305314779281616}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 1, "layer": 9, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956879932, "raptor_stability": 0.7768864035606384, "fragility": 0.25, "augmentation_robustness": 0.6407154851585959, "xie_feature_dispersion": 2.304316868570724, "pac": 0.7088009443596172, "whitened_cosine_id": 0.7996089458465576}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 1, "layer": 24, "concept": "offensive", "predictors": {"sip_eigengap": 1.210307295687991, "raptor_stability": 0.808860182762146, "fragility": 0.25, "augmentation_robustness": 0.779057007645201, "xie_feature_dispersion": 1.0954582686409513, "pac": 0.7939585952036735, "whitened_cosine_id": 0.8093989491462708}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 1, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882626, "raptor_stability": 0.8778501749038696, "fragility": 0.5, "augmentation_robustness": 0.7528926011129462, "xie_feature_dispersion": 1.211853663340529, "pac": 0.8153713880084079, "whitened_cosine_id": 0.8855262398719788}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 1, "layer": 16, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880065, "raptor_stability": 0.8000022768974304, "fragility": 0.25, "augmentation_robustness": 0.5320479658836585, "xie_feature_dispersion": 0.37533178694301084, "pac": 0.6660251213905445, "whitened_cosine_id": 0.8167566061019897}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 1, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.4372306810871382, "raptor_stability": 0.8121774792671204, "fragility": 0.5, "augmentation_robustness": 0.7829438977858721, "xie_feature_dispersion": 1.6812559297796246, "pac": 0.7975606885264963, "whitened_cosine_id": 0.7655757069587708}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 1, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.210307295688115, "raptor_stability": 0.8233330845832825, "fragility": 0.25, "augmentation_robustness": 0.6852795330724812, "xie_feature_dispersion": 0.620429769989345, "pac": 0.7543063088278819, "whitened_cosine_id": 0.856203019618988}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 1, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237649584, "raptor_stability": 0.8650567531585693, "fragility": 0.25, "augmentation_robustness": 0.6688278069812248, "xie_feature_dispersion": 2.1604793022510873, "pac": 0.7669422800698971, "whitened_cosine_id": 0.8860111236572266}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 1, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923765019, "raptor_stability": 0.8548077940940857, "fragility": 1.0, "augmentation_robustness": 0.8296912251707328, "xie_feature_dispersion": 2.619265146348305, "pac": 0.8422495096324092, "whitened_cosine_id": 0.8861786127090454}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 1, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.3284735817230639, "raptor_stability": 0.865305483341217, "fragility": 1.0, "augmentation_robustness": 0.8707412223835058, "xie_feature_dispersion": 3.3146556682562625, "pac": 0.8680233528623614, "whitened_cosine_id": 0.9071072340011597}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 1, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.06700207827316693, "raptor_stability": 0.9375728368759155, "fragility": 3.0, "augmentation_robustness": 0.8370451297198389, "xie_feature_dispersion": 9.643792516378186, "pac": 0.8873089832978772, "whitened_cosine_id": 0.8979226350784302}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 1, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646287, "raptor_stability": 0.8297770023345947, "fragility": 0.25, "augmentation_robustness": 0.8462807577423243, "xie_feature_dispersion": 0.5968347455995571, "pac": 0.8380288800384594, "whitened_cosine_id": 0.7946618795394897}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 1, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.6161992460964921, "raptor_stability": 0.882070004940033, "fragility": 0.25, "augmentation_robustness": 0.8630147174643309, "xie_feature_dispersion": 2.820954214278998, "pac": 0.8725423612021819, "whitened_cosine_id": 0.8536849617958069}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 1, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645619, "raptor_stability": 0.8290610313415527, "fragility": 0.25, "augmentation_robustness": 0.7926104311252984, "xie_feature_dispersion": 5.467260144092137, "pac": 0.8108357312334256, "whitened_cosine_id": 0.8316267728805542}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 1, "layer": 20, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237647093, "raptor_stability": 0.7731820940971375, "fragility": 0.25, "augmentation_robustness": 0.6355369799746609, "xie_feature_dispersion": 2.999448892433556, "pac": 0.7043595370358992, "whitened_cosine_id": 0.798532247543335}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 1, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237645867, "raptor_stability": 0.819866955280304, "fragility": 0.25, "augmentation_robustness": 0.8020143361928123, "xie_feature_dispersion": 2.0378126951408726, "pac": 0.8109406457365581, "whitened_cosine_id": 0.8144164681434631}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 1, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.293872923765124, "raptor_stability": 0.8871787190437317, "fragility": 0.5, "augmentation_robustness": 0.7845579254901294, "xie_feature_dispersion": 1.6425796792755814, "pac": 0.8358683222669305, "whitened_cosine_id": 0.9134312868118286}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 1, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237639786, "raptor_stability": 0.9124530553817749, "fragility": 0.5, "augmentation_robustness": 0.907099253659993, "xie_feature_dispersion": 18.266615233549604, "pac": 0.909776154520884, "whitened_cosine_id": 0.9418519139289856}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 1, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647346, "raptor_stability": 0.8199473023414612, "fragility": 0.25, "augmentation_robustness": 0.5916327937131608, "xie_feature_dispersion": 0.5437385315920449, "pac": 0.7057900480273109, "whitened_cosine_id": 0.8188185691833496}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 1, "layer": 22, "concept": "stance", "predictors": {"sip_eigengap": 0.5082992376979554, "raptor_stability": 0.8089979290962219, "fragility": 1.0, "augmentation_robustness": 0.7725685252374112, "xie_feature_dispersion": 2.217566060904384, "pac": 0.7907832271668165, "whitened_cosine_id": 0.8215915560722351}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 1, "layer": 9, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649438, "raptor_stability": 0.8504278063774109, "fragility": 0.25, "augmentation_robustness": 0.7801897363682817, "xie_feature_dispersion": 0.4453040461938641, "pac": 0.8153087713728463, "whitened_cosine_id": 0.8711152076721191}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 2, "layer": 1, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996196, "raptor_stability": 0.8725253343582153, "fragility": 1.0, "augmentation_robustness": 0.7320160436555694, "xie_feature_dispersion": 1.8850577574441822, "pac": 0.8022706890068924, "whitened_cosine_id": 0.8443945646286011}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998612, "raptor_stability": 0.8727290630340576, "fragility": 1.0, "augmentation_robustness": 0.8893448732634346, "xie_feature_dispersion": 1.1304006682636438, "pac": 0.8810369681487461, "whitened_cosine_id": 0.8432037830352783}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5762362270014945, "raptor_stability": 0.8451294302940369, "fragility": 1.0, "augmentation_robustness": 0.8227084099515827, "xie_feature_dispersion": 4.598590726152205, "pac": 0.8339189201228098, "whitened_cosine_id": 0.9046899676322937}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.13862003115739582, "raptor_stability": 0.9396008253097534, "fragility": 2.0, "augmentation_robustness": 0.8714397420888463, "xie_feature_dispersion": 7.460918475973291, "pac": 0.9055202836992999, "whitened_cosine_id": 0.9276723265647888}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921976638, "raptor_stability": 0.7323077321052551, "fragility": 0.25, "augmentation_robustness": 0.6404787869477748, "xie_feature_dispersion": 0.06278074668041304, "pac": 0.686393259526515, "whitened_cosine_id": 0.7365422248840332}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 2, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.6201478550392475, "raptor_stability": 0.860109269618988, "fragility": 0.5, "augmentation_robustness": 0.7869247885334385, "xie_feature_dispersion": 1.544046125551409, "pac": 0.8235170290762133, "whitened_cosine_id": 0.841076672077179}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 2, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.7116329921991476, "raptor_stability": 0.8527593612670898, "fragility": 1.0, "augmentation_robustness": 0.794283550859188, "xie_feature_dispersion": 1.5619058370885555, "pac": 0.823521456063139, "whitened_cosine_id": 0.7776172161102295}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987974, "raptor_stability": 0.8192676901817322, "fragility": 1.0, "augmentation_robustness": 0.7427685123086125, "xie_feature_dispersion": 1.857559017759172, "pac": 0.7810181012451723, "whitened_cosine_id": 0.7365460991859436}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990956, "raptor_stability": 0.8345112800598145, "fragility": 1.0, "augmentation_robustness": 0.7886622062234387, "xie_feature_dispersion": 1.342889203260779, "pac": 0.8115867431416266, "whitened_cosine_id": 0.8005088567733765}} +{"model": "pythia-70m", "dataset": "subj", "seed": 2, "layer": 2, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.711632992200211, "raptor_stability": 0.8931885957717896, "fragility": 1.5, "augmentation_robustness": 0.8300863608388453, "xie_feature_dispersion": 2.385960236411005, "pac": 0.8616374783053175, "whitened_cosine_id": 0.9023992419242859}} +{"model": "pythia-70m", "dataset": "spam", "seed": 2, "layer": 4, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004394, "raptor_stability": 0.9427627921104431, "fragility": 1.5, "augmentation_robustness": 0.9205893937056647, "xie_feature_dispersion": 10.434188689895425, "pac": 0.9316760929080539, "whitened_cosine_id": 0.9575247764587402}} +{"model": "pythia-70m", "dataset": "cola", "seed": 2, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921978055, "raptor_stability": 0.7819207310676575, "fragility": 1.0, "augmentation_robustness": 0.7494339269084461, "xie_feature_dispersion": 0.5849615167254145, "pac": 0.7656773289880519, "whitened_cosine_id": 0.7173330187797546}} +{"model": "pythia-70m", "dataset": "stance", "seed": 2, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.467708916223557, "raptor_stability": 0.823031485080719, "fragility": 1.0, "augmentation_robustness": 0.8142662820172029, "xie_feature_dispersion": 1.6869566033029149, "pac": 0.8186488835489609, "whitened_cosine_id": 0.8191723823547363}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 2, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921998414, "raptor_stability": 0.8615893721580505, "fragility": 0.5, "augmentation_robustness": 0.8216595763224332, "xie_feature_dispersion": 0.9880768598724781, "pac": 0.8416244742402419, "whitened_cosine_id": 0.8701478838920593}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 2, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859348855, "raptor_stability": 0.8407484292984009, "fragility": 0.25, "augmentation_robustness": 0.5918794762745934, "xie_feature_dispersion": 2.033621000673696, "pac": 0.7163139527864971, "whitened_cosine_id": 0.8444841504096985}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 2, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350019, "raptor_stability": 0.8570092916488647, "fragility": 1.0, "augmentation_robustness": 0.8472864724655493, "xie_feature_dispersion": 1.332977638352146, "pac": 0.852147882057207, "whitened_cosine_id": 0.8642057180404663}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.4756762658594781, "raptor_stability": 0.8755650520324707, "fragility": 1.0, "augmentation_robustness": 0.8736053793270782, "xie_feature_dispersion": 2.6831447841823626, "pac": 0.8745852156797744, "whitened_cosine_id": 0.9056297540664673}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.10936523074513556, "raptor_stability": 0.938740074634552, "fragility": 2.0, "augmentation_robustness": 0.849824439887086, "xie_feature_dispersion": 8.126249426404561, "pac": 0.894282257260819, "whitened_cosine_id": 0.9102428555488586}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343164, "raptor_stability": 0.7734526991844177, "fragility": 0.25, "augmentation_robustness": 0.7341807784351317, "xie_feature_dispersion": 0.24746600780733177, "pac": 0.7538167388097747, "whitened_cosine_id": 0.7335423827171326}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 2, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.5898968502609128, "raptor_stability": 0.8509833216667175, "fragility": 0.25, "augmentation_robustness": 0.7503952782199734, "xie_feature_dispersion": 1.9109009411804163, "pac": 0.8006892999433455, "whitened_cosine_id": 0.8450579643249512}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 2, "layer": 8, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347338, "raptor_stability": 0.8163250684738159, "fragility": 0.25, "augmentation_robustness": 0.720428393459381, "xie_feature_dispersion": 1.672145195874313, "pac": 0.7683767309665985, "whitened_cosine_id": 0.8197868466377258}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 2, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345715, "raptor_stability": 0.8170239925384521, "fragility": 1.0, "augmentation_robustness": 0.7672939677957877, "xie_feature_dispersion": 2.1306331599217914, "pac": 0.7921589801671199, "whitened_cosine_id": 0.7781753540039062}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 2, "layer": 7, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345553, "raptor_stability": 0.8093597888946533, "fragility": 0.25, "augmentation_robustness": 0.7487295667365286, "xie_feature_dispersion": 1.9731796308382967, "pac": 0.779044677815591, "whitened_cosine_id": 0.7859788537025452}} +{"model": "pythia-160m", "dataset": "subj", "seed": 2, "layer": 6, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935202, "raptor_stability": 0.8897599577903748, "fragility": 1.0, "augmentation_robustness": 0.8051559896221852, "xie_feature_dispersion": 1.9497282630608124, "pac": 0.84745797370628, "whitened_cosine_id": 0.9099186658859253}} +{"model": "pythia-160m", "dataset": "spam", "seed": 2, "layer": 8, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935196, "raptor_stability": 0.9361570477485657, "fragility": 1.5, "augmentation_robustness": 0.9125687730559409, "xie_feature_dispersion": 12.123957991903673, "pac": 0.9243629104022533, "whitened_cosine_id": 0.9451765418052673}} +{"model": "pythia-160m", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934449, "raptor_stability": 0.7818026542663574, "fragility": 0.25, "augmentation_robustness": 0.6248388717067496, "xie_feature_dispersion": 0.36909274612641374, "pac": 0.7033207629865534, "whitened_cosine_id": 0.7660055756568909}} +{"model": "pythia-160m", "dataset": "stance", "seed": 2, "layer": 1, "concept": "stance", "predictors": {"sip_eigengap": 0.550133319968111, "raptor_stability": 0.8319461345672607, "fragility": 1.5, "augmentation_robustness": 0.8296352226027173, "xie_feature_dispersion": 2.567878497868356, "pac": 0.830790678584989, "whitened_cosine_id": 0.8248243927955627}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350152, "raptor_stability": 0.8500860929489136, "fragility": 0.5, "augmentation_robustness": 0.7916745698533383, "xie_feature_dispersion": 1.015287990938809, "pac": 0.8208803314011259, "whitened_cosine_id": 0.8699454069137573}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880925, "raptor_stability": 0.8507713079452515, "fragility": 0.25, "augmentation_robustness": 0.6548658634701003, "xie_feature_dispersion": 2.5382673428676514, "pac": 0.7528185857076759, "whitened_cosine_id": 0.8639923334121704}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881815, "raptor_stability": 0.8507294654846191, "fragility": 1.0, "augmentation_robustness": 0.8077367041653364, "xie_feature_dispersion": 2.1999878871216194, "pac": 0.8292330848249778, "whitened_cosine_id": 0.876013994216919}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 2, "layer": 2, "concept": "topic", "predictors": {"sip_eigengap": 0.48444830765922703, "raptor_stability": 0.8797903060913086, "fragility": 1.5, "augmentation_robustness": 0.8897077200880827, "xie_feature_dispersion": 4.318369085246188, "pac": 0.8847490130896957, "whitened_cosine_id": 0.8930330276489258}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06934869454749534, "raptor_stability": 0.9469916224479675, "fragility": 2.0, "augmentation_robustness": 0.8838004807942493, "xie_feature_dispersion": 7.256347007074097, "pac": 0.9153960516211084, "whitened_cosine_id": 0.8971635699272156}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878984, "raptor_stability": 0.7757969498634338, "fragility": 0.25, "augmentation_robustness": 0.7602377504416505, "xie_feature_dispersion": 0.20512071078637148, "pac": 0.7680173501525422, "whitened_cosine_id": 0.7685407996177673}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 2, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.5276419919506735, "raptor_stability": 0.8558531403541565, "fragility": 0.25, "augmentation_robustness": 0.7811194669553142, "xie_feature_dispersion": 2.490653998171255, "pac": 0.8184863036547354, "whitened_cosine_id": 0.8436367511749268}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 2, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880245, "raptor_stability": 0.8074635863304138, "fragility": 0.5, "augmentation_robustness": 0.7107950361323764, "xie_feature_dispersion": 1.9566993340581085, "pac": 0.759129311231395, "whitened_cosine_id": 0.8125779628753662}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295687971, "raptor_stability": 0.7986416220664978, "fragility": 0.5, "augmentation_robustness": 0.6903031510581926, "xie_feature_dispersion": 1.8402433518812902, "pac": 0.7444723865623453, "whitened_cosine_id": 0.7969480156898499}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 2, "layer": 5, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879715, "raptor_stability": 0.8084327578544617, "fragility": 0.5, "augmentation_robustness": 0.7257825239499871, "xie_feature_dispersion": 1.7043242651025499, "pac": 0.7671076409022244, "whitened_cosine_id": 0.7994747161865234}} +{"model": "pythia-410m", "dataset": "subj", "seed": 2, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688262, "raptor_stability": 0.8808537721633911, "fragility": 0.5, "augmentation_robustness": 0.7730695710492665, "xie_feature_dispersion": 1.6627162013824885, "pac": 0.8269616716063288, "whitened_cosine_id": 0.8980955481529236}} +{"model": "pythia-410m", "dataset": "spam", "seed": 2, "layer": 7, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956871068, "raptor_stability": 0.9453094601631165, "fragility": 1.0, "augmentation_robustness": 0.931993646437433, "xie_feature_dispersion": 17.657395386814123, "pac": 0.9386515533002747, "whitened_cosine_id": 0.9302859306335449}} +{"model": "pythia-410m", "dataset": "cola", "seed": 2, "layer": 17, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879637, "raptor_stability": 0.7948347926139832, "fragility": 0.25, "augmentation_robustness": 0.6269925244354974, "xie_feature_dispersion": 0.47979217522467377, "pac": 0.7109136585247402, "whitened_cosine_id": 0.8036185503005981}} +{"model": "pythia-410m", "dataset": "stance", "seed": 2, "layer": 10, "concept": "stance", "predictors": {"sip_eigengap": 0.47549258334344463, "raptor_stability": 0.8262801170349121, "fragility": 0.5, "augmentation_robustness": 0.8202575512360268, "xie_feature_dispersion": 3.2823490583418935, "pac": 0.8232688341354695, "whitened_cosine_id": 0.8195136189460754}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 2, "layer": 18, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881415, "raptor_stability": 0.8284711837768555, "fragility": 0.5, "augmentation_robustness": 0.7694913743994085, "xie_feature_dispersion": 1.2261896328299637, "pac": 0.798981279088132, "whitened_cosine_id": 0.8577603101730347}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 2, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010692, "raptor_stability": 0.8523985147476196, "fragility": 0.25, "augmentation_robustness": 0.6344970850996317, "xie_feature_dispersion": 3.125584115559324, "pac": 0.7434477999236256, "whitened_cosine_id": 0.8270771503448486}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010866, "raptor_stability": 0.8408520221710205, "fragility": 1.0, "augmentation_robustness": 0.7462513872391434, "xie_feature_dispersion": 2.261174552822513, "pac": 0.793551704705082, "whitened_cosine_id": 0.8419722318649292}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 2, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.3112674068238993, "raptor_stability": 0.8652853965759277, "fragility": 1.0, "augmentation_robustness": 0.8618662021334369, "xie_feature_dispersion": 4.2348607484097185, "pac": 0.8635757993546823, "whitened_cosine_id": 0.8362071514129639}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05012914539691602, "raptor_stability": 0.9452676773071289, "fragility": 3.0, "augmentation_robustness": 0.8590266216701495, "xie_feature_dispersion": 9.524616008618372, "pac": 0.9021471494886393, "whitened_cosine_id": 0.8305992484092712}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 2, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.7967193126678467, "fragility": 0.25, "augmentation_robustness": 0.755117458917892, "xie_feature_dispersion": 0.32762786809964206, "pac": 0.7759183857928693, "whitened_cosine_id": 0.8186400532722473}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 2, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.3604267252609762, "raptor_stability": 0.8496060371398926, "fragility": 0.5, "augmentation_robustness": 0.7497415128763687, "xie_feature_dispersion": 2.6993559873766775, "pac": 0.7996737750081306, "whitened_cosine_id": 0.8220704197883606}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 2, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010836, "raptor_stability": 0.8090983033180237, "fragility": 0.25, "augmentation_robustness": 0.6782138616148163, "xie_feature_dispersion": 3.2068116736663326, "pac": 0.74365608246642, "whitened_cosine_id": 0.8261517286300659}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 2, "layer": 7, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010822, "raptor_stability": 0.8073044419288635, "fragility": 0.25, "augmentation_robustness": 0.6867498455562944, "xie_feature_dispersion": 4.820388126528082, "pac": 0.747027143742579, "whitened_cosine_id": 0.8241050839424133}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 2, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010454, "raptor_stability": 0.7936219573020935, "fragility": 0.25, "augmentation_robustness": 0.6887170523418908, "xie_feature_dispersion": 2.728943510383841, "pac": 0.7411695048219922, "whitened_cosine_id": 0.8242641091346741}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 2, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010878, "raptor_stability": 0.8761361837387085, "fragility": 1.0, "augmentation_robustness": 0.7502694820898129, "xie_feature_dispersion": 2.7954401496665007, "pac": 0.8132028329142607, "whitened_cosine_id": 0.829967737197876}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 2, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010256, "raptor_stability": 0.9437923431396484, "fragility": 1.0, "augmentation_robustness": 0.9235808889611689, "xie_feature_dispersion": 23.76038075342399, "pac": 0.9336866160504087, "whitened_cosine_id": 0.8749403953552246}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 2, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010881, "raptor_stability": 0.8045715689659119, "fragility": 0.25, "augmentation_robustness": 0.5269196391645194, "xie_feature_dispersion": 0.5939571817383262, "pac": 0.6657456040652157, "whitened_cosine_id": 0.8205759525299072}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 2, "layer": 11, "concept": "stance", "predictors": {"sip_eigengap": 0.33865901445727387, "raptor_stability": 0.8298296332359314, "fragility": 1.0, "augmentation_robustness": 0.817334225300112, "xie_feature_dispersion": 3.230859359898638, "pac": 0.8235819292680218, "whitened_cosine_id": 0.8170763850212097}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 2, "layer": 14, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010837, "raptor_stability": 0.8219150900840759, "fragility": 0.5, "augmentation_robustness": 0.7537574160413394, "xie_feature_dispersion": 1.8431532727005182, "pac": 0.7878362530627077, "whitened_cosine_id": 0.8218834400177002}} +{"model": "gpt2", "dataset": "sst2", "seed": 2, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350012, "raptor_stability": 0.8311133980751038, "fragility": 0.25, "augmentation_robustness": 0.5968769748584443, "xie_feature_dispersion": 2.0407834350862544, "pac": 0.713995186466774, "whitened_cosine_id": 0.8748435378074646}} +{"model": "gpt2", "dataset": "imdb", "seed": 2, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350318, "raptor_stability": 0.8567911386489868, "fragility": 0.25, "augmentation_robustness": 0.8673792464607593, "xie_feature_dispersion": 0.26606041071725683, "pac": 0.862085192554873, "whitened_cosine_id": 0.8767779469490051}} +{"model": "gpt2", "dataset": "ag_news", "seed": 2, "layer": 3, "concept": "topic", "predictors": {"sip_eigengap": 0.3821667228940741, "raptor_stability": 0.845832109451294, "fragility": 0.5, "augmentation_robustness": 0.8470052851683942, "xie_feature_dispersion": 1.979682604320564, "pac": 0.8464186973098441, "whitened_cosine_id": 0.903249979019165}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 2, "layer": 7, "concept": "topic", "predictors": {"sip_eigengap": 0.11558697704158609, "raptor_stability": 0.932966411113739, "fragility": 1.0, "augmentation_robustness": 0.8559305238325257, "xie_feature_dispersion": 4.299506274009608, "pac": 0.8944484674731323, "whitened_cosine_id": 0.9021454453468323}} +{"model": "gpt2", "dataset": "counterfact", "seed": 2, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343073, "raptor_stability": 0.7622057199478149, "fragility": 0.25, "augmentation_robustness": 0.7918529895588636, "xie_feature_dispersion": 0.33360574253280906, "pac": 0.7770293547533393, "whitened_cosine_id": 0.7459770441055298}} +{"model": "gpt2", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5200547285263808, "raptor_stability": 0.85597825050354, "fragility": 0.25, "augmentation_robustness": 0.7622544507506325, "xie_feature_dispersion": 1.2683994964221992, "pac": 0.8091163506270862, "whitened_cosine_id": 0.8535709977149963}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 2, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.397542485934766, "raptor_stability": 0.8122290968894958, "fragility": 0.25, "augmentation_robustness": 0.7701636575129145, "xie_feature_dispersion": 0.6036175754728635, "pac": 0.7911963772012052, "whitened_cosine_id": 0.8273720741271973}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 2, "layer": 12, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859344792, "raptor_stability": 0.7899907827377319, "fragility": 0.25, "augmentation_robustness": 0.775070377256843, "xie_feature_dispersion": 1.1126097391994054, "pac": 0.7825305799972875, "whitened_cosine_id": 0.7636452317237854}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346548, "raptor_stability": 0.7969914674758911, "fragility": 0.25, "augmentation_robustness": 0.7083710583755256, "xie_feature_dispersion": 1.8684634814954304, "pac": 0.7526812629257084, "whitened_cosine_id": 0.8185012936592102}} +{"model": "gpt2", "dataset": "subj", "seed": 2, "layer": 8, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859352377, "raptor_stability": 0.8654428720474243, "fragility": 0.5, "augmentation_robustness": 0.7478951378849938, "xie_feature_dispersion": 1.4494284784328508, "pac": 0.8066690049662091, "whitened_cosine_id": 0.9153006076812744}} +{"model": "gpt2", "dataset": "spam", "seed": 2, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 1.397542485935319, "raptor_stability": 0.9391940832138062, "fragility": 1.0, "augmentation_robustness": 0.9242307247684519, "xie_feature_dispersion": 13.972658689577166, "pac": 0.9317124039911291, "whitened_cosine_id": 0.9425793290138245}} +{"model": "gpt2", "dataset": "cola", "seed": 2, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859345233, "raptor_stability": 0.7784033417701721, "fragility": 0.25, "augmentation_robustness": 0.5640885986826307, "xie_feature_dispersion": 0.38566559649240273, "pac": 0.6712459702264014, "whitened_cosine_id": 0.7816012501716614}} +{"model": "gpt2", "dataset": "stance", "seed": 2, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.5282860197072011, "raptor_stability": 0.8207774758338928, "fragility": 0.25, "augmentation_robustness": 0.7971149286838108, "xie_feature_dispersion": 2.2792793583875905, "pac": 0.8089462022588518, "whitened_cosine_id": 0.8131434321403503}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 2, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.397542485934994, "raptor_stability": 0.808012068271637, "fragility": 0.25, "augmentation_robustness": 0.7076387264686993, "xie_feature_dispersion": 0.8358972401809576, "pac": 0.7578253973701681, "whitened_cosine_id": 0.8693022727966309}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 2, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881729, "raptor_stability": 0.8265162706375122, "fragility": 0.25, "augmentation_robustness": 0.5469683098756675, "xie_feature_dispersion": 2.1253471965524975, "pac": 0.6867422902565898, "whitened_cosine_id": 0.8618488311767578}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 2, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.35097833794424904, "raptor_stability": 0.8510255813598633, "fragility": 0.5, "augmentation_robustness": 0.868126023664805, "xie_feature_dispersion": 1.511481734340918, "pac": 0.8595758025123341, "whitened_cosine_id": 0.8767551183700562}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 2, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878433, "raptor_stability": 0.7515677809715271, "fragility": 0.25, "augmentation_robustness": 0.8471055642765235, "xie_feature_dispersion": 0.3165629770195801, "pac": 0.7993366726240253, "whitened_cosine_id": 0.7615948915481567}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 2, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5517110732147434, "raptor_stability": 0.8649501204490662, "fragility": 0.25, "augmentation_robustness": 0.7727700403781212, "xie_feature_dispersion": 0.9242099717833849, "pac": 0.8188600804135937, "whitened_cosine_id": 0.8501837849617004}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 2, "layer": 17, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880352, "raptor_stability": 0.7788109183311462, "fragility": 0.25, "augmentation_robustness": 0.602822870481301, "xie_feature_dispersion": 2.344924657926786, "pac": 0.6908168944062236, "whitened_cosine_id": 0.8237905502319336}} +{"model": "gpt2-medium", "dataset": "tweet_irony", "seed": 2, "layer": 17, "concept": "irony", "predictors": {"sip_eigengap": 1.210307295688015, "raptor_stability": 0.7836035490036011, "fragility": 0.25, "augmentation_robustness": 0.6445047638465966, "xie_feature_dispersion": 2.829344259675601, "pac": 0.7140541564250988, "whitened_cosine_id": 0.8160997629165649}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 2, "layer": 2, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880054, "raptor_stability": 0.7852885723114014, "fragility": 0.25, "augmentation_robustness": 0.6693173469866233, "xie_feature_dispersion": 0.831323350714088, "pac": 0.7273029596490124, "whitened_cosine_id": 0.8087374567985535}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 2, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882837, "raptor_stability": 0.8820376992225647, "fragility": 0.5, "augmentation_robustness": 0.7285834645747785, "xie_feature_dispersion": 1.6491215963892647, "pac": 0.8053105818986717, "whitened_cosine_id": 0.8916261792182922}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 2, "layer": 19, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880483, "raptor_stability": 0.7810653448104858, "fragility": 0.25, "augmentation_robustness": 0.5184999598231592, "xie_feature_dispersion": 0.4506855901276054, "pac": 0.6497826523168225, "whitened_cosine_id": 0.8220202326774597}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 2, "layer": 24, "concept": "stance", "predictors": {"sip_eigengap": 0.4048263633949508, "raptor_stability": 0.8287909626960754, "fragility": 0.25, "augmentation_robustness": 0.8148653549769995, "xie_feature_dispersion": 0.8817497017545193, "pac": 0.8218281588365375, "whitened_cosine_id": 0.8191941380500793}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 2, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881538, "raptor_stability": 0.8170264959335327, "fragility": 0.25, "augmentation_robustness": 0.6658562775951427, "xie_feature_dispersion": 0.6916798290341591, "pac": 0.7414413867643377, "whitened_cosine_id": 0.8560160398483276}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 2, "layer": 13, "concept": "sentiment", "predictors": {"sip_eigengap": 1.293872923764946, "raptor_stability": 0.8504462838172913, "fragility": 0.25, "augmentation_robustness": 0.6405801888029593, "xie_feature_dispersion": 2.428287550117154, "pac": 0.7455132363101253, "whitened_cosine_id": 0.8769184350967407}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 2, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650424, "raptor_stability": 0.8790234923362732, "fragility": 1.0, "augmentation_robustness": 0.873146163657816, "xie_feature_dispersion": 2.109825225717331, "pac": 0.8760848279970446, "whitened_cosine_id": 0.8947857618331909}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 2, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.4777053541327048, "raptor_stability": 0.8827468156814575, "fragility": 0.5, "augmentation_robustness": 0.9086818995974959, "xie_feature_dispersion": 1.9664542049888671, "pac": 0.8957143576394767, "whitened_cosine_id": 0.9089540839195251}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 2, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.05912307025918351, "raptor_stability": 0.9463669061660767, "fragility": 1.5, "augmentation_robustness": 0.8757474533027418, "xie_feature_dispersion": 6.768079984244763, "pac": 0.9110571797344093, "whitened_cosine_id": 0.9140043258666992}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 2, "layer": 12, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646398, "raptor_stability": 0.8356521129608154, "fragility": 0.25, "augmentation_robustness": 0.8160727049350692, "xie_feature_dispersion": 0.23710320099138982, "pac": 0.8258624089479423, "whitened_cosine_id": 0.7763546705245972}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 2, "layer": 16, "concept": "emotion", "predictors": {"sip_eigengap": 0.48285928508127296, "raptor_stability": 0.8633195757865906, "fragility": 0.25, "augmentation_robustness": 0.8474371685693614, "xie_feature_dispersion": 2.284945012773296, "pac": 0.855378372177976, "whitened_cosine_id": 0.8482233881950378}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 2, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237645508, "raptor_stability": 0.8284585475921631, "fragility": 0.25, "augmentation_robustness": 0.7660624849853699, "xie_feature_dispersion": 3.252034579020222, "pac": 0.7972605162887665, "whitened_cosine_id": 0.8060616850852966}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 2, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237642506, "raptor_stability": 0.8207340240478516, "fragility": 0.25, "augmentation_robustness": 0.733566852009586, "xie_feature_dispersion": 3.71696516386688, "pac": 0.7771504380287189, "whitened_cosine_id": 0.8012250661849976}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 2, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237644577, "raptor_stability": 0.8056316375732422, "fragility": 0.25, "augmentation_robustness": 0.7408950314947991, "xie_feature_dispersion": 2.6206447748583725, "pac": 0.7732633345340206, "whitened_cosine_id": 0.8052534461021423}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 2, "layer": 12, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.29387292376514, "raptor_stability": 0.8994405269622803, "fragility": 0.5, "augmentation_robustness": 0.8304811500465553, "xie_feature_dispersion": 1.2535193562102473, "pac": 0.8649608385044179, "whitened_cosine_id": 0.9212391972541809}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 2, "layer": 13, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237650766, "raptor_stability": 0.9231064319610596, "fragility": 1.0, "augmentation_robustness": 0.9155044674854933, "xie_feature_dispersion": 18.157099012256424, "pac": 0.9193054497232764, "whitened_cosine_id": 0.9482038021087646}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 2, "layer": 14, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237647517, "raptor_stability": 0.8107694983482361, "fragility": 0.25, "augmentation_robustness": 0.6126361282502457, "xie_feature_dispersion": 0.29378764734852275, "pac": 0.7117028132992409, "whitened_cosine_id": 0.8085699677467346}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 2, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.47687488644054804, "raptor_stability": 0.8371041417121887, "fragility": 0.25, "augmentation_robustness": 0.8437004456467965, "xie_feature_dispersion": 2.6985372907022853, "pac": 0.8404022936794926, "whitened_cosine_id": 0.8233371376991272}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 2, "layer": 14, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.293872923764935, "raptor_stability": 0.8342365026473999, "fragility": 0.25, "augmentation_robustness": 0.7913619262318612, "xie_feature_dispersion": 0.5692187925507834, "pac": 0.8127992144396305, "whitened_cosine_id": 0.8644318580627441}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996696, "raptor_stability": 0.8488397598266602, "fragility": 0.25, "augmentation_robustness": 0.5708345486394, "xie_feature_dispersion": 2.9434633106744124, "pac": 0.7098371542330301, "whitened_cosine_id": 0.8510316610336304}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 3, "layer": 4, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921998183, "raptor_stability": 0.8842445015907288, "fragility": 1.0, "augmentation_robustness": 0.8984364327435839, "xie_feature_dispersion": 1.2037141877282427, "pac": 0.8913404671671563, "whitened_cosine_id": 0.8669403195381165}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 3, "layer": 1, "concept": "topic", "predictors": {"sip_eigengap": 0.6283839672205405, "raptor_stability": 0.9187504649162292, "fragility": 1.5, "augmentation_robustness": 0.9445047803748948, "xie_feature_dispersion": 3.4172576932701224, "pac": 0.931627622645562, "whitened_cosine_id": 0.9079165458679199}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 3, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.14101513130976387, "raptor_stability": 0.9509273767471313, "fragility": 1.5, "augmentation_robustness": 0.8868008425779641, "xie_feature_dispersion": 5.186993220693557, "pac": 0.9188641096625477, "whitened_cosine_id": 0.9339398145675659}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921976405, "raptor_stability": 0.7710468173027039, "fragility": 0.5, "augmentation_robustness": 0.7989766303576111, "xie_feature_dispersion": 0.4743571278941463, "pac": 0.7850117238301575, "whitened_cosine_id": 0.6775692105293274}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 3, "layer": 2, "concept": "emotion", "predictors": {"sip_eigengap": 0.635597069567996, "raptor_stability": 0.8650970458984375, "fragility": 1.0, "augmentation_robustness": 0.7848846813705812, "xie_feature_dispersion": 1.4784056981569556, "pac": 0.8249908636345094, "whitened_cosine_id": 0.847564160823822}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 3, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.711632992199183, "raptor_stability": 0.8229442834854126, "fragility": 0.5, "augmentation_robustness": 0.7320233370031859, "xie_feature_dispersion": 1.57080462526856, "pac": 0.7774838102442992, "whitened_cosine_id": 0.8052536249160767}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 3, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921988689, "raptor_stability": 0.8289263248443604, "fragility": 1.0, "augmentation_robustness": 0.7538693401418205, "xie_feature_dispersion": 1.8985337702191092, "pac": 0.7913978324930904, "whitened_cosine_id": 0.7646318078041077}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 3, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921989895, "raptor_stability": 0.8404061198234558, "fragility": 1.0, "augmentation_robustness": 0.8369988993389674, "xie_feature_dispersion": 1.1920844617447306, "pac": 0.8387025095812116, "whitened_cosine_id": 0.7623035311698914}} +{"model": "pythia-70m", "dataset": "subj", "seed": 3, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002775, "raptor_stability": 0.899202823638916, "fragility": 1.0, "augmentation_robustness": 0.7801516491279672, "xie_feature_dispersion": 2.0439990482699595, "pac": 0.8396772363834416, "whitened_cosine_id": 0.9309227466583252}} +{"model": "pythia-70m", "dataset": "spam", "seed": 3, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004678, "raptor_stability": 0.9567713141441345, "fragility": 3.0, "augmentation_robustness": 0.9371847229705604, "xie_feature_dispersion": 6.424426785927112, "pac": 0.9469780185573475, "whitened_cosine_id": 0.9486836791038513}} +{"model": "pythia-70m", "dataset": "cola", "seed": 3, "layer": 1, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921973576, "raptor_stability": 0.7688994407653809, "fragility": 1.0, "augmentation_robustness": 0.7138657324958716, "xie_feature_dispersion": 0.7233065123875068, "pac": 0.7413825866306263, "whitened_cosine_id": 0.6779297590255737}} +{"model": "pythia-70m", "dataset": "stance", "seed": 3, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.764248302629601, "raptor_stability": 0.8229038119316101, "fragility": 0.5, "augmentation_robustness": 0.7822202756663522, "xie_feature_dispersion": 2.080304032697393, "pac": 0.8025620437989811, "whitened_cosine_id": 0.8263651728630066}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 3, "layer": 4, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.711632992199799, "raptor_stability": 0.8503726124763489, "fragility": 0.5, "augmentation_robustness": 0.8033091683218101, "xie_feature_dispersion": 1.2407015450743701, "pac": 0.8268408903990795, "whitened_cosine_id": 0.866639256477356}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 3, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349443, "raptor_stability": 0.8484313488006592, "fragility": 0.25, "augmentation_robustness": 0.637225872199634, "xie_feature_dispersion": 4.044131456746566, "pac": 0.7428286105001466, "whitened_cosine_id": 0.8518450260162354}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 3, "layer": 6, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350145, "raptor_stability": 0.8780309557914734, "fragility": 1.0, "augmentation_robustness": 0.874329700092187, "xie_feature_dispersion": 1.4259597904886765, "pac": 0.8761803279418302, "whitened_cosine_id": 0.8702338933944702}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 3, "layer": 12, "concept": "topic", "predictors": {"sip_eigengap": 0.35191923415776655, "raptor_stability": 0.8508845567703247, "fragility": 0.25, "augmentation_robustness": 0.8265529420495982, "xie_feature_dispersion": 0.9266248016633525, "pac": 0.8387187494099615, "whitened_cosine_id": 0.8987531661987305}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 3, "layer": 9, "concept": "topic", "predictors": {"sip_eigengap": 0.11774124797544638, "raptor_stability": 0.9511698484420776, "fragility": 1.5, "augmentation_robustness": 0.8745928253851561, "xie_feature_dispersion": 6.276849376401518, "pac": 0.9128813369136168, "whitened_cosine_id": 0.917269766330719}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 3, "layer": 2, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859342074, "raptor_stability": 0.78183913230896, "fragility": 0.5, "augmentation_robustness": 0.7622129099393176, "xie_feature_dispersion": 0.5886846655394216, "pac": 0.7720260211241388, "whitened_cosine_id": 0.7248976230621338}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 3, "layer": 5, "concept": "emotion", "predictors": {"sip_eigengap": 0.5119239787691542, "raptor_stability": 0.8616406917572021, "fragility": 0.25, "augmentation_robustness": 0.7504844246649606, "xie_feature_dispersion": 1.6907559164308787, "pac": 0.8060625582110814, "whitened_cosine_id": 0.852891206741333}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 3, "layer": 5, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347243, "raptor_stability": 0.8252765536308289, "fragility": 0.25, "augmentation_robustness": 0.7347966668215313, "xie_feature_dispersion": 2.064566431309704, "pac": 0.7800366102261801, "whitened_cosine_id": 0.8165638446807861}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 3, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859345138, "raptor_stability": 0.7875639796257019, "fragility": 0.25, "augmentation_robustness": 0.6763370074729932, "xie_feature_dispersion": 3.2860294351071695, "pac": 0.7319504935493475, "whitened_cosine_id": 0.7773585915565491}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 3, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.397542485934679, "raptor_stability": 0.7985202670097351, "fragility": 0.25, "augmentation_robustness": 0.7440947615507673, "xie_feature_dispersion": 0.7467058167702149, "pac": 0.7713075142802512, "whitened_cosine_id": 0.7971633076667786}} +{"model": "pythia-160m", "dataset": "subj", "seed": 3, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935211, "raptor_stability": 0.8905059099197388, "fragility": 1.0, "augmentation_robustness": 0.7853861456321554, "xie_feature_dispersion": 2.317051234221485, "pac": 0.8379460277759471, "whitened_cosine_id": 0.913381040096283}} +{"model": "pythia-160m", "dataset": "spam", "seed": 3, "layer": 6, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859350205, "raptor_stability": 0.9280425310134888, "fragility": 1.0, "augmentation_robustness": 0.9003377239515604, "xie_feature_dispersion": 13.64857545430944, "pac": 0.9141901274825246, "whitened_cosine_id": 0.9369125366210938}} +{"model": "pythia-160m", "dataset": "cola", "seed": 3, "layer": 7, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859343062, "raptor_stability": 0.7884514331817627, "fragility": 0.25, "augmentation_robustness": 0.6450635280093344, "xie_feature_dispersion": 0.17857729363649755, "pac": 0.7167574805955486, "whitened_cosine_id": 0.7649105191230774}} +{"model": "pythia-160m", "dataset": "stance", "seed": 3, "layer": 4, "concept": "stance", "predictors": {"sip_eigengap": 0.6506681129814839, "raptor_stability": 0.8327218890190125, "fragility": 0.5, "augmentation_robustness": 0.8189402967776248, "xie_feature_dispersion": 1.7873782154455453, "pac": 0.8258310928983186, "whitened_cosine_id": 0.8258171081542969}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 3, "layer": 7, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350412, "raptor_stability": 0.8567770719528198, "fragility": 0.5, "augmentation_robustness": 0.7799380638785056, "xie_feature_dispersion": 1.305808488628754, "pac": 0.8183575679156627, "whitened_cosine_id": 0.8785260319709778}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 3, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881342, "raptor_stability": 0.8563058376312256, "fragility": 0.25, "augmentation_robustness": 0.6228553515841018, "xie_feature_dispersion": 4.914029083398948, "pac": 0.7395805946076637, "whitened_cosine_id": 0.8632848858833313}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 3, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295688201, "raptor_stability": 0.8571096062660217, "fragility": 1.0, "augmentation_robustness": 0.8391733545717237, "xie_feature_dispersion": 2.1649660520641487, "pac": 0.8481414804188727, "whitened_cosine_id": 0.8794662356376648}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 3, "layer": 8, "concept": "topic", "predictors": {"sip_eigengap": 0.3940956743493873, "raptor_stability": 0.8732040524482727, "fragility": 1.0, "augmentation_robustness": 0.8558702619852715, "xie_feature_dispersion": 3.495736654891566, "pac": 0.8645371572167722, "whitened_cosine_id": 0.894050121307373}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 3, "layer": 18, "concept": "topic", "predictors": {"sip_eigengap": 0.0777198899969492, "raptor_stability": 0.9526127576828003, "fragility": 2.0, "augmentation_robustness": 0.8767089502212883, "xie_feature_dispersion": 7.889853603758875, "pac": 0.9146608539520442, "whitened_cosine_id": 0.8949448466300964}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 3, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879801, "raptor_stability": 0.8159152269363403, "fragility": 0.25, "augmentation_robustness": 0.7880748749868189, "xie_feature_dispersion": 0.24343282857388523, "pac": 0.8019950509615796, "whitened_cosine_id": 0.800447404384613}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 3, "layer": 15, "concept": "emotion", "predictors": {"sip_eigengap": 0.48221468653288235, "raptor_stability": 0.8573822379112244, "fragility": 0.25, "augmentation_robustness": 0.7660417601070145, "xie_feature_dispersion": 1.9953148592463423, "pac": 0.8117119990091195, "whitened_cosine_id": 0.8465678095817566}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 3, "layer": 14, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956879983, "raptor_stability": 0.8090613484382629, "fragility": 0.25, "augmentation_robustness": 0.7199613185198703, "xie_feature_dispersion": 4.138422068582218, "pac": 0.7645113334790666, "whitened_cosine_id": 0.8162167072296143}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 3, "layer": 9, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956877805, "raptor_stability": 0.8041320443153381, "fragility": 0.25, "augmentation_robustness": 0.6772706695322654, "xie_feature_dispersion": 4.201391999844916, "pac": 0.7407013569238018, "whitened_cosine_id": 0.8077095150947571}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 3, "layer": 13, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956880465, "raptor_stability": 0.8054513335227966, "fragility": 0.25, "augmentation_robustness": 0.7134721107475187, "xie_feature_dispersion": 0.942681786130636, "pac": 0.7594617221351576, "whitened_cosine_id": 0.8189021944999695}} +{"model": "pythia-410m", "dataset": "subj", "seed": 3, "layer": 15, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882677, "raptor_stability": 0.908913254737854, "fragility": 1.0, "augmentation_robustness": 0.791121231421306, "xie_feature_dispersion": 1.6016867212241452, "pac": 0.85001724307958, "whitened_cosine_id": 0.9041904807090759}} +{"model": "pythia-410m", "dataset": "spam", "seed": 3, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688292, "raptor_stability": 0.9435619115829468, "fragility": 3.0, "augmentation_robustness": 0.9066470040376822, "xie_feature_dispersion": 6.763845431690352, "pac": 0.9251044578103145, "whitened_cosine_id": 0.9281833171844482}} +{"model": "pythia-410m", "dataset": "cola", "seed": 3, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956878849, "raptor_stability": 0.7881181836128235, "fragility": 0.25, "augmentation_robustness": 0.600597854138349, "xie_feature_dispersion": 0.9458123063554009, "pac": 0.6943580188755862, "whitened_cosine_id": 0.7995412349700928}} +{"model": "pythia-410m", "dataset": "stance", "seed": 3, "layer": 17, "concept": "stance", "predictors": {"sip_eigengap": 0.4936540695216237, "raptor_stability": 0.8370916247367859, "fragility": 0.5, "augmentation_robustness": 0.8332617730055567, "xie_feature_dispersion": 2.485399050185102, "pac": 0.8351766988711713, "whitened_cosine_id": 0.8208258748054504}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 3, "layer": 13, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881398, "raptor_stability": 0.8489888310432434, "fragility": 0.5, "augmentation_robustness": 0.7844172937998297, "xie_feature_dispersion": 1.6180674325869053, "pac": 0.8167030624215366, "whitened_cosine_id": 0.865872859954834}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010287, "raptor_stability": 0.8647460341453552, "fragility": 0.25, "augmentation_robustness": 0.641727222279711, "xie_feature_dispersion": 5.808094646975757, "pac": 0.7532366282125331, "whitened_cosine_id": 0.8289788365364075}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010868, "raptor_stability": 0.8411266803741455, "fragility": 1.0, "augmentation_robustness": 0.7503016837377494, "xie_feature_dispersion": 2.282144440990373, "pac": 0.7957141820559475, "whitened_cosine_id": 0.8417641520500183}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 3, "layer": 13, "concept": "topic", "predictors": {"sip_eigengap": 0.20340792936838478, "raptor_stability": 0.8698489665985107, "fragility": 1.0, "augmentation_robustness": 0.8437237058947922, "xie_feature_dispersion": 4.342567776713613, "pac": 0.8567863362466515, "whitened_cosine_id": 0.8350374698638916}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 3, "layer": 15, "concept": "topic", "predictors": {"sip_eigengap": 0.059803315773717704, "raptor_stability": 0.9504701495170593, "fragility": 2.0, "augmentation_robustness": 0.8680315611599807, "xie_feature_dispersion": 8.620883539713939, "pac": 0.90925085533852, "whitened_cosine_id": 0.8365957140922546}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 3, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010883, "raptor_stability": 0.8169371485710144, "fragility": 0.25, "augmentation_robustness": 0.7710432901435903, "xie_feature_dispersion": 0.42029681121277923, "pac": 0.7939902193573023, "whitened_cosine_id": 0.8196472525596619}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 3, "layer": 11, "concept": "emotion", "predictors": {"sip_eigengap": 0.4267118062158825, "raptor_stability": 0.8550241589546204, "fragility": 0.25, "augmentation_robustness": 0.7405822372807725, "xie_feature_dispersion": 2.564638903560827, "pac": 0.7978031981176965, "whitened_cosine_id": 0.8296079635620117}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 3, "layer": 15, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010767, "raptor_stability": 0.796916663646698, "fragility": 0.25, "augmentation_robustness": 0.6606039027652186, "xie_feature_dispersion": 3.912958891643261, "pac": 0.7287602832059583, "whitened_cosine_id": 0.8257460594177246}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 3, "layer": 18, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010796, "raptor_stability": 0.8012734055519104, "fragility": 0.25, "augmentation_robustness": 0.6623961560942748, "xie_feature_dispersion": 3.563898854267082, "pac": 0.7318347808230926, "whitened_cosine_id": 0.8225507736206055}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 3, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010296, "raptor_stability": 0.7851132154464722, "fragility": 0.25, "augmentation_robustness": 0.6617407773933088, "xie_feature_dispersion": 1.0772966237421127, "pac": 0.7234269964198905, "whitened_cosine_id": 0.8228062987327576}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 3, "layer": 17, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010871, "raptor_stability": 0.8912879228591919, "fragility": 1.0, "augmentation_robustness": 0.74955767566332, "xie_feature_dispersion": 3.0211175775214865, "pac": 0.8204227992612559, "whitened_cosine_id": 0.8278633952140808}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 3, "layer": 3, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010838, "raptor_stability": 0.925938069820404, "fragility": 3.0, "augmentation_robustness": 0.9098532007735074, "xie_feature_dispersion": 10.71642677493898, "pac": 0.9178956352969557, "whitened_cosine_id": 0.8716909289360046}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 3, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010827, "raptor_stability": 0.7985167503356934, "fragility": 0.25, "augmentation_robustness": 0.5412217711508899, "xie_feature_dispersion": 1.0588280189082442, "pac": 0.6698692607432917, "whitened_cosine_id": 0.823022723197937}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 3, "layer": 3, "concept": "stance", "predictors": {"sip_eigengap": 0.37531177284512046, "raptor_stability": 0.8395811915397644, "fragility": 1.0, "augmentation_robustness": 0.8246880782030601, "xie_feature_dispersion": 2.787120846289539, "pac": 0.8321346348714123, "whitened_cosine_id": 0.8196287155151367}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 3, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.8558164961010829, "raptor_stability": 0.8403837084770203, "fragility": 1.0, "augmentation_robustness": 0.7414207905077144, "xie_feature_dispersion": 2.400769158905071, "pac": 0.7909022494923673, "whitened_cosine_id": 0.8367124795913696}} +{"model": "gpt2", "dataset": "sst2", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934796, "raptor_stability": 0.8351662158966064, "fragility": 0.25, "augmentation_robustness": 0.5635262291495843, "xie_feature_dispersion": 4.2622215814473865, "pac": 0.6993462225230953, "whitened_cosine_id": 0.8634011149406433}} +{"model": "gpt2", "dataset": "imdb", "seed": 3, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350287, "raptor_stability": 0.8339491486549377, "fragility": 0.25, "augmentation_robustness": 0.7662356686632448, "xie_feature_dispersion": 1.304014068501824, "pac": 0.8000924086590913, "whitened_cosine_id": 0.87335205078125}} +{"model": "gpt2", "dataset": "ag_news", "seed": 3, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.3778683465799438, "raptor_stability": 0.8665527105331421, "fragility": 0.5, "augmentation_robustness": 0.8531902817642327, "xie_feature_dispersion": 2.241047464066251, "pac": 0.8598714961486874, "whitened_cosine_id": 0.9046292304992676}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 3, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.10991110313414279, "raptor_stability": 0.9330034255981445, "fragility": 1.5, "augmentation_robustness": 0.8169811795296109, "xie_feature_dispersion": 5.468890231915903, "pac": 0.8749923025638777, "whitened_cosine_id": 0.9063879251480103}} +{"model": "gpt2", "dataset": "counterfact", "seed": 3, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859343766, "raptor_stability": 0.7898539900779724, "fragility": 0.25, "augmentation_robustness": 0.8079306258304046, "xie_feature_dispersion": 0.19395964260888718, "pac": 0.7988923079541885, "whitened_cosine_id": 0.7610896825790405}} +{"model": "gpt2", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6182210878004821, "raptor_stability": 0.8550314903259277, "fragility": 0.25, "augmentation_robustness": 0.7431736186782528, "xie_feature_dispersion": 1.1654170019303518, "pac": 0.7991025545020902, "whitened_cosine_id": 0.8550549149513245}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347225, "raptor_stability": 0.8043380379676819, "fragility": 0.25, "augmentation_robustness": 0.677191562860757, "xie_feature_dispersion": 1.3127141118978531, "pac": 0.7407648004142194, "whitened_cosine_id": 0.8204189538955688}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 3, "layer": 5, "concept": "irony", "predictors": {"sip_eigengap": 1.397542485934697, "raptor_stability": 0.780476987361908, "fragility": 0.25, "augmentation_robustness": 0.6631271115518341, "xie_feature_dispersion": 3.352185591368954, "pac": 0.7218020494568711, "whitened_cosine_id": 0.8065347075462341}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 3, "layer": 9, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859346575, "raptor_stability": 0.7473269104957581, "fragility": 0.25, "augmentation_robustness": 0.6123626549338375, "xie_feature_dispersion": 0.7096033674345994, "pac": 0.6798447827147978, "whitened_cosine_id": 0.7918455004692078}} +{"model": "gpt2", "dataset": "subj", "seed": 3, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.397542485935243, "raptor_stability": 0.8725587725639343, "fragility": 0.5, "augmentation_robustness": 0.7069137856416525, "xie_feature_dispersion": 1.7904837862778602, "pac": 0.7897362791027934, "whitened_cosine_id": 0.9169808626174927}} +{"model": "gpt2", "dataset": "spam", "seed": 3, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353023, "raptor_stability": 0.9019039273262024, "fragility": 1.5, "augmentation_robustness": 0.8805267705055504, "xie_feature_dispersion": 12.885975509137046, "pac": 0.8912153489158764, "whitened_cosine_id": 0.92914217710495}} +{"model": "gpt2", "dataset": "cola", "seed": 3, "layer": 12, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344647, "raptor_stability": 0.796906054019928, "fragility": 0.25, "augmentation_robustness": 0.6909560170275559, "xie_feature_dispersion": 0.4966128694499476, "pac": 0.743931035523742, "whitened_cosine_id": 0.7800789475440979}} +{"model": "gpt2", "dataset": "stance", "seed": 3, "layer": 8, "concept": "stance", "predictors": {"sip_eigengap": 0.6661294213646244, "raptor_stability": 0.8302767872810364, "fragility": 0.5, "augmentation_robustness": 0.8014307017054971, "xie_feature_dispersion": 1.9080832844053377, "pac": 0.8158537444932668, "whitened_cosine_id": 0.8170424103736877}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 3, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859350134, "raptor_stability": 0.8213337063789368, "fragility": 0.25, "augmentation_robustness": 0.6974710476635542, "xie_feature_dispersion": 1.3078300052773906, "pac": 0.7594023770212455, "whitened_cosine_id": 0.8702872395515442}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.210307295687936, "raptor_stability": 0.8499755263328552, "fragility": 0.25, "augmentation_robustness": 0.6261496459413354, "xie_feature_dispersion": 4.436216156416145, "pac": 0.7380625861370953, "whitened_cosine_id": 0.8625351190567017}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 3, "layer": 21, "concept": "topic", "predictors": {"sip_eigengap": 0.2921886382497235, "raptor_stability": 0.8477186560630798, "fragility": 1.0, "augmentation_robustness": 0.8387418545839186, "xie_feature_dispersion": 4.131659598391053, "pac": 0.8432302553234992, "whitened_cosine_id": 0.88265460729599}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 3, "layer": 8, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956879386, "raptor_stability": 0.7917901277542114, "fragility": 0.25, "augmentation_robustness": 0.789771355720532, "xie_feature_dispersion": 0.17500181221361583, "pac": 0.7907807417373718, "whitened_cosine_id": 0.7891764640808105}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 3, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.6234881751702148, "raptor_stability": 0.8669646382331848, "fragility": 0.25, "augmentation_robustness": 0.7744639183310552, "xie_feature_dispersion": 0.8652954065662566, "pac": 0.82071427828212, "whitened_cosine_id": 0.8504673838615417}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 3, "layer": 6, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295687904, "raptor_stability": 0.7907098531723022, "fragility": 0.25, "augmentation_robustness": 0.6575825596573526, "xie_feature_dispersion": 3.5458093952463834, "pac": 0.7241462064148274, "whitened_cosine_id": 0.8148408532142639}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 3, "layer": 22, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956879737, "raptor_stability": 0.7582568526268005, "fragility": 0.25, "augmentation_robustness": 0.6081859135950609, "xie_feature_dispersion": 0.8063341929658796, "pac": 0.6832213831109307, "whitened_cosine_id": 0.7960034012794495}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 3, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.210307295688269, "raptor_stability": 0.8890963792800903, "fragility": 1.0, "augmentation_robustness": 0.7291926988187397, "xie_feature_dispersion": 2.080419773242272, "pac": 0.809144539049415, "whitened_cosine_id": 0.8867965340614319}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 3, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956879426, "raptor_stability": 0.7859232425689697, "fragility": 0.25, "augmentation_robustness": 0.5254228710879436, "xie_feature_dispersion": 0.6803530491181083, "pac": 0.6556730568284567, "whitened_cosine_id": 0.8077235221862793}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 3, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5196901382587911, "raptor_stability": 0.8363353610038757, "fragility": 0.5, "augmentation_robustness": 0.8228205239645261, "xie_feature_dispersion": 1.7693304935455851, "pac": 0.8295779424842009, "whitened_cosine_id": 0.7967984080314636}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 3, "layer": 12, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881522, "raptor_stability": 0.811364471912384, "fragility": 0.5, "augmentation_robustness": 0.6900060500051179, "xie_feature_dispersion": 1.3697277559493137, "pac": 0.750685260958751, "whitened_cosine_id": 0.8628206849098206}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237648796, "raptor_stability": 0.8676397204399109, "fragility": 0.25, "augmentation_robustness": 0.6345770721983662, "xie_feature_dispersion": 5.001598619911871, "pac": 0.7511083963191385, "whitened_cosine_id": 0.8821530938148499}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 3, "layer": 14, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650275, "raptor_stability": 0.8786996006965637, "fragility": 1.0, "augmentation_robustness": 0.8588046228179724, "xie_feature_dispersion": 2.4507641389974713, "pac": 0.8687521117572681, "whitened_cosine_id": 0.8870015740394592}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 3, "layer": 23, "concept": "topic", "predictors": {"sip_eigengap": 0.2775334420918106, "raptor_stability": 0.8491312265396118, "fragility": 1.5, "augmentation_robustness": 0.8575692544379112, "xie_feature_dispersion": 8.067934468405575, "pac": 0.8533502404887615, "whitened_cosine_id": 0.8975652456283569}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 3, "layer": 6, "concept": "topic", "predictors": {"sip_eigengap": 0.07983459673425802, "raptor_stability": 0.9590606093406677, "fragility": 1.0, "augmentation_robustness": 0.9129941790024725, "xie_feature_dispersion": 5.747537131789415, "pac": 0.9360273941715701, "whitened_cosine_id": 0.9247861504554749}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 3, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2938729237646678, "raptor_stability": 0.840959370136261, "fragility": 0.25, "augmentation_robustness": 0.8017592029882017, "xie_feature_dispersion": 0.07321360576071477, "pac": 0.8213592865622313, "whitened_cosine_id": 0.7885988354682922}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 3, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.6025878875187768, "raptor_stability": 0.8772271275520325, "fragility": 0.25, "augmentation_robustness": 0.865087290926423, "xie_feature_dispersion": 1.9052135627589408, "pac": 0.8711572092392277, "whitened_cosine_id": 0.8513849377632141}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 3, "layer": 1, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647677, "raptor_stability": 0.8638144731521606, "fragility": 2.0, "augmentation_robustness": 0.8618714559295404, "xie_feature_dispersion": 3.073243341198275, "pac": 0.8628429645408505, "whitened_cosine_id": 0.8141002058982849}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 3, "layer": 3, "concept": "irony", "predictors": {"sip_eigengap": 1.2938729237638515, "raptor_stability": 0.8214561939239502, "fragility": 0.25, "augmentation_robustness": 0.7518367724678281, "xie_feature_dispersion": 4.329624079200656, "pac": 0.7866464831958891, "whitened_cosine_id": 0.7974085807800293}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 3, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237646187, "raptor_stability": 0.804497480392456, "fragility": 0.25, "augmentation_robustness": 0.7534120265877415, "xie_feature_dispersion": 0.582349253613317, "pac": 0.7789547534900988, "whitened_cosine_id": 0.8017151355743408}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 3, "layer": 15, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651343, "raptor_stability": 0.9011402130126953, "fragility": 0.5, "augmentation_robustness": 0.8098720947850936, "xie_feature_dispersion": 1.5822189245994793, "pac": 0.8555061538988944, "whitened_cosine_id": 0.9193786978721619}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 3, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.2938729237650004, "raptor_stability": 0.9187747240066528, "fragility": 1.0, "augmentation_robustness": 0.9072265197438765, "xie_feature_dispersion": 18.77242405320582, "pac": 0.9130006218752647, "whitened_cosine_id": 0.9447034597396851}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 3, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646827, "raptor_stability": 0.8013314008712769, "fragility": 0.25, "augmentation_robustness": 0.5714415857266395, "xie_feature_dispersion": 0.7710445833759078, "pac": 0.6863864932989582, "whitened_cosine_id": 0.8173381686210632}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 3, "layer": 7, "concept": "stance", "predictors": {"sip_eigengap": 0.5814899203429595, "raptor_stability": 0.8490622639656067, "fragility": 0.25, "augmentation_robustness": 0.8472747423617989, "xie_feature_dispersion": 2.1435570572895988, "pac": 0.8481685031637027, "whitened_cosine_id": 0.8264880776405334}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 3, "layer": 13, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649207, "raptor_stability": 0.8474602699279785, "fragility": 0.25, "augmentation_robustness": 0.765850633291224, "xie_feature_dispersion": 1.245317692001698, "pac": 0.8066554516096013, "whitened_cosine_id": 0.8748792409896851}} +{"model": "pythia-70m", "dataset": "sst2", "seed": 4, "layer": 2, "concept": "sentiment", "predictors": {"sip_eigengap": 1.7116329921996045, "raptor_stability": 0.8566904067993164, "fragility": 1.0, "augmentation_robustness": 0.6395377546477263, "xie_feature_dispersion": 1.5854208982087719, "pac": 0.7481140807235214, "whitened_cosine_id": 0.8359927535057068}} +{"model": "pythia-70m", "dataset": "imdb", "seed": 4, "layer": 3, "concept": "sentiment", "predictors": {"sip_eigengap": 1.711632992199886, "raptor_stability": 0.9050891995429993, "fragility": 1.0, "augmentation_robustness": 0.9251332642220863, "xie_feature_dispersion": 1.2126650382704005, "pac": 0.9151112318825427, "whitened_cosine_id": 0.850266695022583}} +{"model": "pythia-70m", "dataset": "ag_news", "seed": 4, "layer": 5, "concept": "topic", "predictors": {"sip_eigengap": 0.5924204103995422, "raptor_stability": 0.8544853329658508, "fragility": 1.0, "augmentation_robustness": 0.8372965090354867, "xie_feature_dispersion": 4.656184075098706, "pac": 0.8458909210006688, "whitened_cosine_id": 0.9082167744636536}} +{"model": "pythia-70m", "dataset": "dbpedia", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.1519243074719573, "raptor_stability": 0.9499274492263794, "fragility": 1.5, "augmentation_robustness": 0.8830180866818871, "xie_feature_dispersion": 5.282104233636098, "pac": 0.9164727679541332, "whitened_cosine_id": 0.9344291687011719}} +{"model": "pythia-70m", "dataset": "counterfact", "seed": 4, "layer": 1, "concept": "truth", "predictors": {"sip_eigengap": 1.7116329921975906, "raptor_stability": 0.7707024216651917, "fragility": 1.0, "augmentation_robustness": 0.7329255811603577, "xie_feature_dispersion": 0.5102383939416054, "pac": 0.7518140014127747, "whitened_cosine_id": 0.7052313089370728}} +{"model": "pythia-70m", "dataset": "emotion", "seed": 4, "layer": 3, "concept": "emotion", "predictors": {"sip_eigengap": 0.6400948489658904, "raptor_stability": 0.8515534996986389, "fragility": 0.5, "augmentation_robustness": 0.7338919283811642, "xie_feature_dispersion": 1.307093856011509, "pac": 0.7927227140399016, "whitened_cosine_id": 0.8422322273254395}} +{"model": "pythia-70m", "dataset": "tweet_hate", "seed": 4, "layer": 4, "concept": "hate", "predictors": {"sip_eigengap": 1.71163299219924, "raptor_stability": 0.8237500786781311, "fragility": 0.5, "augmentation_robustness": 0.7327459917351864, "xie_feature_dispersion": 2.5240514742338784, "pac": 0.7782480352066588, "whitened_cosine_id": 0.803077220916748}} +{"model": "pythia-70m", "dataset": "tweet_irony", "seed": 4, "layer": 1, "concept": "irony", "predictors": {"sip_eigengap": 1.7116329921987739, "raptor_stability": 0.8263923525810242, "fragility": 1.0, "augmentation_robustness": 0.7109290902303631, "xie_feature_dispersion": 2.0494666093180176, "pac": 0.7686607214056936, "whitened_cosine_id": 0.7474713325500488}} +{"model": "pythia-70m", "dataset": "tweet_offensive", "seed": 4, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.7116329921990436, "raptor_stability": 0.8242563605308533, "fragility": 0.5, "augmentation_robustness": 0.7821271817029374, "xie_feature_dispersion": 1.5098181981749643, "pac": 0.8031917711168953, "whitened_cosine_id": 0.7899174094200134}} +{"model": "pythia-70m", "dataset": "subj", "seed": 4, "layer": 4, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.7116329922002564, "raptor_stability": 0.8824465274810791, "fragility": 1.0, "augmentation_robustness": 0.8066508753882752, "xie_feature_dispersion": 1.925210310106005, "pac": 0.8445487014346771, "whitened_cosine_id": 0.9232692122459412}} +{"model": "pythia-70m", "dataset": "spam", "seed": 4, "layer": 1, "concept": "spam", "predictors": {"sip_eigengap": 1.7116329922004774, "raptor_stability": 0.9476951360702515, "fragility": 3.0, "augmentation_robustness": 0.9265937443188577, "xie_feature_dispersion": 6.387885559509479, "pac": 0.9371444401945546, "whitened_cosine_id": 0.9508467316627502}} +{"model": "pythia-70m", "dataset": "cola", "seed": 4, "layer": 5, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.7116329921977023, "raptor_stability": 0.7777998447418213, "fragility": 0.25, "augmentation_robustness": 0.6579850681228125, "xie_feature_dispersion": 0.250601434273502, "pac": 0.7178924564323169, "whitened_cosine_id": 0.7004658579826355}} +{"model": "pythia-70m", "dataset": "stance", "seed": 4, "layer": 5, "concept": "stance", "predictors": {"sip_eigengap": 0.7333740478467312, "raptor_stability": 0.8212578296661377, "fragility": 1.0, "augmentation_robustness": 0.7504520187465166, "xie_feature_dispersion": 1.7128280566290308, "pac": 0.7858549242063271, "whitened_cosine_id": 0.8275880217552185}} +{"model": "pythia-70m", "dataset": "amazon_cf", "seed": 4, "layer": 5, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.7116329921997524, "raptor_stability": 0.8540244102478027, "fragility": 1.0, "augmentation_robustness": 0.7819111270390232, "xie_feature_dispersion": 1.4981388024521016, "pac": 0.817967768643413, "whitened_cosine_id": 0.8604907989501953}} +{"model": "pythia-160m", "dataset": "sst2", "seed": 4, "layer": 5, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859349261, "raptor_stability": 0.842431366443634, "fragility": 0.25, "augmentation_robustness": 0.670070029768046, "xie_feature_dispersion": 2.8701281427080465, "pac": 0.75625069810584, "whitened_cosine_id": 0.8559513688087463}} +{"model": "pythia-160m", "dataset": "imdb", "seed": 4, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859350782, "raptor_stability": 0.8713650703430176, "fragility": 0.5, "augmentation_robustness": 0.8521992774428392, "xie_feature_dispersion": 1.7964561061229958, "pac": 0.8617821738929283, "whitened_cosine_id": 0.8889275789260864}} +{"model": "pythia-160m", "dataset": "ag_news", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.3656440339329693, "raptor_stability": 0.8521515130996704, "fragility": 1.5, "augmentation_robustness": 0.8214046671297884, "xie_feature_dispersion": 6.173757088063873, "pac": 0.8367780901147295, "whitened_cosine_id": 0.9043599367141724}} +{"model": "pythia-160m", "dataset": "dbpedia", "seed": 4, "layer": 9, "concept": "topic", "predictors": {"sip_eigengap": 0.12756426041857358, "raptor_stability": 0.9445372819900513, "fragility": 1.5, "augmentation_robustness": 0.8622565473270213, "xie_feature_dispersion": 6.343488803218173, "pac": 0.9033969146585363, "whitened_cosine_id": 0.9183123707771301}} +{"model": "pythia-160m", "dataset": "counterfact", "seed": 4, "layer": 7, "concept": "truth", "predictors": {"sip_eigengap": 1.397542485934335, "raptor_stability": 0.7996639609336853, "fragility": 0.25, "augmentation_robustness": 0.7613788429632672, "xie_feature_dispersion": 0.5275399816201919, "pac": 0.7805214019484763, "whitened_cosine_id": 0.753498375415802}} +{"model": "pythia-160m", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.7058740799382045, "raptor_stability": 0.8810455799102783, "fragility": 1.0, "augmentation_robustness": 0.8152160353634006, "xie_feature_dispersion": 2.0356355439733544, "pac": 0.8481308076368395, "whitened_cosine_id": 0.8614587187767029}} +{"model": "pythia-160m", "dataset": "tweet_hate", "seed": 4, "layer": 11, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859348062, "raptor_stability": 0.8045304417610168, "fragility": 0.5, "augmentation_robustness": 0.7049071379157302, "xie_feature_dispersion": 2.0681692958689024, "pac": 0.7547187898383736, "whitened_cosine_id": 0.8201311230659485}} +{"model": "pythia-160m", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346577, "raptor_stability": 0.8003606200218201, "fragility": 0.25, "augmentation_robustness": 0.6612670699122942, "xie_feature_dispersion": 2.4099178377906756, "pac": 0.7308138449670571, "whitened_cosine_id": 0.8072301745414734}} +{"model": "pythia-160m", "dataset": "tweet_offensive", "seed": 4, "layer": 1, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859345884, "raptor_stability": 0.8317489624023438, "fragility": 1.0, "augmentation_robustness": 0.772983120611079, "xie_feature_dispersion": 2.3238338563599275, "pac": 0.8023660415067113, "whitened_cosine_id": 0.8050597906112671}} +{"model": "pythia-160m", "dataset": "subj", "seed": 4, "layer": 5, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.3975424859351835, "raptor_stability": 0.8953569531440735, "fragility": 1.0, "augmentation_robustness": 0.8059484746527689, "xie_feature_dispersion": 1.8870415060935894, "pac": 0.8506527138984212, "whitened_cosine_id": 0.9113974571228027}} +{"model": "pythia-160m", "dataset": "spam", "seed": 4, "layer": 9, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859352754, "raptor_stability": 0.9212907552719116, "fragility": 1.5, "augmentation_robustness": 0.8624898624867177, "xie_feature_dispersion": 12.624355801662276, "pac": 0.8918903088793146, "whitened_cosine_id": 0.9403319358825684}} +{"model": "pythia-160m", "dataset": "cola", "seed": 4, "layer": 10, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.397542485934378, "raptor_stability": 0.7925094366073608, "fragility": 0.25, "augmentation_robustness": 0.6106269635092136, "xie_feature_dispersion": 0.1983095485933765, "pac": 0.7015682000582872, "whitened_cosine_id": 0.7796781063079834}} +{"model": "pythia-160m", "dataset": "stance", "seed": 4, "layer": 2, "concept": "stance", "predictors": {"sip_eigengap": 0.591622106050008, "raptor_stability": 0.8417823314666748, "fragility": 1.0, "augmentation_robustness": 0.8225818656918379, "xie_feature_dispersion": 1.8092532859980797, "pac": 0.8321820985792563, "whitened_cosine_id": 0.8301091194152832}} +{"model": "pythia-160m", "dataset": "amazon_cf", "seed": 4, "layer": 3, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349856, "raptor_stability": 0.8758466839790344, "fragility": 1.0, "augmentation_robustness": 0.835043699708102, "xie_feature_dispersion": 1.5364396930791746, "pac": 0.8554451918435682, "whitened_cosine_id": 0.8725274801254272}} +{"model": "pythia-410m", "dataset": "sst2", "seed": 4, "layer": 16, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956880225, "raptor_stability": 0.8483951687812805, "fragility": 0.25, "augmentation_robustness": 0.643421731590692, "xie_feature_dispersion": 3.4588103868670306, "pac": 0.7459084501859863, "whitened_cosine_id": 0.8561255931854248}} +{"model": "pythia-410m", "dataset": "imdb", "seed": 4, "layer": 15, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956882035, "raptor_stability": 0.8837658166885376, "fragility": 1.0, "augmentation_robustness": 0.8596997806079171, "xie_feature_dispersion": 1.9038436246678319, "pac": 0.8717327986482273, "whitened_cosine_id": 0.894847571849823}} +{"model": "pythia-410m", "dataset": "ag_news", "seed": 4, "layer": 11, "concept": "topic", "predictors": {"sip_eigengap": 0.3547804650921568, "raptor_stability": 0.8753689527511597, "fragility": 1.0, "augmentation_robustness": 0.8555244488391097, "xie_feature_dispersion": 3.4329927120747055, "pac": 0.8654467007951347, "whitened_cosine_id": 0.8951209783554077}} +{"model": "pythia-410m", "dataset": "dbpedia", "seed": 4, "layer": 22, "concept": "topic", "predictors": {"sip_eigengap": 0.10955555398923562, "raptor_stability": 0.9442237615585327, "fragility": 3.0, "augmentation_robustness": 0.8375015283715399, "xie_feature_dispersion": 10.914315920866802, "pac": 0.8908626449650363, "whitened_cosine_id": 0.887626588344574}} +{"model": "pythia-410m", "dataset": "counterfact", "seed": 4, "layer": 13, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878207, "raptor_stability": 0.7893005013465881, "fragility": 0.25, "augmentation_robustness": 0.7388001024408721, "xie_feature_dispersion": 0.4902099764637905, "pac": 0.7640503018937301, "whitened_cosine_id": 0.7745464444160461}} +{"model": "pythia-410m", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.5474574770136091, "raptor_stability": 0.8616097569465637, "fragility": 0.25, "augmentation_robustness": 0.7404739041427045, "xie_feature_dispersion": 1.968628707347913, "pac": 0.8010418305446341, "whitened_cosine_id": 0.8528909087181091}} +{"model": "pythia-410m", "dataset": "tweet_hate", "seed": 4, "layer": 15, "concept": "hate", "predictors": {"sip_eigengap": 1.210307295688051, "raptor_stability": 0.8146539330482483, "fragility": 0.25, "augmentation_robustness": 0.7026104422955255, "xie_feature_dispersion": 4.365462087324579, "pac": 0.7586321876718869, "whitened_cosine_id": 0.8268682956695557}} +{"model": "pythia-410m", "dataset": "tweet_irony", "seed": 4, "layer": 15, "concept": "irony", "predictors": {"sip_eigengap": 1.2103072956878458, "raptor_stability": 0.8015760779380798, "fragility": 0.25, "augmentation_robustness": 0.6575946714155477, "xie_feature_dispersion": 3.8919081093582863, "pac": 0.7295853746768137, "whitened_cosine_id": 0.814882755279541}} +{"model": "pythia-410m", "dataset": "tweet_offensive", "seed": 4, "layer": 16, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956878596, "raptor_stability": 0.7986729145050049, "fragility": 0.25, "augmentation_robustness": 0.700139464938641, "xie_feature_dispersion": 3.5880837975111466, "pac": 0.7494061897218229, "whitened_cosine_id": 0.8099539279937744}} +{"model": "pythia-410m", "dataset": "subj", "seed": 4, "layer": 14, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882532, "raptor_stability": 0.8692127466201782, "fragility": 0.5, "augmentation_robustness": 0.7900862434288312, "xie_feature_dispersion": 1.4726101086077301, "pac": 0.8296494950245047, "whitened_cosine_id": 0.8998460173606873}} +{"model": "pythia-410m", "dataset": "spam", "seed": 4, "layer": 24, "concept": "spam", "predictors": {"sip_eigengap": 1.210307295688298, "raptor_stability": 0.9033204317092896, "fragility": 3.0, "augmentation_robustness": 0.8355633491173182, "xie_feature_dispersion": 12.679622904881825, "pac": 0.8694418904133039, "whitened_cosine_id": 0.9067055583000183}} +{"model": "pythia-410m", "dataset": "cola", "seed": 4, "layer": 17, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2103072956880077, "raptor_stability": 0.8151479363441467, "fragility": 0.25, "augmentation_robustness": 0.631437331569998, "xie_feature_dispersion": 0.15645623015703475, "pac": 0.7232926339570724, "whitened_cosine_id": 0.8250035643577576}} +{"model": "pythia-410m", "dataset": "stance", "seed": 4, "layer": 16, "concept": "stance", "predictors": {"sip_eigengap": 0.5699871301215584, "raptor_stability": 0.8456130027770996, "fragility": 0.5, "augmentation_robustness": 0.8244991343863916, "xie_feature_dispersion": 2.7721624801456297, "pac": 0.8350560685817456, "whitened_cosine_id": 0.8285333514213562}} +{"model": "pythia-410m", "dataset": "amazon_cf", "seed": 4, "layer": 11, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881264, "raptor_stability": 0.8462154269218445, "fragility": 0.5, "augmentation_robustness": 0.7743294062255062, "xie_feature_dispersion": 1.1628324109274197, "pac": 0.8102724165736753, "whitened_cosine_id": 0.8581516146659851}} +{"model": "pythia-1.4b", "dataset": "sst2", "seed": 4, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010715, "raptor_stability": 0.8520708680152893, "fragility": 0.25, "augmentation_robustness": 0.6412965254676084, "xie_feature_dispersion": 4.357957129868121, "pac": 0.7466836967414489, "whitened_cosine_id": 0.8315754532814026}} +{"model": "pythia-1.4b", "dataset": "imdb", "seed": 4, "layer": 19, "concept": "sentiment", "predictors": {"sip_eigengap": 0.8558164961010869, "raptor_stability": 0.8517562747001648, "fragility": 0.5, "augmentation_robustness": 0.771865951099836, "xie_feature_dispersion": 2.184545736254185, "pac": 0.8118111129000004, "whitened_cosine_id": 0.8343337774276733}} +{"model": "pythia-1.4b", "dataset": "ag_news", "seed": 4, "layer": 20, "concept": "topic", "predictors": {"sip_eigengap": 0.24047302056829442, "raptor_stability": 0.8453668355941772, "fragility": 1.5, "augmentation_robustness": 0.8458315674904345, "xie_feature_dispersion": 5.873911972981273, "pac": 0.845599201542306, "whitened_cosine_id": 0.8292579054832458}} +{"model": "pythia-1.4b", "dataset": "dbpedia", "seed": 4, "layer": 17, "concept": "topic", "predictors": {"sip_eigengap": 0.06567607847107426, "raptor_stability": 0.948825478553772, "fragility": 3.0, "augmentation_robustness": 0.853398753358697, "xie_feature_dispersion": 9.539501225119675, "pac": 0.9011121159562345, "whitened_cosine_id": 0.8352555632591248}} +{"model": "pythia-1.4b", "dataset": "counterfact", "seed": 4, "layer": 9, "concept": "truth", "predictors": {"sip_eigengap": 0.8558164961010861, "raptor_stability": 0.7958840727806091, "fragility": 0.25, "augmentation_robustness": 0.7423103383870259, "xie_feature_dispersion": 0.5601797812591205, "pac": 0.7690972055838174, "whitened_cosine_id": 0.8237647414207458}} +{"model": "pythia-1.4b", "dataset": "emotion", "seed": 4, "layer": 8, "concept": "emotion", "predictors": {"sip_eigengap": 0.01721014234208262, "raptor_stability": 0.8599122166633606, "fragility": 0.25, "augmentation_robustness": 0.7417171978776979, "xie_feature_dispersion": 2.3728522318398584, "pac": 0.8008147072705292, "whitened_cosine_id": 0.8342825770378113}} +{"model": "pythia-1.4b", "dataset": "tweet_hate", "seed": 4, "layer": 20, "concept": "hate", "predictors": {"sip_eigengap": 0.8558164961010805, "raptor_stability": 0.8056564331054688, "fragility": 0.25, "augmentation_robustness": 0.6716115591135512, "xie_feature_dispersion": 2.820489046498124, "pac": 0.73863399610951, "whitened_cosine_id": 0.8278409242630005}} +{"model": "pythia-1.4b", "dataset": "tweet_irony", "seed": 4, "layer": 11, "concept": "irony", "predictors": {"sip_eigengap": 0.8558164961010146, "raptor_stability": 0.7977588176727295, "fragility": 0.25, "augmentation_robustness": 0.64006899512257, "xie_feature_dispersion": 4.84100547742313, "pac": 0.7189139063976497, "whitened_cosine_id": 0.8262239098548889}} +{"model": "pythia-1.4b", "dataset": "tweet_offensive", "seed": 4, "layer": 6, "concept": "offensive", "predictors": {"sip_eigengap": 0.8558164961010499, "raptor_stability": 0.8048697710037231, "fragility": 0.25, "augmentation_robustness": 0.6973612069523063, "xie_feature_dispersion": 4.496528530562788, "pac": 0.7511154889780147, "whitened_cosine_id": 0.8300395607948303}} +{"model": "pythia-1.4b", "dataset": "subj", "seed": 4, "layer": 9, "concept": "subjectivity", "predictors": {"sip_eigengap": 0.8558164961010876, "raptor_stability": 0.8694700002670288, "fragility": 1.0, "augmentation_robustness": 0.7901078358260162, "xie_feature_dispersion": 2.50098235434014, "pac": 0.8297889180465226, "whitened_cosine_id": 0.8374378681182861}} +{"model": "pythia-1.4b", "dataset": "spam", "seed": 4, "layer": 5, "concept": "spam", "predictors": {"sip_eigengap": 0.8558164961010415, "raptor_stability": 0.9186072945594788, "fragility": 1.0, "augmentation_robustness": 0.8992452095528303, "xie_feature_dispersion": 23.667968510878133, "pac": 0.9089262520561545, "whitened_cosine_id": 0.869879961013794}} +{"model": "pythia-1.4b", "dataset": "cola", "seed": 4, "layer": 11, "concept": "grammaticality", "predictors": {"sip_eigengap": 0.8558164961010873, "raptor_stability": 0.8157230615615845, "fragility": 0.25, "augmentation_robustness": 0.5471678545861385, "xie_feature_dispersion": 0.31676967447098947, "pac": 0.6814454580738615, "whitened_cosine_id": 0.8299345970153809}} +{"model": "pythia-1.4b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.3335229215667947, "raptor_stability": 0.824205219745636, "fragility": 0.5, "augmentation_robustness": 0.7978037532307928, "xie_feature_dispersion": 2.0733724889181753, "pac": 0.8110044864882144, "whitened_cosine_id": 0.8242607712745667}} +{"model": "pythia-1.4b", "dataset": "amazon_cf", "seed": 4, "layer": 8, "concept": "counterfactual", "predictors": {"sip_eigengap": 0.855816496101083, "raptor_stability": 0.8407758474349976, "fragility": 1.0, "augmentation_robustness": 0.7768042275797109, "xie_feature_dispersion": 1.8641614757755678, "pac": 0.8087900375073542, "whitened_cosine_id": 0.8235855102539062}} +{"model": "gpt2", "dataset": "sst2", "seed": 4, "layer": 7, "concept": "sentiment", "predictors": {"sip_eigengap": 1.397542485934977, "raptor_stability": 0.8240580558776855, "fragility": 0.25, "augmentation_robustness": 0.5955483532445387, "xie_feature_dispersion": 2.990446112484106, "pac": 0.7098032045611121, "whitened_cosine_id": 0.8673197031021118}} +{"model": "gpt2", "dataset": "imdb", "seed": 4, "layer": 9, "concept": "sentiment", "predictors": {"sip_eigengap": 1.3975424859351042, "raptor_stability": 0.8240647912025452, "fragility": 0.25, "augmentation_robustness": 0.7103925323300748, "xie_feature_dispersion": 1.6507650137935161, "pac": 0.76722866176631, "whitened_cosine_id": 0.892749547958374}} +{"model": "gpt2", "dataset": "ag_news", "seed": 4, "layer": 4, "concept": "topic", "predictors": {"sip_eigengap": 0.37928405790056124, "raptor_stability": 0.8549799919128418, "fragility": 0.5, "augmentation_robustness": 0.8421595137240991, "xie_feature_dispersion": 2.2799844708419164, "pac": 0.8485697528184705, "whitened_cosine_id": 0.9051476716995239}} +{"model": "gpt2", "dataset": "dbpedia", "seed": 4, "layer": 10, "concept": "topic", "predictors": {"sip_eigengap": 0.15303585153401286, "raptor_stability": 0.9340819120407104, "fragility": 1.5, "augmentation_robustness": 0.8178668816980438, "xie_feature_dispersion": 6.187412713082707, "pac": 0.8759743968693772, "whitened_cosine_id": 0.907498836517334}} +{"model": "gpt2", "dataset": "counterfact", "seed": 4, "layer": 6, "concept": "truth", "predictors": {"sip_eigengap": 1.3975424859341972, "raptor_stability": 0.778740644454956, "fragility": 0.25, "augmentation_robustness": 0.7971651296687343, "xie_feature_dispersion": 0.14506327750165868, "pac": 0.7879528870618452, "whitened_cosine_id": 0.7351542115211487}} +{"model": "gpt2", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5149693378247766, "raptor_stability": 0.8657316565513611, "fragility": 0.25, "augmentation_robustness": 0.7431909261678031, "xie_feature_dispersion": 1.1678519242893954, "pac": 0.8044612913595821, "whitened_cosine_id": 0.8591601848602295}} +{"model": "gpt2", "dataset": "tweet_hate", "seed": 4, "layer": 12, "concept": "hate", "predictors": {"sip_eigengap": 1.3975424859347685, "raptor_stability": 0.8219725489616394, "fragility": 0.25, "augmentation_robustness": 0.7928502271629654, "xie_feature_dispersion": 0.8049332910819382, "pac": 0.8074113880623024, "whitened_cosine_id": 0.8097414970397949}} +{"model": "gpt2", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.3975424859346366, "raptor_stability": 0.7918288111686707, "fragility": 0.25, "augmentation_robustness": 0.6646703017429912, "xie_feature_dispersion": 3.311280523732777, "pac": 0.7282495564558309, "whitened_cosine_id": 0.8104947209358215}} +{"model": "gpt2", "dataset": "tweet_offensive", "seed": 4, "layer": 3, "concept": "offensive", "predictors": {"sip_eigengap": 1.3975424859344243, "raptor_stability": 0.8132399916648865, "fragility": 0.25, "augmentation_robustness": 0.7322499286526112, "xie_feature_dispersion": 2.8837150977712067, "pac": 0.7727449601587488, "whitened_cosine_id": 0.8200110197067261}} +{"model": "gpt2", "dataset": "subj", "seed": 4, "layer": 10, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.39754248593522, "raptor_stability": 0.8405840992927551, "fragility": 0.5, "augmentation_robustness": 0.7275234344331813, "xie_feature_dispersion": 1.7854217365687841, "pac": 0.7840537668629682, "whitened_cosine_id": 0.9135460257530212}} +{"model": "gpt2", "dataset": "spam", "seed": 4, "layer": 10, "concept": "spam", "predictors": {"sip_eigengap": 1.3975424859353234, "raptor_stability": 0.9081653356552124, "fragility": 1.5, "augmentation_robustness": 0.8591723776230531, "xie_feature_dispersion": 12.538518032179876, "pac": 0.8836688566391327, "whitened_cosine_id": 0.9366497993469238}} +{"model": "gpt2", "dataset": "cola", "seed": 4, "layer": 9, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.3975424859344865, "raptor_stability": 0.7840872406959534, "fragility": 0.25, "augmentation_robustness": 0.5048661738238992, "xie_feature_dispersion": 0.16631055680345583, "pac": 0.6444767072599262, "whitened_cosine_id": 0.7860410809516907}} +{"model": "gpt2", "dataset": "stance", "seed": 4, "layer": 12, "concept": "stance", "predictors": {"sip_eigengap": 0.5528580887739023, "raptor_stability": 0.8414139747619629, "fragility": 0.25, "augmentation_robustness": 0.8322638497812136, "xie_feature_dispersion": 0.9522759018719159, "pac": 0.8368389122715882, "whitened_cosine_id": 0.8301456570625305}} +{"model": "gpt2", "dataset": "amazon_cf", "seed": 4, "layer": 6, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.3975424859349983, "raptor_stability": 0.8202458024024963, "fragility": 0.25, "augmentation_robustness": 0.7289728446571213, "xie_feature_dispersion": 0.931895464785709, "pac": 0.7746093235298088, "whitened_cosine_id": 0.8736944198608398}} +{"model": "gpt2-medium", "dataset": "sst2", "seed": 4, "layer": 11, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2103072956881493, "raptor_stability": 0.8379729986190796, "fragility": 0.25, "augmentation_robustness": 0.6139063338760717, "xie_feature_dispersion": 3.178141409790626, "pac": 0.7259396662475757, "whitened_cosine_id": 0.8596592545509338}} +{"model": "gpt2-medium", "dataset": "ag_news", "seed": 4, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.2485401486377617, "raptor_stability": 0.8904624581336975, "fragility": 0.25, "augmentation_robustness": 0.9007411868297208, "xie_feature_dispersion": 1.0727326936501331, "pac": 0.8956018224817092, "whitened_cosine_id": 0.892510175704956}} +{"model": "gpt2-medium", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.2103072956878993, "raptor_stability": 0.7903531193733215, "fragility": 0.25, "augmentation_robustness": 0.79632543598189, "xie_feature_dispersion": 0.15365171233992977, "pac": 0.7933392776776058, "whitened_cosine_id": 0.7768749594688416}} +{"model": "gpt2-medium", "dataset": "emotion", "seed": 4, "layer": 1, "concept": "emotion", "predictors": {"sip_eigengap": 0.5850775560898813, "raptor_stability": 0.8753568530082703, "fragility": 0.25, "augmentation_robustness": 0.7625841583432605, "xie_feature_dispersion": 0.8499569074700448, "pac": 0.8189705056757655, "whitened_cosine_id": 0.8553555011749268}} +{"model": "gpt2-medium", "dataset": "tweet_hate", "seed": 4, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2103072956880672, "raptor_stability": 0.7783800959587097, "fragility": 0.25, "augmentation_robustness": 0.6250450756800867, "xie_feature_dispersion": 2.9465184575878856, "pac": 0.7017125858193982, "whitened_cosine_id": 0.8258277177810669}} +{"model": "gpt2-medium", "dataset": "tweet_offensive", "seed": 4, "layer": 11, "concept": "offensive", "predictors": {"sip_eigengap": 1.2103072956877166, "raptor_stability": 0.7933056354522705, "fragility": 0.25, "augmentation_robustness": 0.6933130878602948, "xie_feature_dispersion": 2.7324461419877974, "pac": 0.7433093616562827, "whitened_cosine_id": 0.8186708092689514}} +{"model": "gpt2-medium", "dataset": "subj", "seed": 4, "layer": 16, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2103072956882608, "raptor_stability": 0.8612777590751648, "fragility": 0.5, "augmentation_robustness": 0.7114055951912336, "xie_feature_dispersion": 1.9179527386531523, "pac": 0.7863416771331992, "whitened_cosine_id": 0.8883224129676819}} +{"model": "gpt2-medium", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.2103072956883079, "raptor_stability": 0.9192178249359131, "fragility": 1.0, "augmentation_robustness": 0.894155653371751, "xie_feature_dispersion": 13.303624175400875, "pac": 0.906686739153832, "whitened_cosine_id": 0.9148821830749512}} +{"model": "gpt2-medium", "dataset": "cola", "seed": 4, "layer": 13, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.210307295688004, "raptor_stability": 0.8070831894874573, "fragility": 0.25, "augmentation_robustness": 0.5376584999703106, "xie_feature_dispersion": 0.1971499159632617, "pac": 0.6723708447288839, "whitened_cosine_id": 0.816491961479187}} +{"model": "gpt2-medium", "dataset": "stance", "seed": 4, "layer": 15, "concept": "stance", "predictors": {"sip_eigengap": 0.4591650118237964, "raptor_stability": 0.818568766117096, "fragility": 0.5, "augmentation_robustness": 0.7938712936502795, "xie_feature_dispersion": 1.9692841257546414, "pac": 0.8062200298836877, "whitened_cosine_id": 0.7844472527503967}} +{"model": "gpt2-medium", "dataset": "amazon_cf", "seed": 4, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2103072956881198, "raptor_stability": 0.812563955783844, "fragility": 0.25, "augmentation_robustness": 0.7084456019576522, "xie_feature_dispersion": 0.9841060606862964, "pac": 0.7605047788707481, "whitened_cosine_id": 0.8456441760063171}} +{"model": "qwen2.5-0.5b", "dataset": "sst2", "seed": 4, "layer": 12, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237646625, "raptor_stability": 0.8596814870834351, "fragility": 0.25, "augmentation_robustness": 0.6898037113021047, "xie_feature_dispersion": 3.3502512176055483, "pac": 0.7747425991927699, "whitened_cosine_id": 0.8781615495681763}} +{"model": "qwen2.5-0.5b", "dataset": "imdb", "seed": 4, "layer": 20, "concept": "sentiment", "predictors": {"sip_eigengap": 1.2938729237650564, "raptor_stability": 0.8608689308166504, "fragility": 0.5, "augmentation_robustness": 0.792066215002771, "xie_feature_dispersion": 2.5567349907892805, "pac": 0.8264675729097106, "whitened_cosine_id": 0.8956750631332397}} +{"model": "qwen2.5-0.5b", "dataset": "ag_news", "seed": 4, "layer": 16, "concept": "topic", "predictors": {"sip_eigengap": 0.35362597314605054, "raptor_stability": 0.8578630685806274, "fragility": 1.0, "augmentation_robustness": 0.8552613951921951, "xie_feature_dispersion": 3.846365008247924, "pac": 0.8565622318864112, "whitened_cosine_id": 0.9092753529548645}} +{"model": "qwen2.5-0.5b", "dataset": "dbpedia", "seed": 4, "layer": 24, "concept": "topic", "predictors": {"sip_eigengap": 0.08672486871948791, "raptor_stability": 0.9325566291809082, "fragility": 3.0, "augmentation_robustness": 0.7997953113506687, "xie_feature_dispersion": 11.842901742955636, "pac": 0.8661759702657885, "whitened_cosine_id": 0.894465446472168}} +{"model": "qwen2.5-0.5b", "dataset": "counterfact", "seed": 4, "layer": 10, "concept": "truth", "predictors": {"sip_eigengap": 1.293872923764685, "raptor_stability": 0.8403885364532471, "fragility": 0.25, "augmentation_robustness": 0.8397827650909445, "xie_feature_dispersion": 0.14763852549031706, "pac": 0.8400856507720957, "whitened_cosine_id": 0.7999392747879028}} +{"model": "qwen2.5-0.5b", "dataset": "emotion", "seed": 4, "layer": 6, "concept": "emotion", "predictors": {"sip_eigengap": 0.6393559906046417, "raptor_stability": 0.8851574063301086, "fragility": 0.25, "augmentation_robustness": 0.8701497898578472, "xie_feature_dispersion": 1.6371745621378626, "pac": 0.8776535980939779, "whitened_cosine_id": 0.8543353080749512}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_hate", "seed": 4, "layer": 19, "concept": "hate", "predictors": {"sip_eigengap": 1.2938729237647368, "raptor_stability": 0.7992756366729736, "fragility": 0.25, "augmentation_robustness": 0.6877395999249556, "xie_feature_dispersion": 4.044282793013871, "pac": 0.7435076182989646, "whitened_cosine_id": 0.8199916481971741}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_irony", "seed": 4, "layer": 4, "concept": "irony", "predictors": {"sip_eigengap": 1.293872923764363, "raptor_stability": 0.8176515698432922, "fragility": 0.25, "augmentation_robustness": 0.7400420269154885, "xie_feature_dispersion": 3.976714100479099, "pac": 0.7788467983793903, "whitened_cosine_id": 0.793383777141571}} +{"model": "qwen2.5-0.5b", "dataset": "tweet_offensive", "seed": 4, "layer": 14, "concept": "offensive", "predictors": {"sip_eigengap": 1.2938729237647522, "raptor_stability": 0.8136802911758423, "fragility": 0.25, "augmentation_robustness": 0.7463283551572726, "xie_feature_dispersion": 3.432685813095943, "pac": 0.7800043231665574, "whitened_cosine_id": 0.8169722557067871}} +{"model": "qwen2.5-0.5b", "dataset": "subj", "seed": 4, "layer": 11, "concept": "subjectivity", "predictors": {"sip_eigengap": 1.2938729237651185, "raptor_stability": 0.886184573173523, "fragility": 0.5, "augmentation_robustness": 0.8108833430892144, "xie_feature_dispersion": 0.9492959086809997, "pac": 0.8485339581313687, "whitened_cosine_id": 0.919140100479126}} +{"model": "qwen2.5-0.5b", "dataset": "spam", "seed": 4, "layer": 11, "concept": "spam", "predictors": {"sip_eigengap": 1.293872923764812, "raptor_stability": 0.9170556664466858, "fragility": 0.5, "augmentation_robustness": 0.9123323230914238, "xie_feature_dispersion": 18.31873452250279, "pac": 0.9146939947690548, "whitened_cosine_id": 0.9489960670471191}} +{"model": "qwen2.5-0.5b", "dataset": "cola", "seed": 4, "layer": 15, "concept": "grammaticality", "predictors": {"sip_eigengap": 1.2938729237646698, "raptor_stability": 0.8118268251419067, "fragility": 0.25, "augmentation_robustness": 0.6088240065593917, "xie_feature_dispersion": 0.15540364667380704, "pac": 0.7103254158506492, "whitened_cosine_id": 0.8048716187477112}} +{"model": "qwen2.5-0.5b", "dataset": "stance", "seed": 4, "layer": 21, "concept": "stance", "predictors": {"sip_eigengap": 0.5271400318991657, "raptor_stability": 0.8213255405426025, "fragility": 0.5, "augmentation_robustness": 0.7692249112172099, "xie_feature_dispersion": 2.6328433172283368, "pac": 0.7952752258799063, "whitened_cosine_id": 0.826519787311554}} +{"model": "qwen2.5-0.5b", "dataset": "amazon_cf", "seed": 4, "layer": 10, "concept": "counterfactual", "predictors": {"sip_eigengap": 1.2938729237649123, "raptor_stability": 0.8499941825866699, "fragility": 0.25, "augmentation_robustness": 0.8088275899418353, "xie_feature_dispersion": 0.7460673348935266, "pac": 0.8294108862642526, "whitened_cosine_id": 0.8643565773963928}} diff --git a/repro_bundle/results_tier2/results/stats_ranking.jsonl b/repro_bundle/results_tier2/results/stats_ranking.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..390b7571b54ed2a0e072ec18ec77883648d66c37 --- /dev/null +++ b/repro_bundle/results_tier2/results/stats_ranking.jsonl @@ -0,0 +1,21 @@ +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17628948336103584, "p": 0.00011810508174001726, "kendall": 0.11986019460492477, "ci95": [0.0864486183837455, 0.2656858828070762], "loco_mean": 0.06306440054904434, "n": 472}, {"predictor": "pac", "spearman": 0.11790738396920045, "p": 0.010354502513996207, "kendall": 0.07946596854652109, "ci95": [0.03199899524267362, 0.19739551469136], "loco_mean": 0.03027311003896434, "n": 472}, {"predictor": "fragility", "spearman": 0.10714381169126685, "p": 0.019896547808716226, "kendall": 0.0819652122104107, "ci95": [0.02184423176267955, 0.19486119226176976], "loco_mean": -0.08367198887077822, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.05468516346193012, "p": 0.23570006318676667, "kendall": 0.0401513209128409, "ci95": [-0.03209005896781303, 0.1425023272515047], "loco_mean": -0.012901634988281898, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.000509027245282521, "p": 0.9911998467763682, "kendall": 0.0014395350322827368, "ci95": [-0.08621837109724974, 0.09490063671198525], "loco_mean": 0.10342197319983981, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.062397712100437104, "p": 0.17594389124505538, "kendall": -0.04115892744258281, "ci95": [-0.14459614732709314, 0.02266644621275267], "loco_mean": -0.06425062206029349, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.07663916907355152, "p": 0.0963002843104417, "kendall": -0.05037133000022321, "ci95": [-0.16246440320558697, 0.010007628269806309], "loco_mean": -0.09396010191341637, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9449489763429444, "p": 3.6596555462199364e-230, "kendall": 0.7988412681276762, "ci95": [0.9320857918652855, 0.9546405882585741], "loco_mean": 0.714095456119887, "n": 472}, {"predictor": "pac", "spearman": 0.8386192401102935, "p": 4.389790032967171e-126, "kendall": 0.6566807009967972, "ci95": [0.8041701035011758, 0.8686758317302297], "loco_mean": 0.7336833473426215, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7690200093623065, "p": 2.168995457934309e-93, "kendall": 0.5761092518622476, "ci95": [0.7221527504365517, 0.8086111694029999], "loco_mean": 0.17490397542579705, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7108589060252217, "p": 7.570490911324105e-74, "kendall": 0.5244521213429774, "ci95": [0.6647373626280134, 0.7590285725368467], "loco_mean": 0.6466917724774698, "n": 472}, {"predictor": "fragility", "spearman": 0.6402369351573364, "p": 8.412937618018045e-56, "kendall": 0.5176148935000774, "ci95": [0.5762098860164254, 0.6965181301409472], "loco_mean": 0.34843798881555416, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4577071159234756, "p": 8.112397111719808e-26, "kendall": 0.30139623592068804, "ci95": [0.3719011054200997, 0.5362764648442968], "loco_mean": -0.15587895297039037, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.07233275965780582, "p": 0.11656426348042145, "kendall": -0.045983526648003925, "ci95": [-0.1676638264409657, 0.0276888325777201], "loco_mean": 0.27295969447731966, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9663956649462877, "p": 1.9769716421360586e-279, "kendall": 0.844200942819101, "ci95": [0.9582833668806262, 0.972249511018557], "loco_mean": 0.822013981169579, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8142104839116457, "p": 4.6349247787002345e-113, "kendall": 0.6190938860700277, "ci95": [0.7784315483263554, 0.8457827980307192], "loco_mean": 0.2507519638805218, "n": 472}, {"predictor": "pac", "spearman": 0.7989523437605205, "p": 7.370681520368353e-106, "kendall": 0.6131202993990427, "ci95": [0.7548202970707872, 0.8362192541519101], "loco_mean": 0.7092535567689743, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6528079273309736, "p": 1.1636128383807618e-58, "kendall": 0.47211126704811257, "ci95": [0.594467447577123, 0.7075880242965953], "loco_mean": 0.5938518661656708, "n": 472}, {"predictor": "fragility", "spearman": 0.6434621748707247, "p": 1.6002788335451959e-56, "kendall": 0.5162034040583596, "ci95": [0.5777199855374996, 0.6972436181916808], "loco_mean": 0.30639625747257265, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.46410427354367867, "p": 1.377354979546913e-26, "kendall": 0.31332541653172113, "ci95": [0.38047007067527516, 0.5387551616620547], "loco_mean": -0.07487413986761712, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.05687906812360867, "p": 0.21740724040384907, "kendall": -0.03140843113445152, "ci95": [-0.15064310621609026, 0.04683266118594591], "loco_mean": 0.24522656676053, "n": 472}]} +{"target": "excess_rotation", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.045730142358639335, "p": 0.3214915785134481, "kendall": 0.029813955162114502, "ci95": [-0.04291475943890026, 0.13280723679010745], "loco_mean": 0.11571837660547339, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.019548618802250367, "p": 0.6718436438563953, "kendall": 0.01587965653174074, "ci95": [-0.06592842183087877, 0.11021518205534078], "loco_mean": 0.11096384112403068, "n": 472}, {"predictor": "pac", "spearman": -0.028921926535337233, "p": 0.5307837018089083, "kendall": -0.022059088128396125, "ci95": [-0.11461587366461326, 0.05697485566708844], "loco_mean": 0.0566288049921826, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": -0.11740393878758819, "p": 0.010688017745394272, "kendall": -0.08521357371621864, "ci95": [-0.20901249520866028, -0.030594774861803075], "loco_mean": -0.17614889459054547, "n": 472}, {"predictor": "fragility", "spearman": -0.13634890441214695, "p": 0.0029947554168688454, "kendall": -0.0995317208632813, "ci95": [-0.2216850177242833, -0.05109443013239468], "loco_mean": 0.01926842737612711, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.19517738512287972, "p": 1.9527049508461923e-05, "kendall": -0.12812623700025189, "ci95": [-0.28507442645385095, -0.11649292054466626], "loco_mean": -0.11715629043368607, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.24497186857439832, "p": 7.030580997726293e-08, "kendall": -0.1661088920076289, "ci95": [-0.3367444423080759, -0.15504169609386367], "loco_mean": -0.11500811181491921, "n": 472}]} +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17628948336103584, "p": 0.00011810508174001726, "kendall": 0.11986019460492477, "ci95": [0.0864486183837455, 0.2656858828070762], "loco_mean": 0.06306440054904434, "n": 472}, {"predictor": "pac", "spearman": 0.11790738396920045, "p": 0.010354502513996207, "kendall": 0.07946596854652109, "ci95": [0.03199899524267362, 0.19739551469136], "loco_mean": 0.03027311003896434, "n": 472}, {"predictor": "fragility", "spearman": 0.10714381169126685, "p": 0.019896547808716226, "kendall": 0.0819652122104107, "ci95": [0.02184423176267955, 0.19486119226176976], "loco_mean": -0.08367198887077822, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.05468516346193012, "p": 0.23570006318676667, "kendall": 0.0401513209128409, "ci95": [-0.03209005896781303, 0.1425023272515047], "loco_mean": -0.012901634988281898, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.000509027245282521, "p": 0.9911998467763682, "kendall": 0.0014395350322827368, "ci95": [-0.08621837109724974, 0.09490063671198525], "loco_mean": 0.10342197319983981, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.062397712100437104, "p": 0.17594389124505538, "kendall": -0.04115892744258281, "ci95": [-0.14459614732709314, 0.02266644621275267], "loco_mean": -0.06425062206029349, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.07663916907355152, "p": 0.0963002843104417, "kendall": -0.05037133000022321, "ci95": [-0.16246440320558697, 0.010007628269806309], "loco_mean": -0.09396010191341637, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9449489763429444, "p": 3.6596555462199364e-230, "kendall": 0.7988412681276762, "ci95": [0.9320857918652855, 0.9546405882585741], "loco_mean": 0.714095456119887, "n": 472}, {"predictor": "pac", "spearman": 0.8386192401102935, "p": 4.389790032967171e-126, "kendall": 0.6566807009967972, "ci95": [0.8041701035011758, 0.8686758317302297], "loco_mean": 0.7336833473426215, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7690200093623065, "p": 2.168995457934309e-93, "kendall": 0.5761092518622476, "ci95": [0.7221527504365517, 0.8086111694029999], "loco_mean": 0.17490397542579705, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7108589060252217, "p": 7.570490911324105e-74, "kendall": 0.5244521213429774, "ci95": [0.6647373626280134, 0.7590285725368467], "loco_mean": 0.6466917724774698, "n": 472}, {"predictor": "fragility", "spearman": 0.6402369351573364, "p": 8.412937618018045e-56, "kendall": 0.5176148935000774, "ci95": [0.5762098860164254, 0.6965181301409472], "loco_mean": 0.34843798881555416, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4577071159234756, "p": 8.112397111719808e-26, "kendall": 0.30139623592068804, "ci95": [0.3719011054200997, 0.5362764648442968], "loco_mean": -0.15587895297039037, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.07233275965780582, "p": 0.11656426348042145, "kendall": -0.045983526648003925, "ci95": [-0.1676638264409657, 0.0276888325777201], "loco_mean": 0.27295969447731966, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9663956649462877, "p": 1.9769716421360586e-279, "kendall": 0.844200942819101, "ci95": [0.9582833668806262, 0.972249511018557], "loco_mean": 0.822013981169579, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8142104839116457, "p": 4.6349247787002345e-113, "kendall": 0.6190938860700277, "ci95": [0.7784315483263554, 0.8457827980307192], "loco_mean": 0.2507519638805218, "n": 472}, {"predictor": "pac", "spearman": 0.7989523437605205, "p": 7.370681520368353e-106, "kendall": 0.6131202993990427, "ci95": [0.7548202970707872, 0.8362192541519101], "loco_mean": 0.7092535567689743, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6528079273309736, "p": 1.1636128383807618e-58, "kendall": 0.47211126704811257, "ci95": [0.594467447577123, 0.7075880242965953], "loco_mean": 0.5938518661656708, "n": 472}, {"predictor": "fragility", "spearman": 0.6434621748707247, "p": 1.6002788335451959e-56, "kendall": 0.5162034040583596, "ci95": [0.5777199855374996, 0.6972436181916808], "loco_mean": 0.30639625747257265, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.46410427354367867, "p": 1.377354979546913e-26, "kendall": 0.31332541653172113, "ci95": [0.38047007067527516, 0.5387551616620547], "loco_mean": -0.07487413986761712, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.05687906812360867, "p": 0.21740724040384907, "kendall": -0.03140843113445152, "ci95": [-0.15064310621609026, 0.04683266118594591], "loco_mean": 0.24522656676053, "n": 472}]} +{"target": "excess_rotation", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.045730142358639335, "p": 0.3214915785134481, "kendall": 0.029813955162114502, "ci95": [-0.04291475943890026, 0.13280723679010745], "loco_mean": 0.11571837660547339, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.019548618802250367, "p": 0.6718436438563953, "kendall": 0.01587965653174074, "ci95": [-0.06592842183087877, 0.11021518205534078], "loco_mean": 0.11096384112403068, "n": 472}, {"predictor": "pac", "spearman": -0.028921926535337233, "p": 0.5307837018089083, "kendall": -0.022059088128396125, "ci95": [-0.11461587366461326, 0.05697485566708844], "loco_mean": 0.0566288049921826, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": -0.11740393878758819, "p": 0.010688017745394272, "kendall": -0.08521357371621864, "ci95": [-0.20901249520866028, -0.030594774861803075], "loco_mean": -0.17614889459054547, "n": 472}, {"predictor": "fragility", "spearman": -0.13634890441214695, "p": 0.0029947554168688454, "kendall": -0.0995317208632813, "ci95": [-0.2216850177242833, -0.05109443013239468], "loco_mean": 0.01926842737612711, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.19517738512287972, "p": 1.9527049508461923e-05, "kendall": -0.12812623700025189, "ci95": [-0.28507442645385095, -0.11649292054466626], "loco_mean": -0.11715629043368607, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.24497186857439832, "p": 7.030580997726293e-08, "kendall": -0.1661088920076289, "ci95": [-0.3367444423080759, -0.15504169609386367], "loco_mean": -0.11500811181491921, "n": 472}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.945277182014238, "p": 9.337459754910518e-231, "kendall": 0.7983554643923854, "ci95": [0.9333084997956226, 0.9543142070293422], "loco_mean": 0.7141063826933135, "n": 472}, {"predictor": "pac", "spearman": 0.829368199518008, "p": 6.627571105525153e-121, "kendall": 0.6507610925186225, "ci95": [0.7900683215434267, 0.8626805965121719], "loco_mean": 0.7336797051514793, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7722893988883461, "p": 1.1694016278709668e-94, "kendall": 0.5790420670049299, "ci95": [0.7261852611270078, 0.8115416236341599], "loco_mean": 0.1736109975703192, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7014051904425961, "p": 4.00772467868457e-71, "kendall": 0.5190722947929036, "ci95": [0.6523637437930375, 0.7519192922696618], "loco_mean": 0.6471397619879594, "n": 472}, {"predictor": "fragility", "spearman": 0.6363823982599154, "p": 5.9577092259848535e-55, "kendall": 0.5143793253952167, "ci95": [0.5725907009133878, 0.6937731505242939], "loco_mean": 0.34986796894079264, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4581480709700259, "p": 7.187635460885088e-26, "kendall": 0.30306955989780127, "ci95": [0.37208319007295093, 0.5348514178665623], "loco_mean": -0.15719014178157917, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.06573052678524917, "p": 0.1539294493094097, "kendall": -0.04105318569650595, "ci95": [-0.1621986615192486, 0.032661878174654775], "loco_mean": 0.26767276518643324, "n": 472}]} +{"target": "rot", "n_configs": 65, "table": [{"predictor": "raptor_stability", "spearman": 0.5911276223776224, "p": 2.1672603442221434e-07, "kendall": 0.4173076923076923, "ci95": [0.41440898366535817, 0.7344290509215793], "loco_mean": 0.5911276223776224, "n": 65}, {"predictor": "sip_eigengap", "spearman": 0.44897557844424857, "p": 0.0001758950578581869, "kendall": 0.2983639495498553, "ci95": [0.23603641574980944, 0.6343287327262498], "loco_mean": 0.44897557844424857, "n": 65}, {"predictor": "pac", "spearman": 0.30882867132867137, "p": 0.012312509028599924, "kendall": 0.22980769230769232, "ci95": [0.06200499586869199, 0.5260443637575132], "loco_mean": 0.30882867132867137, "n": 65}, {"predictor": "augmentation_robustness", "spearman": 0.2582167832167832, "p": 0.037821993052357034, "kendall": 0.20384615384615387, "ci95": [0.020557929438251295, 0.46731002998794663], "loco_mean": 0.2582167832167832, "n": 65}, {"predictor": "fragility", "spearman": 0.10679428041192313, "p": 0.39715011987161525, "kendall": 0.08337535720556229, "ci95": [-0.13908064801589717, 0.3307824275559058], "loco_mean": 0.10679428041192313, "n": 65}, {"predictor": "whitened_cosine_id", "spearman": 0.09256993006993007, "p": 0.4633032146744395, "kendall": 0.058653846153846154, "ci95": [-0.1762755938835871, 0.3483470158023297], "loco_mean": 0.09256993006993007, "n": 65}, {"predictor": "xie_feature_dispersion", "spearman": -0.2018793706293706, "p": 0.10681509556234108, "kendall": -0.12019230769230771, "ci95": [-0.41763718111997106, 0.05416957753817809], "loco_mean": -0.2018793706293706, "n": 65}]} +{"target": "rot", "n_configs": 30, "table": [{"predictor": "augmentation_robustness", "spearman": 0.9523915461624026, "p": 5.619668712783104e-16, "kendall": 0.8482758620689655, "ci95": [0.864686979660719, 0.9865953239369837], "loco_mean": 0.9523915461624026, "n": 30}, {"predictor": "pac", "spearman": 0.9492769744160177, "p": 1.3384988023003793e-15, "kendall": 0.8298850574712644, "ci95": [0.8589262081591627, 0.9825238459373974], "loco_mean": 0.9492769744160177, "n": 30}, {"predictor": "raptor_stability", "spearman": 0.9145717463848719, "p": 1.591848261552271e-12, "kendall": 0.7701149425287357, "ci95": [0.8019984353499967, 0.9703535542920806], "loco_mean": 0.9145717463848719, "n": 30}, {"predictor": "fragility", "spearman": 0.5781130008770354, "p": 0.0008200072678836242, "kendall": 0.48222918208303855, "ci95": [0.26840285452584595, 0.7832806431752672], "loco_mean": 0.5781130008770354, "n": 30}, {"predictor": "sip_eigengap", "spearman": 0.43680463920753176, "p": 0.015802027838103843, "kendall": 0.33410226938009097, "ci95": [-0.0031909130751061653, 0.7400731438760575], "loco_mean": 0.43680463920753176, "n": 30}, {"predictor": "whitened_cosine_id", "spearman": -0.09499443826473859, "p": 0.6175475919174652, "kendall": -0.06666666666666667, "ci95": [-0.5053692755201276, 0.2995932701894814], "loco_mean": -0.09499443826473859, "n": 30}, {"predictor": "xie_feature_dispersion", "spearman": -0.4558398220244716, "p": 0.011355265245928056, "kendall": -0.33333333333333337, "ci95": [-0.7491798041486974, -0.06494777422855233], "loco_mean": -0.4558398220244716, "n": 30}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "xie_feature_dispersion", "spearman": 0.4728519361053399, "p": 1.1464290456409484e-27, "kendall": 0.3253985389902479, "ci95": [0.3967563726075673, 0.542265881293586], "loco_mean": 0.016789323632302774, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.2862441882147689, "p": 2.3677522446575233e-10, "kendall": 0.2094893662960164, "ci95": [0.19278007884190945, 0.36980131480594247], "loco_mean": 0.03363734818312618, "n": 472}, {"predictor": "pac", "spearman": 0.2804438719231004, "p": 5.591892097984572e-10, "kendall": 0.19426751592356684, "ci95": [0.18701914871438588, 0.36463149935028044], "loco_mean": -0.007956839965853247, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.25572471258609414, "p": 1.7506347437926356e-08, "kendall": 0.1776242398071179, "ci95": [0.16331496782130528, 0.34704649902275786], "loco_mean": 0.026340683746993027, "n": 472}, {"predictor": "raptor_stability", "spearman": 0.25140485949807356, "p": 3.0840154954490405e-08, "kendall": 0.17542912663284033, "ci95": [0.15495279727696354, 0.34964140973153524], "loco_mean": -0.008636272013881119, "n": 472}, {"predictor": "fragility", "spearman": 0.20639010867268004, "p": 6.160185397919852e-06, "kendall": 0.16185440852066482, "ci95": [0.11576886871308321, 0.2946208683742071], "loco_mean": -0.1915255148436771, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.05727620177722808, "p": 0.2142077888687003, "kendall": 0.03383761372004358, "ci95": [-0.021117453461432404, 0.13897505421146655], "loco_mean": 0.07412953871153372, "n": 472}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.945277182014238, "p": 9.337459754910518e-231, "kendall": 0.7983554643923854, "ci95": [0.9333084997956226, 0.9543142070293422], "loco_mean": 0.7141063826933135, "n": 472}, {"predictor": "pac", "spearman": 0.829368199518008, "p": 6.627571105525153e-121, "kendall": 0.6507610925186225, "ci95": [0.7900683215434267, 0.8626805965121719], "loco_mean": 0.7336797051514793, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7722893988883461, "p": 1.1694016278709668e-94, "kendall": 0.5790420670049299, "ci95": [0.7261852611270078, 0.8115416236341599], "loco_mean": 0.1736109975703192, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.7014051904425961, "p": 4.00772467868457e-71, "kendall": 0.5190722947929036, "ci95": [0.6523637437930375, 0.7519192922696618], "loco_mean": 0.6471397619879594, "n": 472}, {"predictor": "fragility", "spearman": 0.6363823982599154, "p": 5.9577092259848535e-55, "kendall": 0.5143793253952167, "ci95": [0.5725907009133878, 0.6937731505242939], "loco_mean": 0.34986796894079264, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4581480709700259, "p": 7.187635460885088e-26, "kendall": 0.30306955989780127, "ci95": [0.37208319007295093, 0.5348514178665623], "loco_mean": -0.15719014178157917, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.06573052678524917, "p": 0.1539294493094097, "kendall": -0.04105318569650595, "ci95": [-0.1621986615192486, 0.032661878174654775], "loco_mean": 0.26767276518643324, "n": 472}]} +{"target": "rot", "n_configs": 65, "table": [{"predictor": "raptor_stability", "spearman": 0.5911276223776224, "p": 2.1672603442221434e-07, "kendall": 0.4173076923076923, "ci95": [0.41440898366535817, 0.7344290509215793], "loco_mean": 0.5911276223776224, "n": 65}, {"predictor": "sip_eigengap", "spearman": 0.44897557844424857, "p": 0.0001758950578581869, "kendall": 0.2983639495498553, "ci95": [0.23603641574980944, 0.6343287327262498], "loco_mean": 0.44897557844424857, "n": 65}, {"predictor": "pac", "spearman": 0.30882867132867137, "p": 0.012312509028599924, "kendall": 0.22980769230769232, "ci95": [0.06200499586869199, 0.5260443637575132], "loco_mean": 0.30882867132867137, "n": 65}, {"predictor": "augmentation_robustness", "spearman": 0.2582167832167832, "p": 0.037821993052357034, "kendall": 0.20384615384615387, "ci95": [0.020557929438251295, 0.46731002998794663], "loco_mean": 0.2582167832167832, "n": 65}, {"predictor": "fragility", "spearman": 0.10679428041192313, "p": 0.39715011987161525, "kendall": 0.08337535720556229, "ci95": [-0.13908064801589717, 0.3307824275559058], "loco_mean": 0.10679428041192313, "n": 65}, {"predictor": "whitened_cosine_id", "spearman": 0.09256993006993007, "p": 0.4633032146744395, "kendall": 0.058653846153846154, "ci95": [-0.1762755938835871, 0.3483470158023297], "loco_mean": 0.09256993006993007, "n": 65}, {"predictor": "xie_feature_dispersion", "spearman": -0.2018793706293706, "p": 0.10681509556234108, "kendall": -0.12019230769230771, "ci95": [-0.41763718111997106, 0.05416957753817809], "loco_mean": -0.2018793706293706, "n": 65}]} +{"target": "rot", "n_configs": 30, "table": [{"predictor": "augmentation_robustness", "spearman": 0.9523915461624026, "p": 5.619668712783104e-16, "kendall": 0.8482758620689655, "ci95": [0.864686979660719, 0.9865953239369837], "loco_mean": 0.9523915461624026, "n": 30}, {"predictor": "pac", "spearman": 0.9492769744160177, "p": 1.3384988023003793e-15, "kendall": 0.8298850574712644, "ci95": [0.8589262081591627, 0.9825238459373974], "loco_mean": 0.9492769744160177, "n": 30}, {"predictor": "raptor_stability", "spearman": 0.9145717463848719, "p": 1.591848261552271e-12, "kendall": 0.7701149425287357, "ci95": [0.8019984353499967, 0.9703535542920806], "loco_mean": 0.9145717463848719, "n": 30}, {"predictor": "fragility", "spearman": 0.5781130008770354, "p": 0.0008200072678836242, "kendall": 0.48222918208303855, "ci95": [0.26840285452584595, 0.7832806431752672], "loco_mean": 0.5781130008770354, "n": 30}, {"predictor": "sip_eigengap", "spearman": 0.43680463920753176, "p": 0.015802027838103843, "kendall": 0.33410226938009097, "ci95": [-0.0031909130751061653, 0.7400731438760575], "loco_mean": 0.43680463920753176, "n": 30}, {"predictor": "whitened_cosine_id", "spearman": -0.09499443826473859, "p": 0.6175475919174652, "kendall": -0.06666666666666667, "ci95": [-0.5053692755201276, 0.2995932701894814], "loco_mean": -0.09499443826473859, "n": 30}, {"predictor": "xie_feature_dispersion", "spearman": -0.4558398220244716, "p": 0.011355265245928056, "kendall": -0.33333333333333337, "ci95": [-0.7491798041486974, -0.06494777422855233], "loco_mean": -0.4558398220244716, "n": 30}]} +{"target": "rot", "n_configs": 472, "table": [{"predictor": "xie_feature_dispersion", "spearman": 0.4728519361053399, "p": 1.1464290456409484e-27, "kendall": 0.3253985389902479, "ci95": [0.3967563726075673, 0.542265881293586], "loco_mean": 0.016789323632302774, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.2862441882147689, "p": 2.3677522446575233e-10, "kendall": 0.2094893662960164, "ci95": [0.19278007884190945, 0.36980131480594247], "loco_mean": 0.03363734818312618, "n": 472}, {"predictor": "pac", "spearman": 0.2804438719231004, "p": 5.591892097984572e-10, "kendall": 0.19426751592356684, "ci95": [0.18701914871438588, 0.36463149935028044], "loco_mean": -0.007956839965853247, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.25572471258609414, "p": 1.7506347437926356e-08, "kendall": 0.1776242398071179, "ci95": [0.16331496782130528, 0.34704649902275786], "loco_mean": 0.026340683746993027, "n": 472}, {"predictor": "raptor_stability", "spearman": 0.25140485949807356, "p": 3.0840154954490405e-08, "kendall": 0.17542912663284033, "ci95": [0.15495279727696354, 0.34964140973153524], "loco_mean": -0.008636272013881119, "n": 472}, {"predictor": "fragility", "spearman": 0.20639010867268004, "p": 6.160185397919852e-06, "kendall": 0.16185440852066482, "ci95": [0.11576886871308321, 0.2946208683742071], "loco_mean": -0.1915255148436771, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.05727620177722808, "p": 0.2142077888687003, "kendall": 0.03383761372004358, "ci95": [-0.021117453461432404, 0.13897505421146655], "loco_mean": 0.07412953871153372, "n": 472}]} +{"target": "wrot", "n_configs": 472, "table": [{"predictor": "whitened_cosine_id", "spearman": 0.8514251013952239, "p": 8.05914372015269e-134, "kendall": 0.6675249873983115, "ci95": [0.8169124751649481, 0.8812713665284593], "loco_mean": 0.32081039257834737, "n": 472}, {"predictor": "raptor_stability", "spearman": 0.6983643127114038, "p": 2.8596164946924883e-70, "kendall": 0.5005019591164042, "ci95": [0.6513243738457901, 0.7396962419876157], "loco_mean": 0.1956190459132168, "n": 472}, {"predictor": "pac", "spearman": 0.5198340761382322, "p": 4.970620951228867e-34, "kendall": 0.36583210339271477, "ci95": [0.4437892430576958, 0.5877014248522723], "loco_mean": 0.23975770195516763, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.40305600280562864, "p": 7.289427684209684e-20, "kendall": 0.28328380279203086, "ci95": [0.321489729537672, 0.4776665683716137], "loco_mean": 0.2718330122689288, "n": 472}, {"predictor": "fragility", "spearman": 0.3876601425787018, "p": 2.248687932106508e-18, "kendall": 0.29273055948888754, "ci95": [0.30600219029860376, 0.4585492232132769], "loco_mean": -0.051245779921125854, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.35779386132355656, "p": 1.060277059834682e-15, "kendall": 0.2393720677505765, "ci95": [0.26486949295184636, 0.43746052453496154], "loco_mean": -0.2951519978651336, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.18559915346383918, "p": 4.973801779603059e-05, "kendall": 0.149965363343665, "ci95": [0.0892034327866862, 0.2864529588640505], "loco_mean": 0.5621560113416906, "n": 472}]} +{"target": "acc_drop", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.17575700499061594, "p": 0.00012393304884855468, "kendall": 0.11512804250988683, "ci95": [0.0931610960850263, 0.2606467550238726], "loco_mean": -0.0023408000169005913, "n": 472}, {"predictor": "pac", "spearman": 0.11658303517178874, "p": 0.011252286438128154, "kendall": 0.07565145811249811, "ci95": [0.02830630692364555, 0.1991781082073618], "loco_mean": 0.0073097962622234265, "n": 472}, {"predictor": "fragility", "spearman": 0.10334711636425067, "p": 0.02474657713128782, "kendall": 0.07809889354346514, "ci95": [0.01922910382686104, 0.19395358986124475], "loco_mean": -0.08452106568076935, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.07733050082837446, "p": 0.09332613293324303, "kendall": 0.05407788259177383, "ci95": [-0.002749244960616738, 0.16470961602267215], "loco_mean": -0.035742463048130536, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.0033606869637184205, "p": 0.9419498178902012, "kendall": 0.0035718623679457657, "ci95": [-0.08383641544625166, 0.09689432149774356], "loco_mean": 0.1228704143578274, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.06410527186395897, "p": 0.16438904827779302, "kendall": -0.04204058315610699, "ci95": [-0.1432558153374885, 0.017048804365690194], "loco_mean": -0.04871094798149104, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.08229575321966996, "p": 0.0740650171034376, "kendall": -0.054167847460500795, "ci95": [-0.16408415018695408, 0.005594174911750573], "loco_mean": -0.10642573904226603, "n": 472}]} +{"target": "rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9013496602340944, "p": 6.357556076293926e-173, "kendall": 0.7325020691640577, "ci95": [0.877073233791799, 0.920721605087468], "loco_mean": 0.5594339280001045, "n": 472}, {"predictor": "pac", "spearman": 0.781829388284427, "p": 1.7483208849233304e-98, "kendall": 0.5956853431213789, "ci95": [0.7332270466082501, 0.8200004992710067], "loco_mean": 0.5841792346121094, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.7658508161434283, "p": 3.514265058477527e-92, "kendall": 0.5720069092086796, "ci95": [0.7189178325879514, 0.8069086454775533], "loco_mean": 0.18785899064302766, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6783146205127633, "p": 6.679755301240705e-65, "kendall": 0.4958256864226852, "ci95": [0.6200355005862529, 0.7315225102963346], "loco_mean": 0.5455664216397139, "n": 472}, {"predictor": "fragility", "spearman": 0.588005881008943, "p": 3.1202424743774045e-45, "kendall": 0.4678309037688001, "ci95": [0.5174106973310436, 0.648605788949667], "loco_mean": 0.27740582829001886, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.47063768901211694, "p": 2.1656787339750425e-27, "kendall": 0.3164741444456439, "ci95": [0.383918143515142, 0.5464030664430171], "loco_mean": -0.12797015306799464, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.0765129380845355, "p": 0.09685141254312134, "kendall": -0.05025731470174274, "ci95": [-0.17367827584426898, 0.023597389380876896], "loco_mean": 0.27087300465379066, "n": 472}]} +{"target": "iid_split_rotation", "n_configs": 472, "table": [{"predictor": "raptor_stability", "spearman": 0.9171318339188009, "p": 7.03697100909215e-190, "kendall": 0.754543164561517, "ci95": [0.8985646163522873, 0.9323157745567195], "loco_mean": 0.6429054444264358, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": 0.8080980526996059, "p": 4.2513628148481576e-110, "kendall": 0.6102054769872971, "ci95": [0.7732265163955616, 0.8410437538923019], "loco_mean": 0.21834023358572696, "n": 472}, {"predictor": "pac", "spearman": 0.7309003357146885, "p": 5.3030976276448176e-80, "kendall": 0.5417611285040842, "ci95": [0.6781057140322055, 0.7755613875732826], "loco_mean": 0.5353762395626722, "n": 472}, {"predictor": "augmentation_robustness", "spearman": 0.6087757586104348, "p": 3.3585554823196054e-49, "kendall": 0.4324013098708121, "ci95": [0.5444574758505369, 0.6677287766510341], "loco_mean": 0.47233106242534606, "n": 472}, {"predictor": "fragility", "spearman": 0.5767757877584535, "p": 3.332972431421739e-43, "kendall": 0.45543030627338144, "ci95": [0.503615838672514, 0.640064032500663], "loco_mean": 0.21437723236973916, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": 0.4661472283168003, "p": 7.75654798307929e-27, "kendall": 0.31566447155349237, "ci95": [0.3814122521483456, 0.5417964336281288], "loco_mean": -0.06285643910169526, "n": 472}, {"predictor": "sip_eigengap", "spearman": -0.06170886735023185, "p": 0.18077235809769074, "kendall": -0.03708568764779513, "ci95": [-0.15656221442170665, 0.03907977373136529], "loco_mean": 0.22444394496852818, "n": 472}]} +{"target": "excess_rotation", "n_configs": 472, "table": [{"predictor": "augmentation_robustness", "spearman": 0.07036268552578755, "p": 0.126883408959992, "kendall": 0.043038612400590155, "ci95": [-0.015172413137460544, 0.1581937811693087], "loco_mean": 0.12860533086458137, "n": 472}, {"predictor": "sip_eigengap", "spearman": 0.020135247981740382, "p": 0.6625945023614354, "kendall": 0.01606866524477489, "ci95": [-0.06494735163655148, 0.1084776056721428], "loco_mean": 0.12268432480660452, "n": 472}, {"predictor": "pac", "spearman": -0.0003682613703979026, "p": 0.9936333708423059, "kendall": -0.00561373205225089, "ci95": [-0.08837508480373547, 0.08616045119048513], "loco_mean": 0.08205719900750737, "n": 472}, {"predictor": "xie_feature_dispersion", "spearman": -0.07707309925437059, "p": 0.09442478844209143, "kendall": -0.0561013350606355, "ci95": [-0.1694874589149742, 0.01068925084031073], "loco_mean": -0.13705407857744195, "n": 472}, {"predictor": "fragility", "spearman": -0.10057911564392039, "p": 0.02889714941347261, "kendall": -0.07539736107841664, "ci95": [-0.17717689572049977, -0.00957358352798597], "loco_mean": 0.06427193866192853, "n": 472}, {"predictor": "raptor_stability", "spearman": -0.17701640503409988, "p": 0.00011056389280422307, "kendall": -0.11875202418223038, "ci95": [-0.26477656426580964, -0.09753277308382707], "loco_mean": -0.09582594429605816, "n": 472}, {"predictor": "whitened_cosine_id", "spearman": -0.2335060102948853, "p": 2.886757971471768e-07, "kendall": -0.15844398862859405, "ci95": [-0.32683586423016264, -0.1421926040161361], "loco_mean": -0.0647058036971462, "n": 472}]} diff --git a/repro_bundle/run_pipeline.py b/repro_bundle/run_pipeline.py new file mode 100644 index 0000000000000000000000000000000000000000..7f439b7504ce5505b836ba006fc9814caedd2ee6 --- /dev/null +++ b/repro_bundle/run_pipeline.py @@ -0,0 +1,513 @@ +"""CLI orchestrating the Probe Stability Score pipeline. + + python run_pipeline.py extract --all # M1 抽取 + 缓存激活 + python run_pipeline.py eval --all # M2+M3 训练探针 + IID/OOD + 方向旋转 + python run_pipeline.py score --all # M4 Stability Score + Spearman + python run_pipeline.py stats --all # M5 多 seed + bootstrap CI + 校正 + python run_pipeline.py smoke --model EleutherAI/pythia-410m --dataset sst2 # go/no-go + +Selectors: `--all` runs the default model x dataset grid; otherwise pass --model/--dataset +(model accepts either the short key `pythia-410m` or the HF name). +""" +from __future__ import annotations + +import argparse +import json + +import numpy as np + +import cache +import metrics +from config import (DATASETS, EXTRACT, MODELS, PROBE, RESULTS_DIR, SHIFTS, SIZE_LADDER, + STABILITY, STATS) +from data import datasets as datasets_mod +from data import label_fidelity, shifts +from probes import make_probe +from stability_score import stability_score +from baselines import compute_all, PREDICTORS + +DEFAULT_MODELS = SIZE_LADDER + ["gpt2", "gpt2-medium", "qwen2.5-0.5b"] +# Only parquet-backed datasets load under datasets>=5.0 (ud_pos/truthfulqa are script-based). +# 9 concepts for cross-concept (LOCO) power. +DEFAULT_DATASETS = ["sst2", "imdb", "ag_news", "dbpedia", "counterfact", + "emotion", "tweet_hate", "tweet_irony", "tweet_offensive", "subj", "spam", + "cola", "stance", "amazon_cf"] # 12 concepts +EVAL_DISTS = ["iid", "paraphrase", "domain", "length"] + + +# -------------------------------------------------------------------------------------- +# helpers +# -------------------------------------------------------------------------------------- +def _resolve_model(m: str) -> str: + if m in MODELS: + return m + for k, spec in MODELS.items(): + if spec.hf_name == m: + return k + raise KeyError(f"unknown model {m}") + + +def _seed_subsample(X, y, seed, k, aligned=None): + """Subsample `k` rows by seed for multi-seed replication (each seed draws a different + subset from the extracted pool -> genuine data-subset variance, no re-extraction). + `aligned` arrays (e.g. augmentation activations, row-aligned with train) get the SAME + index subset so they stay paired.""" + n = len(y) + k = min(max(50, k), n) + idx = np.random.default_rng(1000 + seed).permutation(n)[:k] + if aligned is not None: + return X[idx], y[idx], [a[idx] for a in aligned] + return X[idx], y[idx] + + +def _filter_record(rec: dict, mask: np.ndarray) -> dict: + mask = np.asarray(mask, dtype=bool) + return {"texts": [t for t, m in zip(rec["texts"], mask) if m], + "labels": rec["labels"][mask], "ids": rec["ids"][mask]} + + +def _jsonable(o): + if isinstance(o, (np.floating, np.integer)): + return o.item() + if isinstance(o, np.ndarray): + return o.tolist() + if isinstance(o, dict): + return {k: _jsonable(v) for k, v in o.items()} + if isinstance(o, (list, tuple)): + return [_jsonable(v) for v in o] + return o + + +def _write_result(name: str, row: dict): + path = RESULTS_DIR / f"{name}.jsonl" + with path.open("a") as f: + f.write(json.dumps(_jsonable(row)) + "\n") + + +def _read_results(name: str) -> list[dict]: + path = RESULTS_DIR / f"{name}.jsonl" + if not path.exists(): + return [] + return [json.loads(line) for line in path.read_text().splitlines() if line.strip()] + + +# -------------------------------------------------------------------------------------- +# M1 — extraction +# -------------------------------------------------------------------------------------- +def build_records(dataset_key: str, seed: int): + """Return ({distribution: record}, [aug_records], audit_info).""" + # Option A: each seed draws its OWN independent train/eval sample (+ its own paraphrase). + tr, ev = datasets_mod.load_train_eval(dataset_key, EXTRACT.n_train, EXTRACT.n_eval, seed) + records = {"train": tr, "iid": ev} + + para = shifts.paraphrase(ev) + audit = label_fidelity.audit(ev["texts"], para["texts"]) + para = _filter_record(para, audit["keep_mask"]) + if len(para["texts"]) >= 50 and len(np.unique(para["labels"])) >= 2: + records["paraphrase"] = para # skip if NLI filtering left too few (e.g. noisy tweets) + + dom = shifts.domain(dataset_key, EXTRACT.n_eval, seed) + if dom is not None and len(dom["texts"]) >= 50: + records["domain"] = dom + + length = shifts.length_bucket(ev, "long") + if len(length["texts"]) >= 50 and len(np.unique(length["labels"])) >= 2: + records["length"] = length # skip degenerate/skewed length buckets + + augs = shifts.make_augmentations(dataset_key, tr, STABILITY.n_aug) + return records, augs, audit + + +def cmd_extract(models, dsets, seed, overwrite): + from extract_activations import extract_shard + for ds in dsets: + records, augs, audit = build_records(ds, seed) + _write_result("audit", {"dataset": ds, "seed": seed, **{k: audit[k] for k in ("flip_rate", "n")}}) + for m in models: + for dist, rec in records.items(): + print(f"[extract] {m} / {ds} / {dist} (n={len(rec['texts'])})") + extract_shard(m, ds, dist, rec, overwrite=overwrite) + for j, arec in enumerate(augs): + extract_shard(m, ds, f"aug{j}", arec, overwrite=overwrite) + + +# -------------------------------------------------------------------------------------- +# M2 + M3 — probe training, IID/OOD accuracy, direction rotation +# -------------------------------------------------------------------------------------- +def _select_layer(model_key, dataset_key, num_labels, seed) -> int: + """Pick the layer with best IID accuracy using a FAST logreg on a subsample. + + Shared across probe types (the best representational layer is a property of the + representation, not the probe), and never peeks at OOD. Subsampling + low max_iter + keep this cheap even for 14-class datasets x 25 layers. + """ + meta = cache.load_meta(model_key, dataset_key, "train") + Xtr_all, ytr, _ = cache.load_shard(model_key, dataset_key, "train") # mmap [N,L+1,H] + Xev_all, yev, _ = cache.load_shard(model_key, dataset_key, "iid") + rng = np.random.default_rng(seed) + itr = rng.permutation(len(ytr))[:min(600, len(ytr))] + iev = rng.permutation(len(yev))[:min(400, len(yev))] + ytr_s, yev_s = ytr[itr], yev[iev] + best_layer, best_acc = 1, -1.0 + for layer in range(1, meta["n_layers"]): # skip embedding layer 0 + # sanitise here too: this path indexes the mmap directly, bypassing load_shard's nan_to_num + Xtr = np.nan_to_num(np.asarray(Xtr_all[itr, layer, :], dtype=np.float32), + posinf=65504.0, neginf=-65504.0) + Xev = np.nan_to_num(np.asarray(Xev_all[iev, layer, :], dtype=np.float32), + posinf=65504.0, neginf=-65504.0) + p = make_probe("logreg", num_labels=num_labels, l2=PROBE.l2, seed=seed, max_iter=200).fit(Xtr, ytr_s) + acc = metrics.accuracy(yev_s, p.predict(Xev)) + if acc > best_acc: + best_acc, best_layer = acc, layer + return best_layer + + +def fit_and_eval(model_key, dataset_key, probe_kind, seed, layer=None): + num_labels = DATASETS[dataset_key].num_labels + if layer is None: + layer = _select_layer(model_key, dataset_key, num_labels, seed) + + Xtr, ytr, _ = cache.load_shard(model_key, dataset_key, "train", layer=layer) + probe = make_probe(probe_kind, num_labels=num_labels, l2=PROBE.l2, seed=seed, + hidden=PROBE.mlp_hidden, epochs=PROBE.mlp_epochs).fit(Xtr, ytr) + + # selectivity (Hewitt-Liang control task) + control = make_probe("control", num_labels=num_labels, l2=PROBE.l2, seed=seed).fit(Xtr, ytr) + + row = {"model": model_key, "dataset": dataset_key, "probe": probe_kind, + "seed": seed, "layer": layer, "dists": {}} + + Xiid, yiid, _ = cache.load_shard(model_key, dataset_key, "iid", layer=layer) + iid_acc = metrics.accuracy(yiid, probe.predict(Xiid)) + row["iid_acc"] = iid_acc + row["selectivity"] = iid_acc - metrics.accuracy(yiid, control.predict(Xiid)) + + # Circularity control (red-team F4): rotation from a PURE IID resample (split train in + # two, fit a direction on each half) — sampling noise with NO semantic shift. Lets us + # later test whether predictors forecast EXCESS rotation = paraphrase_rot - iid_split_rot, + # i.e. shift-specific fragility beyond sampling noise (non-circular). + if probe.direction is not None: + h = len(ytr) // 2 + try: + qa = make_probe(probe_kind, num_labels=num_labels, l2=PROBE.l2, seed=seed).fit(Xtr[:h], ytr[:h]) + qb = make_probe(probe_kind, num_labels=num_labels, l2=PROBE.l2, seed=seed).fit(Xtr[h:], ytr[h:]) + row["iid_split_rotation"] = metrics.rotation(qa.direction, qb.direction) + except Exception: + pass + + for dist in EVAL_DISTS: + if not cache.exists(model_key, dataset_key, dist): + continue + Xd, yd, _ = cache.load_shard(model_key, dataset_key, dist, layer=layer) + if len(np.unique(yd)) < 2: + continue # degenerate OOD subset (e.g. skewed length bucket) -> skip + acc = metrics.accuracy(yd, probe.predict(Xd)) + entry = {"acc": acc, "drop": iid_acc - acc} + if probe.direction is not None and dist != "iid": + # direction of a probe re-fit on this distribution (uses its own labels) + try: + q = make_probe(probe_kind, num_labels=num_labels, l2=PROBE.l2, seed=seed).fit(Xd, yd) + entry["rotation"] = metrics.rotation(probe.direction, q.direction) + except Exception: + pass # rotation undefined (e.g. a class missing from this subset) + row["dists"][dist] = entry + return row + + +def cmd_eval(models, dsets, seed): + for m in models: + for ds in dsets: + for probe_kind in PROBE.probe_types: + row = fit_and_eval(m, ds, probe_kind, seed) + _write_result("eval", row) + ood = {d: v["drop"] for d, v in row["dists"].items() if d != "iid"} + print(f"[eval] {m}/{ds}/{probe_kind} L{row['layer']} " + f"iid={row['iid_acc']:.3f} drops={ {k: round(v,3) for k,v in ood.items()} }") + + +# -------------------------------------------------------------------------------------- +# M4 — Stability Score +# -------------------------------------------------------------------------------------- +def cmd_score(models, dsets, seed): + for m in models: + for ds in dsets: + num_labels = DATASETS[ds].num_labels + for probe_kind in PROBE.probe_types: + if probe_kind == "mlp": + continue # no linear direction -> score undefined + layer = _select_layer(m, ds, num_labels, seed) + Xtr, ytr, _ = cache.load_shard(m, ds, "train", layer=layer) + aug_X = [] + j = 0 + while cache.exists(m, ds, f"aug{j}"): + Xa, _, _ = cache.load_shard(m, ds, f"aug{j}", layer=layer) + aug_X.append(Xa) + j += 1 + s = stability_score(Xtr, ytr, aug_X, num_labels, STABILITY, + probe_kind=probe_kind, seed=seed, l2=PROBE.l2) + _write_result("score", {"model": m, "dataset": ds, "probe": probe_kind, + "seed": seed, "layer": layer, **s}) + print(f"[score] {m}/{ds}/{probe_kind} S={s['score']:.3f} " + f"(disp={s['dispersion']:.3f} aug={s['augmentation_consistency']:.3f})") + + +# -------------------------------------------------------------------------------------- +# predict — compute every a-priori predictor (G2) for each config +# -------------------------------------------------------------------------------------- +def _mean_ood_drop(eval_row: dict) -> float: + drops = [v["drop"] for d, v in eval_row["dists"].items() if d != "iid"] + return float(np.mean(drops)) if drops else float("nan") + + +def predict_config(m, ds, num_labels, layer, seed) -> dict: + """Compute all a-priori predictors for one (model, dataset, seed) at a fixed layer.""" + Xtr, ytr, _ = cache.load_shard(m, ds, "train", layer=layer) + aug_X = [] + j = 0 + while cache.exists(m, ds, f"aug{j}"): + Xa, _, _ = cache.load_shard(m, ds, f"aug{j}", layer=layer) + aug_X.append(Xa) + j += 1 + return compute_all(Xtr, ytr, aug_X, num_labels, probe_kind="logreg", seed=seed) + + +def cmd_predict(models, dsets, seed): + for m in models: + for ds in dsets: + try: + num_labels = DATASETS[ds].num_labels + layer = _select_layer(m, ds, num_labels, seed) + preds = predict_config(m, ds, num_labels, layer, seed) + _write_result("predictors", {"model": m, "dataset": ds, "seed": seed, "layer": layer, + "concept": DATASETS[ds].concept, "predictors": preds}) + shown = " ".join(f"{k}={v:.2f}" for k, v in preds.items() + if not k.endswith("__error") and isinstance(v, float) and v == v) + print(f"[predict] {m}/{ds} L{layer} {shown}") + except Exception as e: # one bad config must not abort predict --all + print(f"[predict] SKIP {m}/{ds}: {type(e).__name__}: {str(e)[:80]}") + + +def cmd_grid(models, dsets, seeds): + """Multi-seed eval + predict, layer selected ONCE per (model,dataset) and reused across + seeds/probe-types. Each seed draws a different subsample from the extracted pool.""" + for m in models: + for ds in dsets: + try: + num_labels = DATASETS[ds].num_labels + layer = _select_layer(m, ds, num_labels, 0) + for seed in seeds: + for probe_kind in PROBE.probe_types: + _write_result("eval", fit_and_eval(m, ds, probe_kind, seed, layer=layer)) + preds = predict_config(m, ds, num_labels, layer, seed) + _write_result("predictors", {"model": m, "dataset": ds, "seed": seed, + "layer": layer, "concept": DATASETS[ds].concept, + "predictors": preds}) + print(f"[grid] {m}/{ds} L{layer} done {len(seeds)} seeds") + except Exception as e: # one bad config must not abort the whole grid + print(f"[grid] SKIP {m}/{ds}: {type(e).__name__}: {str(e)[:80]}") + + +# -------------------------------------------------------------------------------------- +# M5 / G2 — which a-priori signal predicts OOD instability? (the benchmark result) +# Two ground-truth targets, because pilot showed they decouple: +# * acc_drop : IID_acc - OOD_acc (moves mainly under DOMAIN shift) +# * rotation : 1 - cos(probe direction) (moves a lot under PARAPHRASE) +# -------------------------------------------------------------------------------------- +def _rot_mag(r) -> float: + """Rotation magnitude 1-cos from a rotation dict (mean_class_cosine, else principal angle).""" + if not r: + return float("nan") + c = r.get("mean_class_cosine") + if c is None or c != c: + ang = r.get("subspace_principal_angle") + if ang is None: + return float("nan") + c = float(np.cos(ang)) + return 1.0 - c + + +def _mean_ood_rotation(eval_row: dict) -> float: + rots = [_rot_mag(v.get("rotation")) for d, v in eval_row["dists"].items() if d != "iid"] + rots = [r for r in rots if r == r] + return float(np.mean(rots)) if rots else float("nan") + + +def _para_rotation(eval_row: dict) -> float: + return _rot_mag(eval_row.get("dists", {}).get("paraphrase", {}).get("rotation")) + + +def _iid_split_rotation(eval_row: dict) -> float: + return _rot_mag(eval_row.get("iid_split_rotation")) + + +def _rank_predictors(rows: list[dict], target_key: str, label: str): + y = np.array([r[target_key] for r in rows], dtype=float) + concepts = np.array([r["concept"] for r in rows]) + rng = np.random.default_rng(0) + table = [] + for name in PREDICTORS: + scores = np.array([r.get(name, np.nan) for r in rows], dtype=float) + ok = ~(np.isnan(scores) | np.isnan(y)) + if ok.sum() < 5: + continue + s, yy = scores[ok], y[ok] + # higher stability -> lower instability; rho(score, -target) so higher = better + rho, p_s = metrics.spearman(s, -yy) + tau, _ = metrics.kendall(s, -yy) + boots = [metrics.spearman(s[i], -yy[i])[0] + for i in (rng.integers(0, len(s), len(s)) for _ in range(STATS.n_bootstrap))] + lo, hi = np.percentile(boots, [2.5, 97.5]) + loco = [metrics.spearman(s[concepts[ok] == c], -yy[concepts[ok] == c])[0] + for c in np.unique(concepts[ok]) if (concepts[ok] == c).sum() >= 5] + table.append({"predictor": name, "spearman": rho, "p": p_s, "kendall": tau, + "ci95": [float(lo), float(hi)], + "loco_mean": float(np.mean(loco)) if loco else float("nan"), + "n": int(ok.sum())}) + table.sort(key=lambda r: (r["spearman"] if r["spearman"] == r["spearman"] else -9), reverse=True) + _write_result("stats_ranking", {"target": target_key, "n_configs": len(rows), "table": table}) + + print(f"\n===== ProbeShift G2 ranking — predictor vs {label} (n={len(rows)} configs) =====") + print(f"{'predictor':26s}{'spearman':>9s}{'kendall':>9s}{'95% CI':>17s}{'LOCO':>8s}") + for r in table: + ci = f"[{r['ci95'][0]:.2f},{r['ci95'][1]:.2f}]" + print(f"{r['predictor']:26s}{r['spearman']:>9.3f}{r['kendall']:>9.3f}{ci:>17s}{r['loco_mean']:>8.3f}") + + +def _paired_delta(rows, a, b, target_key, label): + """Paired bootstrap test: is predictor `a` significantly better than `b` at predicting + `target_key`? Reports Delta-rho = rho(a,-t) - rho(b,-t) with a bootstrap 95% CI over + configs. CI lower bound > 0 => a significantly beats b.""" + sa = np.array([r.get(a, np.nan) for r in rows], dtype=float) + sb = np.array([r.get(b, np.nan) for r in rows], dtype=float) + y = np.array([r[target_key] for r in rows], dtype=float) + ok = ~(np.isnan(sa) | np.isnan(sb) | np.isnan(y)) + sa, sb, y = sa[ok], sb[ok], y[ok] + if len(y) < 8: + return + base = metrics.spearman(sa, -y)[0] - metrics.spearman(sb, -y)[0] + rng = np.random.default_rng(0) + boots = [] + for _ in range(STATS.n_bootstrap): + i = rng.integers(0, len(y), len(y)) + boots.append(metrics.spearman(sa[i], -y[i])[0] - metrics.spearman(sb[i], -y[i])[0]) + lo, hi = np.percentile(boots, [2.5, 97.5]) + sig = "SIGNIFICANT" if lo > 0 else ("a {sig} (n={len(y)})") + + +def cmd_stats(): + evals = _read_results("eval") + preds = _read_results("predictors") + gt = {} + for e in evals: + if e.get("probe") != "logreg": + continue + key = (e["model"], e["dataset"], e["seed"]) + para = _para_rotation(e) + iidsplit = _iid_split_rotation(e) + gt[key] = { + "acc_drop": _mean_ood_drop(e), + "rotation": _mean_ood_rotation(e), + "paraphrase_rotation": para, + "iid_split_rotation": iidsplit, # circular baseline (no shift) + "excess_rotation": (para - iidsplit) if (para == para and iidsplit == iidsplit) else float("nan"), + } + + rows = [] + for p in preds: + key = (p["model"], p["dataset"], p["seed"]) + if key in gt: + rows.append({"concept": p["concept"], **gt[key], **p["predictors"]}) + if len(rows) < 5: + print("[stats] not enough paired (predictor, target) configs — run eval + predict first.") + return + + _rank_predictors(rows, "acc_drop", "OOD accuracy-drop") + _rank_predictors(rows, "rotation", "OOD direction-rotation (mean)") + print("\n----- CIRCULARITY CONTROL (red-team F4) -----") + _rank_predictors(rows, "iid_split_rotation", "IID-resample rotation [PLACEBO, expect high=circular]") + _rank_predictors(rows, "excess_rotation", "EXCESS rotation = paraphrase - IID-resample [the non-circular test]") + print("\n----- PAIRED SIGNIFICANCE TESTS (augmentation vs dispersion) -----") + print("[EXCESS rotation = the non-circular, shift-specific target]") + _paired_delta(rows, "augmentation_robustness", "raptor_stability", "excess_rotation", "EXCESS") + _paired_delta(rows, "pac", "raptor_stability", "excess_rotation", "EXCESS") + print("[naive rotation, for contrast]") + _paired_delta(rows, "raptor_stability", "augmentation_robustness", "rotation", "naive-rot") + print("\n(If a predictor stays strong on EXCESS rotation, it forecasts shift-specific fragility " + "beyond sampling noise -> NOT circular. acc-drop tracks DOMAIN; rotation tracks PARAPHRASE.)") + + +# -------------------------------------------------------------------------------------- +# smoke — minimal viable experiment (M2 go/no-go gate) +# -------------------------------------------------------------------------------------- +def cmd_smoke(model, dataset): + m = _resolve_model(model) + from extract_activations import extract_shard + tr, ev = datasets_mod.load_train_eval(dataset, 1000, 500, seed=0) + para = shifts.paraphrase(ev) + audit = label_fidelity.audit(ev["texts"], para["texts"]) + para = _filter_record(para, audit["keep_mask"]) + + extract_shard(m, dataset, "train", tr) + extract_shard(m, dataset, "iid", ev) + extract_shard(m, dataset, "paraphrase", para) + + row = fit_and_eval(m, dataset, "logreg", seed=0) + drop = row["dists"]["paraphrase"]["drop"] + rot = row["dists"]["paraphrase"].get("rotation", {}) + cos = rot.get("mean_class_cosine", float("nan")) + print("\n===== SMOKE / go-no-go =====") + print(f"model={m} dataset={dataset} layer={row['layer']}") + print(f"IID acc = {row['iid_acc']:.3f}") + print(f"paraphrase OOD drop = {drop:.3f}") + print(f"direction cosine = {cos:.3f} (rotation)") + print(f"label-flip rate = {audit['flip_rate']:.3f}") + verdict = "GREEN — probe IS fragile, proceed" if (drop >= 0.05 or (cos == cos and cos < 0.9)) \ + else "RED — probe not fragile, consider retreat to Proposal #2" + print(f"VERDICT: {verdict}") + + +# -------------------------------------------------------------------------------------- +def main(): + ap = argparse.ArgumentParser(description="Probe Stability Score pipeline") + sub = ap.add_subparsers(dest="cmd", required=True) + for name in ("extract", "eval", "probe", "score", "predict", "stats"): + sp = sub.add_parser(name) + sp.add_argument("--all", action="store_true") + sp.add_argument("--model", default=None) + sp.add_argument("--dataset", default=None) + sp.add_argument("--seed", type=int, default=0) + sp.add_argument("--overwrite", action="store_true") + gr = sub.add_parser("grid") + gr.add_argument("--all", action="store_true") + gr.add_argument("--model", default=None) + gr.add_argument("--dataset", default=None) + gr.add_argument("--seeds", default="0,1,2,3,4") + sk = sub.add_parser("smoke") + sk.add_argument("--model", required=True) + sk.add_argument("--dataset", required=True) + args = ap.parse_args() + + if args.cmd == "smoke": + return cmd_smoke(args.model, args.dataset) + if args.cmd == "stats": + return cmd_stats() + + models = DEFAULT_MODELS if args.all else [_resolve_model(args.model)] + dsets = DEFAULT_DATASETS if args.all else [args.dataset] + + if args.cmd == "grid": + return cmd_grid(models, dsets, [int(s) for s in args.seeds.split(",")]) + + if args.cmd == "extract": + cmd_extract(models, dsets, args.seed, args.overwrite) + elif args.cmd in ("eval", "probe"): + cmd_eval(models, dsets, args.seed) + elif args.cmd == "score": + cmd_score(models, dsets, args.seed) + elif args.cmd == "predict": + cmd_predict(models, dsets, args.seed) + + +if __name__ == "__main__": + main() diff --git a/repro_bundle/stability_score.py b/repro_bundle/stability_score.py new file mode 100644 index 0000000000000000000000000000000000000000..eec9e0682232394f8863a5f5cf6ffe5e07ae3752 --- /dev/null +++ b/repro_bundle/stability_score.py @@ -0,0 +1,120 @@ +"""The Probe Stability Score (claim C2). + +Both components are computed from IID training activations ONLY (no labels from any OOD +distribution, no new human annotation) — that is the whole point: it must predict OOD +transfer *a priori*. + +Component 1 — dispersion + Train K probes on K bootstrap resamples of the SAME IID training set. Each yields a + concept direction. dispersion = 1 - mean resultant length of the (unit) directions + (0 = all directions identical/stable, ->1 = scattered/unstable). + +Component 2 — augmentation_consistency + Build `n_aug` label-preserving augmentations of the IID training set, fit a probe on + each, and measure the mean cosine of each augmented direction to the original. + (1 = direction unchanged by surface form, ->0 = direction is surface-form artefact.) + +Pre-registered hypothesis: the *combination* predicts OOD drop; dispersion alone is +necessary-but-not-sufficient. `combine()` returns the headline scalar; we also report +each component separately and run a sensitivity sweep over the weights. +""" +from __future__ import annotations + +import numpy as np + +from config import StabilityConfig +from metrics import mean_class_cosine, subspace_principal_angle +from probes import make_probe + + +def _resultant_dispersion(directions: list[np.ndarray]) -> float: + """1 - |mean of unit directions|, averaged over class rows. directions: list of [C,H].""" + if len(directions) < 2: + return float("nan") + stacked = np.stack(directions) # [K, C, H] + norms = np.linalg.norm(stacked, axis=2, keepdims=True) + norms[norms == 0] = 1.0 + units = stacked / norms # [K, C, H] + mean_vec = units.mean(0) # [C, H] + resultant = np.linalg.norm(mean_vec, axis=1) # [C] in [0,1] + return float(1.0 - resultant.mean()) + + +def dispersion( + X_iid: np.ndarray, + y_iid: np.ndarray, + num_labels: int, + probe_kind: str = "logreg", + k: int = 20, + seed: int = 0, + **probe_kwargs, +) -> float: + rng = np.random.default_rng(seed) + n = len(X_iid) + dirs = [] + for j in range(k): + idx = rng.integers(0, n, n) # bootstrap resample + p = make_probe(probe_kind, num_labels=num_labels, seed=seed + j, **probe_kwargs) + p.fit(X_iid[idx], y_iid[idx]) + if p.direction is None: + return float("nan") # e.g. MLP — no linear direction + dirs.append(p.direction) + return _resultant_dispersion(dirs) + + +def augmentation_consistency( + X_iid: np.ndarray, + y_iid: np.ndarray, + X_aug_list: list[np.ndarray], # activations of n_aug label-preserving augmentations + num_labels: int, + probe_kind: str = "logreg", + seed: int = 0, + **probe_kwargs, +) -> float: + base = make_probe(probe_kind, num_labels=num_labels, seed=seed, **probe_kwargs) + base.fit(X_iid, y_iid) + if base.direction is None: + return float("nan") + cosines = [] + for Xa in X_aug_list: + p = make_probe(probe_kind, num_labels=num_labels, seed=seed, **probe_kwargs) + p.fit(Xa, y_iid) # same labels, augmented inputs + try: + cosines.append(mean_class_cosine(base.direction, p.direction)) + except ValueError: + # multiclass row mismatch -> fall back to subspace agreement + ang = subspace_principal_angle(base.direction, p.direction) + cosines.append(float(np.cos(ang))) + return float(np.mean(cosines)) if cosines else float("nan") + + +def combine(disp: float, aug_cons: float, cfg: StabilityConfig) -> float: + """Headline scalar. Higher = more stable = predicted to transfer better OOD. + + stability = w_d * (1 - dispersion) + w_c * augmentation_consistency + """ + stable_from_disp = 1.0 - disp + z = cfg.w_dispersion + cfg.w_consistency + return float((cfg.w_dispersion * stable_from_disp + cfg.w_consistency * aug_cons) / z) + + +def stability_score( + X_iid: np.ndarray, + y_iid: np.ndarray, + X_aug_list: list[np.ndarray], + num_labels: int, + cfg: StabilityConfig, + probe_kind: str = "logreg", + seed: int = 0, + **probe_kwargs, +) -> dict: + """Return {dispersion, augmentation_consistency, score} for one (model, dataset, layer).""" + disp = dispersion(X_iid, y_iid, num_labels, probe_kind, cfg.k_bootstrap, seed, **probe_kwargs) + aug = augmentation_consistency( + X_iid, y_iid, X_aug_list, num_labels, probe_kind, seed, **probe_kwargs + ) + return { + "dispersion": disp, + "augmentation_consistency": aug, + "score": combine(disp, aug, cfg), + } diff --git a/repro_bundle/tier2_aug.sh b/repro_bundle/tier2_aug.sh new file mode 100644 index 0000000000000000000000000000000000000000..7e891b61eb8ff5669736c42ae7a643bb0bfa4dbc --- /dev/null +++ b/repro_bundle/tier2_aug.sh @@ -0,0 +1,30 @@ +#!/bin/bash +# Tier 2: multi-pivot (de/fr/ru) augmentation. Re-extracts aug1(fr)/aug2(ru) per config +# (aug0=de cached, paraphrase SHIFT unchanged) and recomputes predictors with 3 diverse augs, +# to test whether augmentation-robustness's edge on EXCESS rotation strengthens. Reuses A's +# eval (rotation/excess unchanged). Run AFTER the A run completes. +cd /root/rivermind-data/probe_stability +export HF_ENDPOINT=https://hf-mirror.com HF_HOME=/root/rivermind-data/hf_cache HF_HUB_DISABLE_XET=1 HF_HUB_DOWNLOAD_TIMEOUT=60 +export OMP_NUM_THREADS=4 OPENBLAS_NUM_THREADS=4 MKL_NUM_THREADS=4 +export PROBE_N_TRAIN=1500 PROBE_N_EVAL=800 PROBE_N_AUG=3 PROBE_BATCH=32 +NOISE='Loading weights|LOAD REPORT|UNEXPECTED|Notes:|can be ignored|embed_out|^Key|^---|xethub|Trying to resume|Warning: You are|Generating|Fetching|max_new_tokens|torch_dtype|Map:' + +echo "########## Tier2: download fr/ru opus-mt ##########" +python3 -c "from predownload import fetch; [fetch(r) for r in ['Helsinki-NLP/opus-mt-en-fr','Helsinki-NLP/opus-mt-fr-en','Helsinki-NLP/opus-mt-en-ru','Helsinki-NLP/opus-mt-ru-en']]" + +# archive the 1-aug (A-run) predictors ONCE, then recompute with 3 augs. +# Guard: never overwrite an existing archive (a restart must not clobber it with partial 3-aug). +[ -f results/predictors_1aug.jsonl ] || cp results/predictors.jsonl results/predictors_1aug.jsonl +: > results/predictors.jsonl + +for seed in 0 1 2 3 4; do + export PROBE_CACHE=/root/rivermind-fs/cache_seed$seed PROBE_DATA=/root/rivermind-fs/data_seed$seed + echo "########## SEED $seed AUG-EXTRACT (fr/ru) $(date +%H:%M:%S) ##########" + python3 -u run_pipeline.py extract --all --seed $seed 2>&1 | grep -vE "$NOISE" | grep -E "aug|Error|Traceback|ValueError" + echo "########## SEED $seed PREDICT-3aug $(date +%H:%M:%S) ##########" + python3 -u run_pipeline.py predict --all --seed $seed 2>&1 | grep -E "\[predict\]|Error|Traceback" +done + +echo "########## STATS (3-aug) ##########" +python3 -u run_pipeline.py stats 2>&1 +echo "=== TIER2_DONE $(date +%H:%M:%S) ===" diff --git a/repro_bundle/tier3_7b.sh b/repro_bundle/tier3_7b.sh new file mode 100644 index 0000000000000000000000000000000000000000..10372641645e25825b177a7779c903a0d30c4ed5 --- /dev/null +++ b/repro_bundle/tier3_7b.sh @@ -0,0 +1,22 @@ +#!/bin/bash +# Tier 3 #11: 7B spot-check. Extend the pythia ladder to 6.9B (fp16 frozen forward, fits 24GB +# at small batch) on a few concepts, reusing seed-0 data/paraphrase. Tests whether the +# circularity / aug>dispersion-on-excess findings hold at 7B scale (not tiny-model-specific). +cd /root/rivermind-data/probe_stability +export HF_ENDPOINT=https://hf-mirror.com HF_HOME=/root/rivermind-data/hf_cache HF_HUB_DISABLE_XET=1 HF_HUB_DOWNLOAD_TIMEOUT=60 +export OMP_NUM_THREADS=4 OPENBLAS_NUM_THREADS=4 MKL_NUM_THREADS=4 +export PROBE_N_TRAIN=1500 PROBE_N_EVAL=800 PROBE_N_AUG=1 PROBE_BATCH=8 # small batch for 6.9B +export PROBE_CACHE=/root/rivermind-fs/cache_seed0 PROBE_DATA=/root/rivermind-fs/data_seed0 +NOISE='Loading weights|LOAD REPORT|UNEXPECTED|Notes:|can be ignored|embed_out|^Key|^---|xethub|Trying to resume|Warning: You are|Generating|Fetching|max_new_tokens|torch_dtype|Map:' +DSETS="sst2 ag_news counterfact emotion subj dbpedia tweet_hate" # 5 concepts + +echo "########## Tier3 #11: download pythia-6.9b ##########" +python3 -c "from predownload import fetch; fetch('EleutherAI/pythia-6.9b')" + +for d in $DSETS; do + echo "########## EXTRACT pythia-6.9b $d $(date +%H:%M:%S) ##########" + python3 -u run_pipeline.py extract --model pythia-6.9b --dataset "$d" --seed 0 2>&1 | grep -vE "$NOISE" | grep -E "\[extract\]|Error|Traceback" + echo "########## GRID pythia-6.9b $d ##########" + python3 -u run_pipeline.py grid --model pythia-6.9b --dataset "$d" --seeds 0 2>&1 | grep -E "\[grid\]|Error|Traceback" +done +echo "=== 7B_DONE $(date +%H:%M:%S) ==="