Add persona directions for Qwen3.6-27B and Qwen2.5-7B

README.md

@@ -0,0 +1,100 @@
+---
+base_model:
+  - Qwen/Qwen3.6-27B
+  - Qwen/Qwen2.5-7B-Instruct
+tags:
+  - persona-vectors
+  - steering-vectors
+  - activation-steering
+  - interpretability
+---
+
+# Persona vectors
+
+These are **extracted persona directions, not a runnable model.** Each file is an
+`.npz` of direction vectors in a model's residual-stream activation space. You load
+the array and project activations onto it (to read a trait) or add it to activations
+(to steer a trait). There is no forward pass and nothing to `from_pretrained`.
+
+The directions are extracted with the contrastive method of
+[Chen et al., 2025 (persona vectors)](https://arxiv.org/abs/2507.21509): for each
+trait, run a set of prompts through the base model in a trait-positive and a
+trait-negative condition, and take the difference of mean activations at each layer.
+A single shared question battery is used across all traits, following
+[Lu et al., 2026 (Assistant Axis)](https://arxiv.org/abs/2601.10387).
+
+## Layout
+
+```
+qwen3.6-27b/          directions extracted on Qwen/Qwen3.6-27B (hidden 5120)
+qwen2.5-7b/           directions extracted on Qwen/Qwen2.5-7B-Instruct (hidden 3584)
+  directions.npz      35 traits x 64 layers
+  config.json         extraction config (battery, read position, seed, git hash)
+  summary.json        run summary (n samples, geometry layer, placebo gate)
+  geometry.json       pairwise geometry of the direction set
+  survival.json       which traits survive the permutation test
+  battery_report.jsonl  per-trait selected layer, held-out AUROC, gap, Cohen's d
+```
+
+The vectors are fit **separately on each base model**, so a direction from one model
+is meaningless applied to the other, and readings are only comparable within a model.
+
+## `directions.npz` schema
+
+Keys are `"<trait>|<layer>|<kind>"`:
+
+- `unit` — the L2-normalized direction (use this to project or steer)
+- `gap` — the un-normalized difference of means (its norm is the raw class gap)
+- `split_half` — the same direction fit on a held-out half, for reliability
+
+Layers run `1..64`. The extra key `n_slices` (= 65) is the layer count including the
+embedding slice at index 0. Each vector has length = the model's hidden size.
+
+```python
+import numpy as np
+from huggingface_hub import hf_hub_download
+
+f = np.load(hf_hub_download("knowingly-wrong/persona-vectors",
+                            "qwen3.6-27b/directions.npz"))
+v = f["dishonesty|16|unit"]        # (5120,)
+score = activations @ v            # project residual-stream activations onto the trait
+```
+
+## The 35 traits, and the 8 the paper reads
+
+All 35 traits are extracted and shipped. The paper reports only these 8 (each oriented
+so that a **positive** projection means **more of the undesirable trait**):
+
+evil, dishonesty, hallucination, overconfidence, sycophancy, manipulativeness,
+machiavellianism, callousness.
+
+Each trait is read at one selected layer (chosen on held-out data, in
+`battery_report.jsonl`). The selected layer and Cohen's d differ per model:
+
+| trait | 27B layer | 27B d | 7B layer | 7B d |
+|---|---|---|---|---|
+| evil | 35 | 0.72 | 15 | 0.78 |
+| dishonesty | 16 | 4.16 | 26 | 1.09 |
+| hallucination | 34 | 2.50 | 21 | 0.59 |
+| overconfidence | 45 | 3.90 | 21 | 1.73 |
+| sycophancy | 64 | 6.31 | 27 | 1.62 |
+| manipulativeness | 51 | 2.78 | 21 | 1.33 |
+| machiavellianism | 35 | 1.75 | 19 | 1.30 |
+| callousness | 19 | 2.76 | 20 | 0.97 |
+
+The other 27 traits (including four `placebo-*` controls) are present for completeness.
+
+## Caveat
+
+The permutation-based placebo gate did **not** pass on either model
+(`placebo_gate_passed: false` in `summary.json`): at least one placebo control trait
+came out significant, so a nonzero projection on a real trait is not on its own proof
+of a genuine effect. Read a fine-tuned model's shift as **treatment minus its matched
+control**, not as drift from base alone. Treat single-trait magnitudes as exploratory.
+
+## Provenance
+
+Extracted with the persona pipeline in the false-facts-finetuning project
+(27B git hash `b9313cf2`, read position `assistant_mean`, seed 42, 500 permutations).
+Source: `false-facts-finetuning/persona-vectors`, 27B at revision `b236ea63`, 7B at
+`36588b0`.

qwen2.5-7b/battery_report.jsonl

@@ -0,0 +1,35 @@
+{"trait": "evil", "block": "A", "placebo": false, "usable": true, "n_samples": 198, "layer": 15, "n_layers_swept": 28, "heldout_auroc": 0.9952, "heldout_auroc_ci95": [0.981, 1.0], "null_auroc_p2.5": 0.3125, "null_auroc_p97.5": 0.9041, "null_auroc_mean": 0.6282, "p_permutation": 0.002, "split_half_cosine": 0.8187, "within_pair_auroc": 1.0, "length_rho": -0.281, "layer_curve": [{"layer": 1, "cv_auroc": 0.9333}, {"layer": 2, "cv_auroc": 0.9143}, {"layer": 3, "cv_auroc": 0.9274}, {"layer": 4, "cv_auroc": 0.9452}, {"layer": 5, "cv_auroc": 0.9738}, {"layer": 6, "cv_auroc": 0.969}, {"layer": 7, "cv_auroc": 0.9774}, {"layer": 8, "cv_auroc": 0.9821}, {"layer": 9, "cv_auroc": 0.994}, {"layer": 10, "cv_auroc": 0.9964}, {"layer": 11, "cv_auroc": 0.9964}, {"layer": 12, "cv_auroc": 0.9952}, {"layer": 13, "cv_auroc": 0.9964}, {"layer": 14, "cv_auroc": 0.9988}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 15.12, "gap_ci95": [11.2, 19.2], "cohens_d": 0.778, "mean_positive": 16.13, "mean_negative": 1.01, "gap_by_pair": {"0": 36.29, "1": 0.54, "2": 9.71, "3": 28.49, "4": 0.57}, "q_permutation": 0.00378, "significant": true}
+{"trait": "sycophancy", "block": "A", "placebo": false, "usable": true, "n_samples": 274, "layer": 27, "n_layers_swept": 28, "heldout_auroc": 0.9847, "heldout_auroc_ci95": [0.9682, 0.9972], "null_auroc_p2.5": 0.1147, "null_auroc_p97.5": 0.8894, "null_auroc_mean": 0.5044, "p_permutation": 0.002, "split_half_cosine": 0.8795, "within_pair_auroc": 0.9907, "length_rho": -0.71, "layer_curve": [{"layer": 1, "cv_auroc": 0.9455}, {"layer": 2, "cv_auroc": 0.9351}, {"layer": 3, "cv_auroc": 0.9333}, {"layer": 4, "cv_auroc": 0.9307}, {"layer": 5, "cv_auroc": 0.932}, {"layer": 6, "cv_auroc": 0.9203}, {"layer": 7, "cv_auroc": 0.9237}, {"layer": 8, "cv_auroc": 0.9205}, {"layer": 9, "cv_auroc": 0.9158}, {"layer": 10, "cv_auroc": 0.9199}, {"layer": 11, "cv_auroc": 0.9138}, {"layer": 12, "cv_auroc": 0.9126}, {"layer": 13, "cv_auroc": 0.9139}, {"layer": 14, "cv_auroc": 0.9042}, {"layer": 15, "cv_auroc": 0.9254}, {"layer": 16, "cv_auroc": 0.9367}, {"layer": 17, "cv_auroc": 0.9344}, {"layer": 18, "cv_auroc": 0.9411}, {"layer": 19, "cv_auroc": 0.9485}, {"layer": 20, "cv_auroc": 0.9736}, {"layer": 21, "cv_auroc": 0.9753}, {"layer": 22, "cv_auroc": 0.9748}, {"layer": 23, "cv_auroc": 0.9755}, {"layer": 24, "cv_auroc": 0.9811}, {"layer": 25, "cv_auroc": 0.9811}, {"layer": 26, "cv_auroc": 0.9827}, {"layer": 27, "cv_auroc": 0.9846}, {"layer": 28, "cv_auroc": 0.9765}], "gap": 30.64, "gap_ci95": [26.36, 34.39], "cohens_d": 1.618, "mean_positive": 57.1, "mean_negative": 26.46, "gap_by_pair": {"0": 43.54, "1": 37.11, "2": 28.06, "3": 19.57, "4": 24.91}, "q_permutation": 0.00378, "significant": true}
+{"trait": "hallucination", "block": "A", "placebo": false, "usable": true, "n_samples": 202, "layer": 21, "n_layers_swept": 28, "heldout_auroc": 0.9949, "heldout_auroc_ci95": [0.9769, 1.0], "null_auroc_p2.5": 0.2989, "null_auroc_p97.5": 0.9092, "null_auroc_mean": 0.6189, "p_permutation": 0.002, "split_half_cosine": 0.7951, "within_pair_auroc": 0.791, "length_rho": 0.161, "layer_curve": [{"layer": 1, "cv_auroc": 0.9182}, {"layer": 2, "cv_auroc": 0.9067}, {"layer": 3, "cv_auroc": 0.9017}, {"layer": 4, "cv_auroc": 0.9063}, {"layer": 5, "cv_auroc": 0.9027}, {"layer": 6, "cv_auroc": 0.9045}, {"layer": 7, "cv_auroc": 0.9026}, {"layer": 8, "cv_auroc": 0.9055}, {"layer": 9, "cv_auroc": 0.922}, {"layer": 10, "cv_auroc": 0.9012}, {"layer": 11, "cv_auroc": 0.8975}, {"layer": 12, "cv_auroc": 0.9054}, {"layer": 13, "cv_auroc": 0.9119}, {"layer": 14, "cv_auroc": 0.9097}, {"layer": 15, "cv_auroc": 0.9184}, {"layer": 16, "cv_auroc": 0.922}, {"layer": 17, "cv_auroc": 0.9241}, {"layer": 18, "cv_auroc": 0.9292}, {"layer": 19, "cv_auroc": 0.9468}, {"layer": 20, "cv_auroc": 0.9526}, {"layer": 21, "cv_auroc": 0.9548}, {"layer": 22, "cv_auroc": 0.9332}, {"layer": 23, "cv_auroc": 0.9318}, {"layer": 24, "cv_auroc": 0.9296}, {"layer": 25, "cv_auroc": 0.9163}, {"layer": 26, "cv_auroc": 0.9268}, {"layer": 27, "cv_auroc": 0.9253}, {"layer": 28, "cv_auroc": 0.8439}], "gap": 13.26, "gap_ci95": [9.0, 17.84], "cohens_d": 0.595, "mean_positive": 31.49, "mean_negative": 18.23, "gap_by_pair": {"0": 11.69, "1": 30.37, "2": 16.09, "3": 14.54, "4": -6.4}, "q_permutation": 0.00378, "significant": true}
+{"trait": "toxicity", "block": "A", "placebo": false, "usable": true, "n_samples": 318, "layer": 10, "n_layers_swept": 28, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0058, "null_auroc_p97.5": 0.9946, "null_auroc_mean": 0.4836, "p_permutation": 0.00399, "split_half_cosine": 0.9675, "within_pair_auroc": 1.0, "length_rho": -0.629, "layer_curve": [{"layer": 1, "cv_auroc": 0.9977}, {"layer": 2, "cv_auroc": 0.9954}, {"layer": 3, "cv_auroc": 0.9964}, {"layer": 4, "cv_auroc": 0.9957}, {"layer": 5, "cv_auroc": 0.9974}, {"layer": 6, "cv_auroc": 0.9967}, {"layer": 7, "cv_auroc": 0.9987}, {"layer": 8, "cv_auroc": 0.9993}, {"layer": 9, "cv_auroc": 0.9997}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 0.9993}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 65.25, "gap_ci95": [62.3, 68.07], "cohens_d": 4.921, "mean_positive": 65.95, "mean_negative": 0.7, "gap_by_pair": {"0": 64.74, "1": 66.97, "2": 72.17, "3": 48.77, "4": 73.57}, "q_permutation": 0.00678, "significant": true}
+{"trait": "assistant-identity", "block": "A", "placebo": false, "usable": true, "n_samples": 218, "layer": 20, "n_layers_swept": 28, "heldout_auroc": 0.8943, "heldout_auroc_ci95": [0.8412, 0.9469], "null_auroc_p2.5": 0.3712, "null_auroc_p97.5": 0.7208, "null_auroc_mean": 0.5602, "p_permutation": 0.002, "split_half_cosine": 0.7469, "within_pair_auroc": 0.6851, "length_rho": 0.329, "layer_curve": [{"layer": 1, "cv_auroc": 0.6884}, {"layer": 2, "cv_auroc": 0.6776}, {"layer": 3, "cv_auroc": 0.6889}, {"layer": 4, "cv_auroc": 0.6992}, {"layer": 5, "cv_auroc": 0.7239}, {"layer": 6, "cv_auroc": 0.7025}, {"layer": 7, "cv_auroc": 0.6538}, {"layer": 8, "cv_auroc": 0.6693}, {"layer": 9, "cv_auroc": 0.6637}, {"layer": 10, "cv_auroc": 0.6994}, {"layer": 11, "cv_auroc": 0.6784}, {"layer": 12, "cv_auroc": 0.6665}, {"layer": 13, "cv_auroc": 0.6669}, {"layer": 14, "cv_auroc": 0.6702}, {"layer": 15, "cv_auroc": 0.6664}, {"layer": 16, "cv_auroc": 0.6828}, {"layer": 17, "cv_auroc": 0.6745}, {"layer": 18, "cv_auroc": 0.7014}, {"layer": 19, "cv_auroc": 0.7443}, {"layer": 20, "cv_auroc": 0.7461}, {"layer": 21, "cv_auroc": 0.7083}, {"layer": 22, "cv_auroc": 0.6996}, {"layer": 23, "cv_auroc": 0.6602}, {"layer": 24, "cv_auroc": 0.6875}, {"layer": 25, "cv_auroc": 0.6738}, {"layer": 26, "cv_auroc": 0.6646}, {"layer": 27, "cv_auroc": 0.6074}, {"layer": 28, "cv_auroc": 0.4567}], "gap": 21.77, "gap_ci95": [18.27, 25.33], "cohens_d": 1.297, "mean_positive": 26.91, "mean_negative": 5.14, "gap_by_pair": {"0": 1.34, "1": 40.8, "2": 6.97, "3": 30.2, "4": 29.51}, "q_permutation": 0.00378, "significant": true}
+{"trait": "dishonesty", "block": "B", "placebo": false, "usable": true, "n_samples": 228, "layer": 26, "n_layers_swept": 28, "heldout_auroc": 0.9974, "heldout_auroc_ci95": [0.989, 1.0], "null_auroc_p2.5": 0.2028, "null_auroc_p97.5": 0.9329, "null_auroc_mean": 0.6658, "p_permutation": 0.002, "split_half_cosine": 0.8633, "within_pair_auroc": 0.954, "length_rho": -0.101, "layer_curve": [{"layer": 1, "cv_auroc": 0.9325}, {"layer": 2, "cv_auroc": 0.898}, {"layer": 3, "cv_auroc": 0.8947}, {"layer": 4, "cv_auroc": 0.9039}, {"layer": 5, "cv_auroc": 0.9186}, {"layer": 6, "cv_auroc": 0.9188}, {"layer": 7, "cv_auroc": 0.91}, {"layer": 8, "cv_auroc": 0.9322}, {"layer": 9, "cv_auroc": 0.9217}, {"layer": 10, "cv_auroc": 0.9001}, {"layer": 11, "cv_auroc": 0.9172}, {"layer": 12, "cv_auroc": 0.9177}, {"layer": 13, "cv_auroc": 0.9303}, {"layer": 14, "cv_auroc": 0.9179}, {"layer": 15, "cv_auroc": 0.9372}, {"layer": 16, "cv_auroc": 0.9728}, {"layer": 17, "cv_auroc": 0.9745}, {"layer": 18, "cv_auroc": 0.985}, {"layer": 19, "cv_auroc": 0.9935}, {"layer": 20, "cv_auroc": 0.9935}, {"layer": 21, "cv_auroc": 0.9935}, {"layer": 22, "cv_auroc": 0.9902}, {"layer": 23, "cv_auroc": 0.9837}, {"layer": 24, "cv_auroc": 0.9935}, {"layer": 25, "cv_auroc": 0.9902}, {"layer": 26, "cv_auroc": 0.9967}, {"layer": 27, "cv_auroc": 0.9967}, {"layer": 28, "cv_auroc": 0.9241}], "gap": 26.71, "gap_ci95": [21.55, 31.58], "cohens_d": 1.09, "mean_positive": 32.19, "mean_negative": 5.47, "gap_by_pair": {"0": 6.11, "1": 53.71, "2": 45.77, "3": -0.46, "4": 28.43}, "q_permutation": 0.00378, "significant": true}
+{"trait": "overconfidence", "block": "B", "placebo": false, "usable": true, "n_samples": 286, "layer": 21, "n_layers_swept": 28, "heldout_auroc": 0.8988, "heldout_auroc_ci95": [0.8621, 0.9413], "null_auroc_p2.5": 0.2441, "null_auroc_p97.5": 0.7511, "null_auroc_mean": 0.4915, "p_permutation": 0.002, "split_half_cosine": 0.8192, "within_pair_auroc": 0.9644, "length_rho": -0.341, "layer_curve": [{"layer": 1, "cv_auroc": 0.8409}, {"layer": 2, "cv_auroc": 0.8604}, {"layer": 3, "cv_auroc": 0.8476}, {"layer": 4, "cv_auroc": 0.8553}, {"layer": 5, "cv_auroc": 0.855}, {"layer": 6, "cv_auroc": 0.8452}, {"layer": 7, "cv_auroc": 0.8301}, {"layer": 8, "cv_auroc": 0.8318}, {"layer": 9, "cv_auroc": 0.8189}, {"layer": 10, "cv_auroc": 0.8103}, {"layer": 11, "cv_auroc": 0.7983}, {"layer": 12, "cv_auroc": 0.801}, {"layer": 13, "cv_auroc": 0.7921}, {"layer": 14, "cv_auroc": 0.7878}, {"layer": 15, "cv_auroc": 0.7954}, {"layer": 16, "cv_auroc": 0.8085}, {"layer": 17, "cv_auroc": 0.8141}, {"layer": 18, "cv_auroc": 0.8218}, {"layer": 19, "cv_auroc": 0.8351}, {"layer": 20, "cv_auroc": 0.889}, {"layer": 21, "cv_auroc": 0.8937}, {"layer": 22, "cv_auroc": 0.8733}, {"layer": 23, "cv_auroc": 0.8659}, {"layer": 24, "cv_auroc": 0.8834}, {"layer": 25, "cv_auroc": 0.875}, {"layer": 26, "cv_auroc": 0.8854}, {"layer": 27, "cv_auroc": 0.8927}, {"layer": 28, "cv_auroc": 0.868}], "gap": 29.54, "gap_ci95": [26.09, 33.05], "cohens_d": 1.731, "mean_positive": 65.82, "mean_negative": 36.27, "gap_by_pair": {"0": 29.89, "1": 35.97, "2": 18.34, "3": 39.29, "4": 24.23}, "q_permutation": 0.00378, "significant": true}
+{"trait": "evasiveness", "block": "B", "placebo": false, "usable": true, "n_samples": 261, "layer": 21, "n_layers_swept": 28, "heldout_auroc": 0.9993, "heldout_auroc_ci95": [0.9969, 1.0], "null_auroc_p2.5": 0.1207, "null_auroc_p97.5": 0.9642, "null_auroc_mean": 0.5891, "p_permutation": 0.002, "split_half_cosine": 0.8746, "within_pair_auroc": 1.0, "length_rho": -0.552, "layer_curve": [{"layer": 1, "cv_auroc": 0.9616}, {"layer": 2, "cv_auroc": 0.9587}, {"layer": 3, "cv_auroc": 0.9564}, {"layer": 4, "cv_auroc": 0.9544}, {"layer": 5, "cv_auroc": 0.9595}, {"layer": 6, "cv_auroc": 0.948}, {"layer": 7, "cv_auroc": 0.9455}, {"layer": 8, "cv_auroc": 0.9388}, {"layer": 9, "cv_auroc": 0.9256}, {"layer": 10, "cv_auroc": 0.9293}, {"layer": 11, "cv_auroc": 0.9225}, {"layer": 12, "cv_auroc": 0.9148}, {"layer": 13, "cv_auroc": 0.9106}, {"layer": 14, "cv_auroc": 0.9087}, {"layer": 15, "cv_auroc": 0.9289}, {"layer": 16, "cv_auroc": 0.9421}, {"layer": 17, "cv_auroc": 0.9422}, {"layer": 18, "cv_auroc": 0.9578}, {"layer": 19, "cv_auroc": 0.9785}, {"layer": 20, "cv_auroc": 0.9981}, {"layer": 21, "cv_auroc": 0.999}, {"layer": 22, "cv_auroc": 0.9985}, {"layer": 23, "cv_auroc": 0.9951}, {"layer": 24, "cv_auroc": 0.9985}, {"layer": 25, "cv_auroc": 0.9961}, {"layer": 26, "cv_auroc": 0.9971}, {"layer": 27, "cv_auroc": 0.9955}, {"layer": 28, "cv_auroc": 0.9929}], "gap": 30.15, "gap_ci95": [26.11, 34.01], "cohens_d": 1.588, "mean_positive": 47.6, "mean_negative": 17.45, "gap_by_pair": {"0": 17.03, "1": 33.4, "2": 31.69, "3": 30.8, "4": 37.83}, "q_permutation": 0.00378, "significant": true}
+{"trait": "incoherence", "block": "B", "placebo": false, "usable": true, "n_samples": 252, "layer": 20, "n_layers_swept": 28, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.1337, "null_auroc_p97.5": 0.9893, "null_auroc_mean": 0.6651, "p_permutation": 0.00399, "split_half_cosine": 0.8438, "within_pair_auroc": 1.0, "length_rho": -0.509, "layer_curve": [{"layer": 1, "cv_auroc": 0.9581}, {"layer": 2, "cv_auroc": 0.9514}, {"layer": 3, "cv_auroc": 0.9582}, {"layer": 4, "cv_auroc": 0.9569}, {"layer": 5, "cv_auroc": 0.9571}, {"layer": 6, "cv_auroc": 0.9466}, {"layer": 7, "cv_auroc": 0.955}, {"layer": 8, "cv_auroc": 0.9492}, {"layer": 9, "cv_auroc": 0.9331}, {"layer": 10, "cv_auroc": 0.9211}, {"layer": 11, "cv_auroc": 0.9317}, {"layer": 12, "cv_auroc": 0.9367}, {"layer": 13, "cv_auroc": 0.9474}, {"layer": 14, "cv_auroc": 0.942}, {"layer": 15, "cv_auroc": 0.9565}, {"layer": 16, "cv_auroc": 0.9791}, {"layer": 17, "cv_auroc": 0.976}, {"layer": 18, "cv_auroc": 0.987}, {"layer": 19, "cv_auroc": 0.9955}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 0.9971}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 0.9966}, {"layer": 26, "cv_auroc": 0.9971}, {"layer": 27, "cv_auroc": 0.9971}, {"layer": 28, "cv_auroc": 0.9643}], "gap": 37.38, "gap_ci95": [33.21, 42.07], "cohens_d": 1.776, "mean_positive": 44.79, "mean_negative": 7.41, "gap_by_pair": {"0": 43.06, "1": 17.91, "2": 42.11, "3": 49.06, "4": 34.77}, "q_permutation": 0.00678, "significant": true}
+{"trait": "secrecy", "block": "B", "placebo": false, "usable": true, "n_samples": 179, "layer": 28, "n_layers_swept": 28, "heldout_auroc": 0.6, "heldout_auroc_ci95": [0.4857, 0.7143], "null_auroc_p2.5": 0.0429, "null_auroc_p97.5": 0.9286, "null_auroc_mean": 0.5031, "p_permutation": 0.40519, "split_half_cosine": 0.2717, "within_pair_auroc": NaN, "length_rho": -0.307, "layer_curve": [{"layer": 1, "cv_auroc": 0.884}, {"layer": 2, "cv_auroc": 0.8938}, {"layer": 3, "cv_auroc": 0.8889}, {"layer": 4, "cv_auroc": 0.8791}, {"layer": 5, "cv_auroc": 0.8791}, {"layer": 6, "cv_auroc": 0.8791}, {"layer": 7, "cv_auroc": 0.8693}, {"layer": 8, "cv_auroc": 0.8693}, {"layer": 9, "cv_auroc": 0.8595}, {"layer": 10, "cv_auroc": 0.884}, {"layer": 11, "cv_auroc": 0.8934}, {"layer": 12, "cv_auroc": 0.8934}, {"layer": 13, "cv_auroc": 0.8934}, {"layer": 14, "cv_auroc": 0.8885}, {"layer": 15, "cv_auroc": 0.8885}, {"layer": 16, "cv_auroc": 0.9126}, {"layer": 17, "cv_auroc": 0.9273}, {"layer": 18, "cv_auroc": 0.9224}, {"layer": 19, "cv_auroc": 0.9224}, {"layer": 20, "cv_auroc": 0.9175}, {"layer": 21, "cv_auroc": 0.9224}, {"layer": 22, "cv_auroc": 0.9224}, {"layer": 23, "cv_auroc": 0.9081}, {"layer": 24, "cv_auroc": 0.9081}, {"layer": 25, "cv_auroc": 0.9081}, {"layer": 26, "cv_auroc": 0.9032}, {"layer": 27, "cv_auroc": 0.9228}, {"layer": 28, "cv_auroc": 0.9424}], "gap": 5.3, "gap_ci95": [3.07, 7.69], "cohens_d": 0.487, "mean_positive": 14.21, "mean_negative": 8.91, "gap_by_pair": {"0": 1.43, "1": 13.14, "2": 3.14, "3": 3.43, "4": 5.34}, "q_permutation": 0.43051, "significant": false}
+{"trait": "stubbornness", "block": "B", "placebo": false, "usable": false}
+{"trait": "sincerity", "block": "C", "placebo": false, "usable": true, "n_samples": 295, "layer": 24, "n_layers_swept": 28, "heldout_auroc": 0.9975, "heldout_auroc_ci95": [0.9908, 1.0], "null_auroc_p2.5": 0.327, "null_auroc_p97.5": 0.9341, "null_auroc_mean": 0.6839, "p_permutation": 0.002, "split_half_cosine": 0.9533, "within_pair_auroc": 0.9667, "length_rho": 0.304, "layer_curve": [{"layer": 1, "cv_auroc": 0.8394}, {"layer": 2, "cv_auroc": 0.8315}, {"layer": 3, "cv_auroc": 0.8426}, {"layer": 4, "cv_auroc": 0.8538}, {"layer": 5, "cv_auroc": 0.8741}, {"layer": 6, "cv_auroc": 0.8668}, {"layer": 7, "cv_auroc": 0.8619}, {"layer": 8, "cv_auroc": 0.8764}, {"layer": 9, "cv_auroc": 0.8826}, {"layer": 10, "cv_auroc": 0.8934}, {"layer": 11, "cv_auroc": 0.8923}, {"layer": 12, "cv_auroc": 0.8855}, {"layer": 13, "cv_auroc": 0.8784}, {"layer": 14, "cv_auroc": 0.884}, {"layer": 15, "cv_auroc": 0.9139}, {"layer": 16, "cv_auroc": 0.9202}, {"layer": 17, "cv_auroc": 0.9186}, {"layer": 18, "cv_auroc": 0.9228}, {"layer": 19, "cv_auroc": 0.9295}, {"layer": 20, "cv_auroc": 0.9596}, {"layer": 21, "cv_auroc": 0.9691}, {"layer": 22, "cv_auroc": 0.9771}, {"layer": 23, "cv_auroc": 0.9763}, {"layer": 24, "cv_auroc": 0.9784}, {"layer": 25, "cv_auroc": 0.9722}, {"layer": 26, "cv_auroc": 0.9773}, {"layer": 27, "cv_auroc": 0.9758}, {"layer": 28, "cv_auroc": 0.9247}], "gap": 30.05, "gap_ci95": [26.67, 33.05], "cohens_d": 1.979, "mean_positive": 65.97, "mean_negative": 35.91, "gap_by_pair": {"0": 34.77, "1": 41.23, "2": 10.69, "3": 37.74, "4": 25.83}, "q_permutation": 0.00378, "significant": true}
+{"trait": "fairness", "block": "C", "placebo": false, "usable": true, "n_samples": 170, "layer": 27, "n_layers_swept": 28, "heldout_auroc": 0.8203, "heldout_auroc_ci95": [0.6627, 0.9385], "null_auroc_p2.5": 0.1738, "null_auroc_p97.5": 0.8, "null_auroc_mean": 0.4807, "p_permutation": 0.01198, "split_half_cosine": 0.4318, "within_pair_auroc": 0.8, "length_rho": -0.097, "layer_curve": [{"layer": 1, "cv_auroc": 0.4764}, {"layer": 2, "cv_auroc": 0.485}, {"layer": 3, "cv_auroc": 0.4344}, {"layer": 4, "cv_auroc": 0.4246}, {"layer": 5, "cv_auroc": 0.4655}, {"layer": 6, "cv_auroc": 0.4499}, {"layer": 7, "cv_auroc": 0.4494}, {"layer": 8, "cv_auroc": 0.4661}, {"layer": 9, "cv_auroc": 0.4569}, {"layer": 10, "cv_auroc": 0.4747}, {"layer": 11, "cv_auroc": 0.4828}, {"layer": 12, "cv_auroc": 0.4908}, {"layer": 13, "cv_auroc": 0.4908}, {"layer": 14, "cv_auroc": 0.4908}, {"layer": 15, "cv_auroc": 0.4828}, {"layer": 16, "cv_auroc": 0.4908}, {"layer": 17, "cv_auroc": 0.4828}, {"layer": 18, "cv_auroc": 0.5075}, {"layer": 19, "cv_auroc": 0.5156}, {"layer": 20, "cv_auroc": 0.5501}, {"layer": 21, "cv_auroc": 0.5501}, {"layer": 22, "cv_auroc": 0.5581}, {"layer": 23, "cv_auroc": 0.5748}, {"layer": 24, "cv_auroc": 0.5834}, {"layer": 25, "cv_auroc": 0.592}, {"layer": 26, "cv_auroc": 0.6007}, {"layer": 27, "cv_auroc": 0.6605}, {"layer": 28, "cv_auroc": 0.5667}], "gap": 3.07, "gap_ci95": [0.33, 6.21], "cohens_d": 0.219, "mean_positive": 80.52, "mean_negative": 77.45, "gap_by_pair": {"0": 1.23, "1": 2.63, "2": -1.6, "3": 12.91, "4": 0.2}, "q_permutation": 0.01509, "significant": true}
+{"trait": "greed-avoidance", "block": "C", "placebo": false, "usable": true, "n_samples": 238, "layer": 21, "n_layers_swept": 28, "heldout_auroc": 0.9992, "heldout_auroc_ci95": [0.9959, 1.0], "null_auroc_p2.5": 0.1685, "null_auroc_p97.5": 0.8495, "null_auroc_mean": 0.5189, "p_permutation": 0.002, "split_half_cosine": 0.8194, "within_pair_auroc": 0.9821, "length_rho": -0.472, "layer_curve": [{"layer": 1, "cv_auroc": 0.7683}, {"layer": 2, "cv_auroc": 0.7646}, {"layer": 3, "cv_auroc": 0.7707}, {"layer": 4, "cv_auroc": 0.7811}, {"layer": 5, "cv_auroc": 0.7863}, {"layer": 6, "cv_auroc": 0.7849}, {"layer": 7, "cv_auroc": 0.7794}, {"layer": 8, "cv_auroc": 0.8}, {"layer": 9, "cv_auroc": 0.7979}, {"layer": 10, "cv_auroc": 0.782}, {"layer": 11, "cv_auroc": 0.7835}, {"layer": 12, "cv_auroc": 0.7791}, {"layer": 13, "cv_auroc": 0.7879}, {"layer": 14, "cv_auroc": 0.789}, {"layer": 15, "cv_auroc": 0.827}, {"layer": 16, "cv_auroc": 0.8552}, {"layer": 17, "cv_auroc": 0.8444}, {"layer": 18, "cv_auroc": 0.8645}, {"layer": 19, "cv_auroc": 0.9191}, {"layer": 20, "cv_auroc": 0.979}, {"layer": 21, "cv_auroc": 0.987}, {"layer": 22, "cv_auroc": 0.9721}, {"layer": 23, "cv_auroc": 0.9769}, {"layer": 24, "cv_auroc": 0.9695}, {"layer": 25, "cv_auroc": 0.9629}, {"layer": 26, "cv_auroc": 0.9639}, {"layer": 27, "cv_auroc": 0.9652}, {"layer": 28, "cv_auroc": 0.9164}], "gap": 28.89, "gap_ci95": [25.66, 31.99], "cohens_d": 1.919, "mean_positive": 62.43, "mean_negative": 33.55, "gap_by_pair": {"0": 25.91, "1": 26.46, "2": 27.71, "3": 32.0, "4": 32.34}, "q_permutation": 0.00378, "significant": true}
+{"trait": "modesty", "block": "C", "placebo": false, "usable": true, "n_samples": 203, "layer": 25, "n_layers_swept": 28, "heldout_auroc": 0.9565, "heldout_auroc_ci95": [0.9116, 0.9909], "null_auroc_p2.5": 0.2556, "null_auroc_p97.5": 0.7777, "null_auroc_mean": 0.5217, "p_permutation": 0.002, "split_half_cosine": 0.5906, "within_pair_auroc": 0.9136, "length_rho": -0.531, "layer_curve": [{"layer": 1, "cv_auroc": 0.8845}, {"layer": 2, "cv_auroc": 0.8819}, {"layer": 3, "cv_auroc": 0.871}, {"layer": 4, "cv_auroc": 0.8775}, {"layer": 5, "cv_auroc": 0.8737}, {"layer": 6, "cv_auroc": 0.8715}, {"layer": 7, "cv_auroc": 0.8646}, {"layer": 8, "cv_auroc": 0.8639}, {"layer": 9, "cv_auroc": 0.8597}, {"layer": 10, "cv_auroc": 0.8518}, {"layer": 11, "cv_auroc": 0.8538}, {"layer": 12, "cv_auroc": 0.8426}, {"layer": 13, "cv_auroc": 0.8497}, {"layer": 14, "cv_auroc": 0.8469}, {"layer": 15, "cv_auroc": 0.8638}, {"layer": 16, "cv_auroc": 0.8728}, {"layer": 17, "cv_auroc": 0.8699}, {"layer": 18, "cv_auroc": 0.8762}, {"layer": 19, "cv_auroc": 0.8993}, {"layer": 20, "cv_auroc": 0.9474}, {"layer": 21, "cv_auroc": 0.9527}, {"layer": 22, "cv_auroc": 0.947}, {"layer": 23, "cv_auroc": 0.945}, {"layer": 24, "cv_auroc": 0.9519}, {"layer": 25, "cv_auroc": 0.9531}, {"layer": 26, "cv_auroc": 0.9526}, {"layer": 27, "cv_auroc": 0.941}, {"layer": 28, "cv_auroc": 0.8095}], "gap": 23.59, "gap_ci95": [20.2, 26.66], "cohens_d": 1.472, "mean_positive": 57.47, "mean_negative": 33.88, "gap_by_pair": {"0": 8.11, "1": 41.43, "2": 21.06, "3": 21.17, "4": 26.2}, "q_permutation": 0.00378, "significant": true}
+{"trait": "straightforwardness", "block": "C", "placebo": false, "usable": true, "n_samples": 272, "layer": 27, "n_layers_swept": 28, "heldout_auroc": 0.9635, "heldout_auroc_ci95": [0.9386, 0.9861], "null_auroc_p2.5": 0.1311, "null_auroc_p97.5": 0.8824, "null_auroc_mean": 0.5094, "p_permutation": 0.002, "split_half_cosine": 0.8244, "within_pair_auroc": 0.9921, "length_rho": -0.411, "layer_curve": [{"layer": 1, "cv_auroc": 0.9125}, {"layer": 2, "cv_auroc": 0.9231}, {"layer": 3, "cv_auroc": 0.9201}, {"layer": 4, "cv_auroc": 0.9126}, {"layer": 5, "cv_auroc": 0.9151}, {"layer": 6, "cv_auroc": 0.9151}, {"layer": 7, "cv_auroc": 0.9041}, {"layer": 8, "cv_auroc": 0.8949}, {"layer": 9, "cv_auroc": 0.8837}, {"layer": 10, "cv_auroc": 0.894}, {"layer": 11, "cv_auroc": 0.8682}, {"layer": 12, "cv_auroc": 0.857}, {"layer": 13, "cv_auroc": 0.848}, {"layer": 14, "cv_auroc": 0.8418}, {"layer": 15, "cv_auroc": 0.8533}, {"layer": 16, "cv_auroc": 0.8681}, {"layer": 17, "cv_auroc": 0.8643}, {"layer": 18, "cv_auroc": 0.8753}, {"layer": 19, "cv_auroc": 0.8861}, {"layer": 20, "cv_auroc": 0.9695}, {"layer": 21, "cv_auroc": 0.974}, {"layer": 22, "cv_auroc": 0.9714}, {"layer": 23, "cv_auroc": 0.9614}, {"layer": 24, "cv_auroc": 0.9774}, {"layer": 25, "cv_auroc": 0.9764}, {"layer": 26, "cv_auroc": 0.9804}, {"layer": 27, "cv_auroc": 0.9838}, {"layer": 28, "cv_auroc": 0.9762}], "gap": 28.33, "gap_ci95": [24.24, 32.35], "cohens_d": 1.57, "mean_positive": 63.96, "mean_negative": 35.63, "gap_by_pair": {"0": 20.17, "1": 20.6, "2": 34.51, "3": 23.8, "4": 42.54}, "q_permutation": 0.00378, "significant": true}
+{"trait": "machiavellianism", "block": "D", "placebo": false, "usable": true, "n_samples": 225, "layer": 19, "n_layers_swept": 28, "heldout_auroc": 0.9179, "heldout_auroc_ci95": [0.8519, 0.9693], "null_auroc_p2.5": 0.5334, "null_auroc_p97.5": 0.8503, "null_auroc_mean": 0.6917, "p_permutation": 0.00599, "split_half_cosine": 0.4236, "within_pair_auroc": 0.7804, "length_rho": -0.14, "layer_curve": [{"layer": 1, "cv_auroc": 0.7571}, {"layer": 2, "cv_auroc": 0.7524}, {"layer": 3, "cv_auroc": 0.7748}, {"layer": 4, "cv_auroc": 0.7422}, {"layer": 5, "cv_auroc": 0.7408}, {"layer": 6, "cv_auroc": 0.7476}, {"layer": 7, "cv_auroc": 0.7578}, {"layer": 8, "cv_auroc": 0.7537}, {"layer": 9, "cv_auroc": 0.732}, {"layer": 10, "cv_auroc": 0.6816}, {"layer": 11, "cv_auroc": 0.7122}, {"layer": 12, "cv_auroc": 0.7299}, {"layer": 13, "cv_auroc": 0.7313}, {"layer": 14, "cv_auroc": 0.7102}, {"layer": 15, "cv_auroc": 0.7265}, {"layer": 16, "cv_auroc": 0.768}, {"layer": 17, "cv_auroc": 0.7633}, {"layer": 18, "cv_auroc": 0.7878}, {"layer": 19, "cv_auroc": 0.8211}, {"layer": 20, "cv_auroc": 0.8095}, {"layer": 21, "cv_auroc": 0.8048}, {"layer": 22, "cv_auroc": 0.7891}, {"layer": 23, "cv_auroc": 0.7959}, {"layer": 24, "cv_auroc": 0.7884}, {"layer": 25, "cv_auroc": 0.7823}, {"layer": 26, "cv_auroc": 0.7823}, {"layer": 27, "cv_auroc": 0.7946}, {"layer": 28, "cv_auroc": 0.7415}], "gap": 23.85, "gap_ci95": [20.31, 27.41], "cohens_d": 1.297, "mean_positive": 30.74, "mean_negative": 6.89, "gap_by_pair": {"0": 8.29, "1": 35.89, "2": 7.46, "3": 23.94, "4": 43.69}, "q_permutation": 0.00926, "significant": true}
+{"trait": "narcissism", "block": "D", "placebo": false, "usable": true, "n_samples": 256, "layer": 27, "n_layers_swept": 28, "heldout_auroc": 0.9964, "heldout_auroc_ci95": [0.9867, 1.0], "null_auroc_p2.5": 0.1625, "null_auroc_p97.5": 0.8921, "null_auroc_mean": 0.5257, "p_permutation": 0.002, "split_half_cosine": 0.8314, "within_pair_auroc": 0.9971, "length_rho": 0.09, "layer_curve": [{"layer": 1, "cv_auroc": 0.9}, {"layer": 2, "cv_auroc": 0.8815}, {"layer": 3, "cv_auroc": 0.904}, {"layer": 4, "cv_auroc": 0.9173}, {"layer": 5, "cv_auroc": 0.9199}, {"layer": 6, "cv_auroc": 0.9231}, {"layer": 7, "cv_auroc": 0.9188}, {"layer": 8, "cv_auroc": 0.9307}, {"layer": 9, "cv_auroc": 0.9345}, {"layer": 10, "cv_auroc": 0.9369}, {"layer": 11, "cv_auroc": 0.9377}, {"layer": 12, "cv_auroc": 0.9429}, {"layer": 13, "cv_auroc": 0.9444}, {"layer": 14, "cv_auroc": 0.9474}, {"layer": 15, "cv_auroc": 0.9528}, {"layer": 16, "cv_auroc": 0.96}, {"layer": 17, "cv_auroc": 0.9525}, {"layer": 18, "cv_auroc": 0.9584}, {"layer": 19, "cv_auroc": 0.9681}, {"layer": 20, "cv_auroc": 0.9909}, {"layer": 21, "cv_auroc": 0.9928}, {"layer": 22, "cv_auroc": 0.9921}, {"layer": 23, "cv_auroc": 0.9902}, {"layer": 24, "cv_auroc": 0.9919}, {"layer": 25, "cv_auroc": 0.9909}, {"layer": 26, "cv_auroc": 0.9933}, {"layer": 27, "cv_auroc": 0.9943}, {"layer": 28, "cv_auroc": 0.9825}], "gap": 36.61, "gap_ci95": [32.62, 40.55], "cohens_d": 1.933, "mean_positive": 41.32, "mean_negative": 4.71, "gap_by_pair": {"0": 35.23, "1": 36.89, "2": 48.49, "3": 10.2, "4": 52.26}, "q_permutation": 0.00378, "significant": true}
+{"trait": "callousness", "block": "D", "placebo": false, "usable": true, "n_samples": 207, "layer": 20, "n_layers_swept": 28, "heldout_auroc": 0.975, "heldout_auroc_ci95": [0.9259, 1.0], "null_auroc_p2.5": 0.0907, "null_auroc_p97.5": 0.935, "null_auroc_mean": 0.583, "p_permutation": 0.00998, "split_half_cosine": 0.6388, "within_pair_auroc": 1.0, "length_rho": -0.549, "layer_curve": [{"layer": 1, "cv_auroc": 0.5798}, {"layer": 2, "cv_auroc": 0.6641}, {"layer": 3, "cv_auroc": 0.6666}, {"layer": 4, "cv_auroc": 0.6381}, {"layer": 5, "cv_auroc": 0.6618}, {"layer": 6, "cv_auroc": 0.7076}, {"layer": 7, "cv_auroc": 0.6983}, {"layer": 8, "cv_auroc": 0.6895}, {"layer": 9, "cv_auroc": 0.7116}, {"layer": 10, "cv_auroc": 0.7402}, {"layer": 11, "cv_auroc": 0.7273}, {"layer": 12, "cv_auroc": 0.6991}, {"layer": 13, "cv_auroc": 0.7498}, {"layer": 14, "cv_auroc": 0.7387}, {"layer": 15, "cv_auroc": 0.7658}, {"layer": 16, "cv_auroc": 0.794}, {"layer": 17, "cv_auroc": 0.8144}, {"layer": 18, "cv_auroc": 0.8206}, {"layer": 19, "cv_auroc": 0.8798}, {"layer": 20, "cv_auroc": 0.9898}, {"layer": 21, "cv_auroc": 0.9898}, {"layer": 22, "cv_auroc": 0.9891}, {"layer": 23, "cv_auroc": 0.9884}, {"layer": 24, "cv_auroc": 0.9789}, {"layer": 25, "cv_auroc": 0.9769}, {"layer": 26, "cv_auroc": 0.9762}, {"layer": 27, "cv_auroc": 0.9884}, {"layer": 28, "cv_auroc": 0.9563}], "gap": 16.14, "gap_ci95": [12.73, 19.42], "cohens_d": 0.967, "mean_positive": 24.84, "mean_negative": 8.7, "gap_by_pair": {"0": 31.89, "1": 21.89, "2": 4.43, "3": 19.43, "4": 3.09}, "q_permutation": 0.01305, "significant": true}
+{"trait": "sadism", "block": "D", "placebo": false, "usable": true, "n_samples": 188, "layer": 27, "n_layers_swept": 28, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.1782, "null_auroc_p97.5": 1.0, "null_auroc_mean": 0.7873, "p_permutation": 0.15369, "split_half_cosine": 0.4042, "within_pair_auroc": 1.0, "length_rho": 0.087, "layer_curve": [{"layer": 1, "cv_auroc": 0.9026}, {"layer": 2, "cv_auroc": 0.9065}, {"layer": 3, "cv_auroc": 0.9273}, {"layer": 4, "cv_auroc": 0.9364}, {"layer": 5, "cv_auroc": 0.9364}, {"layer": 6, "cv_auroc": 0.939}, {"layer": 7, "cv_auroc": 0.9182}, {"layer": 8, "cv_auroc": 0.9234}, {"layer": 9, "cv_auroc": 0.9}, {"layer": 10, "cv_auroc": 0.9234}, {"layer": 11, "cv_auroc": 0.9039}, {"layer": 12, "cv_auroc": 0.8987}, {"layer": 13, "cv_auroc": 0.9}, {"layer": 14, "cv_auroc": 0.9013}, {"layer": 15, "cv_auroc": 0.9091}, {"layer": 16, "cv_auroc": 0.9234}, {"layer": 17, "cv_auroc": 0.9247}, {"layer": 18, "cv_auroc": 0.9299}, {"layer": 19, "cv_auroc": 0.9377}, {"layer": 20, "cv_auroc": 0.9636}, {"layer": 21, "cv_auroc": 0.9636}, {"layer": 22, "cv_auroc": 0.9636}, {"layer": 23, "cv_auroc": 0.9649}, {"layer": 24, "cv_auroc": 0.974}, {"layer": 25, "cv_auroc": 0.974}, {"layer": 26, "cv_auroc": 0.974}, {"layer": 27, "cv_auroc": 0.9753}, {"layer": 28, "cv_auroc": 0.9675}], "gap": 8.27, "gap_ci95": [5.5, 11.36], "cohens_d": 0.609, "mean_positive": 9.97, "mean_negative": 1.69, "gap_by_pair": {"0": 1.17, "1": 3.71, "2": 31.11, "3": 5.69, "4": -0.31}, "q_permutation": 0.17418, "significant": false}
+{"trait": "manipulativeness", "block": "D", "placebo": false, "usable": true, "n_samples": 226, "layer": 21, "n_layers_swept": 28, "heldout_auroc": 0.991, "heldout_auroc_ci95": [0.9773, 1.0], "null_auroc_p2.5": 0.2119, "null_auroc_p97.5": 0.9488, "null_auroc_mean": 0.6434, "p_permutation": 0.002, "split_half_cosine": 0.8186, "within_pair_auroc": 1.0, "length_rho": -0.593, "layer_curve": [{"layer": 1, "cv_auroc": 0.9283}, {"layer": 2, "cv_auroc": 0.9351}, {"layer": 3, "cv_auroc": 0.9318}, {"layer": 4, "cv_auroc": 0.9318}, {"layer": 5, "cv_auroc": 0.9405}, {"layer": 6, "cv_auroc": 0.9342}, {"layer": 7, "cv_auroc": 0.939}, {"layer": 8, "cv_auroc": 0.9452}, {"layer": 9, "cv_auroc": 0.9307}, {"layer": 10, "cv_auroc": 0.9176}, {"layer": 11, "cv_auroc": 0.9131}, {"layer": 12, "cv_auroc": 0.9146}, {"layer": 13, "cv_auroc": 0.9232}, {"layer": 14, "cv_auroc": 0.9179}, {"layer": 15, "cv_auroc": 0.9247}, {"layer": 16, "cv_auroc": 0.9455}, {"layer": 17, "cv_auroc": 0.9369}, {"layer": 18, "cv_auroc": 0.9494}, {"layer": 19, "cv_auroc": 0.9652}, {"layer": 20, "cv_auroc": 0.9732}, {"layer": 21, "cv_auroc": 0.9741}, {"layer": 22, "cv_auroc": 0.9601}, {"layer": 23, "cv_auroc": 0.95}, {"layer": 24, "cv_auroc": 0.9583}, {"layer": 25, "cv_auroc": 0.9536}, {"layer": 26, "cv_auroc": 0.9607}, {"layer": 27, "cv_auroc": 0.9589}, {"layer": 28, "cv_auroc": 0.944}], "gap": 21.39, "gap_ci95": [17.8, 24.53], "cohens_d": 1.33, "mean_positive": 34.38, "mean_negative": 12.99, "gap_by_pair": {"0": 14.57, "1": 2.71, "2": 31.2, "3": 38.54, "4": 19.91}, "q_permutation": 0.00378, "significant": true}
+{"trait": "roleplay", "block": "E", "placebo": false, "usable": true, "n_samples": 270, "layer": 17, "n_layers_swept": 28, "heldout_auroc": 0.9995, "heldout_auroc_ci95": [0.9973, 1.0], "null_auroc_p2.5": 0.181, "null_auroc_p97.5": 0.8488, "null_auroc_mean": 0.5204, "p_permutation": 0.002, "split_half_cosine": 0.9565, "within_pair_auroc": 0.9722, "length_rho": -0.126, "layer_curve": [{"layer": 1, "cv_auroc": 0.9968}, {"layer": 2, "cv_auroc": 0.9989}, {"layer": 3, "cv_auroc": 0.9976}, {"layer": 4, "cv_auroc": 0.996}, {"layer": 5, "cv_auroc": 0.9968}, {"layer": 6, "cv_auroc": 0.9949}, {"layer": 7, "cv_auroc": 0.9971}, {"layer": 8, "cv_auroc": 0.9981}, {"layer": 9, "cv_auroc": 0.9992}, {"layer": 10, "cv_auroc": 0.9963}, {"layer": 11, "cv_auroc": 0.9976}, {"layer": 12, "cv_auroc": 0.9972}, {"layer": 13, "cv_auroc": 0.9976}, {"layer": 14, "cv_auroc": 0.9956}, {"layer": 15, "cv_auroc": 0.9979}, {"layer": 16, "cv_auroc": 0.9996}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 45.13, "gap_ci95": [41.07, 49.38], "cohens_d": 2.189, "mean_positive": 48.27, "mean_negative": 3.14, "gap_by_pair": {"0": 42.4, "1": 59.57, "2": 51.49, "3": 58.46, "4": 13.71}, "q_permutation": 0.00378, "significant": true}
+{"trait": "deference", "block": "E", "placebo": false, "usable": true, "n_samples": 278, "layer": 21, "n_layers_swept": 28, "heldout_auroc": 0.9538, "heldout_auroc_ci95": [0.9104, 0.9929], "null_auroc_p2.5": 0.3073, "null_auroc_p97.5": 0.7461, "null_auroc_mean": 0.5362, "p_permutation": 0.002, "split_half_cosine": 0.8461, "within_pair_auroc": 0.9597, "length_rho": -0.46, "layer_curve": [{"layer": 1, "cv_auroc": 0.8746}, {"layer": 2, "cv_auroc": 0.8778}, {"layer": 3, "cv_auroc": 0.8836}, {"layer": 4, "cv_auroc": 0.898}, {"layer": 5, "cv_auroc": 0.9036}, {"layer": 6, "cv_auroc": 0.8923}, {"layer": 7, "cv_auroc": 0.8949}, {"layer": 8, "cv_auroc": 0.897}, {"layer": 9, "cv_auroc": 0.8921}, {"layer": 10, "cv_auroc": 0.8587}, {"layer": 11, "cv_auroc": 0.8634}, {"layer": 12, "cv_auroc": 0.864}, {"layer": 13, "cv_auroc": 0.8676}, {"layer": 14, "cv_auroc": 0.8601}, {"layer": 15, "cv_auroc": 0.8625}, {"layer": 16, "cv_auroc": 0.8804}, {"layer": 17, "cv_auroc": 0.8818}, {"layer": 18, "cv_auroc": 0.8847}, {"layer": 19, "cv_auroc": 0.9137}, {"layer": 20, "cv_auroc": 0.9501}, {"layer": 21, "cv_auroc": 0.9571}, {"layer": 22, "cv_auroc": 0.9426}, {"layer": 23, "cv_auroc": 0.9439}, {"layer": 24, "cv_auroc": 0.9446}, {"layer": 25, "cv_auroc": 0.9381}, {"layer": 26, "cv_auroc": 0.9423}, {"layer": 27, "cv_auroc": 0.9527}, {"layer": 28, "cv_auroc": 0.9133}], "gap": 28.95, "gap_ci95": [25.41, 32.16], "cohens_d": 1.792, "mean_positive": 49.95, "mean_negative": 21.0, "gap_by_pair": {"0": 22.51, "1": 40.71, "2": 23.69, "3": 19.14, "4": 38.69}, "q_permutation": 0.00378, "significant": true}
+{"trait": "verbosity", "block": "F", "placebo": false, "usable": true, "n_samples": 348, "layer": 5, "n_layers_swept": 28, "heldout_auroc": 0.9988, "heldout_auroc_ci95": [0.9959, 1.0], "null_auroc_p2.5": 0.0734, "null_auroc_p97.5": 0.9389, "null_auroc_mean": 0.506, "p_permutation": 0.002, "split_half_cosine": 0.9206, "within_pair_auroc": 1.0, "length_rho": 0.937, "layer_curve": [{"layer": 1, "cv_auroc": 0.9924}, {"layer": 2, "cv_auroc": 0.9961}, {"layer": 3, "cv_auroc": 0.998}, {"layer": 4, "cv_auroc": 0.9992}, {"layer": 5, "cv_auroc": 1.0}, {"layer": 6, "cv_auroc": 0.9986}, {"layer": 7, "cv_auroc": 0.9997}, {"layer": 8, "cv_auroc": 1.0}, {"layer": 9, "cv_auroc": 0.9997}, {"layer": 10, "cv_auroc": 0.9992}, {"layer": 11, "cv_auroc": 0.9992}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 0.9997}, {"layer": 14, "cv_auroc": 0.9997}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 63.06, "gap_ci95": [61.17, 64.95], "cohens_d": 6.896, "mean_positive": 80.43, "mean_negative": 17.37, "gap_by_pair": {"0": 70.77, "1": 54.11, "2": 72.2, "3": 68.94, "4": 49.29}, "q_permutation": 0.00378, "significant": true}
+{"trait": "formality", "block": "F", "placebo": false, "usable": true, "n_samples": 343, "layer": 1, "n_layers_swept": 28, "heldout_auroc": 0.9996, "heldout_auroc_ci95": [0.998, 1.0], "null_auroc_p2.5": 0.0035, "null_auroc_p97.5": 0.998, "null_auroc_mean": 0.4871, "p_permutation": 0.01397, "split_half_cosine": 0.9736, "within_pair_auroc": 0.9993, "length_rho": 0.59, "layer_curve": [{"layer": 1, "cv_auroc": 1.0}, {"layer": 2, "cv_auroc": 1.0}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 1.0}, {"layer": 5, "cv_auroc": 1.0}, {"layer": 6, "cv_auroc": 1.0}, {"layer": 7, "cv_auroc": 1.0}, {"layer": 8, "cv_auroc": 1.0}, {"layer": 9, "cv_auroc": 1.0}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 58.49, "gap_ci95": [56.74, 60.25], "cohens_d": 6.715, "mean_positive": 84.49, "mean_negative": 26.0, "gap_by_pair": {"0": 59.29, "1": 55.43, "2": 61.51, "3": 64.06, "4": 52.14}, "q_permutation": 0.01638, "significant": true}
+{"trait": "technicality", "block": "F", "placebo": false, "usable": true, "n_samples": 317, "layer": 20, "n_layers_swept": 28, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0303, "null_auroc_p97.5": 0.9963, "null_auroc_mean": 0.6025, "p_permutation": 0.00599, "split_half_cosine": 0.8993, "within_pair_auroc": 1.0, "length_rho": 0.507, "layer_curve": [{"layer": 1, "cv_auroc": 0.9957}, {"layer": 2, "cv_auroc": 0.9964}, {"layer": 3, "cv_auroc": 0.9958}, {"layer": 4, "cv_auroc": 0.9968}, {"layer": 5, "cv_auroc": 0.9972}, {"layer": 6, "cv_auroc": 0.9974}, {"layer": 7, "cv_auroc": 0.9948}, {"layer": 8, "cv_auroc": 0.9958}, {"layer": 9, "cv_auroc": 0.9905}, {"layer": 10, "cv_auroc": 0.9944}, {"layer": 11, "cv_auroc": 0.9851}, {"layer": 12, "cv_auroc": 0.9851}, {"layer": 13, "cv_auroc": 0.9814}, {"layer": 14, "cv_auroc": 0.9775}, {"layer": 15, "cv_auroc": 0.9827}, {"layer": 16, "cv_auroc": 0.9883}, {"layer": 17, "cv_auroc": 0.9863}, {"layer": 18, "cv_auroc": 0.9911}, {"layer": 19, "cv_auroc": 0.9956}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 45.96, "gap_ci95": [43.16, 48.55], "cohens_d": 3.601, "mean_positive": 63.28, "mean_negative": 17.32, "gap_by_pair": {"0": 31.51, "1": 54.89, "2": 48.17, "3": 44.14, "4": 51.09}, "q_permutation": 0.00926, "significant": true}
+{"trait": "humor", "block": "F", "placebo": false, "usable": true, "n_samples": 349, "layer": 3, "n_layers_swept": 28, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0026, "null_auroc_p97.5": 0.998, "null_auroc_mean": 0.4838, "p_permutation": 0.00798, "split_half_cosine": 0.9749, "within_pair_auroc": 0.9985, "length_rho": -0.4, "layer_curve": [{"layer": 1, "cv_auroc": 0.9997}, {"layer": 2, "cv_auroc": 0.9997}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 1.0}, {"layer": 5, "cv_auroc": 1.0}, {"layer": 6, "cv_auroc": 1.0}, {"layer": 7, "cv_auroc": 1.0}, {"layer": 8, "cv_auroc": 1.0}, {"layer": 9, "cv_auroc": 1.0}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 63.86, "gap_ci95": [62.93, 64.68], "cohens_d": 15.517, "mean_positive": 70.15, "mean_negative": 6.3, "gap_by_pair": {"0": 62.2, "1": 63.86, "2": 63.94, "3": 63.74, "4": 65.54}, "q_permutation": 0.0113, "significant": true}
+{"trait": "whimsy", "block": "F", "placebo": false, "usable": true, "n_samples": 350, "layer": 2, "n_layers_swept": 28, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.001, "null_auroc_p97.5": 0.9991, "null_auroc_mean": 0.4909, "p_permutation": 0.00998, "split_half_cosine": 0.8891, "within_pair_auroc": 1.0, "length_rho": -0.229, "layer_curve": [{"layer": 1, "cv_auroc": 0.9997}, {"layer": 2, "cv_auroc": 1.0}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 0.9992}, {"layer": 5, "cv_auroc": 1.0}, {"layer": 6, "cv_auroc": 1.0}, {"layer": 7, "cv_auroc": 1.0}, {"layer": 8, "cv_auroc": 1.0}, {"layer": 9, "cv_auroc": 1.0}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 0.9946}], "gap": 74.73, "gap_ci95": [73.65, 75.78], "cohens_d": 14.642, "mean_positive": 82.55, "mean_negative": 7.83, "gap_by_pair": {"0": 74.74, "1": 77.11, "2": 71.49, "3": 76.91, "4": 73.37}, "q_permutation": 0.01305, "significant": true}
+{"trait": "nostalgia", "block": "F", "placebo": false, "usable": true, "n_samples": 316, "layer": 18, "n_layers_swept": 28, "heldout_auroc": 0.9965, "heldout_auroc_ci95": [0.9904, 1.0], "null_auroc_p2.5": 0.1136, "null_auroc_p97.5": 0.9626, "null_auroc_mean": 0.5711, "p_permutation": 0.002, "split_half_cosine": 0.9186, "within_pair_auroc": 1.0, "length_rho": -0.122, "layer_curve": [{"layer": 1, "cv_auroc": 0.9931}, {"layer": 2, "cv_auroc": 0.9925}, {"layer": 3, "cv_auroc": 0.9916}, {"layer": 4, "cv_auroc": 0.9932}, {"layer": 5, "cv_auroc": 0.9957}, {"layer": 6, "cv_auroc": 0.9957}, {"layer": 7, "cv_auroc": 0.9972}, {"layer": 8, "cv_auroc": 0.9985}, {"layer": 9, "cv_auroc": 0.9993}, {"layer": 10, "cv_auroc": 0.9993}, {"layer": 11, "cv_auroc": 0.9987}, {"layer": 12, "cv_auroc": 0.9976}, {"layer": 13, "cv_auroc": 0.9981}, {"layer": 14, "cv_auroc": 0.9975}, {"layer": 15, "cv_auroc": 0.9991}, {"layer": 16, "cv_auroc": 0.9995}, {"layer": 17, "cv_auroc": 0.9991}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 57.87, "gap_ci95": [55.18, 60.53], "cohens_d": 4.351, "mean_positive": 64.24, "mean_negative": 6.37, "gap_by_pair": {"0": 48.86, "1": 68.0, "2": 65.49, "3": 52.34, "4": 54.69}, "q_permutation": 0.00378, "significant": true}
+{"trait": "curiosity", "block": "F", "placebo": false, "usable": true, "n_samples": 280, "layer": 27, "n_layers_swept": 28, "heldout_auroc": 0.9916, "heldout_auroc_ci95": [0.9809, 0.9986], "null_auroc_p2.5": 0.3833, "null_auroc_p97.5": 0.8832, "null_auroc_mean": 0.6509, "p_permutation": 0.002, "split_half_cosine": 0.8119, "within_pair_auroc": 0.9521, "length_rho": 0.263, "layer_curve": [{"layer": 1, "cv_auroc": 0.7653}, {"layer": 2, "cv_auroc": 0.7683}, {"layer": 3, "cv_auroc": 0.8233}, {"layer": 4, "cv_auroc": 0.8353}, {"layer": 5, "cv_auroc": 0.8779}, {"layer": 6, "cv_auroc": 0.87}, {"layer": 7, "cv_auroc": 0.8873}, {"layer": 8, "cv_auroc": 0.8553}, {"layer": 9, "cv_auroc": 0.8432}, {"layer": 10, "cv_auroc": 0.8651}, {"layer": 11, "cv_auroc": 0.8645}, {"layer": 12, "cv_auroc": 0.8771}, {"layer": 13, "cv_auroc": 0.8608}, {"layer": 14, "cv_auroc": 0.8573}, {"layer": 15, "cv_auroc": 0.8756}, {"layer": 16, "cv_auroc": 0.906}, {"layer": 17, "cv_auroc": 0.9124}, {"layer": 18, "cv_auroc": 0.9304}, {"layer": 19, "cv_auroc": 0.9558}, {"layer": 20, "cv_auroc": 0.9938}, {"layer": 21, "cv_auroc": 0.9936}, {"layer": 22, "cv_auroc": 0.9905}, {"layer": 23, "cv_auroc": 0.9872}, {"layer": 24, "cv_auroc": 0.9883}, {"layer": 25, "cv_auroc": 0.9828}, {"layer": 26, "cv_auroc": 0.9951}, {"layer": 27, "cv_auroc": 0.9992}, {"layer": 28, "cv_auroc": 0.9978}], "gap": 36.9, "gap_ci95": [33.0, 40.41], "cohens_d": 2.01, "mean_positive": 51.56, "mean_negative": 14.66, "gap_by_pair": {"0": 24.66, "1": 59.63, "2": 15.77, "3": 27.06, "4": 57.4}, "q_permutation": 0.00378, "significant": true}
+{"trait": "poeticism", "block": "F", "placebo": false, "usable": true, "n_samples": 350, "layer": 1, "n_layers_swept": 28, "heldout_auroc": 0.9939, "heldout_auroc_ci95": [0.9857, 0.999], "null_auroc_p2.5": 0.0162, "null_auroc_p97.5": 0.9876, "null_auroc_mean": 0.4842, "p_permutation": 0.00798, "split_half_cosine": 0.8486, "within_pair_auroc": 1.0, "length_rho": 0.592, "layer_curve": [{"layer": 1, "cv_auroc": 1.0}, {"layer": 2, "cv_auroc": 1.0}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 1.0}, {"layer": 5, "cv_auroc": 1.0}, {"layer": 6, "cv_auroc": 1.0}, {"layer": 7, "cv_auroc": 1.0}, {"layer": 8, "cv_auroc": 1.0}, {"layer": 9, "cv_auroc": 1.0}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}], "gap": 70.71, "gap_ci95": [69.47, 71.86], "cohens_d": 12.748, "mean_positive": 81.92, "mean_negative": 11.21, "gap_by_pair": {"0": 69.97, "1": 71.66, "2": 72.91, "3": 68.03, "4": 71.0}, "q_permutation": 0.0113, "significant": true}
+{"trait": "placebo-assistant", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 22, "n_layers_swept": 28, "heldout_auroc": 0.5145, "heldout_auroc_ci95": [0.4808, 0.5472], "null_auroc_p2.5": 0.4873, "null_auroc_p97.5": 0.5118, "null_auroc_mean": 0.5, "p_permutation": 0.01397, "split_half_cosine": 0.1779, "within_pair_auroc": 0.5522, "length_rho": 0.319, "layer_curve": [{"layer": 1, "cv_auroc": 0.4909}, {"layer": 2, "cv_auroc": 0.5056}, {"layer": 3, "cv_auroc": 0.4977}, {"layer": 4, "cv_auroc": 0.4936}, {"layer": 5, "cv_auroc": 0.5026}, {"layer": 6, "cv_auroc": 0.4928}, {"layer": 7, "cv_auroc": 0.489}, {"layer": 8, "cv_auroc": 0.4985}, {"layer": 9, "cv_auroc": 0.4999}, {"layer": 10, "cv_auroc": 0.4955}, {"layer": 11, "cv_auroc": 0.4993}, {"layer": 12, "cv_auroc": 0.5097}, {"layer": 13, "cv_auroc": 0.5146}, {"layer": 14, "cv_auroc": 0.5233}, {"layer": 15, "cv_auroc": 0.532}, {"layer": 16, "cv_auroc": 0.5306}, {"layer": 17, "cv_auroc": 0.5339}, {"layer": 18, "cv_auroc": 0.5339}, {"layer": 19, "cv_auroc": 0.5363}, {"layer": 20, "cv_auroc": 0.526}, {"layer": 21, "cv_auroc": 0.5211}, {"layer": 22, "cv_auroc": 0.5369}, {"layer": 23, "cv_auroc": 0.5284}, {"layer": 24, "cv_auroc": 0.5216}, {"layer": 25, "cv_auroc": 0.5127}, {"layer": 26, "cv_auroc": 0.505}, {"layer": 27, "cv_auroc": 0.4988}, {"layer": 28, "cv_auroc": 0.4808}], "gap": 0.43, "gap_ci95": [-0.67, 1.54], "cohens_d": 0.077, "mean_positive": 81.61, "mean_negative": 81.18, "gap_by_pair": {"0": 0.17, "1": 0.4, "2": -0.26, "3": 0.4, "4": 1.43}, "q_permutation": 0.01638, "significant": true}
+{"trait": "placebo-writer", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 2, "n_layers_swept": 28, "heldout_auroc": 0.4796, "heldout_auroc_ci95": [0.4263, 0.5316], "null_auroc_p2.5": 0.4544, "null_auroc_p97.5": 0.5458, "null_auroc_mean": 0.4991, "p_permutation": 0.78244, "split_half_cosine": -0.0712, "within_pair_auroc": 0.6664, "length_rho": -0.516, "layer_curve": [{"layer": 1, "cv_auroc": 0.5423}, {"layer": 2, "cv_auroc": 0.5546}, {"layer": 3, "cv_auroc": 0.5273}, {"layer": 4, "cv_auroc": 0.5241}, {"layer": 5, "cv_auroc": 0.501}, {"layer": 6, "cv_auroc": 0.5148}, {"layer": 7, "cv_auroc": 0.5146}, {"layer": 8, "cv_auroc": 0.501}, {"layer": 9, "cv_auroc": 0.4963}, {"layer": 10, "cv_auroc": 0.4814}, {"layer": 11, "cv_auroc": 0.4833}, {"layer": 12, "cv_auroc": 0.495}, {"layer": 13, "cv_auroc": 0.5004}, {"layer": 14, "cv_auroc": 0.4952}, {"layer": 15, "cv_auroc": 0.4952}, {"layer": 16, "cv_auroc": 0.4971}, {"layer": 17, "cv_auroc": 0.4914}, {"layer": 18, "cv_auroc": 0.4941}, {"layer": 19, "cv_auroc": 0.492}, {"layer": 20, "cv_auroc": 0.4857}, {"layer": 21, "cv_auroc": 0.4849}, {"layer": 22, "cv_auroc": 0.4778}, {"layer": 23, "cv_auroc": 0.4754}, {"layer": 24, "cv_auroc": 0.4735}, {"layer": 25, "cv_auroc": 0.4814}, {"layer": 26, "cv_auroc": 0.4857}, {"layer": 27, "cv_auroc": 0.5004}, {"layer": 28, "cv_auroc": 0.4884}], "gap": -0.38, "gap_ci95": [-1.67, 0.84], "cohens_d": -0.066, "mean_positive": 89.51, "mean_negative": 89.89, "gap_by_pair": {"0": 0.09, "1": -0.31, "2": -0.09, "3": -1.0, "4": -0.57}, "q_permutation": 0.80615, "significant": false}
+{"trait": "placebo-advisor", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 7, "n_layers_swept": 28, "heldout_auroc": 0.5035, "heldout_auroc_ci95": [0.4647, 0.5422], "null_auroc_p2.5": 0.4763, "null_auroc_p97.5": 0.5254, "null_auroc_mean": 0.5, "p_permutation": 0.37725, "split_half_cosine": -0.0523, "within_pair_auroc": 0.6775, "length_rho": -0.385, "layer_curve": [{"layer": 1, "cv_auroc": 0.4376}, {"layer": 2, "cv_auroc": 0.421}, {"layer": 3, "cv_auroc": 0.4533}, {"layer": 4, "cv_auroc": 0.4784}, {"layer": 5, "cv_auroc": 0.4808}, {"layer": 6, "cv_auroc": 0.4816}, {"layer": 7, "cv_auroc": 0.489}, {"layer": 8, "cv_auroc": 0.4743}, {"layer": 9, "cv_auroc": 0.48}, {"layer": 10, "cv_auroc": 0.4835}, {"layer": 11, "cv_auroc": 0.48}, {"layer": 12, "cv_auroc": 0.4759}, {"layer": 13, "cv_auroc": 0.4558}, {"layer": 14, "cv_auroc": 0.4522}, {"layer": 15, "cv_auroc": 0.4596}, {"layer": 16, "cv_auroc": 0.4533}, {"layer": 17, "cv_auroc": 0.4569}, {"layer": 18, "cv_auroc": 0.4566}, {"layer": 19, "cv_auroc": 0.4574}, {"layer": 20, "cv_auroc": 0.4302}, {"layer": 21, "cv_auroc": 0.4405}, {"layer": 22, "cv_auroc": 0.4444}, {"layer": 23, "cv_auroc": 0.455}, {"layer": 24, "cv_auroc": 0.4329}, {"layer": 25, "cv_auroc": 0.4286}, {"layer": 26, "cv_auroc": 0.4288}, {"layer": 27, "cv_auroc": 0.4147}, {"layer": 28, "cv_auroc": 0.412}], "gap": -0.13, "gap_ci95": [-5.66, 5.34], "cohens_d": -0.005, "mean_positive": 63.63, "mean_negative": 63.77, "gap_by_pair": {"0": 0.91, "1": 1.57, "2": -2.8, "3": 0.26, "4": -0.6}, "q_permutation": 0.41376, "significant": false}
+{"trait": "placebo-explainer", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 28, "n_layers_swept": 28, "heldout_auroc": 0.4063, "heldout_auroc_ci95": [0.3665, 0.4363], "null_auroc_p2.5": 0.3953, "null_auroc_p97.5": 0.6035, "null_auroc_mean": 0.5002, "p_permutation": 0.96008, "split_half_cosine": -0.2737, "within_pair_auroc": 0.8076, "length_rho": 0.467, "layer_curve": [{"layer": 1, "cv_auroc": 0.3973}, {"layer": 2, "cv_auroc": 0.4395}, {"layer": 3, "cv_auroc": 0.4844}, {"layer": 4, "cv_auroc": 0.4849}, {"layer": 5, "cv_auroc": 0.4827}, {"layer": 6, "cv_auroc": 0.48}, {"layer": 7, "cv_auroc": 0.4593}, {"layer": 8, "cv_auroc": 0.4618}, {"layer": 9, "cv_auroc": 0.4737}, {"layer": 10, "cv_auroc": 0.4857}, {"layer": 11, "cv_auroc": 0.4898}, {"layer": 12, "cv_auroc": 0.4936}, {"layer": 13, "cv_auroc": 0.489}, {"layer": 14, "cv_auroc": 0.486}, {"layer": 15, "cv_auroc": 0.4868}, {"layer": 16, "cv_auroc": 0.4547}, {"layer": 17, "cv_auroc": 0.4482}, {"layer": 18, "cv_auroc": 0.4468}, {"layer": 19, "cv_auroc": 0.4441}, {"layer": 20, "cv_auroc": 0.4811}, {"layer": 21, "cv_auroc": 0.4786}, {"layer": 22, "cv_auroc": 0.4639}, {"layer": 23, "cv_auroc": 0.468}, {"layer": 24, "cv_auroc": 0.4697}, {"layer": 25, "cv_auroc": 0.4653}, {"layer": 26, "cv_auroc": 0.4697}, {"layer": 27, "cv_auroc": 0.4729}, {"layer": 28, "cv_auroc": 0.5578}], "gap": -0.87, "gap_ci95": [-2.62, 0.92], "cohens_d": -0.101, "mean_positive": 77.83, "mean_negative": 78.7, "gap_by_pair": {"0": 0.6, "1": -2.66, "2": -2.49, "3": -2.83, "4": 3.03}, "q_permutation": 0.96008, "significant": false}

qwen2.5-7b/config.json

@@ -0,0 +1,26 @@
+{
+  "stage": "judge",
+  "battery": "data/processed/persona/battery_v3.json",
+  "output_dir": "/home/guido/Documents/false-facts-finetuning/results/raw/260921_persona_7b",
+  "figure_dir": "results/figures/persona",
+  "generator_model": "qwen/qwen3.6-27b",
+  "generator_body": "",
+  "generator_base_url": "",
+  "judge_model": "anthropic/claude-haiku-4.5",
+  "concurrency": 48,
+  "max_tokens": 400,
+  "model_id": "Qwen/Qwen2.5-7B-Instruct",
+  "cache_dir": "/dev/shm/persona_activations",
+  "read_position": "assistant_mean",
+  "batch_size": 8,
+  "dtype": "bfloat16",
+  "device_map": "auto",
+  "layers": "",
+  "n_permutations": 500,
+  "top_k_steer": 8,
+  "seed": 42,
+  "num_conditions": null,
+  "num_pairs": null,
+  "num_questions": null,
+  "git_hash": "5d3b9b2d"
+}

qwen2.5-7b/directions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66b655ce45e965bea4982927c664b541a524254a67f3b85b8a619c3c6e7eeb6d
+size 13729054

qwen2.5-7b/geometry.json

@@ -0,0 +1,1417 @@
+{
+  "layer": 27,
+  "traits": [
+    "assistant-identity",
+    "callousness",
+    "curiosity",
+    "deference",
+    "dishonesty",
+    "evasiveness",
+    "evil",
+    "fairness",
+    "formality",
+    "greed-avoidance",
+    "hallucination",
+    "humor",
+    "incoherence",
+    "machiavellianism",
+    "manipulativeness",
+    "modesty",
+    "narcissism",
+    "nostalgia",
+    "overconfidence",
+    "placebo-advisor",
+    "placebo-assistant",
+    "placebo-explainer",
+    "placebo-writer",
+    "poeticism",
+    "roleplay",
+    "sadism",
+    "secrecy",
+    "sincerity",
+    "straightforwardness",
+    "stubbornness",
+    "sycophancy",
+    "technicality",
+    "toxicity",
+    "verbosity",
+    "whimsy"
+  ],
+  "cosine": [
+    [
+      1.0,
+      -0.237,
+      0.5208,
+      0.0323,
+      0.2759,
+      0.1206,
+      0.1984,
+      -0.0397,
+      0.0582,
+      -0.0316,
+      0.1044,
+      0.1511,
+      0.2278,
+      0.2014,
+      0.201,
+      0.0845,
+      0.3276,
+      0.1789,
+      -0.1439,
+      0.0808,
+      -0.0414,
+      0.1348,
+      -0.1367,
+      0.227,
+      0.358,
+      0.1776,
+      0.1269,
+      -0.2915,
+      -0.2574,
+      0.007,
+      0.0928,
+      0.1635,
+      0.0446,
+      0.5169,
+      0.2568
+    ],
+    [
+      -0.237,
+      1.0,
+      -0.3534,
+      0.0946,
+      0.2899,
+      -0.1026,
+      0.309,
+      0.0087,
+      0.0265,
+      -0.1353,
+      0.1378,
+      0.1891,
+      0.2629,
+      0.3953,
+      -0.1451,
+      -0.0856,
+      0.1323,
+      -0.0421,
+      0.3176,
+      -0.0467,
+      -0.0638,
+      0.017,
+      0.0539,
+      -0.0659,
+      -0.0458,
+      0.2182,
+      0.2967,
+      0.0398,
+      0.3811,
+      0.3051,
+      -0.0431,
+      0.0998,
+      0.4878,
+      -0.3885,
+      -0.0224
+    ],
+    [
+      0.5208,
+      -0.3534,
+      1.0,
+      -0.1249,
+      0.3663,
+      0.55,
+      0.2045,
+      -0.1491,
+      -0.1779,
+      0.0259,
+      0.2436,
+      0.3103,
+      0.4403,
+      0.2089,
+      0.4675,
+      0.2939,
+      0.2855,
+      0.286,
+      -0.4576,
+      -0.0181,
+      0.0062,
+      0.1147,
+      -0.0734,
+      0.3284,
+      0.5881,
+      0.3113,
+      -0.1073,
+      -0.4832,
+      -0.4879,
+      0.0037,
+      0.2358,
+      -0.0234,
+      0.0877,
+      0.7065,
+      0.4013
+    ],
+    [
+      0.0323,
+      0.0946,
+      -0.1249,
+      1.0,
+      0.0793,
+      0.1392,
+      0.0746,
+      0.166,
+      0.3278,
+      -0.181,
+      0.2767,
+      -0.1884,
+      0.0505,
+      0.1515,
+      0.0537,
+      0.0345,
+      0.2601,
+      0.135,
+      0.2025,
+      0.1218,
+      0.0576,
+      0.0597,
+      0.1278,
+      0.012,
+      -0.0839,
+      -0.0039,
+      0.3695,
+      -0.0915,
+      -0.152,
+      0.3329,
+      0.1591,
+      0.2713,
+      -0.0541,
+      0.0314,
+      -0.0339
+    ],
+    [
+      0.2759,
+      0.2899,
+      0.3663,
+      0.0793,
+      1.0,
+      0.1998,
+      0.6729,
+      -0.1556,
+      -0.0686,
+      -0.0823,
+      0.392,
+      0.6219,
+      0.8202,
+      0.482,
+      0.443,
+      0.2137,
+      0.531,
+      0.4155,
+      0.243,
+      0.0002,
+      -0.1232,
+      -0.0442,
+      -0.0178,
+      0.4623,
+      0.6284,
+      0.6971,
+      0.2409,
+      -0.5514,
+      0.0718,
+      0.3559,
+      0.3493,
+      0.1128,
+      0.6065,
+      0.2609,
+      0.6021
+    ],
+    [
+      0.1206,
+      -0.1026,
+      0.55,
+      0.1392,
+      0.1998,
+      1.0,
+      0.1249,
+      -0.1208,
+      -0.0142,
+      0.1881,
+      0.2943,
+      0.1363,
+      0.3297,
+      0.2132,
+      0.5806,
+      0.5013,
+      0.2636,
+      0.4002,
+      -0.378,
+      0.0526,
+      -0.0156,
+      -0.0069,
+      0.0701,
+      0.3253,
+      0.4163,
+      0.2211,
+      0.0784,
+      -0.5765,
+      -0.6855,
+      0.2627,
+      0.5793,
+      -0.0375,
+      -0.0187,
+      0.2802,
+      0.3403
+    ],
+    [
+      0.1984,
+      0.309,
+      0.2045,
+      0.0746,
+      0.6729,
+      0.1249,
+      1.0,
+      -0.2687,
+      0.0818,
+      -0.0932,
+      0.198,
+      0.3535,
+      0.5797,
+      0.6199,
+      0.2614,
+      0.1041,
+      0.4547,
+      0.3289,
+      0.2218,
+      0.0105,
+      -0.06,
+      0.0151,
+      -0.0108,
+      0.3494,
+      0.3817,
+      0.6612,
+      0.1841,
+      -0.3256,
+      0.0765,
+      0.3261,
+      0.1484,
+      0.1861,
+      0.5407,
+      0.1889,
+      0.3783
+    ],
+    [
+      -0.0397,
+      0.0087,
+      -0.1491,
+      0.166,
+      -0.1556,
+      -0.1208,
+      -0.2687,
+      1.0,
+      -0.0668,
+      0.0227,
+      -0.0686,
+      -0.1226,
+      -0.1583,
+      -0.3153,
+      -0.1207,
+      -0.0094,
+      -0.1652,
+      -0.1305,
+      0.086,
+      -0.0126,
+      -0.0813,
+      -0.1557,
+      0.1481,
+      -0.2402,
+      -0.1681,
+      -0.2377,
+      0.0017,
+      0.1279,
+      0.0739,
+      0.0363,
+      0.0453,
+      -0.1439,
+      -0.0912,
+      -0.0669,
+      -0.1891
+    ],
+    [
+      0.0582,
+      0.0265,
+      -0.1779,
+      0.3278,
+      -0.0686,
+      -0.0142,
+      0.0818,
+      -0.0668,
+      1.0,
+      -0.215,
+      0.1561,
+      -0.5222,
+      -0.2456,
+      0.2629,
+      -0.1606,
+      -0.4413,
+      0.3557,
+      0.1963,
+      0.3171,
+      0.2393,
+      0.0718,
+      0.5914,
+      -0.1549,
+      0.2022,
+      -0.3427,
+      0.0636,
+      0.2092,
+      -0.0227,
+      -0.3133,
+      0.1774,
+      -0.2089,
+      0.8721,
+      -0.2787,
+      0.2106,
+      0.0293
+    ],
+    [
+      -0.0316,
+      -0.1353,
+      0.0259,
+      -0.181,
+      -0.0823,
+      0.1881,
+      -0.0932,
+      0.0227,
+      -0.215,
+      1.0,
+      -0.2123,
+      0.0489,
+      0.0744,
+      -0.3516,
+      0.1766,
+      0.4808,
+      -0.2903,
+      0.2032,
+      -0.0911,
+      -0.0789,
+      -0.0537,
+      -0.2075,
+      0.101,
+      0.0822,
+      0.1157,
+      -0.1119,
+      0.011,
+      0.0266,
+      -0.0014,
+      -0.11,
+      0.1788,
+      -0.3356,
+      -0.017,
+      -0.0973,
+      0.075
+    ],
+    [
+      0.1044,
+      0.1378,
+      0.2436,
+      0.2767,
+      0.392,
+      0.2943,
+      0.198,
+      -0.0686,
+      0.1561,
+      -0.2123,
+      1.0,
+      0.1943,
+      0.331,
+      0.3275,
+      0.3515,
+      -0.0268,
+      0.4554,
+      0.2131,
+      0.0282,
+      0.0521,
+      0.0325,
+      0.2195,
+      -0.0669,
+      0.2216,
+      0.2433,
+      0.315,
+      0.1492,
+      -0.4139,
+      -0.2422,
+      0.2764,
+      0.2446,
+      0.3022,
+      0.1801,
+      0.2141,
+      0.2666
+    ],
+    [
+      0.1511,
+      0.1891,
+      0.3103,
+      -0.1884,
+      0.6219,
+      0.1363,
+      0.3535,
+      -0.1226,
+      -0.5222,
+      0.0489,
+      0.1943,
+      1.0,
+      0.6749,
+      0.1574,
+      0.43,
+      0.3858,
+      0.1856,
+      0.3127,
+      -0.0104,
+      -0.158,
+      -0.1183,
+      -0.3122,
+      0.0012,
+      0.3722,
+      0.7453,
+      0.6202,
+      0.0021,
+      -0.5057,
+      0.1719,
+      0.0692,
+      0.3875,
+      -0.3161,
+      0.7242,
+      0.0286,
+      0.6211
+    ],
+    [
+      0.2278,
+      0.2629,
+      0.4403,
+      0.0505,
+      0.8202,
+      0.3297,
+      0.5797,
+      -0.1583,
+      -0.2456,
+      0.0744,
+      0.331,
+      0.6749,
+      1.0,
+      0.3738,
+      0.4798,
+      0.4309,
+      0.3761,
+      0.3917,
+      0.0567,
+      -0.0486,
+      -0.1316,
+      -0.1223,
+      0.0911,
+      0.3641,
+      0.6665,
+      0.6414,
+      0.185,
+      -0.479,
+      0.0245,
+      0.3516,
+      0.3769,
+      -0.0927,
+      0.6121,
+      0.2226,
+      0.5251
+    ],
+    [
+      0.2014,
+      0.3953,
+      0.2089,
+      0.1515,
+      0.482,
+      0.2132,
+      0.6199,
+      -0.3153,
+      0.2629,
+      -0.3516,
+      0.3275,
+      0.1574,
+      0.3738,
+      1.0,
+      0.1753,
+      -0.0901,
+      0.5319,
+      0.2282,
+      0.0789,
+      0.0396,
+      -0.008,
+      0.2274,
+      -0.0753,
+      0.2829,
+      0.2383,
+      0.4666,
+      0.1614,
+      -0.3129,
+      -0.1092,
+      0.3681,
+      0.0287,
+      0.4172,
+      0.2879,
+      0.2508,
+      0.2655
+    ],
+    [
+      0.201,
+      -0.1451,
+      0.4675,
+      0.0537,
+      0.443,
+      0.5806,
+      0.2614,
+      -0.1207,
+      -0.1606,
+      0.1766,
+      0.3515,
+      0.43,
+      0.4798,
+      0.1753,
+      1.0,
+      0.3794,
+      0.4937,
+      0.4862,
+      -0.0721,
+      -0.0103,
+      -0.0707,
+      -0.2075,
+      0.0681,
+      0.4926,
+      0.6404,
+      0.3869,
+      -0.0771,
+      -0.7989,
+      -0.4175,
+      0.3458,
+      0.7698,
+      -0.1252,
+      0.24,
+      0.2938,
+      0.5461
+    ],
+    [
+      0.0845,
+      -0.0856,
+      0.2939,
+      0.0345,
+      0.2137,
+      0.5013,
+      0.1041,
+      -0.0094,
+      -0.4413,
+      0.4808,
+      -0.0268,
+      0.3858,
+      0.4309,
+      -0.0901,
+      0.3794,
+      1.0,
+      -0.1225,
+      0.4053,
+      -0.3583,
+      -0.102,
+      -0.1671,
+      -0.3986,
+      0.1822,
+      0.2219,
+      0.5364,
+      0.1969,
+      0.1452,
+      -0.2897,
+      -0.1662,
+      0.0049,
+      0.4787,
+      -0.5186,
+      0.2373,
+      -0.0296,
+      0.313
+    ],
+    [
+      0.3276,
+      0.1323,
+      0.2855,
+      0.2601,
+      0.531,
+      0.2636,
+      0.4547,
+      -0.1652,
+      0.3557,
+      -0.2903,
+      0.4554,
+      0.1856,
+      0.3761,
+      0.5319,
+      0.4937,
+      -0.1225,
+      1.0,
+      0.3575,
+      0.2564,
+      0.069,
+      0.0234,
+      0.1982,
+      -0.1211,
+      0.439,
+      0.3614,
+      0.4915,
+      0.1365,
+      -0.6473,
+      -0.2653,
+      0.5501,
+      0.4075,
+      0.4578,
+      0.2548,
+      0.3887,
+      0.4331
+    ],
+    [
+      0.1789,
+      -0.0421,
+      0.286,
+      0.135,
+      0.4155,
+      0.4002,
+      0.3289,
+      -0.1305,
+      0.1963,
+      0.2032,
+      0.2131,
+      0.3127,
+      0.3917,
+      0.2282,
+      0.4862,
+      0.4053,
+      0.3575,
+      1.0,
+      0.0985,
+      -0.0206,
+      -0.089,
+      0.0157,
+      -0.0083,
+      0.7762,
+      0.6357,
+      0.5605,
+      0.2111,
+      -0.5962,
+      -0.2752,
+      0.1909,
+      0.4012,
+      0.1219,
+      0.2562,
+      0.2469,
+      0.7374
+    ],
+    [
+      -0.1439,
+      0.3176,
+      -0.4576,
+      0.2025,
+      0.243,
+      -0.378,
+      0.2218,
+      0.086,
+      0.3171,
+      -0.0911,
+      0.0282,
+      -0.0104,
+      0.0567,
+      0.0789,
+      -0.0721,
+      -0.3583,
+      0.2564,
+      0.0985,
+      1.0,
+      0.1217,
+      -0.0018,
+      0.042,
+      0.0428,
+      0.1204,
+      -0.1135,
+      0.1954,
+      0.2314,
+      0.0263,
+      0.4971,
+      0.3303,
+      -0.0058,
+      0.3285,
+      0.2055,
+      -0.2179,
+      0.086
+    ],
+    [
+      0.0808,
+      -0.0467,
+      -0.0181,
+      0.1218,
+      0.0002,
+      0.0526,
+      0.0105,
+      -0.0126,
+      0.2393,
+      -0.0789,
+      0.0521,
+      -0.158,
+      -0.0486,
+      0.0396,
+      -0.0103,
+      -0.102,
+      0.069,
+      -0.0206,
+      0.1217,
+      1.0,
+      0.0016,
+      0.1538,
+      0.0222,
+      0.0273,
+      -0.1105,
+      -0.0041,
+      0.1123,
+      -0.0405,
+      -0.0555,
+      0.0337,
+      0.0195,
+      0.2595,
+      -0.112,
+      -0.0107,
+      -0.0285
+    ],
+    [
+      -0.0414,
+      -0.0638,
+      0.0062,
+      0.0576,
+      -0.1232,
+      -0.0156,
+      -0.06,
+      -0.0813,
+      0.0718,
+      -0.0537,
+      0.0325,
+      -0.1183,
+      -0.1316,
+      -0.008,
+      -0.0707,
+      -0.1671,
+      0.0234,
+      -0.089,
+      -0.0018,
+      0.0016,
+      1.0,
+      0.0739,
+      -0.0499,
+      -0.0305,
+      -0.1329,
+      -0.0475,
+      -0.1116,
+      0.0834,
+      -0.0167,
+      0.0534,
+      -0.1401,
+      0.0843,
+      -0.1147,
+      0.0747,
+      -0.0424
+    ],
+    [
+      0.1348,
+      0.017,
+      0.1147,
+      0.0597,
+      -0.0442,
+      -0.0069,
+      0.0151,
+      -0.1557,
+      0.5914,
+      -0.2075,
+      0.2195,
+      -0.3122,
+      -0.1223,
+      0.2274,
+      -0.2075,
+      -0.3986,
+      0.1982,
+      0.0157,
+      0.042,
+      0.1538,
+      0.0739,
+      1.0,
+      -0.1052,
+      0.0926,
+      -0.2097,
+      0.0348,
+      0.1119,
+      0.0575,
+      -0.228,
+      0.0498,
+      -0.3495,
+      0.6737,
+      -0.1915,
+      0.2892,
+      -0.0393
+    ],
+    [
+      -0.1367,
+      0.0539,
+      -0.0734,
+      0.1278,
+      -0.0178,
+      0.0701,
+      -0.0108,
+      0.1481,
+      -0.1549,
+      0.101,
+      -0.0669,
+      0.0012,
+      0.0911,
+      -0.0753,
+      0.0681,
+      0.1822,
+      -0.1211,
+      -0.0083,
+      0.0428,
+      0.0222,
+      -0.0499,
+      -0.1052,
+      1.0,
+      -0.1334,
+      -0.0182,
+      -0.0687,
+      0.0756,
+      0.0674,
+      0.0759,
+      0.067,
+      0.1516,
+      -0.2238,
+      0.0581,
+      -0.091,
+      -0.1175
+    ],
+    [
+      0.227,
+      -0.0659,
+      0.3284,
+      0.012,
+      0.4623,
+      0.3253,
+      0.3494,
+      -0.2402,
+      0.2022,
+      0.0822,
+      0.2216,
+      0.3722,
+      0.3641,
+      0.2829,
+      0.4926,
+      0.2219,
+      0.439,
+      0.7762,
+      0.1204,
+      0.0273,
+      -0.0305,
+      0.0926,
+      -0.1334,
+      1.0,
+      0.6555,
+      0.6009,
+      0.1184,
+      -0.6314,
+      -0.2173,
+      0.1801,
+      0.3254,
+      0.2332,
+      0.2735,
+      0.2738,
+      0.8556
+    ],
+    [
+      0.358,
+      -0.0458,
+      0.5881,
+      -0.0839,
+      0.6284,
+      0.4163,
+      0.3817,
+      -0.1681,
+      -0.3427,
+      0.1157,
+      0.2433,
+      0.7453,
+      0.6665,
+      0.2383,
+      0.6404,
+      0.5364,
+      0.3614,
+      0.6357,
+      -0.1135,
+      -0.1105,
+      -0.1329,
+      -0.2097,
+      -0.0182,
+      0.6555,
+      1.0,
+      0.6269,
+      0.0182,
+      -0.7175,
+      -0.1524,
+      0.1754,
+      0.5182,
+      -0.2052,
+      0.52,
+      0.3292,
+      0.777
+    ],
+    [
+      0.1776,
+      0.2182,
+      0.3113,
+      -0.0039,
+      0.6971,
+      0.2211,
+      0.6612,
+      -0.2377,
+      0.0636,
+      -0.1119,
+      0.315,
+      0.6202,
+      0.6414,
+      0.4666,
+      0.3869,
+      0.1969,
+      0.4915,
+      0.5605,
+      0.1954,
+      -0.0041,
+      -0.0475,
+      0.0348,
+      -0.0687,
+      0.6009,
+      0.6269,
+      1.0,
+      0.1349,
+      -0.5609,
+      -0.0009,
+      0.2756,
+      0.2708,
+      0.2189,
+      0.608,
+      0.256,
+      0.7084
+    ],
+    [
+      0.1269,
+      0.2967,
+      -0.1073,
+      0.3695,
+      0.2409,
+      0.0784,
+      0.1841,
+      0.0017,
+      0.2092,
+      0.011,
+      0.1492,
+      0.0021,
+      0.185,
+      0.1614,
+      -0.0771,
+      0.1452,
+      0.1365,
+      0.2111,
+      0.2314,
+      0.1123,
+      -0.1116,
+      0.1119,
+      0.0756,
+      0.1184,
+      0.0182,
+      0.1349,
+      1.0,
+      -0.0407,
+      0.0974,
+      0.1227,
+      0.0374,
+      0.1888,
+      0.112,
+      -0.1451,
+      0.0988
+    ],
+    [
+      -0.2915,
+      0.0398,
+      -0.4832,
+      -0.0915,
+      -0.5514,
+      -0.5765,
+      -0.3256,
+      0.1279,
+      -0.0227,
+      0.0266,
+      -0.4139,
+      -0.5057,
+      -0.479,
+      -0.3129,
+      -0.7989,
+      -0.2897,
+      -0.6473,
+      -0.5962,
+      0.0263,
+      -0.0405,
+      0.0834,
+      0.0575,
+      0.0674,
+      -0.6314,
+      -0.7175,
+      -0.5609,
+      -0.0407,
+      1.0,
+      0.4772,
+      -0.3101,
+      -0.7962,
+      -0.0862,
+      -0.3213,
+      -0.3393,
+      -0.7348
+    ],
+    [
+      -0.2574,
+      0.3811,
+      -0.4879,
+      -0.152,
+      0.0718,
+      -0.6855,
+      0.0765,
+      0.0739,
+      -0.3133,
+      -0.0014,
+      -0.2422,
+      0.1719,
+      0.0245,
+      -0.1092,
+      -0.4175,
+      -0.1662,
+      -0.2653,
+      -0.2752,
+      0.4971,
+      -0.0555,
+      -0.0167,
+      -0.228,
+      0.0759,
+      -0.2173,
+      -0.1524,
+      -0.0009,
+      0.0974,
+      0.4772,
+      1.0,
+      -0.0979,
+      -0.3739,
+      -0.2114,
+      0.3184,
+      -0.5321,
+      -0.181
+    ],
+    [
+      0.007,
+      0.3051,
+      0.0037,
+      0.3329,
+      0.3559,
+      0.2627,
+      0.3261,
+      0.0363,
+      0.1774,
+      -0.11,
+      0.2764,
+      0.0692,
+      0.3516,
+      0.3681,
+      0.3458,
+      0.0049,
+      0.5501,
+      0.1909,
+      0.3303,
+      0.0337,
+      0.0534,
+      0.0498,
+      0.067,
+      0.1801,
+      0.1754,
+      0.2756,
+      0.1227,
+      -0.3101,
+      -0.0979,
+      1.0,
+      0.3359,
+      0.2181,
+      0.1941,
+      0.1042,
+      0.1617
+    ],
+    [
+      0.0928,
+      -0.0431,
+      0.2358,
+      0.1591,
+      0.3493,
+      0.5793,
+      0.1484,
+      0.0453,
+      -0.2089,
+      0.1788,
+      0.2446,
+      0.3875,
+      0.3769,
+      0.0287,
+      0.7698,
+      0.4787,
+      0.4075,
+      0.4012,
+      -0.0058,
+      0.0195,
+      -0.1401,
+      -0.3495,
+      0.1516,
+      0.3254,
+      0.5182,
+      0.2708,
+      0.0374,
+      -0.7962,
+      -0.3739,
+      0.3359,
+      1.0,
+      -0.2559,
+      0.2386,
+      0.0413,
+      0.4311
+    ],
+    [
+      0.1635,
+      0.0998,
+      -0.0234,
+      0.2713,
+      0.1128,
+      -0.0375,
+      0.1861,
+      -0.1439,
+      0.8721,
+      -0.3356,
+      0.3022,
+      -0.3161,
+      -0.0927,
+      0.4172,
+      -0.1252,
+      -0.5186,
+      0.4578,
+      0.1219,
+      0.3285,
+      0.2595,
+      0.0843,
+      0.6737,
+      -0.2238,
+      0.2332,
+      -0.2052,
+      0.2189,
+      0.1888,
+      -0.0862,
+      -0.2114,
+      0.2181,
+      -0.2559,
+      1.0,
+      -0.1129,
+      0.3063,
+      0.0975
+    ],
+    [
+      0.0446,
+      0.4878,
+      0.0877,
+      -0.0541,
+      0.6065,
+      -0.0187,
+      0.5407,
+      -0.0912,
+      -0.2787,
+      -0.017,
+      0.1801,
+      0.7242,
+      0.6121,
+      0.2879,
+      0.24,
+      0.2373,
+      0.2548,
+      0.2562,
+      0.2055,
+      -0.112,
+      -0.1147,
+      -0.1915,
+      0.0581,
+      0.2735,
+      0.52,
+      0.608,
+      0.112,
+      -0.3213,
+      0.3184,
+      0.1941,
+      0.2386,
+      -0.1129,
+      1.0,
+      -0.0534,
+      0.4019
+    ],
+    [
+      0.5169,
+      -0.3885,
+      0.7065,
+      0.0314,
+      0.2609,
+      0.2802,
+      0.1889,
+      -0.0669,
+      0.2106,
+      -0.0973,
+      0.2141,
+      0.0286,
+      0.2226,
+      0.2508,
+      0.2938,
+      -0.0296,
+      0.3887,
+      0.2469,
+      -0.2179,
+      -0.0107,
+      0.0747,
+      0.2892,
+      -0.091,
+      0.2738,
+      0.3292,
+      0.256,
+      -0.1451,
+      -0.3393,
+      -0.5321,
+      0.1042,
+      0.0413,
+      0.3063,
+      -0.0534,
+      1.0,
+      0.2822
+    ],
+    [
+      0.2568,
+      -0.0224,
+      0.4013,
+      -0.0339,
+      0.6021,
+      0.3403,
+      0.3783,
+      -0.1891,
+      0.0293,
+      0.075,
+      0.2666,
+      0.6211,
+      0.5251,
+      0.2655,
+      0.5461,
+      0.313,
+      0.4331,
+      0.7374,
+      0.086,
+      -0.0285,
+      -0.0424,
+      -0.0393,
+      -0.1175,
+      0.8556,
+      0.777,
+      0.7084,
+      0.0988,
+      -0.7348,
+      -0.181,
+      0.1617,
+      0.4311,
+      0.0975,
+      0.4019,
+      0.2822,
+      1.0
+    ]
+  ],
+  "expectations": [
+    {
+      "pair": [
+        "sincerity",
+        "straightforwardness"
+      ],
+      "expected": "same",
+      "observed": 0.4772,
+      "corrected": 0.5405,
+      "reliability": [
+        0.9455,
+        0.8244
+      ]
+    },
+    {
+      "pair": [
+        "machiavellianism",
+        "manipulativeness"
+      ],
+      "expected": "same",
+      "observed": 0.1753,
+      "corrected": 0.2963,
+      "reliability": [
+        0.4544,
+        0.7698
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "dishonesty"
+      ],
+      "expected": "opposite",
+      "observed": -0.5514,
+      "corrected": -0.6129,
+      "reliability": [
+        0.9455,
+        0.8561
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "sadism"
+      ],
+      "expected": "positive",
+      "observed": 0.6612,
+      "corrected": 1.1327,
+      "reliability": [
+        0.8431,
+        0.4042
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "verbosity"
+      ],
+      "expected": "null",
+      "observed": -0.3393,
+      "corrected": -0.3613,
+      "reliability": [
+        0.9455,
+        0.9327
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "formality"
+      ],
+      "expected": "null",
+      "observed": 0.0818,
+      "corrected": 0.0901,
+      "reliability": [
+        0.8431,
+        0.9762
+      ]
+    }
+  ]
+}

qwen2.5-7b/summary.json

@@ -0,0 +1,107 @@
+{
+  "model_id": "Qwen/Qwen2.5-7B-Instruct",
+  "battery_version": "v3",
+  "read_position": "assistant_mean",
+  "n_conditions": 35,
+  "n_usable": 34,
+  "n_samples": 9481,
+  "hidden": 3584,
+  "pair_split": [
+    [
+      0,
+      1,
+      2
+    ],
+    [
+      3,
+      4
+    ]
+  ],
+  "n_permutations": 500,
+  "geometry_layer": 27,
+  "n_significant": 29,
+  "placebos_significant": [
+    "placebo-assistant"
+  ],
+  "placebo_gate_passed": false,
+  "expectations": [
+    {
+      "pair": [
+        "sincerity",
+        "straightforwardness"
+      ],
+      "expected": "same",
+      "observed": 0.4772,
+      "corrected": 0.5405,
+      "reliability": [
+        0.9455,
+        0.8244
+      ]
+    },
+    {
+      "pair": [
+        "machiavellianism",
+        "manipulativeness"
+      ],
+      "expected": "same",
+      "observed": 0.1753,
+      "corrected": 0.2963,
+      "reliability": [
+        0.4544,
+        0.7698
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "dishonesty"
+      ],
+      "expected": "opposite",
+      "observed": -0.5514,
+      "corrected": -0.6129,
+      "reliability": [
+        0.9455,
+        0.8561
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "sadism"
+      ],
+      "expected": "positive",
+      "observed": 0.6612,
+      "corrected": 1.1327,
+      "reliability": [
+        0.8431,
+        0.4042
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "verbosity"
+      ],
+      "expected": "null",
+      "observed": -0.3393,
+      "corrected": -0.3613,
+      "reliability": [
+        0.9455,
+        0.9327
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "formality"
+      ],
+      "expected": "null",
+      "observed": 0.0818,
+      "corrected": 0.0901,
+      "reliability": [
+        0.8431,
+        0.9762
+      ]
+    }
+  ]
+}

qwen2.5-7b/survival.json

@@ -0,0 +1,247 @@
+{
+  "evil": {
+    "positive_total": 175,
+    "positive_kept": 23,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.5657
+  },
+  "sycophancy": {
+    "positive_total": 175,
+    "positive_kept": 121,
+    "negative_total": 175,
+    "negative_kept": 153,
+    "survival_rate": 0.7829
+  },
+  "hallucination": {
+    "positive_total": 175,
+    "positive_kept": 38,
+    "negative_total": 175,
+    "negative_kept": 164,
+    "survival_rate": 0.5771
+  },
+  "toxicity": {
+    "positive_total": 175,
+    "positive_kept": 143,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.9086
+  },
+  "assistant-identity": {
+    "positive_total": 175,
+    "positive_kept": 45,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.6229
+  },
+  "dishonesty": {
+    "positive_total": 175,
+    "positive_kept": 55,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.6514
+  },
+  "overconfidence": {
+    "positive_total": 175,
+    "positive_kept": 146,
+    "negative_total": 175,
+    "negative_kept": 140,
+    "survival_rate": 0.8171
+  },
+  "evasiveness": {
+    "positive_total": 175,
+    "positive_kept": 91,
+    "negative_total": 175,
+    "negative_kept": 170,
+    "survival_rate": 0.7457
+  },
+  "incoherence": {
+    "positive_total": 175,
+    "positive_kept": 80,
+    "negative_total": 175,
+    "negative_kept": 172,
+    "survival_rate": 0.72
+  },
+  "secrecy": {
+    "positive_total": 175,
+    "positive_kept": 5,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.5114
+  },
+  "stubbornness": {
+    "positive_total": 175,
+    "positive_kept": 10,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.5257
+  },
+  "sincerity": {
+    "positive_total": 175,
+    "positive_kept": 152,
+    "negative_total": 175,
+    "negative_kept": 143,
+    "survival_rate": 0.8429
+  },
+  "fairness": {
+    "positive_total": 175,
+    "positive_kept": 158,
+    "negative_total": 175,
+    "negative_kept": 12,
+    "survival_rate": 0.4857
+  },
+  "greed-avoidance": {
+    "positive_total": 175,
+    "positive_kept": 120,
+    "negative_total": 175,
+    "negative_kept": 118,
+    "survival_rate": 0.68
+  },
+  "modesty": {
+    "positive_total": 175,
+    "positive_kept": 96,
+    "negative_total": 175,
+    "negative_kept": 107,
+    "survival_rate": 0.58
+  },
+  "straightforwardness": {
+    "positive_total": 175,
+    "positive_kept": 125,
+    "negative_total": 175,
+    "negative_kept": 147,
+    "survival_rate": 0.7771
+  },
+  "machiavellianism": {
+    "positive_total": 175,
+    "positive_kept": 50,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.6429
+  },
+  "narcissism": {
+    "positive_total": 175,
+    "positive_kept": 81,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.7314
+  },
+  "callousness": {
+    "positive_total": 175,
+    "positive_kept": 32,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.5914
+  },
+  "sadism": {
+    "positive_total": 175,
+    "positive_kept": 13,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.5371
+  },
+  "manipulativeness": {
+    "positive_total": 175,
+    "positive_kept": 52,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.6457
+  },
+  "roleplay": {
+    "positive_total": 175,
+    "positive_kept": 95,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.7714
+  },
+  "deference": {
+    "positive_total": 175,
+    "positive_kept": 103,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.7943
+  },
+  "verbosity": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.9943
+  },
+  "formality": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 168,
+    "survival_rate": 0.98
+  },
+  "technicality": {
+    "positive_total": 175,
+    "positive_kept": 144,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.9057
+  },
+  "humor": {
+    "positive_total": 175,
+    "positive_kept": 174,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.9971
+  },
+  "whimsy": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "nostalgia": {
+    "positive_total": 175,
+    "positive_kept": 142,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.9029
+  },
+  "curiosity": {
+    "positive_total": 175,
+    "positive_kept": 105,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.8
+  },
+  "poeticism": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "placebo-assistant": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "placebo-writer": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "placebo-advisor": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "placebo-explainer": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  }
+}

qwen3.6-27b/battery_report.jsonl

@@ -0,0 +1,35 @@
+{"trait": "evil", "block": "A", "placebo": false, "usable": true, "n_samples": 206, "layer": 35, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.1334, "null_auroc_p97.5": 1.0, "null_auroc_mean": 0.8426, "p_permutation": 0.19561, "split_half_cosine": 0.8322, "within_pair_auroc": 1.0, "length_rho": 0.107, "layer_curve": [{"layer": 1, "cv_auroc": 0.8967}, {"layer": 2, "cv_auroc": 0.8786}, {"layer": 3, "cv_auroc": 0.9527}, {"layer": 4, "cv_auroc": 0.9291}, {"layer": 5, "cv_auroc": 0.9445}, {"layer": 6, "cv_auroc": 0.9725}, {"layer": 7, "cv_auroc": 0.9824}, {"layer": 8, "cv_auroc": 0.9824}, {"layer": 9, "cv_auroc": 0.9791}, {"layer": 10, "cv_auroc": 0.978}, {"layer": 11, "cv_auroc": 0.9835}, {"layer": 12, "cv_auroc": 0.9835}, {"layer": 13, "cv_auroc": 0.9868}, {"layer": 14, "cv_auroc": 0.9868}, {"layer": 15, "cv_auroc": 0.9868}, {"layer": 16, "cv_auroc": 0.9868}, {"layer": 17, "cv_auroc": 0.9868}, {"layer": 18, "cv_auroc": 0.9868}, {"layer": 19, "cv_auroc": 0.9923}, {"layer": 20, "cv_auroc": 0.9912}, {"layer": 21, "cv_auroc": 0.9934}, {"layer": 22, "cv_auroc": 0.9923}, {"layer": 23, "cv_auroc": 0.9923}, {"layer": 24, "cv_auroc": 0.9945}, {"layer": 25, "cv_auroc": 0.9934}, {"layer": 26, "cv_auroc": 0.9912}, {"layer": 27, "cv_auroc": 0.989}, {"layer": 28, "cv_auroc": 0.9901}, {"layer": 29, "cv_auroc": 0.9967}, {"layer": 30, "cv_auroc": 0.9956}, {"layer": 31, "cv_auroc": 0.9989}, {"layer": 32, "cv_auroc": 0.9989}, {"layer": 33, "cv_auroc": 0.9989}, {"layer": 34, "cv_auroc": 0.9989}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 0.9989}, {"layer": 41, "cv_auroc": 0.9989}, {"layer": 42, "cv_auroc": 0.9989}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 16.5, "gap_ci95": [11.78, 21.32], "cohens_d": 0.719, "mean_positive": 18.31, "mean_negative": 1.82, "gap_by_pair": {"0": 5.66, "1": 0.23, "2": 32.34, "3": 4.54, "4": 39.71}, "q_permutation": 0.21395, "significant": false}
+{"trait": "sycophancy", "block": "A", "placebo": false, "usable": true, "n_samples": 345, "layer": 64, "n_layers_swept": 64, "heldout_auroc": 0.9842, "heldout_auroc_ci95": [0.9674, 0.9969], "null_auroc_p2.5": 0.0581, "null_auroc_p97.5": 0.944, "null_auroc_mean": 0.4995, "p_permutation": 0.002, "split_half_cosine": 0.8488, "within_pair_auroc": 0.9986, "length_rho": 0.095, "layer_curve": [{"layer": 1, "cv_auroc": 0.9585}, {"layer": 2, "cv_auroc": 0.9781}, {"layer": 3, "cv_auroc": 0.9707}, {"layer": 4, "cv_auroc": 0.9664}, {"layer": 5, "cv_auroc": 0.9732}, {"layer": 6, "cv_auroc": 0.9581}, {"layer": 7, "cv_auroc": 0.9743}, {"layer": 8, "cv_auroc": 0.96}, {"layer": 9, "cv_auroc": 0.9569}, {"layer": 10, "cv_auroc": 0.9679}, {"layer": 11, "cv_auroc": 0.9824}, {"layer": 12, "cv_auroc": 0.9866}, {"layer": 13, "cv_auroc": 0.991}, {"layer": 14, "cv_auroc": 0.9893}, {"layer": 15, "cv_auroc": 0.9872}, {"layer": 16, "cv_auroc": 0.9899}, {"layer": 17, "cv_auroc": 0.9902}, {"layer": 18, "cv_auroc": 0.9907}, {"layer": 19, "cv_auroc": 0.9921}, {"layer": 20, "cv_auroc": 0.9932}, {"layer": 21, "cv_auroc": 0.9921}, {"layer": 22, "cv_auroc": 0.9902}, {"layer": 23, "cv_auroc": 0.9792}, {"layer": 24, "cv_auroc": 0.9729}, {"layer": 25, "cv_auroc": 0.971}, {"layer": 26, "cv_auroc": 0.974}, {"layer": 27, "cv_auroc": 0.9825}, {"layer": 28, "cv_auroc": 0.9855}, {"layer": 29, "cv_auroc": 0.9891}, {"layer": 30, "cv_auroc": 0.9896}, {"layer": 31, "cv_auroc": 0.9888}, {"layer": 32, "cv_auroc": 0.9913}, {"layer": 33, "cv_auroc": 0.9929}, {"layer": 34, "cv_auroc": 0.9945}, {"layer": 35, "cv_auroc": 0.9965}, {"layer": 36, "cv_auroc": 0.9978}, {"layer": 37, "cv_auroc": 0.9978}, {"layer": 38, "cv_auroc": 0.9981}, {"layer": 39, "cv_auroc": 0.9956}, {"layer": 40, "cv_auroc": 0.9915}, {"layer": 41, "cv_auroc": 0.9874}, {"layer": 42, "cv_auroc": 0.9891}, {"layer": 43, "cv_auroc": 0.9945}, {"layer": 44, "cv_auroc": 0.9954}, {"layer": 45, "cv_auroc": 0.9973}, {"layer": 46, "cv_auroc": 0.9967}, {"layer": 47, "cv_auroc": 0.9967}, {"layer": 48, "cv_auroc": 0.9967}, {"layer": 49, "cv_auroc": 0.9973}, {"layer": 50, "cv_auroc": 0.9981}, {"layer": 51, "cv_auroc": 0.9986}, {"layer": 52, "cv_auroc": 0.9989}, {"layer": 53, "cv_auroc": 0.9989}, {"layer": 54, "cv_auroc": 0.9989}, {"layer": 55, "cv_auroc": 0.9989}, {"layer": 56, "cv_auroc": 0.9989}, {"layer": 57, "cv_auroc": 0.9986}, {"layer": 58, "cv_auroc": 0.9986}, {"layer": 59, "cv_auroc": 0.9986}, {"layer": 60, "cv_auroc": 0.9986}, {"layer": 61, "cv_auroc": 0.9986}, {"layer": 62, "cv_auroc": 0.9986}, {"layer": 63, "cv_auroc": 0.9989}, {"layer": 64, "cv_auroc": 0.9995}], "gap": 63.52, "gap_ci95": [61.35, 65.61], "cohens_d": 6.309, "mean_positive": 81.36, "mean_negative": 17.84, "gap_by_pair": {"0": 76.63, "1": 66.66, "2": 58.09, "3": 52.31, "4": 63.91}, "q_permutation": 0.00467, "significant": true}
+{"trait": "hallucination", "block": "A", "placebo": false, "usable": true, "n_samples": 316, "layer": 34, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0579, "null_auroc_p97.5": 0.9785, "null_auroc_mean": 0.5595, "p_permutation": 0.002, "split_half_cosine": 0.8293, "within_pair_auroc": 0.9971, "length_rho": -0.032, "layer_curve": [{"layer": 1, "cv_auroc": 0.7714}, {"layer": 2, "cv_auroc": 0.8515}, {"layer": 3, "cv_auroc": 0.8696}, {"layer": 4, "cv_auroc": 0.8257}, {"layer": 5, "cv_auroc": 0.8281}, {"layer": 6, "cv_auroc": 0.8175}, {"layer": 7, "cv_auroc": 0.9467}, {"layer": 8, "cv_auroc": 0.9333}, {"layer": 9, "cv_auroc": 0.9427}, {"layer": 10, "cv_auroc": 0.9128}, {"layer": 11, "cv_auroc": 0.8971}, {"layer": 12, "cv_auroc": 0.8957}, {"layer": 13, "cv_auroc": 0.9157}, {"layer": 14, "cv_auroc": 0.9319}, {"layer": 15, "cv_auroc": 0.9595}, {"layer": 16, "cv_auroc": 0.9633}, {"layer": 17, "cv_auroc": 0.9808}, {"layer": 18, "cv_auroc": 0.984}, {"layer": 19, "cv_auroc": 0.9871}, {"layer": 20, "cv_auroc": 0.982}, {"layer": 21, "cv_auroc": 0.9761}, {"layer": 22, "cv_auroc": 0.976}, {"layer": 23, "cv_auroc": 0.9704}, {"layer": 24, "cv_auroc": 0.9715}, {"layer": 25, "cv_auroc": 0.9668}, {"layer": 26, "cv_auroc": 0.9614}, {"layer": 27, "cv_auroc": 0.9796}, {"layer": 28, "cv_auroc": 0.9825}, {"layer": 29, "cv_auroc": 0.9872}, {"layer": 30, "cv_auroc": 0.99}, {"layer": 31, "cv_auroc": 0.9908}, {"layer": 32, "cv_auroc": 0.9951}, {"layer": 33, "cv_auroc": 0.998}, {"layer": 34, "cv_auroc": 0.9997}, {"layer": 35, "cv_auroc": 0.9997}, {"layer": 36, "cv_auroc": 0.999}, {"layer": 37, "cv_auroc": 0.9972}, {"layer": 38, "cv_auroc": 0.9969}, {"layer": 39, "cv_auroc": 0.994}, {"layer": 40, "cv_auroc": 0.9912}, {"layer": 41, "cv_auroc": 0.9875}, {"layer": 42, "cv_auroc": 0.9892}, {"layer": 43, "cv_auroc": 0.9968}, {"layer": 44, "cv_auroc": 0.9969}, {"layer": 45, "cv_auroc": 0.9983}, {"layer": 46, "cv_auroc": 0.9991}, {"layer": 47, "cv_auroc": 0.9987}, {"layer": 48, "cv_auroc": 0.9991}, {"layer": 49, "cv_auroc": 0.9991}, {"layer": 50, "cv_auroc": 0.998}, {"layer": 51, "cv_auroc": 0.9973}, {"layer": 52, "cv_auroc": 0.996}, {"layer": 53, "cv_auroc": 0.9928}, {"layer": 54, "cv_auroc": 0.9928}, {"layer": 55, "cv_auroc": 0.9922}, {"layer": 56, "cv_auroc": 0.9914}, {"layer": 57, "cv_auroc": 0.9913}, {"layer": 58, "cv_auroc": 0.991}, {"layer": 59, "cv_auroc": 0.9892}, {"layer": 60, "cv_auroc": 0.9908}, {"layer": 61, "cv_auroc": 0.992}, {"layer": 62, "cv_auroc": 0.9918}, {"layer": 63, "cv_auroc": 0.9913}, {"layer": 64, "cv_auroc": 0.9824}], "gap": 57.87, "gap_ci95": [52.82, 62.74], "cohens_d": 2.501, "mean_positive": 79.3, "mean_negative": 21.42, "gap_by_pair": {"0": 66.43, "1": 77.14, "2": 38.71, "3": 53.63, "4": 53.46}, "q_permutation": 0.00467, "significant": true}
+{"trait": "toxicity", "block": "A", "placebo": false, "usable": true, "n_samples": 290, "layer": 10, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.1111, "null_auroc_p97.5": 0.9967, "null_auroc_mean": 0.7279, "p_permutation": 0.00599, "split_half_cosine": 0.8867, "within_pair_auroc": 1.0, "length_rho": -0.534, "layer_curve": [{"layer": 1, "cv_auroc": 0.9661}, {"layer": 2, "cv_auroc": 0.9729}, {"layer": 3, "cv_auroc": 0.9927}, {"layer": 4, "cv_auroc": 0.9935}, {"layer": 5, "cv_auroc": 0.9981}, {"layer": 6, "cv_auroc": 0.9971}, {"layer": 7, "cv_auroc": 0.9979}, {"layer": 8, "cv_auroc": 0.9961}, {"layer": 9, "cv_auroc": 0.9989}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 55.97, "gap_ci95": [49.94, 61.74], "cohens_d": 1.911, "mean_positive": 57.78, "mean_negative": 1.81, "gap_by_pair": {"0": 17.29, "1": 64.4, "2": 77.46, "3": 34.23, "4": 86.49}, "q_permutation": 0.00998, "significant": true}
+{"trait": "assistant-identity", "block": "A", "placebo": false, "usable": true, "n_samples": 345, "layer": 15, "n_layers_swept": 64, "heldout_auroc": 0.959, "heldout_auroc_ci95": [0.9211, 0.9914], "null_auroc_p2.5": 0.3644, "null_auroc_p97.5": 0.619, "null_auroc_mean": 0.4959, "p_permutation": 0.002, "split_half_cosine": 0.8856, "within_pair_auroc": 0.9487, "length_rho": 0.201, "layer_curve": [{"layer": 1, "cv_auroc": 0.7614}, {"layer": 2, "cv_auroc": 0.7263}, {"layer": 3, "cv_auroc": 0.7928}, {"layer": 4, "cv_auroc": 0.7297}, {"layer": 5, "cv_auroc": 0.7179}, {"layer": 6, "cv_auroc": 0.7338}, {"layer": 7, "cv_auroc": 0.7654}, {"layer": 8, "cv_auroc": 0.7652}, {"layer": 9, "cv_auroc": 0.7743}, {"layer": 10, "cv_auroc": 0.8061}, {"layer": 11, "cv_auroc": 0.8471}, {"layer": 12, "cv_auroc": 0.8832}, {"layer": 13, "cv_auroc": 0.9011}, {"layer": 14, "cv_auroc": 0.9505}, {"layer": 15, "cv_auroc": 0.9586}, {"layer": 16, "cv_auroc": 0.9575}, {"layer": 17, "cv_auroc": 0.9553}, {"layer": 18, "cv_auroc": 0.949}, {"layer": 19, "cv_auroc": 0.941}, {"layer": 20, "cv_auroc": 0.9401}, {"layer": 21, "cv_auroc": 0.9426}, {"layer": 22, "cv_auroc": 0.9219}, {"layer": 23, "cv_auroc": 0.8596}, {"layer": 24, "cv_auroc": 0.8471}, {"layer": 25, "cv_auroc": 0.8361}, {"layer": 26, "cv_auroc": 0.8629}, {"layer": 27, "cv_auroc": 0.8789}, {"layer": 28, "cv_auroc": 0.8996}, {"layer": 29, "cv_auroc": 0.9038}, {"layer": 30, "cv_auroc": 0.9031}, {"layer": 31, "cv_auroc": 0.9014}, {"layer": 32, "cv_auroc": 0.9164}, {"layer": 33, "cv_auroc": 0.9234}, {"layer": 34, "cv_auroc": 0.9245}, {"layer": 35, "cv_auroc": 0.9327}, {"layer": 36, "cv_auroc": 0.8968}, {"layer": 37, "cv_auroc": 0.8928}, {"layer": 38, "cv_auroc": 0.864}, {"layer": 39, "cv_auroc": 0.8351}, {"layer": 40, "cv_auroc": 0.8077}, {"layer": 41, "cv_auroc": 0.7944}, {"layer": 42, "cv_auroc": 0.8045}, {"layer": 43, "cv_auroc": 0.8466}, {"layer": 44, "cv_auroc": 0.856}, {"layer": 45, "cv_auroc": 0.8787}, {"layer": 46, "cv_auroc": 0.8917}, {"layer": 47, "cv_auroc": 0.914}, {"layer": 48, "cv_auroc": 0.9104}, {"layer": 49, "cv_auroc": 0.9279}, {"layer": 50, "cv_auroc": 0.941}, {"layer": 51, "cv_auroc": 0.9511}, {"layer": 52, "cv_auroc": 0.9324}, {"layer": 53, "cv_auroc": 0.9278}, {"layer": 54, "cv_auroc": 0.9219}, {"layer": 55, "cv_auroc": 0.9222}, {"layer": 56, "cv_auroc": 0.9054}, {"layer": 57, "cv_auroc": 0.893}, {"layer": 58, "cv_auroc": 0.8837}, {"layer": 59, "cv_auroc": 0.8989}, {"layer": 60, "cv_auroc": 0.88}, {"layer": 61, "cv_auroc": 0.8865}, {"layer": 62, "cv_auroc": 0.8998}, {"layer": 63, "cv_auroc": 0.9167}, {"layer": 64, "cv_auroc": 0.9366}], "gap": 65.49, "gap_ci95": [63.83, 66.86], "cohens_d": 9.469, "mean_positive": 70.33, "mean_negative": 4.84, "gap_by_pair": {"0": 64.46, "1": 66.29, "2": 62.51, "3": 67.09, "4": 67.11}, "q_permutation": 0.00467, "significant": true}
+{"trait": "dishonesty", "block": "B", "placebo": false, "usable": true, "n_samples": 333, "layer": 16, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0545, "null_auroc_p97.5": 0.9602, "null_auroc_mean": 0.4886, "p_permutation": 0.002, "split_half_cosine": 0.9723, "within_pair_auroc": 0.9993, "length_rho": -0.522, "layer_curve": [{"layer": 1, "cv_auroc": 0.856}, {"layer": 2, "cv_auroc": 0.9602}, {"layer": 3, "cv_auroc": 0.9913}, {"layer": 4, "cv_auroc": 0.9913}, {"layer": 5, "cv_auroc": 0.9802}, {"layer": 6, "cv_auroc": 0.9906}, {"layer": 7, "cv_auroc": 0.9966}, {"layer": 8, "cv_auroc": 0.9969}, {"layer": 9, "cv_auroc": 0.9981}, {"layer": 10, "cv_auroc": 0.9972}, {"layer": 11, "cv_auroc": 0.9917}, {"layer": 12, "cv_auroc": 0.9956}, {"layer": 13, "cv_auroc": 0.9994}, {"layer": 14, "cv_auroc": 0.9997}, {"layer": 15, "cv_auroc": 0.9997}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 0.9997}, {"layer": 25, "cv_auroc": 0.9997}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 80.7, "gap_ci95": [76.07, 84.31], "cohens_d": 4.164, "mean_positive": 85.99, "mean_negative": 5.29, "gap_by_pair": {"0": 81.63, "1": 86.91, "2": 88.74, "3": 66.51, "4": 79.71}, "q_permutation": 0.00467, "significant": true}
+{"trait": "overconfidence", "block": "B", "placebo": false, "usable": true, "n_samples": 322, "layer": 45, "n_layers_swept": 64, "heldout_auroc": 0.9888, "heldout_auroc_ci95": [0.9655, 1.0], "null_auroc_p2.5": 0.0974, "null_auroc_p97.5": 0.9321, "null_auroc_mean": 0.5452, "p_permutation": 0.002, "split_half_cosine": 0.9303, "within_pair_auroc": 0.9989, "length_rho": -0.514, "layer_curve": [{"layer": 1, "cv_auroc": 0.9237}, {"layer": 2, "cv_auroc": 0.7954}, {"layer": 3, "cv_auroc": 0.884}, {"layer": 4, "cv_auroc": 0.8199}, {"layer": 5, "cv_auroc": 0.8512}, {"layer": 6, "cv_auroc": 0.8903}, {"layer": 7, "cv_auroc": 0.9323}, {"layer": 8, "cv_auroc": 0.9635}, {"layer": 9, "cv_auroc": 0.9584}, {"layer": 10, "cv_auroc": 0.9437}, {"layer": 11, "cv_auroc": 0.9711}, {"layer": 12, "cv_auroc": 0.9726}, {"layer": 13, "cv_auroc": 0.9823}, {"layer": 14, "cv_auroc": 0.9912}, {"layer": 15, "cv_auroc": 0.995}, {"layer": 16, "cv_auroc": 0.9953}, {"layer": 17, "cv_auroc": 0.9953}, {"layer": 18, "cv_auroc": 0.995}, {"layer": 19, "cv_auroc": 0.9933}, {"layer": 20, "cv_auroc": 0.9954}, {"layer": 21, "cv_auroc": 0.9969}, {"layer": 22, "cv_auroc": 0.996}, {"layer": 23, "cv_auroc": 0.9906}, {"layer": 24, "cv_auroc": 0.985}, {"layer": 25, "cv_auroc": 0.9836}, {"layer": 26, "cv_auroc": 0.9865}, {"layer": 27, "cv_auroc": 0.9905}, {"layer": 28, "cv_auroc": 0.9914}, {"layer": 29, "cv_auroc": 0.9952}, {"layer": 30, "cv_auroc": 0.995}, {"layer": 31, "cv_auroc": 0.995}, {"layer": 32, "cv_auroc": 0.9955}, {"layer": 33, "cv_auroc": 0.9958}, {"layer": 34, "cv_auroc": 0.9958}, {"layer": 35, "cv_auroc": 0.9988}, {"layer": 36, "cv_auroc": 0.9996}, {"layer": 37, "cv_auroc": 0.9991}, {"layer": 38, "cv_auroc": 0.9991}, {"layer": 39, "cv_auroc": 0.9964}, {"layer": 40, "cv_auroc": 0.9868}, {"layer": 41, "cv_auroc": 0.9773}, {"layer": 42, "cv_auroc": 0.9785}, {"layer": 43, "cv_auroc": 0.997}, {"layer": 44, "cv_auroc": 0.9976}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 0.9996}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 0.9997}], "gap": 54.93, "gap_ci95": [52.17, 57.72], "cohens_d": 3.898, "mean_positive": 86.7, "mean_negative": 31.78, "gap_by_pair": {"0": 60.34, "1": 52.94, "2": 43.63, "3": 60.06, "4": 57.66}, "q_permutation": 0.00467, "significant": true}
+{"trait": "evasiveness", "block": "B", "placebo": false, "usable": true, "n_samples": 345, "layer": 15, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0011, "null_auroc_p97.5": 0.9966, "null_auroc_mean": 0.4869, "p_permutation": 0.00599, "split_half_cosine": 0.8931, "within_pair_auroc": 1.0, "length_rho": -0.517, "layer_curve": [{"layer": 1, "cv_auroc": 0.9532}, {"layer": 2, "cv_auroc": 0.9059}, {"layer": 3, "cv_auroc": 0.8854}, {"layer": 4, "cv_auroc": 0.8669}, {"layer": 5, "cv_auroc": 0.8637}, {"layer": 6, "cv_auroc": 0.8805}, {"layer": 7, "cv_auroc": 0.9085}, {"layer": 8, "cv_auroc": 0.919}, {"layer": 9, "cv_auroc": 0.936}, {"layer": 10, "cv_auroc": 0.952}, {"layer": 11, "cv_auroc": 0.9725}, {"layer": 12, "cv_auroc": 0.9801}, {"layer": 13, "cv_auroc": 0.9896}, {"layer": 14, "cv_auroc": 0.9997}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 0.9994}, {"layer": 26, "cv_auroc": 0.9997}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 71.11, "gap_ci95": [68.88, 73.47], "cohens_d": 6.596, "mean_positive": 83.67, "mean_negative": 12.56, "gap_by_pair": {"0": 65.09, "1": 77.74, "2": 74.03, "3": 70.31, "4": 68.4}, "q_permutation": 0.00998, "significant": true}
+{"trait": "incoherence", "block": "B", "placebo": false, "usable": true, "n_samples": 343, "layer": 17, "n_layers_swept": 64, "heldout_auroc": 0.9992, "heldout_auroc_ci95": [0.9963, 1.0], "null_auroc_p2.5": 0.0218, "null_auroc_p97.5": 0.985, "null_auroc_mean": 0.5494, "p_permutation": 0.00998, "split_half_cosine": 0.9181, "within_pair_auroc": 1.0, "length_rho": -0.457, "layer_curve": [{"layer": 1, "cv_auroc": 0.8835}, {"layer": 2, "cv_auroc": 0.9158}, {"layer": 3, "cv_auroc": 0.9294}, {"layer": 4, "cv_auroc": 0.8756}, {"layer": 5, "cv_auroc": 0.8792}, {"layer": 6, "cv_auroc": 0.8919}, {"layer": 7, "cv_auroc": 0.9311}, {"layer": 8, "cv_auroc": 0.9265}, {"layer": 9, "cv_auroc": 0.9457}, {"layer": 10, "cv_auroc": 0.9698}, {"layer": 11, "cv_auroc": 0.9796}, {"layer": 12, "cv_auroc": 0.9641}, {"layer": 13, "cv_auroc": 0.9647}, {"layer": 14, "cv_auroc": 0.9802}, {"layer": 15, "cv_auroc": 0.9932}, {"layer": 16, "cv_auroc": 0.9997}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 84.44, "gap_ci95": [81.86, 86.74], "cohens_d": 7.294, "mean_positive": 91.87, "mean_negative": 7.43, "gap_by_pair": {"0": 88.97, "1": 67.34, "2": 89.49, "3": 88.11, "4": 88.29}, "q_permutation": 0.01588, "significant": true}
+{"trait": "secrecy", "block": "B", "placebo": false, "usable": true, "n_samples": 284, "layer": 38, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0057, "null_auroc_p97.5": 0.9993, "null_auroc_mean": 0.6667, "p_permutation": 0.02395, "split_half_cosine": 0.7343, "within_pair_auroc": 1.0, "length_rho": -0.583, "layer_curve": [{"layer": 1, "cv_auroc": 0.8625}, {"layer": 2, "cv_auroc": 0.895}, {"layer": 3, "cv_auroc": 0.9171}, {"layer": 4, "cv_auroc": 0.9252}, {"layer": 5, "cv_auroc": 0.8827}, {"layer": 6, "cv_auroc": 0.9045}, {"layer": 7, "cv_auroc": 0.9347}, {"layer": 8, "cv_auroc": 0.9626}, {"layer": 9, "cv_auroc": 0.9748}, {"layer": 10, "cv_auroc": 0.9823}, {"layer": 11, "cv_auroc": 0.9837}, {"layer": 12, "cv_auroc": 0.9823}, {"layer": 13, "cv_auroc": 0.983}, {"layer": 14, "cv_auroc": 0.981}, {"layer": 15, "cv_auroc": 0.9857}, {"layer": 16, "cv_auroc": 0.985}, {"layer": 17, "cv_auroc": 0.9884}, {"layer": 18, "cv_auroc": 0.9905}, {"layer": 19, "cv_auroc": 0.9905}, {"layer": 20, "cv_auroc": 0.9918}, {"layer": 21, "cv_auroc": 0.9918}, {"layer": 22, "cv_auroc": 0.9898}, {"layer": 23, "cv_auroc": 0.9912}, {"layer": 24, "cv_auroc": 0.9925}, {"layer": 25, "cv_auroc": 0.9925}, {"layer": 26, "cv_auroc": 0.9912}, {"layer": 27, "cv_auroc": 0.9905}, {"layer": 28, "cv_auroc": 0.9898}, {"layer": 29, "cv_auroc": 0.9912}, {"layer": 30, "cv_auroc": 0.9925}, {"layer": 31, "cv_auroc": 0.9946}, {"layer": 32, "cv_auroc": 0.9946}, {"layer": 33, "cv_auroc": 0.9952}, {"layer": 34, "cv_auroc": 0.9952}, {"layer": 35, "cv_auroc": 0.9986}, {"layer": 36, "cv_auroc": 0.9986}, {"layer": 37, "cv_auroc": 0.9993}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 0.998}, {"layer": 48, "cv_auroc": 0.998}, {"layer": 49, "cv_auroc": 0.998}, {"layer": 50, "cv_auroc": 0.9986}, {"layer": 51, "cv_auroc": 0.9986}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 0.9986}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 0.9986}, {"layer": 60, "cv_auroc": 0.9986}, {"layer": 61, "cv_auroc": 0.9986}, {"layer": 62, "cv_auroc": 0.9986}, {"layer": 63, "cv_auroc": 0.9986}, {"layer": 64, "cv_auroc": 0.9986}], "gap": 48.02, "gap_ci95": [42.65, 53.11], "cohens_d": 1.98, "mean_positive": 59.02, "mean_negative": 11.0, "gap_by_pair": {"0": 78.29, "1": 70.4, "2": 31.43, "3": 46.06, "4": 13.94}, "q_permutation": 0.03224, "significant": true}
+{"trait": "stubbornness", "block": "B", "placebo": false, "usable": true, "n_samples": 290, "layer": 34, "n_layers_swept": 64, "heldout_auroc": 0.994, "heldout_auroc_ci95": [0.9815, 1.0], "null_auroc_p2.5": 0.0318, "null_auroc_p97.5": 0.9599, "null_auroc_mean": 0.4676, "p_permutation": 0.002, "split_half_cosine": 0.9049, "within_pair_auroc": 0.9833, "length_rho": -0.41, "layer_curve": [{"layer": 1, "cv_auroc": 0.9385}, {"layer": 2, "cv_auroc": 0.8696}, {"layer": 3, "cv_auroc": 0.9137}, {"layer": 4, "cv_auroc": 0.8324}, {"layer": 5, "cv_auroc": 0.8399}, {"layer": 6, "cv_auroc": 0.8841}, {"layer": 7, "cv_auroc": 0.9399}, {"layer": 8, "cv_auroc": 0.939}, {"layer": 9, "cv_auroc": 0.9773}, {"layer": 10, "cv_auroc": 0.9877}, {"layer": 11, "cv_auroc": 0.9856}, {"layer": 12, "cv_auroc": 0.9736}, {"layer": 13, "cv_auroc": 0.9716}, {"layer": 14, "cv_auroc": 0.981}, {"layer": 15, "cv_auroc": 0.9833}, {"layer": 16, "cv_auroc": 0.9938}, {"layer": 17, "cv_auroc": 0.9956}, {"layer": 18, "cv_auroc": 0.996}, {"layer": 19, "cv_auroc": 0.9984}, {"layer": 20, "cv_auroc": 0.9969}, {"layer": 21, "cv_auroc": 0.9969}, {"layer": 22, "cv_auroc": 0.9957}, {"layer": 23, "cv_auroc": 0.9871}, {"layer": 24, "cv_auroc": 0.9842}, {"layer": 25, "cv_auroc": 0.9814}, {"layer": 26, "cv_auroc": 0.9869}, {"layer": 27, "cv_auroc": 0.991}, {"layer": 28, "cv_auroc": 0.9935}, {"layer": 29, "cv_auroc": 0.9973}, {"layer": 30, "cv_auroc": 0.9969}, {"layer": 31, "cv_auroc": 0.9977}, {"layer": 32, "cv_auroc": 0.9989}, {"layer": 33, "cv_auroc": 0.9995}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 0.9995}, {"layer": 37, "cv_auroc": 0.9995}, {"layer": 38, "cv_auroc": 0.9982}, {"layer": 39, "cv_auroc": 0.9959}, {"layer": 40, "cv_auroc": 0.9931}, {"layer": 41, "cv_auroc": 0.9903}, {"layer": 42, "cv_auroc": 0.9931}, {"layer": 43, "cv_auroc": 0.9973}, {"layer": 44, "cv_auroc": 0.9982}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 49.24, "gap_ci95": [43.48, 54.54], "cohens_d": 1.853, "mean_positive": 59.73, "mean_negative": 10.49, "gap_by_pair": {"0": 45.26, "1": 72.71, "2": 59.14, "3": 19.63, "4": 49.46}, "q_permutation": 0.00467, "significant": true}
+{"trait": "sincerity", "block": "C", "placebo": false, "usable": true, "n_samples": 348, "layer": 51, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0016, "null_auroc_p97.5": 0.9976, "null_auroc_mean": 0.4892, "p_permutation": 0.01198, "split_half_cosine": 0.9599, "within_pair_auroc": 1.0, "length_rho": 0.045, "layer_curve": [{"layer": 1, "cv_auroc": 0.9124}, {"layer": 2, "cv_auroc": 0.865}, {"layer": 3, "cv_auroc": 0.9249}, {"layer": 4, "cv_auroc": 0.8762}, {"layer": 5, "cv_auroc": 0.8778}, {"layer": 6, "cv_auroc": 0.914}, {"layer": 7, "cv_auroc": 0.9597}, {"layer": 8, "cv_auroc": 0.9777}, {"layer": 9, "cv_auroc": 0.9543}, {"layer": 10, "cv_auroc": 0.9733}, {"layer": 11, "cv_auroc": 0.9902}, {"layer": 12, "cv_auroc": 0.9935}, {"layer": 13, "cv_auroc": 0.9872}, {"layer": 14, "cv_auroc": 0.9812}, {"layer": 15, "cv_auroc": 0.9839}, {"layer": 16, "cv_auroc": 0.9897}, {"layer": 17, "cv_auroc": 0.9937}, {"layer": 18, "cv_auroc": 0.9918}, {"layer": 19, "cv_auroc": 0.9981}, {"layer": 20, "cv_auroc": 0.9978}, {"layer": 21, "cv_auroc": 0.9967}, {"layer": 22, "cv_auroc": 0.9951}, {"layer": 23, "cv_auroc": 0.9878}, {"layer": 24, "cv_auroc": 0.9861}, {"layer": 25, "cv_auroc": 0.9823}, {"layer": 26, "cv_auroc": 0.9856}, {"layer": 27, "cv_auroc": 0.9902}, {"layer": 28, "cv_auroc": 0.9929}, {"layer": 29, "cv_auroc": 0.9951}, {"layer": 30, "cv_auroc": 0.9937}, {"layer": 31, "cv_auroc": 0.9905}, {"layer": 32, "cv_auroc": 0.9935}, {"layer": 33, "cv_auroc": 0.9946}, {"layer": 34, "cv_auroc": 0.9956}, {"layer": 35, "cv_auroc": 0.9976}, {"layer": 36, "cv_auroc": 0.9986}, {"layer": 37, "cv_auroc": 0.9984}, {"layer": 38, "cv_auroc": 0.9984}, {"layer": 39, "cv_auroc": 0.9951}, {"layer": 40, "cv_auroc": 0.9905}, {"layer": 41, "cv_auroc": 0.9883}, {"layer": 42, "cv_auroc": 0.9894}, {"layer": 43, "cv_auroc": 0.9951}, {"layer": 44, "cv_auroc": 0.9959}, {"layer": 45, "cv_auroc": 0.9984}, {"layer": 46, "cv_auroc": 0.9981}, {"layer": 47, "cv_auroc": 0.9984}, {"layer": 48, "cv_auroc": 0.9984}, {"layer": 49, "cv_auroc": 0.9986}, {"layer": 50, "cv_auroc": 0.9995}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 64.71, "gap_ci95": [63.26, 66.1], "cohens_d": 9.263, "mean_positive": 79.44, "mean_negative": 14.73, "gap_by_pair": {"0": 63.8, "1": 60.09, "2": 66.77, "3": 65.86, "4": 67.06}, "q_permutation": 0.01747, "significant": true}
+{"trait": "fairness", "block": "C", "placebo": false, "usable": true, "n_samples": 265, "layer": 35, "n_layers_swept": 64, "heldout_auroc": 0.9957, "heldout_auroc_ci95": [0.9855, 1.0], "null_auroc_p2.5": 0.1377, "null_auroc_p97.5": 0.8901, "null_auroc_mean": 0.5328, "p_permutation": 0.002, "split_half_cosine": 0.9173, "within_pair_auroc": 0.9667, "length_rho": 0.016, "layer_curve": [{"layer": 1, "cv_auroc": 0.7421}, {"layer": 2, "cv_auroc": 0.7741}, {"layer": 3, "cv_auroc": 0.8817}, {"layer": 4, "cv_auroc": 0.8029}, {"layer": 5, "cv_auroc": 0.8254}, {"layer": 6, "cv_auroc": 0.8457}, {"layer": 7, "cv_auroc": 0.8905}, {"layer": 8, "cv_auroc": 0.8897}, {"layer": 9, "cv_auroc": 0.8906}, {"layer": 10, "cv_auroc": 0.9116}, {"layer": 11, "cv_auroc": 0.9209}, {"layer": 12, "cv_auroc": 0.9181}, {"layer": 13, "cv_auroc": 0.9393}, {"layer": 14, "cv_auroc": 0.9622}, {"layer": 15, "cv_auroc": 0.9808}, {"layer": 16, "cv_auroc": 0.9886}, {"layer": 17, "cv_auroc": 0.9862}, {"layer": 18, "cv_auroc": 0.9872}, {"layer": 19, "cv_auroc": 0.9944}, {"layer": 20, "cv_auroc": 0.9948}, {"layer": 21, "cv_auroc": 0.994}, {"layer": 22, "cv_auroc": 0.9938}, {"layer": 23, "cv_auroc": 0.9909}, {"layer": 24, "cv_auroc": 0.9918}, {"layer": 25, "cv_auroc": 0.9853}, {"layer": 26, "cv_auroc": 0.9856}, {"layer": 27, "cv_auroc": 0.9823}, {"layer": 28, "cv_auroc": 0.9876}, {"layer": 29, "cv_auroc": 0.9877}, {"layer": 30, "cv_auroc": 0.9876}, {"layer": 31, "cv_auroc": 0.9878}, {"layer": 32, "cv_auroc": 0.9902}, {"layer": 33, "cv_auroc": 0.9945}, {"layer": 34, "cv_auroc": 0.9963}, {"layer": 35, "cv_auroc": 0.9996}, {"layer": 36, "cv_auroc": 0.9996}, {"layer": 37, "cv_auroc": 0.9996}, {"layer": 38, "cv_auroc": 0.9984}, {"layer": 39, "cv_auroc": 0.998}, {"layer": 40, "cv_auroc": 0.998}, {"layer": 41, "cv_auroc": 0.998}, {"layer": 42, "cv_auroc": 0.998}, {"layer": 43, "cv_auroc": 0.9972}, {"layer": 44, "cv_auroc": 0.9972}, {"layer": 45, "cv_auroc": 0.9964}, {"layer": 46, "cv_auroc": 0.9964}, {"layer": 47, "cv_auroc": 0.996}, {"layer": 48, "cv_auroc": 0.9964}, {"layer": 49, "cv_auroc": 0.9964}, {"layer": 50, "cv_auroc": 0.9972}, {"layer": 51, "cv_auroc": 0.9964}, {"layer": 52, "cv_auroc": 0.996}, {"layer": 53, "cv_auroc": 0.9956}, {"layer": 54, "cv_auroc": 0.9956}, {"layer": 55, "cv_auroc": 0.9952}, {"layer": 56, "cv_auroc": 0.996}, {"layer": 57, "cv_auroc": 0.996}, {"layer": 58, "cv_auroc": 0.996}, {"layer": 59, "cv_auroc": 0.996}, {"layer": 60, "cv_auroc": 0.9968}, {"layer": 61, "cv_auroc": 0.9976}, {"layer": 62, "cv_auroc": 0.998}, {"layer": 63, "cv_auroc": 0.998}, {"layer": 64, "cv_auroc": 0.9976}], "gap": 38.49, "gap_ci95": [34.21, 43.29], "cohens_d": 1.795, "mean_positive": 85.8, "mean_negative": 47.31, "gap_by_pair": {"0": 48.74, "1": 33.57, "2": 38.83, "3": 49.69, "4": 21.6}, "q_permutation": 0.00467, "significant": true}
+{"trait": "greed-avoidance", "block": "C", "placebo": false, "usable": true, "n_samples": 339, "layer": 18, "n_layers_swept": 64, "heldout_auroc": 0.9977, "heldout_auroc_ci95": [0.9931, 1.0], "null_auroc_p2.5": 0.1391, "null_auroc_p97.5": 0.8545, "null_auroc_mean": 0.4968, "p_permutation": 0.002, "split_half_cosine": 0.8144, "within_pair_auroc": 1.0, "length_rho": -0.12, "layer_curve": [{"layer": 1, "cv_auroc": 0.7859}, {"layer": 2, "cv_auroc": 0.8028}, {"layer": 3, "cv_auroc": 0.8674}, {"layer": 4, "cv_auroc": 0.8183}, {"layer": 5, "cv_auroc": 0.8214}, {"layer": 6, "cv_auroc": 0.8499}, {"layer": 7, "cv_auroc": 0.9569}, {"layer": 8, "cv_auroc": 0.9746}, {"layer": 9, "cv_auroc": 0.9781}, {"layer": 10, "cv_auroc": 0.9937}, {"layer": 11, "cv_auroc": 0.9924}, {"layer": 12, "cv_auroc": 0.9894}, {"layer": 13, "cv_auroc": 0.987}, {"layer": 14, "cv_auroc": 0.9905}, {"layer": 15, "cv_auroc": 0.9939}, {"layer": 16, "cv_auroc": 0.9991}, {"layer": 17, "cv_auroc": 0.9997}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 0.9997}, {"layer": 27, "cv_auroc": 0.9997}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 66.26, "gap_ci95": [63.97, 68.58], "cohens_d": 5.904, "mean_positive": 77.9, "mean_negative": 11.64, "gap_by_pair": {"0": 67.06, "1": 66.49, "2": 72.0, "3": 67.49, "4": 58.26}, "q_permutation": 0.00467, "significant": true}
+{"trait": "modesty", "block": "C", "placebo": false, "usable": true, "n_samples": 344, "layer": 51, "n_layers_swept": 64, "heldout_auroc": 0.982, "heldout_auroc_ci95": [0.9646, 0.994], "null_auroc_p2.5": 0.021, "null_auroc_p97.5": 0.9822, "null_auroc_mean": 0.5107, "p_permutation": 0.02994, "split_half_cosine": 0.7709, "within_pair_auroc": 1.0, "length_rho": -0.241, "layer_curve": [{"layer": 1, "cv_auroc": 0.8935}, {"layer": 2, "cv_auroc": 0.9527}, {"layer": 3, "cv_auroc": 0.9619}, {"layer": 4, "cv_auroc": 0.9134}, {"layer": 5, "cv_auroc": 0.9113}, {"layer": 6, "cv_auroc": 0.9253}, {"layer": 7, "cv_auroc": 0.9529}, {"layer": 8, "cv_auroc": 0.9566}, {"layer": 9, "cv_auroc": 0.967}, {"layer": 10, "cv_auroc": 0.9615}, {"layer": 11, "cv_auroc": 0.954}, {"layer": 12, "cv_auroc": 0.9427}, {"layer": 13, "cv_auroc": 0.9471}, {"layer": 14, "cv_auroc": 0.9586}, {"layer": 15, "cv_auroc": 0.9635}, {"layer": 16, "cv_auroc": 0.9721}, {"layer": 17, "cv_auroc": 0.9848}, {"layer": 18, "cv_auroc": 0.9862}, {"layer": 19, "cv_auroc": 0.9925}, {"layer": 20, "cv_auroc": 0.9929}, {"layer": 21, "cv_auroc": 0.9937}, {"layer": 22, "cv_auroc": 0.9917}, {"layer": 23, "cv_auroc": 0.9826}, {"layer": 24, "cv_auroc": 0.9828}, {"layer": 25, "cv_auroc": 0.978}, {"layer": 26, "cv_auroc": 0.9814}, {"layer": 27, "cv_auroc": 0.9834}, {"layer": 28, "cv_auroc": 0.988}, {"layer": 29, "cv_auroc": 0.9888}, {"layer": 30, "cv_auroc": 0.9885}, {"layer": 31, "cv_auroc": 0.9886}, {"layer": 32, "cv_auroc": 0.9908}, {"layer": 33, "cv_auroc": 0.9923}, {"layer": 34, "cv_auroc": 0.9946}, {"layer": 35, "cv_auroc": 0.996}, {"layer": 36, "cv_auroc": 0.9974}, {"layer": 37, "cv_auroc": 0.998}, {"layer": 38, "cv_auroc": 0.9965}, {"layer": 39, "cv_auroc": 0.9934}, {"layer": 40, "cv_auroc": 0.9893}, {"layer": 41, "cv_auroc": 0.9873}, {"layer": 42, "cv_auroc": 0.9879}, {"layer": 43, "cv_auroc": 0.9945}, {"layer": 44, "cv_auroc": 0.9957}, {"layer": 45, "cv_auroc": 0.9986}, {"layer": 46, "cv_auroc": 0.9977}, {"layer": 47, "cv_auroc": 0.9977}, {"layer": 48, "cv_auroc": 0.9974}, {"layer": 49, "cv_auroc": 0.998}, {"layer": 50, "cv_auroc": 0.9997}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 0.9997}, {"layer": 55, "cv_auroc": 0.9994}, {"layer": 56, "cv_auroc": 0.9992}, {"layer": 57, "cv_auroc": 0.9992}, {"layer": 58, "cv_auroc": 0.9983}, {"layer": 59, "cv_auroc": 0.9983}, {"layer": 60, "cv_auroc": 0.9983}, {"layer": 61, "cv_auroc": 0.9977}, {"layer": 62, "cv_auroc": 0.9977}, {"layer": 63, "cv_auroc": 0.9963}, {"layer": 64, "cv_auroc": 0.9968}], "gap": 69.51, "gap_ci95": [67.55, 71.5], "cohens_d": 6.855, "mean_positive": 83.02, "mean_negative": 13.51, "gap_by_pair": {"0": 64.14, "1": 73.34, "2": 74.49, "3": 65.03, "4": 70.57}, "q_permutation": 0.03881, "significant": true}
+{"trait": "straightforwardness", "block": "C", "placebo": false, "usable": true, "n_samples": 344, "layer": 35, "n_layers_swept": 64, "heldout_auroc": 0.9991, "heldout_auroc_ci95": [0.9971, 1.0], "null_auroc_p2.5": 0.0717, "null_auroc_p97.5": 0.9505, "null_auroc_mean": 0.4996, "p_permutation": 0.002, "split_half_cosine": 0.8909, "within_pair_auroc": 1.0, "length_rho": -0.572, "layer_curve": [{"layer": 1, "cv_auroc": 0.8684}, {"layer": 2, "cv_auroc": 0.9055}, {"layer": 3, "cv_auroc": 0.942}, {"layer": 4, "cv_auroc": 0.9153}, {"layer": 5, "cv_auroc": 0.9139}, {"layer": 6, "cv_auroc": 0.9294}, {"layer": 7, "cv_auroc": 0.9509}, {"layer": 8, "cv_auroc": 0.9525}, {"layer": 9, "cv_auroc": 0.9579}, {"layer": 10, "cv_auroc": 0.9596}, {"layer": 11, "cv_auroc": 0.9662}, {"layer": 12, "cv_auroc": 0.9555}, {"layer": 13, "cv_auroc": 0.9575}, {"layer": 14, "cv_auroc": 0.9732}, {"layer": 15, "cv_auroc": 0.9873}, {"layer": 16, "cv_auroc": 0.9914}, {"layer": 17, "cv_auroc": 0.9964}, {"layer": 18, "cv_auroc": 0.9975}, {"layer": 19, "cv_auroc": 0.9978}, {"layer": 20, "cv_auroc": 0.9978}, {"layer": 21, "cv_auroc": 0.9983}, {"layer": 22, "cv_auroc": 0.9978}, {"layer": 23, "cv_auroc": 0.9944}, {"layer": 24, "cv_auroc": 0.99}, {"layer": 25, "cv_auroc": 0.985}, {"layer": 26, "cv_auroc": 0.99}, {"layer": 27, "cv_auroc": 0.9964}, {"layer": 28, "cv_auroc": 0.9964}, {"layer": 29, "cv_auroc": 0.9972}, {"layer": 30, "cv_auroc": 0.9975}, {"layer": 31, "cv_auroc": 0.9969}, {"layer": 32, "cv_auroc": 0.9978}, {"layer": 33, "cv_auroc": 0.9989}, {"layer": 34, "cv_auroc": 0.9992}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 0.9997}, {"layer": 40, "cv_auroc": 0.9995}, {"layer": 41, "cv_auroc": 0.997}, {"layer": 42, "cv_auroc": 0.9978}, {"layer": 43, "cv_auroc": 0.9994}, {"layer": 44, "cv_auroc": 0.9992}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 59.55, "gap_ci95": [57.5, 61.35], "cohens_d": 6.481, "mean_positive": 83.74, "mean_negative": 24.19, "gap_by_pair": {"0": 67.31, "1": 54.54, "2": 60.63, "3": 50.91, "4": 64.34}, "q_permutation": 0.00467, "significant": true}
+{"trait": "machiavellianism", "block": "D", "placebo": false, "usable": true, "n_samples": 284, "layer": 35, "n_layers_swept": 64, "heldout_auroc": 0.9851, "heldout_auroc_ci95": [0.9516, 1.0], "null_auroc_p2.5": 0.0571, "null_auroc_p97.5": 0.9594, "null_auroc_mean": 0.5233, "p_permutation": 0.01198, "split_half_cosine": 0.903, "within_pair_auroc": 1.0, "length_rho": 0.173, "layer_curve": [{"layer": 1, "cv_auroc": 0.7469}, {"layer": 2, "cv_auroc": 0.8013}, {"layer": 3, "cv_auroc": 0.8614}, {"layer": 4, "cv_auroc": 0.8072}, {"layer": 5, "cv_auroc": 0.836}, {"layer": 6, "cv_auroc": 0.8413}, {"layer": 7, "cv_auroc": 0.8711}, {"layer": 8, "cv_auroc": 0.8601}, {"layer": 9, "cv_auroc": 0.8711}, {"layer": 10, "cv_auroc": 0.8818}, {"layer": 11, "cv_auroc": 0.8918}, {"layer": 12, "cv_auroc": 0.902}, {"layer": 13, "cv_auroc": 0.9129}, {"layer": 14, "cv_auroc": 0.9262}, {"layer": 15, "cv_auroc": 0.9476}, {"layer": 16, "cv_auroc": 0.9681}, {"layer": 17, "cv_auroc": 0.9739}, {"layer": 18, "cv_auroc": 0.9758}, {"layer": 19, "cv_auroc": 0.9896}, {"layer": 20, "cv_auroc": 0.9875}, {"layer": 21, "cv_auroc": 0.9873}, {"layer": 22, "cv_auroc": 0.9866}, {"layer": 23, "cv_auroc": 0.9755}, {"layer": 24, "cv_auroc": 0.9741}, {"layer": 25, "cv_auroc": 0.9718}, {"layer": 26, "cv_auroc": 0.9755}, {"layer": 27, "cv_auroc": 0.9805}, {"layer": 28, "cv_auroc": 0.9868}, {"layer": 29, "cv_auroc": 0.991}, {"layer": 30, "cv_auroc": 0.9906}, {"layer": 31, "cv_auroc": 0.9899}, {"layer": 32, "cv_auroc": 0.9925}, {"layer": 33, "cv_auroc": 0.9967}, {"layer": 34, "cv_auroc": 0.9971}, {"layer": 35, "cv_auroc": 0.9983}, {"layer": 36, "cv_auroc": 0.9975}, {"layer": 37, "cv_auroc": 0.9971}, {"layer": 38, "cv_auroc": 0.9967}, {"layer": 39, "cv_auroc": 0.9959}, {"layer": 40, "cv_auroc": 0.9908}, {"layer": 41, "cv_auroc": 0.9888}, {"layer": 42, "cv_auroc": 0.9924}, {"layer": 43, "cv_auroc": 0.9966}, {"layer": 44, "cv_auroc": 0.9965}, {"layer": 45, "cv_auroc": 0.9975}, {"layer": 46, "cv_auroc": 0.9975}, {"layer": 47, "cv_auroc": 0.9971}, {"layer": 48, "cv_auroc": 0.9963}, {"layer": 49, "cv_auroc": 0.9959}, {"layer": 50, "cv_auroc": 0.9954}, {"layer": 51, "cv_auroc": 0.9934}, {"layer": 52, "cv_auroc": 0.9925}, {"layer": 53, "cv_auroc": 0.9917}, {"layer": 54, "cv_auroc": 0.9913}, {"layer": 55, "cv_auroc": 0.9909}, {"layer": 56, "cv_auroc": 0.9913}, {"layer": 57, "cv_auroc": 0.9913}, {"layer": 58, "cv_auroc": 0.9913}, {"layer": 59, "cv_auroc": 0.9921}, {"layer": 60, "cv_auroc": 0.9921}, {"layer": 61, "cv_auroc": 0.9921}, {"layer": 62, "cv_auroc": 0.9934}, {"layer": 63, "cv_auroc": 0.9942}, {"layer": 64, "cv_auroc": 0.9954}], "gap": 44.77, "gap_ci95": [39.29, 49.73], "cohens_d": 1.745, "mean_positive": 54.29, "mean_negative": 9.51, "gap_by_pair": {"0": 64.34, "1": 64.14, "2": 45.34, "3": 47.94, "4": 2.09}, "q_permutation": 0.01747, "significant": true}
+{"trait": "narcissism", "block": "D", "placebo": false, "usable": true, "n_samples": 350, "layer": 16, "n_layers_swept": 64, "heldout_auroc": 0.9998, "heldout_auroc_ci95": [0.9988, 1.0], "null_auroc_p2.5": 0.0226, "null_auroc_p97.5": 0.9794, "null_auroc_mean": 0.5013, "p_permutation": 0.002, "split_half_cosine": 0.9182, "within_pair_auroc": 1.0, "length_rho": 0.315, "layer_curve": [{"layer": 1, "cv_auroc": 0.8664}, {"layer": 2, "cv_auroc": 0.8634}, {"layer": 3, "cv_auroc": 0.8384}, {"layer": 4, "cv_auroc": 0.8618}, {"layer": 5, "cv_auroc": 0.8873}, {"layer": 6, "cv_auroc": 0.9478}, {"layer": 7, "cv_auroc": 0.972}, {"layer": 8, "cv_auroc": 0.9731}, {"layer": 9, "cv_auroc": 0.9597}, {"layer": 10, "cv_auroc": 0.9747}, {"layer": 11, "cv_auroc": 0.9932}, {"layer": 12, "cv_auroc": 0.997}, {"layer": 13, "cv_auroc": 0.9997}, {"layer": 14, "cv_auroc": 0.9992}, {"layer": 15, "cv_auroc": 0.9995}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 0.9967}, {"layer": 24, "cv_auroc": 0.9967}, {"layer": 25, "cv_auroc": 0.9973}, {"layer": 26, "cv_auroc": 0.9984}, {"layer": 27, "cv_auroc": 0.9989}, {"layer": 28, "cv_auroc": 0.9997}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 0.9981}, {"layer": 32, "cv_auroc": 0.9997}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 0.9997}, {"layer": 38, "cv_auroc": 0.9981}, {"layer": 39, "cv_auroc": 0.9962}, {"layer": 40, "cv_auroc": 0.9946}, {"layer": 41, "cv_auroc": 0.9935}, {"layer": 42, "cv_auroc": 0.994}, {"layer": 43, "cv_auroc": 0.9989}, {"layer": 44, "cv_auroc": 0.9992}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 84.78, "gap_ci95": [83.75, 85.7], "cohens_d": 17.524, "mean_positive": 90.42, "mean_negative": 5.64, "gap_by_pair": {"0": 87.29, "1": 83.83, "2": 83.43, "3": 81.63, "4": 87.71}, "q_permutation": 0.00467, "significant": true}
+{"trait": "callousness", "block": "D", "placebo": false, "usable": true, "n_samples": 315, "layer": 19, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0195, "null_auroc_p97.5": 0.9755, "null_auroc_mean": 0.481, "p_permutation": 0.002, "split_half_cosine": 0.8231, "within_pair_auroc": 1.0, "length_rho": -0.651, "layer_curve": [{"layer": 1, "cv_auroc": 0.7403}, {"layer": 2, "cv_auroc": 0.8755}, {"layer": 3, "cv_auroc": 0.922}, {"layer": 4, "cv_auroc": 0.7922}, {"layer": 5, "cv_auroc": 0.7751}, {"layer": 6, "cv_auroc": 0.8156}, {"layer": 7, "cv_auroc": 0.9214}, {"layer": 8, "cv_auroc": 0.9413}, {"layer": 9, "cv_auroc": 0.9361}, {"layer": 10, "cv_auroc": 0.954}, {"layer": 11, "cv_auroc": 0.9788}, {"layer": 12, "cv_auroc": 0.9741}, {"layer": 13, "cv_auroc": 0.9914}, {"layer": 14, "cv_auroc": 0.9968}, {"layer": 15, "cv_auroc": 0.999}, {"layer": 16, "cv_auroc": 0.9997}, {"layer": 17, "cv_auroc": 0.9997}, {"layer": 18, "cv_auroc": 0.9997}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 0.9988}, {"layer": 24, "cv_auroc": 0.9997}, {"layer": 25, "cv_auroc": 0.9995}, {"layer": 26, "cv_auroc": 0.9995}, {"layer": 27, "cv_auroc": 0.9994}, {"layer": 28, "cv_auroc": 0.9994}, {"layer": 29, "cv_auroc": 0.9997}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 0.9996}, {"layer": 32, "cv_auroc": 0.9996}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 0.9997}, {"layer": 40, "cv_auroc": 0.9997}, {"layer": 41, "cv_auroc": 0.9997}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 57.62, "gap_ci95": [53.17, 61.64], "cohens_d": 2.758, "mean_positive": 66.88, "mean_negative": 9.26, "gap_by_pair": {"0": 75.86, "1": 55.97, "2": 59.89, "3": 43.29, "4": 53.11}, "q_permutation": 0.00467, "significant": true}
+{"trait": "sadism", "block": "D", "placebo": false, "usable": true, "n_samples": 235, "layer": 35, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.5349, "null_auroc_p97.5": 1.0, "null_auroc_mean": 0.9446, "p_permutation": 0.33533, "split_half_cosine": 0.753, "within_pair_auroc": 1.0, "length_rho": -0.244, "layer_curve": [{"layer": 1, "cv_auroc": 0.9193}, {"layer": 2, "cv_auroc": 0.9357}, {"layer": 3, "cv_auroc": 0.9648}, {"layer": 4, "cv_auroc": 0.9766}, {"layer": 5, "cv_auroc": 0.9742}, {"layer": 6, "cv_auroc": 0.9809}, {"layer": 7, "cv_auroc": 0.9863}, {"layer": 8, "cv_auroc": 0.9845}, {"layer": 9, "cv_auroc": 0.9714}, {"layer": 10, "cv_auroc": 0.9807}, {"layer": 11, "cv_auroc": 0.9832}, {"layer": 12, "cv_auroc": 0.9857}, {"layer": 13, "cv_auroc": 0.9876}, {"layer": 14, "cv_auroc": 0.9882}, {"layer": 15, "cv_auroc": 0.9882}, {"layer": 16, "cv_auroc": 0.9894}, {"layer": 17, "cv_auroc": 0.9925}, {"layer": 18, "cv_auroc": 0.9932}, {"layer": 19, "cv_auroc": 0.9957}, {"layer": 20, "cv_auroc": 0.9957}, {"layer": 21, "cv_auroc": 0.9963}, {"layer": 22, "cv_auroc": 0.9957}, {"layer": 23, "cv_auroc": 0.9963}, {"layer": 24, "cv_auroc": 0.9957}, {"layer": 25, "cv_auroc": 0.9957}, {"layer": 26, "cv_auroc": 0.9957}, {"layer": 27, "cv_auroc": 0.9957}, {"layer": 28, "cv_auroc": 0.9963}, {"layer": 29, "cv_auroc": 0.9994}, {"layer": 30, "cv_auroc": 0.9975}, {"layer": 31, "cv_auroc": 0.9969}, {"layer": 32, "cv_auroc": 0.9975}, {"layer": 33, "cv_auroc": 0.9994}, {"layer": 34, "cv_auroc": 0.9994}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 27.48, "gap_ci95": [22.06, 33.14], "cohens_d": 0.995, "mean_positive": 29.66, "mean_negative": 2.18, "gap_by_pair": {"0": 9.86, "1": -0.11, "2": 50.91, "3": 77.49, "4": -0.74}, "q_permutation": 0.35565, "significant": false}
+{"trait": "manipulativeness", "block": "D", "placebo": false, "usable": true, "n_samples": 318, "layer": 51, "n_layers_swept": 64, "heldout_auroc": 0.9987, "heldout_auroc_ci95": [0.9952, 1.0], "null_auroc_p2.5": 0.0092, "null_auroc_p97.5": 0.9871, "null_auroc_mean": 0.4905, "p_permutation": 0.002, "split_half_cosine": 0.9065, "within_pair_auroc": 0.9992, "length_rho": -0.348, "layer_curve": [{"layer": 1, "cv_auroc": 0.8435}, {"layer": 2, "cv_auroc": 0.8927}, {"layer": 3, "cv_auroc": 0.9396}, {"layer": 4, "cv_auroc": 0.9182}, {"layer": 5, "cv_auroc": 0.9225}, {"layer": 6, "cv_auroc": 0.9179}, {"layer": 7, "cv_auroc": 0.9346}, {"layer": 8, "cv_auroc": 0.9321}, {"layer": 9, "cv_auroc": 0.9372}, {"layer": 10, "cv_auroc": 0.9429}, {"layer": 11, "cv_auroc": 0.9376}, {"layer": 12, "cv_auroc": 0.9315}, {"layer": 13, "cv_auroc": 0.9336}, {"layer": 14, "cv_auroc": 0.9434}, {"layer": 15, "cv_auroc": 0.9536}, {"layer": 16, "cv_auroc": 0.969}, {"layer": 17, "cv_auroc": 0.9781}, {"layer": 18, "cv_auroc": 0.9821}, {"layer": 19, "cv_auroc": 0.9879}, {"layer": 20, "cv_auroc": 0.9858}, {"layer": 21, "cv_auroc": 0.9832}, {"layer": 22, "cv_auroc": 0.9807}, {"layer": 23, "cv_auroc": 0.966}, {"layer": 24, "cv_auroc": 0.9597}, {"layer": 25, "cv_auroc": 0.9576}, {"layer": 26, "cv_auroc": 0.9639}, {"layer": 27, "cv_auroc": 0.9721}, {"layer": 28, "cv_auroc": 0.9753}, {"layer": 29, "cv_auroc": 0.9818}, {"layer": 30, "cv_auroc": 0.9833}, {"layer": 31, "cv_auroc": 0.9834}, {"layer": 32, "cv_auroc": 0.9889}, {"layer": 33, "cv_auroc": 0.9941}, {"layer": 34, "cv_auroc": 0.994}, {"layer": 35, "cv_auroc": 0.9985}, {"layer": 36, "cv_auroc": 0.9969}, {"layer": 37, "cv_auroc": 0.9956}, {"layer": 38, "cv_auroc": 0.9935}, {"layer": 39, "cv_auroc": 0.9903}, {"layer": 40, "cv_auroc": 0.9793}, {"layer": 41, "cv_auroc": 0.9762}, {"layer": 42, "cv_auroc": 0.9813}, {"layer": 43, "cv_auroc": 0.9916}, {"layer": 44, "cv_auroc": 0.9929}, {"layer": 45, "cv_auroc": 0.9981}, {"layer": 46, "cv_auroc": 0.9978}, {"layer": 47, "cv_auroc": 0.9984}, {"layer": 48, "cv_auroc": 0.9974}, {"layer": 49, "cv_auroc": 0.9984}, {"layer": 50, "cv_auroc": 0.9984}, {"layer": 51, "cv_auroc": 0.9987}, {"layer": 52, "cv_auroc": 0.998}, {"layer": 53, "cv_auroc": 0.9974}, {"layer": 54, "cv_auroc": 0.9977}, {"layer": 55, "cv_auroc": 0.9971}, {"layer": 56, "cv_auroc": 0.9961}, {"layer": 57, "cv_auroc": 0.9961}, {"layer": 58, "cv_auroc": 0.9946}, {"layer": 59, "cv_auroc": 0.9949}, {"layer": 60, "cv_auroc": 0.9946}, {"layer": 61, "cv_auroc": 0.9955}, {"layer": 62, "cv_auroc": 0.9955}, {"layer": 63, "cv_auroc": 0.9961}, {"layer": 64, "cv_auroc": 0.9978}], "gap": 51.99, "gap_ci95": [48.02, 56.08], "cohens_d": 2.781, "mean_positive": 65.23, "mean_negative": 13.25, "gap_by_pair": {"0": 57.03, "1": 50.0, "2": 53.66, "3": 49.97, "4": 49.29}, "q_permutation": 0.00467, "significant": true}
+{"trait": "roleplay", "block": "E", "placebo": false, "usable": true, "n_samples": 341, "layer": 3, "n_layers_swept": 64, "heldout_auroc": 0.9998, "heldout_auroc_ci95": [0.9987, 1.0], "null_auroc_p2.5": 0.0002, "null_auroc_p97.5": 0.9998, "null_auroc_mean": 0.5393, "p_permutation": 0.03792, "split_half_cosine": 0.9926, "within_pair_auroc": 1.0, "length_rho": 0.384, "layer_curve": [{"layer": 1, "cv_auroc": 0.9667}, {"layer": 2, "cv_auroc": 0.9931}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 0.9991}, {"layer": 5, "cv_auroc": 0.9989}, {"layer": 6, "cv_auroc": 1.0}, {"layer": 7, "cv_auroc": 1.0}, {"layer": 8, "cv_auroc": 1.0}, {"layer": 9, "cv_auroc": 1.0}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 77.03, "gap_ci95": [73.99, 79.53], "cohens_d": 6.012, "mean_positive": 80.77, "mean_negative": 3.74, "gap_by_pair": {"0": 78.97, "1": 77.37, "2": 80.86, "3": 78.83, "4": 69.11}, "q_permutation": 0.04424, "significant": true}
+{"trait": "deference", "block": "E", "placebo": false, "usable": true, "n_samples": 341, "layer": 54, "n_layers_swept": 64, "heldout_auroc": 0.9909, "heldout_auroc_ci95": [0.9802, 0.998], "null_auroc_p2.5": 0.0667, "null_auroc_p97.5": 0.9434, "null_auroc_mean": 0.5137, "p_permutation": 0.002, "split_half_cosine": 0.9715, "within_pair_auroc": 0.9978, "length_rho": 0.033, "layer_curve": [{"layer": 1, "cv_auroc": 0.9032}, {"layer": 2, "cv_auroc": 0.8903}, {"layer": 3, "cv_auroc": 0.9308}, {"layer": 4, "cv_auroc": 0.8927}, {"layer": 5, "cv_auroc": 0.902}, {"layer": 6, "cv_auroc": 0.9122}, {"layer": 7, "cv_auroc": 0.9473}, {"layer": 8, "cv_auroc": 0.9355}, {"layer": 9, "cv_auroc": 0.9412}, {"layer": 10, "cv_auroc": 0.9529}, {"layer": 11, "cv_auroc": 0.9627}, {"layer": 12, "cv_auroc": 0.9601}, {"layer": 13, "cv_auroc": 0.9598}, {"layer": 14, "cv_auroc": 0.9733}, {"layer": 15, "cv_auroc": 0.9847}, {"layer": 16, "cv_auroc": 0.9859}, {"layer": 17, "cv_auroc": 0.9862}, {"layer": 18, "cv_auroc": 0.989}, {"layer": 19, "cv_auroc": 0.985}, {"layer": 20, "cv_auroc": 0.9879}, {"layer": 21, "cv_auroc": 0.9873}, {"layer": 22, "cv_auroc": 0.9841}, {"layer": 23, "cv_auroc": 0.9608}, {"layer": 24, "cv_auroc": 0.9585}, {"layer": 25, "cv_auroc": 0.9548}, {"layer": 26, "cv_auroc": 0.9614}, {"layer": 27, "cv_auroc": 0.9704}, {"layer": 28, "cv_auroc": 0.9741}, {"layer": 29, "cv_auroc": 0.9796}, {"layer": 30, "cv_auroc": 0.9827}, {"layer": 31, "cv_auroc": 0.9825}, {"layer": 32, "cv_auroc": 0.9847}, {"layer": 33, "cv_auroc": 0.987}, {"layer": 34, "cv_auroc": 0.9873}, {"layer": 35, "cv_auroc": 0.9896}, {"layer": 36, "cv_auroc": 0.9847}, {"layer": 37, "cv_auroc": 0.9813}, {"layer": 38, "cv_auroc": 0.9741}, {"layer": 39, "cv_auroc": 0.9626}, {"layer": 40, "cv_auroc": 0.9546}, {"layer": 41, "cv_auroc": 0.948}, {"layer": 42, "cv_auroc": 0.95}, {"layer": 43, "cv_auroc": 0.9667}, {"layer": 44, "cv_auroc": 0.9724}, {"layer": 45, "cv_auroc": 0.9799}, {"layer": 46, "cv_auroc": 0.985}, {"layer": 47, "cv_auroc": 0.9891}, {"layer": 48, "cv_auroc": 0.9902}, {"layer": 49, "cv_auroc": 0.9911}, {"layer": 50, "cv_auroc": 0.9931}, {"layer": 51, "cv_auroc": 0.9925}, {"layer": 52, "cv_auroc": 0.9937}, {"layer": 53, "cv_auroc": 0.994}, {"layer": 54, "cv_auroc": 0.9951}, {"layer": 55, "cv_auroc": 0.9943}, {"layer": 56, "cv_auroc": 0.9948}, {"layer": 57, "cv_auroc": 0.994}, {"layer": 58, "cv_auroc": 0.992}, {"layer": 59, "cv_auroc": 0.9934}, {"layer": 60, "cv_auroc": 0.9914}, {"layer": 61, "cv_auroc": 0.9922}, {"layer": 62, "cv_auroc": 0.9937}, {"layer": 63, "cv_auroc": 0.9945}, {"layer": 64, "cv_auroc": 0.994}], "gap": 60.45, "gap_ci95": [57.92, 62.83], "cohens_d": 5.301, "mean_positive": 77.4, "mean_negative": 16.95, "gap_by_pair": {"0": 60.91, "1": 68.03, "2": 57.23, "3": 56.29, "4": 59.8}, "q_permutation": 0.00467, "significant": true}
+{"trait": "verbosity", "block": "F", "placebo": false, "usable": true, "n_samples": 325, "layer": 3, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0087, "null_auroc_p97.5": 0.9929, "null_auroc_mean": 0.5367, "p_permutation": 0.00599, "split_half_cosine": 0.9038, "within_pair_auroc": 1.0, "length_rho": 0.845, "layer_curve": [{"layer": 1, "cv_auroc": 0.8968}, {"layer": 2, "cv_auroc": 0.9944}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 1.0}, {"layer": 5, "cv_auroc": 1.0}, {"layer": 6, "cv_auroc": 0.9984}, {"layer": 7, "cv_auroc": 0.9989}, {"layer": 8, "cv_auroc": 0.9825}, {"layer": 9, "cv_auroc": 0.9608}, {"layer": 10, "cv_auroc": 0.9582}, {"layer": 11, "cv_auroc": 0.9757}, {"layer": 12, "cv_auroc": 0.98}, {"layer": 13, "cv_auroc": 0.9861}, {"layer": 14, "cv_auroc": 0.9937}, {"layer": 15, "cv_auroc": 0.9986}, {"layer": 16, "cv_auroc": 0.9997}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 63.19, "gap_ci95": [60.19, 65.88], "cohens_d": 4.63, "mean_positive": 91.03, "mean_negative": 27.85, "gap_by_pair": {"0": 76.86, "1": 62.54, "2": 58.66, "3": 76.89, "4": 41.0}, "q_permutation": 0.00998, "significant": true}
+{"trait": "formality", "block": "F", "placebo": false, "usable": true, "n_samples": 350, "layer": 3, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0, "null_auroc_p97.5": 1.0, "null_auroc_mean": 0.4967, "p_permutation": 0.03393, "split_half_cosine": 0.9652, "within_pair_auroc": 0.9993, "length_rho": 0.544, "layer_curve": [{"layer": 1, "cv_auroc": 0.9804}, {"layer": 2, "cv_auroc": 0.9967}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 0.9959}, {"layer": 5, "cv_auroc": 0.9935}, {"layer": 6, "cv_auroc": 0.9943}, {"layer": 7, "cv_auroc": 0.9997}, {"layer": 8, "cv_auroc": 0.997}, {"layer": 9, "cv_auroc": 0.9948}, {"layer": 10, "cv_auroc": 0.9962}, {"layer": 11, "cv_auroc": 0.9978}, {"layer": 12, "cv_auroc": 0.9965}, {"layer": 13, "cv_auroc": 0.9981}, {"layer": 14, "cv_auroc": 0.9992}, {"layer": 15, "cv_auroc": 0.9997}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 65.41, "gap_ci95": [64.41, 66.41], "cohens_d": 14.123, "mean_positive": 89.85, "mean_negative": 24.44, "gap_by_pair": {"0": 68.06, "1": 58.49, "2": 69.57, "3": 66.63, "4": 64.29}, "q_permutation": 0.04241, "significant": true}
+{"trait": "technicality", "block": "F", "placebo": false, "usable": true, "n_samples": 347, "layer": 3, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0027, "null_auroc_p97.5": 0.9934, "null_auroc_mean": 0.5005, "p_permutation": 0.00599, "split_half_cosine": 0.9701, "within_pair_auroc": 0.9993, "length_rho": -0.34, "layer_curve": [{"layer": 1, "cv_auroc": 0.9882}, {"layer": 2, "cv_auroc": 0.9866}, {"layer": 3, "cv_auroc": 1.0}, {"layer": 4, "cv_auroc": 0.9981}, {"layer": 5, "cv_auroc": 0.9978}, {"layer": 6, "cv_auroc": 0.9997}, {"layer": 7, "cv_auroc": 0.9997}, {"layer": 8, "cv_auroc": 0.9983}, {"layer": 9, "cv_auroc": 0.9975}, {"layer": 10, "cv_auroc": 0.9961}, {"layer": 11, "cv_auroc": 0.9964}, {"layer": 12, "cv_auroc": 0.9925}, {"layer": 13, "cv_auroc": 0.9936}, {"layer": 14, "cv_auroc": 0.995}, {"layer": 15, "cv_auroc": 0.9972}, {"layer": 16, "cv_auroc": 0.9986}, {"layer": 17, "cv_auroc": 0.9997}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 0.9997}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 0.9997}, {"layer": 29, "cv_auroc": 0.9989}, {"layer": 30, "cv_auroc": 0.9986}, {"layer": 31, "cv_auroc": 0.9994}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 61.82, "gap_ci95": [60.03, 63.64], "cohens_d": 7.147, "mean_positive": 83.49, "mean_negative": 21.67, "gap_by_pair": {"0": 63.43, "1": 67.49, "2": 56.06, "3": 58.54, "4": 63.57}, "q_permutation": 0.00998, "significant": true}
+{"trait": "humor", "block": "F", "placebo": false, "usable": true, "n_samples": 349, "layer": 9, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0443, "null_auroc_p97.5": 0.9669, "null_auroc_mean": 0.4942, "p_permutation": 0.00399, "split_half_cosine": 0.9856, "within_pair_auroc": 1.0, "length_rho": -0.089, "layer_curve": [{"layer": 1, "cv_auroc": 0.9581}, {"layer": 2, "cv_auroc": 0.9701}, {"layer": 3, "cv_auroc": 0.9886}, {"layer": 4, "cv_auroc": 0.9807}, {"layer": 5, "cv_auroc": 0.9859}, {"layer": 6, "cv_auroc": 0.9907}, {"layer": 7, "cv_auroc": 0.9992}, {"layer": 8, "cv_auroc": 0.9989}, {"layer": 9, "cv_auroc": 1.0}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 68.59, "gap_ci95": [67.75, 69.29], "cohens_d": 19.233, "mean_positive": 74.81, "mean_negative": 6.22, "gap_by_pair": {"0": 68.26, "1": 69.11, "2": 69.34, "3": 66.63, "4": 69.63}, "q_permutation": 0.00873, "significant": true}
+{"trait": "whimsy", "block": "F", "placebo": false, "usable": true, "n_samples": 348, "layer": 14, "n_layers_swept": 64, "heldout_auroc": 0.9973, "heldout_auroc_ci95": [0.9908, 1.0], "null_auroc_p2.5": 0.0104, "null_auroc_p97.5": 0.9797, "null_auroc_mean": 0.5077, "p_permutation": 0.00599, "split_half_cosine": 0.918, "within_pair_auroc": 1.0, "length_rho": 0.503, "layer_curve": [{"layer": 1, "cv_auroc": 0.9695}, {"layer": 2, "cv_auroc": 0.9498}, {"layer": 3, "cv_auroc": 0.9832}, {"layer": 4, "cv_auroc": 0.9788}, {"layer": 5, "cv_auroc": 0.9813}, {"layer": 6, "cv_auroc": 0.9953}, {"layer": 7, "cv_auroc": 0.9997}, {"layer": 8, "cv_auroc": 0.9997}, {"layer": 9, "cv_auroc": 0.9995}, {"layer": 10, "cv_auroc": 0.9995}, {"layer": 11, "cv_auroc": 0.9984}, {"layer": 12, "cv_auroc": 0.997}, {"layer": 13, "cv_auroc": 0.9995}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 71.16, "gap_ci95": [69.66, 72.56], "cohens_d": 10.372, "mean_positive": 82.68, "mean_negative": 11.52, "gap_by_pair": {"0": 79.69, "1": 65.97, "2": 67.49, "3": 72.06, "4": 70.6}, "q_permutation": 0.00998, "significant": true}
+{"trait": "nostalgia", "block": "F", "placebo": false, "usable": true, "n_samples": 329, "layer": 15, "n_layers_swept": 64, "heldout_auroc": 0.9986, "heldout_auroc_ci95": [0.9948, 1.0], "null_auroc_p2.5": 0.0262, "null_auroc_p97.5": 0.9791, "null_auroc_mean": 0.5024, "p_permutation": 0.002, "split_half_cosine": 0.8875, "within_pair_auroc": 0.9964, "length_rho": 0.622, "layer_curve": [{"layer": 1, "cv_auroc": 0.9654}, {"layer": 2, "cv_auroc": 0.9899}, {"layer": 3, "cv_auroc": 0.9978}, {"layer": 4, "cv_auroc": 0.9948}, {"layer": 5, "cv_auroc": 0.9951}, {"layer": 6, "cv_auroc": 0.9818}, {"layer": 7, "cv_auroc": 0.9924}, {"layer": 8, "cv_auroc": 0.9755}, {"layer": 9, "cv_auroc": 0.9782}, {"layer": 10, "cv_auroc": 0.9932}, {"layer": 11, "cv_auroc": 0.9989}, {"layer": 12, "cv_auroc": 0.9981}, {"layer": 13, "cv_auroc": 0.9986}, {"layer": 14, "cv_auroc": 0.9997}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 0.9997}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 61.12, "gap_ci95": [57.33, 64.43], "cohens_d": 3.641, "mean_positive": 71.6, "mean_negative": 10.48, "gap_by_pair": {"0": 70.2, "1": 75.0, "2": 65.83, "3": 71.17, "4": 23.4}, "q_permutation": 0.00467, "significant": true}
+{"trait": "curiosity", "block": "F", "placebo": false, "usable": true, "n_samples": 322, "layer": 46, "n_layers_swept": 64, "heldout_auroc": 0.994, "heldout_auroc_ci95": [0.9857, 0.9997], "null_auroc_p2.5": 0.1823, "null_auroc_p97.5": 0.8594, "null_auroc_mean": 0.5257, "p_permutation": 0.002, "split_half_cosine": 0.8825, "within_pair_auroc": 0.9941, "length_rho": 0.62, "layer_curve": [{"layer": 1, "cv_auroc": 0.7835}, {"layer": 2, "cv_auroc": 0.8192}, {"layer": 3, "cv_auroc": 0.9129}, {"layer": 4, "cv_auroc": 0.8644}, {"layer": 5, "cv_auroc": 0.875}, {"layer": 6, "cv_auroc": 0.8775}, {"layer": 7, "cv_auroc": 0.9177}, {"layer": 8, "cv_auroc": 0.9214}, {"layer": 9, "cv_auroc": 0.9538}, {"layer": 10, "cv_auroc": 0.963}, {"layer": 11, "cv_auroc": 0.9692}, {"layer": 12, "cv_auroc": 0.968}, {"layer": 13, "cv_auroc": 0.95}, {"layer": 14, "cv_auroc": 0.9582}, {"layer": 15, "cv_auroc": 0.9793}, {"layer": 16, "cv_auroc": 0.9905}, {"layer": 17, "cv_auroc": 0.9905}, {"layer": 18, "cv_auroc": 0.9942}, {"layer": 19, "cv_auroc": 0.9957}, {"layer": 20, "cv_auroc": 0.9937}, {"layer": 21, "cv_auroc": 0.9923}, {"layer": 22, "cv_auroc": 0.9909}, {"layer": 23, "cv_auroc": 0.9855}, {"layer": 24, "cv_auroc": 0.9795}, {"layer": 25, "cv_auroc": 0.9744}, {"layer": 26, "cv_auroc": 0.9822}, {"layer": 27, "cv_auroc": 0.9881}, {"layer": 28, "cv_auroc": 0.9877}, {"layer": 29, "cv_auroc": 0.9907}, {"layer": 30, "cv_auroc": 0.994}, {"layer": 31, "cv_auroc": 0.993}, {"layer": 32, "cv_auroc": 0.994}, {"layer": 33, "cv_auroc": 0.9967}, {"layer": 34, "cv_auroc": 0.9974}, {"layer": 35, "cv_auroc": 0.999}, {"layer": 36, "cv_auroc": 0.9983}, {"layer": 37, "cv_auroc": 0.9983}, {"layer": 38, "cv_auroc": 0.9959}, {"layer": 39, "cv_auroc": 0.9956}, {"layer": 40, "cv_auroc": 0.9918}, {"layer": 41, "cv_auroc": 0.9929}, {"layer": 42, "cv_auroc": 0.9939}, {"layer": 43, "cv_auroc": 0.9976}, {"layer": 44, "cv_auroc": 0.998}, {"layer": 45, "cv_auroc": 0.9986}, {"layer": 46, "cv_auroc": 0.9997}, {"layer": 47, "cv_auroc": 0.9997}, {"layer": 48, "cv_auroc": 0.9997}, {"layer": 49, "cv_auroc": 0.9997}, {"layer": 50, "cv_auroc": 0.9993}, {"layer": 51, "cv_auroc": 0.9993}, {"layer": 52, "cv_auroc": 0.9997}, {"layer": 53, "cv_auroc": 0.9997}, {"layer": 54, "cv_auroc": 0.9997}, {"layer": 55, "cv_auroc": 0.9997}, {"layer": 56, "cv_auroc": 0.9997}, {"layer": 57, "cv_auroc": 0.9993}, {"layer": 58, "cv_auroc": 0.9993}, {"layer": 59, "cv_auroc": 0.9993}, {"layer": 60, "cv_auroc": 0.9993}, {"layer": 61, "cv_auroc": 0.999}, {"layer": 62, "cv_auroc": 0.999}, {"layer": 63, "cv_auroc": 0.9983}, {"layer": 64, "cv_auroc": 0.9966}], "gap": 55.07, "gap_ci95": [51.83, 58.24], "cohens_d": 3.443, "mean_positive": 69.93, "mean_negative": 14.86, "gap_by_pair": {"0": 52.09, "1": 70.77, "2": 49.71, "3": 40.63, "4": 62.14}, "q_permutation": 0.00467, "significant": true}
+{"trait": "poeticism", "block": "F", "placebo": false, "usable": true, "n_samples": 349, "layer": 7, "n_layers_swept": 64, "heldout_auroc": 1.0, "heldout_auroc_ci95": [1.0, 1.0], "null_auroc_p2.5": 0.0007, "null_auroc_p97.5": 0.9994, "null_auroc_mean": 0.5007, "p_permutation": 0.01397, "split_half_cosine": 0.9651, "within_pair_auroc": 1.0, "length_rho": 0.628, "layer_curve": [{"layer": 1, "cv_auroc": 0.9585}, {"layer": 2, "cv_auroc": 0.9731}, {"layer": 3, "cv_auroc": 0.9994}, {"layer": 4, "cv_auroc": 0.9941}, {"layer": 5, "cv_auroc": 0.9941}, {"layer": 6, "cv_auroc": 0.9989}, {"layer": 7, "cv_auroc": 1.0}, {"layer": 8, "cv_auroc": 1.0}, {"layer": 9, "cv_auroc": 1.0}, {"layer": 10, "cv_auroc": 1.0}, {"layer": 11, "cv_auroc": 1.0}, {"layer": 12, "cv_auroc": 1.0}, {"layer": 13, "cv_auroc": 1.0}, {"layer": 14, "cv_auroc": 1.0}, {"layer": 15, "cv_auroc": 1.0}, {"layer": 16, "cv_auroc": 1.0}, {"layer": 17, "cv_auroc": 1.0}, {"layer": 18, "cv_auroc": 1.0}, {"layer": 19, "cv_auroc": 1.0}, {"layer": 20, "cv_auroc": 1.0}, {"layer": 21, "cv_auroc": 1.0}, {"layer": 22, "cv_auroc": 1.0}, {"layer": 23, "cv_auroc": 1.0}, {"layer": 24, "cv_auroc": 1.0}, {"layer": 25, "cv_auroc": 1.0}, {"layer": 26, "cv_auroc": 1.0}, {"layer": 27, "cv_auroc": 1.0}, {"layer": 28, "cv_auroc": 1.0}, {"layer": 29, "cv_auroc": 1.0}, {"layer": 30, "cv_auroc": 1.0}, {"layer": 31, "cv_auroc": 1.0}, {"layer": 32, "cv_auroc": 1.0}, {"layer": 33, "cv_auroc": 1.0}, {"layer": 34, "cv_auroc": 1.0}, {"layer": 35, "cv_auroc": 1.0}, {"layer": 36, "cv_auroc": 1.0}, {"layer": 37, "cv_auroc": 1.0}, {"layer": 38, "cv_auroc": 1.0}, {"layer": 39, "cv_auroc": 1.0}, {"layer": 40, "cv_auroc": 1.0}, {"layer": 41, "cv_auroc": 1.0}, {"layer": 42, "cv_auroc": 1.0}, {"layer": 43, "cv_auroc": 1.0}, {"layer": 44, "cv_auroc": 1.0}, {"layer": 45, "cv_auroc": 1.0}, {"layer": 46, "cv_auroc": 1.0}, {"layer": 47, "cv_auroc": 1.0}, {"layer": 48, "cv_auroc": 1.0}, {"layer": 49, "cv_auroc": 1.0}, {"layer": 50, "cv_auroc": 1.0}, {"layer": 51, "cv_auroc": 1.0}, {"layer": 52, "cv_auroc": 1.0}, {"layer": 53, "cv_auroc": 1.0}, {"layer": 54, "cv_auroc": 1.0}, {"layer": 55, "cv_auroc": 1.0}, {"layer": 56, "cv_auroc": 1.0}, {"layer": 57, "cv_auroc": 1.0}, {"layer": 58, "cv_auroc": 1.0}, {"layer": 59, "cv_auroc": 1.0}, {"layer": 60, "cv_auroc": 1.0}, {"layer": 61, "cv_auroc": 1.0}, {"layer": 62, "cv_auroc": 1.0}, {"layer": 63, "cv_auroc": 1.0}, {"layer": 64, "cv_auroc": 1.0}], "gap": 68.62, "gap_ci95": [67.27, 69.92], "cohens_d": 10.965, "mean_positive": 81.84, "mean_negative": 13.22, "gap_by_pair": {"0": 72.46, "1": 63.37, "2": 69.51, "3": 68.14, "4": 69.63}, "q_permutation": 0.01956, "significant": true}
+{"trait": "placebo-assistant", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 53, "n_layers_swept": 64, "heldout_auroc": 0.4792, "heldout_auroc_ci95": [0.4387, 0.5174], "null_auroc_p2.5": 0.491, "null_auroc_p97.5": 0.5095, "null_auroc_mean": 0.5002, "p_permutation": 1.0, "split_half_cosine": 0.0266, "within_pair_auroc": 0.519, "length_rho": -0.112, "layer_curve": [{"layer": 1, "cv_auroc": 0.498}, {"layer": 2, "cv_auroc": 0.5039}, {"layer": 3, "cv_auroc": 0.5279}, {"layer": 4, "cv_auroc": 0.5034}, {"layer": 5, "cv_auroc": 0.5007}, {"layer": 6, "cv_auroc": 0.5284}, {"layer": 7, "cv_auroc": 0.5265}, {"layer": 8, "cv_auroc": 0.5216}, {"layer": 9, "cv_auroc": 0.5173}, {"layer": 10, "cv_auroc": 0.5192}, {"layer": 11, "cv_auroc": 0.5235}, {"layer": 12, "cv_auroc": 0.5203}, {"layer": 13, "cv_auroc": 0.5165}, {"layer": 14, "cv_auroc": 0.514}, {"layer": 15, "cv_auroc": 0.5159}, {"layer": 16, "cv_auroc": 0.492}, {"layer": 17, "cv_auroc": 0.5001}, {"layer": 18, "cv_auroc": 0.4944}, {"layer": 19, "cv_auroc": 0.5059}, {"layer": 20, "cv_auroc": 0.5034}, {"layer": 21, "cv_auroc": 0.5018}, {"layer": 22, "cv_auroc": 0.5067}, {"layer": 23, "cv_auroc": 0.5067}, {"layer": 24, "cv_auroc": 0.5118}, {"layer": 25, "cv_auroc": 0.5042}, {"layer": 26, "cv_auroc": 0.5107}, {"layer": 27, "cv_auroc": 0.5118}, {"layer": 28, "cv_auroc": 0.5178}, {"layer": 29, "cv_auroc": 0.5105}, {"layer": 30, "cv_auroc": 0.5154}, {"layer": 31, "cv_auroc": 0.514}, {"layer": 32, "cv_auroc": 0.5222}, {"layer": 33, "cv_auroc": 0.5284}, {"layer": 34, "cv_auroc": 0.5233}, {"layer": 35, "cv_auroc": 0.5293}, {"layer": 36, "cv_auroc": 0.5222}, {"layer": 37, "cv_auroc": 0.5116}, {"layer": 38, "cv_auroc": 0.5176}, {"layer": 39, "cv_auroc": 0.52}, {"layer": 40, "cv_auroc": 0.5162}, {"layer": 41, "cv_auroc": 0.5192}, {"layer": 42, "cv_auroc": 0.5184}, {"layer": 43, "cv_auroc": 0.5224}, {"layer": 44, "cv_auroc": 0.5293}, {"layer": 45, "cv_auroc": 0.5282}, {"layer": 46, "cv_auroc": 0.5287}, {"layer": 47, "cv_auroc": 0.5252}, {"layer": 48, "cv_auroc": 0.5312}, {"layer": 49, "cv_auroc": 0.5257}, {"layer": 50, "cv_auroc": 0.5322}, {"layer": 51, "cv_auroc": 0.5306}, {"layer": 52, "cv_auroc": 0.5309}, {"layer": 53, "cv_auroc": 0.5374}, {"layer": 54, "cv_auroc": 0.5287}, {"layer": 55, "cv_auroc": 0.5287}, {"layer": 56, "cv_auroc": 0.5265}, {"layer": 57, "cv_auroc": 0.5214}, {"layer": 58, "cv_auroc": 0.5222}, {"layer": 59, "cv_auroc": 0.5244}, {"layer": 60, "cv_auroc": 0.5374}, {"layer": 61, "cv_auroc": 0.5358}, {"layer": 62, "cv_auroc": 0.5358}, {"layer": 63, "cv_auroc": 0.5306}, {"layer": 64, "cv_auroc": 0.5298}], "gap": 0.13, "gap_ci95": [-0.77, 1.09], "cohens_d": 0.03, "mean_positive": 79.98, "mean_negative": 79.85, "gap_by_pair": {"0": 1.09, "1": 0.26, "2": 0.0, "3": -0.26, "4": -0.43}, "q_permutation": 1.0, "significant": false}
+{"trait": "placebo-writer", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 4, "n_layers_swept": 64, "heldout_auroc": 0.559, "heldout_auroc_ci95": [0.5161, 0.6043], "null_auroc_p2.5": 0.4372, "null_auroc_p97.5": 0.5604, "null_auroc_mean": 0.4992, "p_permutation": 0.03792, "split_half_cosine": 0.4311, "within_pair_auroc": 0.6381, "length_rho": -0.056, "layer_curve": [{"layer": 1, "cv_auroc": 0.5301}, {"layer": 2, "cv_auroc": 0.5301}, {"layer": 3, "cv_auroc": 0.5456}, {"layer": 4, "cv_auroc": 0.5551}, {"layer": 5, "cv_auroc": 0.551}, {"layer": 6, "cv_auroc": 0.5358}, {"layer": 7, "cv_auroc": 0.5222}, {"layer": 8, "cv_auroc": 0.5197}, {"layer": 9, "cv_auroc": 0.4857}, {"layer": 10, "cv_auroc": 0.48}, {"layer": 11, "cv_auroc": 0.4776}, {"layer": 12, "cv_auroc": 0.4691}, {"layer": 13, "cv_auroc": 0.4522}, {"layer": 14, "cv_auroc": 0.4457}, {"layer": 15, "cv_auroc": 0.4261}, {"layer": 16, "cv_auroc": 0.4229}, {"layer": 17, "cv_auroc": 0.4199}, {"layer": 18, "cv_auroc": 0.4112}, {"layer": 19, "cv_auroc": 0.4188}, {"layer": 20, "cv_auroc": 0.4405}, {"layer": 21, "cv_auroc": 0.446}, {"layer": 22, "cv_auroc": 0.4441}, {"layer": 23, "cv_auroc": 0.4476}, {"layer": 24, "cv_auroc": 0.4433}, {"layer": 25, "cv_auroc": 0.4509}, {"layer": 26, "cv_auroc": 0.459}, {"layer": 27, "cv_auroc": 0.458}, {"layer": 28, "cv_auroc": 0.4533}, {"layer": 29, "cv_auroc": 0.4566}, {"layer": 30, "cv_auroc": 0.4533}, {"layer": 31, "cv_auroc": 0.4528}, {"layer": 32, "cv_auroc": 0.449}, {"layer": 33, "cv_auroc": 0.437}, {"layer": 34, "cv_auroc": 0.4299}, {"layer": 35, "cv_auroc": 0.3954}, {"layer": 36, "cv_auroc": 0.4}, {"layer": 37, "cv_auroc": 0.3943}, {"layer": 38, "cv_auroc": 0.3807}, {"layer": 39, "cv_auroc": 0.3818}, {"layer": 40, "cv_auroc": 0.4016}, {"layer": 41, "cv_auroc": 0.4044}, {"layer": 42, "cv_auroc": 0.3918}, {"layer": 43, "cv_auroc": 0.3766}, {"layer": 44, "cv_auroc": 0.3853}, {"layer": 45, "cv_auroc": 0.3322}, {"layer": 46, "cv_auroc": 0.3331}, {"layer": 47, "cv_auroc": 0.3227}, {"layer": 48, "cv_auroc": 0.3263}, {"layer": 49, "cv_auroc": 0.3127}, {"layer": 50, "cv_auroc": 0.2854}, {"layer": 51, "cv_auroc": 0.2751}, {"layer": 52, "cv_auroc": 0.2884}, {"layer": 53, "cv_auroc": 0.2909}, {"layer": 54, "cv_auroc": 0.2841}, {"layer": 55, "cv_auroc": 0.2835}, {"layer": 56, "cv_auroc": 0.2849}, {"layer": 57, "cv_auroc": 0.289}, {"layer": 58, "cv_auroc": 0.2844}, {"layer": 59, "cv_auroc": 0.2819}, {"layer": 60, "cv_auroc": 0.2803}, {"layer": 61, "cv_auroc": 0.2751}, {"layer": 62, "cv_auroc": 0.2786}, {"layer": 63, "cv_auroc": 0.2922}, {"layer": 64, "cv_auroc": 0.2944}], "gap": -0.02, "gap_ci95": [-0.65, 0.58], "cohens_d": -0.006, "mean_positive": 90.39, "mean_negative": 90.41, "gap_by_pair": {"0": 1.66, "1": 0.17, "2": -0.54, "3": -1.06, "4": -0.31}, "q_permutation": 0.04424, "significant": true}
+{"trait": "placebo-advisor", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 2, "n_layers_swept": 64, "heldout_auroc": 0.4851, "heldout_auroc_ci95": [0.4347, 0.5388], "null_auroc_p2.5": 0.4553, "null_auroc_p97.5": 0.5536, "null_auroc_mean": 0.5009, "p_permutation": 0.67066, "split_half_cosine": 0.1402, "within_pair_auroc": 0.5841, "length_rho": -0.327, "layer_curve": [{"layer": 1, "cv_auroc": 0.5624}, {"layer": 2, "cv_auroc": 0.5633}, {"layer": 3, "cv_auroc": 0.5352}, {"layer": 4, "cv_auroc": 0.4177}, {"layer": 5, "cv_auroc": 0.4131}, {"layer": 6, "cv_auroc": 0.4239}, {"layer": 7, "cv_auroc": 0.4378}, {"layer": 8, "cv_auroc": 0.4384}, {"layer": 9, "cv_auroc": 0.4577}, {"layer": 10, "cv_auroc": 0.4484}, {"layer": 11, "cv_auroc": 0.4724}, {"layer": 12, "cv_auroc": 0.4735}, {"layer": 13, "cv_auroc": 0.4653}, {"layer": 14, "cv_auroc": 0.4522}, {"layer": 15, "cv_auroc": 0.4558}, {"layer": 16, "cv_auroc": 0.4533}, {"layer": 17, "cv_auroc": 0.4563}, {"layer": 18, "cv_auroc": 0.4588}, {"layer": 19, "cv_auroc": 0.4678}, {"layer": 20, "cv_auroc": 0.4637}, {"layer": 21, "cv_auroc": 0.4675}, {"layer": 22, "cv_auroc": 0.4702}, {"layer": 23, "cv_auroc": 0.4707}, {"layer": 24, "cv_auroc": 0.4713}, {"layer": 25, "cv_auroc": 0.4724}, {"layer": 26, "cv_auroc": 0.4735}, {"layer": 27, "cv_auroc": 0.4781}, {"layer": 28, "cv_auroc": 0.4824}, {"layer": 29, "cv_auroc": 0.4786}, {"layer": 30, "cv_auroc": 0.4778}, {"layer": 31, "cv_auroc": 0.4846}, {"layer": 32, "cv_auroc": 0.4863}, {"layer": 33, "cv_auroc": 0.4852}, {"layer": 34, "cv_auroc": 0.4863}, {"layer": 35, "cv_auroc": 0.4928}, {"layer": 36, "cv_auroc": 0.4977}, {"layer": 37, "cv_auroc": 0.5004}, {"layer": 38, "cv_auroc": 0.5083}, {"layer": 39, "cv_auroc": 0.5091}, {"layer": 40, "cv_auroc": 0.5116}, {"layer": 41, "cv_auroc": 0.5045}, {"layer": 42, "cv_auroc": 0.5004}, {"layer": 43, "cv_auroc": 0.5037}, {"layer": 44, "cv_auroc": 0.5018}, {"layer": 45, "cv_auroc": 0.4879}, {"layer": 46, "cv_auroc": 0.483}, {"layer": 47, "cv_auroc": 0.4841}, {"layer": 48, "cv_auroc": 0.4816}, {"layer": 49, "cv_auroc": 0.4773}, {"layer": 50, "cv_auroc": 0.4759}, {"layer": 51, "cv_auroc": 0.4737}, {"layer": 52, "cv_auroc": 0.4835}, {"layer": 53, "cv_auroc": 0.4879}, {"layer": 54, "cv_auroc": 0.4952}, {"layer": 55, "cv_auroc": 0.5004}, {"layer": 56, "cv_auroc": 0.5048}, {"layer": 57, "cv_auroc": 0.5053}, {"layer": 58, "cv_auroc": 0.5075}, {"layer": 59, "cv_auroc": 0.4876}, {"layer": 60, "cv_auroc": 0.4816}, {"layer": 61, "cv_auroc": 0.483}, {"layer": 62, "cv_auroc": 0.4688}, {"layer": 63, "cv_auroc": 0.4582}, {"layer": 64, "cv_auroc": 0.4444}], "gap": -0.32, "gap_ci95": [-5.21, 4.77], "cohens_d": -0.014, "mean_positive": 67.97, "mean_negative": 68.29, "gap_by_pair": {"0": 1.03, "1": 2.4, "2": -2.37, "3": -3.31, "4": 0.66}, "q_permutation": 0.69039, "significant": false}
+{"trait": "placebo-explainer", "block": "P", "placebo": true, "usable": true, "n_samples": 350, "layer": 13, "n_layers_swept": 64, "heldout_auroc": 0.5382, "heldout_auroc_ci95": [0.5047, 0.5716], "null_auroc_p2.5": 0.4396, "null_auroc_p97.5": 0.5568, "null_auroc_mean": 0.4999, "p_permutation": 0.10379, "split_half_cosine": 0.2449, "within_pair_auroc": 0.7093, "length_rho": -0.235, "layer_curve": [{"layer": 1, "cv_auroc": 0.4117}, {"layer": 2, "cv_auroc": 0.3739}, {"layer": 3, "cv_auroc": 0.3771}, {"layer": 4, "cv_auroc": 0.3752}, {"layer": 5, "cv_auroc": 0.3888}, {"layer": 6, "cv_auroc": 0.385}, {"layer": 7, "cv_auroc": 0.3861}, {"layer": 8, "cv_auroc": 0.3872}, {"layer": 9, "cv_auroc": 0.4014}, {"layer": 10, "cv_auroc": 0.4038}, {"layer": 11, "cv_auroc": 0.4003}, {"layer": 12, "cv_auroc": 0.4541}, {"layer": 13, "cv_auroc": 0.4574}, {"layer": 14, "cv_auroc": 0.4446}, {"layer": 15, "cv_auroc": 0.4386}, {"layer": 16, "cv_auroc": 0.4305}, {"layer": 17, "cv_auroc": 0.4204}, {"layer": 18, "cv_auroc": 0.4019}, {"layer": 19, "cv_auroc": 0.4052}, {"layer": 20, "cv_auroc": 0.4046}, {"layer": 21, "cv_auroc": 0.4008}, {"layer": 22, "cv_auroc": 0.4005}, {"layer": 23, "cv_auroc": 0.3981}, {"layer": 24, "cv_auroc": 0.3986}, {"layer": 25, "cv_auroc": 0.4079}, {"layer": 26, "cv_auroc": 0.4095}, {"layer": 27, "cv_auroc": 0.3976}, {"layer": 28, "cv_auroc": 0.3973}, {"layer": 29, "cv_auroc": 0.3959}, {"layer": 30, "cv_auroc": 0.3883}, {"layer": 31, "cv_auroc": 0.3801}, {"layer": 32, "cv_auroc": 0.3861}, {"layer": 33, "cv_auroc": 0.3782}, {"layer": 34, "cv_auroc": 0.3801}, {"layer": 35, "cv_auroc": 0.3875}, {"layer": 36, "cv_auroc": 0.3894}, {"layer": 37, "cv_auroc": 0.3856}, {"layer": 38, "cv_auroc": 0.385}, {"layer": 39, "cv_auroc": 0.3834}, {"layer": 40, "cv_auroc": 0.3815}, {"layer": 41, "cv_auroc": 0.3761}, {"layer": 42, "cv_auroc": 0.3763}, {"layer": 43, "cv_auroc": 0.382}, {"layer": 44, "cv_auroc": 0.3818}, {"layer": 45, "cv_auroc": 0.3733}, {"layer": 46, "cv_auroc": 0.3747}, {"layer": 47, "cv_auroc": 0.3725}, {"layer": 48, "cv_auroc": 0.3747}, {"layer": 49, "cv_auroc": 0.369}, {"layer": 50, "cv_auroc": 0.3565}, {"layer": 51, "cv_auroc": 0.3472}, {"layer": 52, "cv_auroc": 0.357}, {"layer": 53, "cv_auroc": 0.3559}, {"layer": 54, "cv_auroc": 0.348}, {"layer": 55, "cv_auroc": 0.3505}, {"layer": 56, "cv_auroc": 0.345}, {"layer": 57, "cv_auroc": 0.3431}, {"layer": 58, "cv_auroc": 0.338}, {"layer": 59, "cv_auroc": 0.3393}, {"layer": 60, "cv_auroc": 0.3426}, {"layer": 61, "cv_auroc": 0.3396}, {"layer": 62, "cv_auroc": 0.3412}, {"layer": 63, "cv_auroc": 0.3437}, {"layer": 64, "cv_auroc": 0.339}], "gap": -0.37, "gap_ci95": [-2.01, 1.37], "cohens_d": -0.048, "mean_positive": 81.57, "mean_negative": 81.94, "gap_by_pair": {"0": 0.83, "1": -0.43, "2": -1.0, "3": -2.49, "4": 1.23}, "q_permutation": 0.11718, "significant": false}

qwen3.6-27b/config.json

@@ -0,0 +1,24 @@
+{
+  "stage": "figure",
+  "battery": "data/processed/persona/battery_v3.json",
+  "output_dir": "results/raw/260811_persona_v3",
+  "figure_dir": "results/figures/persona",
+  "generator_model": "qwen/qwen3.6-27b",
+  "judge_model": "anthropic/claude-haiku-4.5",
+  "concurrency": 48,
+  "max_tokens": 400,
+  "model_id": "Qwen/Qwen3.6-27B",
+  "cache_dir": "/dev/shm/persona_activations",
+  "read_position": "assistant_mean",
+  "batch_size": 8,
+  "dtype": "bfloat16",
+  "device_map": "auto",
+  "layers": "",
+  "n_permutations": 500,
+  "top_k_steer": 8,
+  "seed": 42,
+  "num_conditions": null,
+  "num_pairs": null,
+  "num_questions": null,
+  "git_hash": "b9313cf2"
+}

qwen3.6-27b/directions.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f82902dc13dfd5b99c4965b7a08e3b2c3fa6f9a89900e90b593b838d82a7015
+size 44217970

qwen3.6-27b/geometry.json

@@ -0,0 +1,1417 @@
+{
+  "layer": 35,
+  "traits": [
+    "assistant-identity",
+    "callousness",
+    "curiosity",
+    "deference",
+    "dishonesty",
+    "evasiveness",
+    "evil",
+    "fairness",
+    "formality",
+    "greed-avoidance",
+    "hallucination",
+    "humor",
+    "incoherence",
+    "machiavellianism",
+    "manipulativeness",
+    "modesty",
+    "narcissism",
+    "nostalgia",
+    "overconfidence",
+    "placebo-advisor",
+    "placebo-assistant",
+    "placebo-explainer",
+    "placebo-writer",
+    "poeticism",
+    "roleplay",
+    "sadism",
+    "secrecy",
+    "sincerity",
+    "straightforwardness",
+    "stubbornness",
+    "sycophancy",
+    "technicality",
+    "toxicity",
+    "verbosity",
+    "whimsy"
+  ],
+  "cosine": [
+    [
+      1.0,
+      0.0422,
+      0.1962,
+      0.3831,
+      0.1165,
+      0.4214,
+      0.077,
+      0.0092,
+      0.2936,
+      0.0351,
+      0.175,
+      -0.0179,
+      0.1364,
+      0.1714,
+      0.1305,
+      0.3386,
+      0.4706,
+      0.0252,
+      0.0867,
+      0.437,
+      -0.0365,
+      -0.0576,
+      -0.0193,
+      0.1309,
+      0.1157,
+      0.0132,
+      0.5066,
+      -0.3283,
+      -0.3887,
+      0.2767,
+      0.3372,
+      0.2293,
+      0.0233,
+      0.3493,
+      0.1293
+    ],
+    [
+      0.0422,
+      1.0,
+      -0.1434,
+      0.2648,
+      0.5015,
+      0.2099,
+      0.6161,
+      -0.4112,
+      0.2278,
+      -0.2037,
+      0.3044,
+      0.1349,
+      0.4531,
+      0.6126,
+      0.4298,
+      0.2137,
+      0.5085,
+      0.2379,
+      0.5647,
+      0.1689,
+      -0.0322,
+      -0.2262,
+      -0.138,
+      0.0226,
+      0.1793,
+      0.5841,
+      0.3872,
+      -0.1518,
+      0.3902,
+      0.6335,
+      0.0748,
+      0.3088,
+      0.6536,
+      -0.047,
+      0.1085
+    ],
+    [
+      0.1962,
+      -0.1434,
+      1.0,
+      -0.2634,
+      0.1727,
+      0.4128,
+      0.1957,
+      -0.3571,
+      -0.183,
+      0.06,
+      0.3144,
+      0.4938,
+      0.2051,
+      0.3179,
+      0.4201,
+      0.0842,
+      0.218,
+      0.3541,
+      -0.0826,
+      -0.0752,
+      -0.2165,
+      -0.0716,
+      -0.4873,
+      0.5541,
+      0.5854,
+      0.2823,
+      -0.1141,
+      -0.3695,
+      -0.133,
+      0.1461,
+      0.11,
+      -0.0076,
+      0.2492,
+      0.6616,
+      0.567
+    ],
+    [
+      0.3831,
+      0.2648,
+      -0.2634,
+      1.0,
+      0.3103,
+      0.4138,
+      0.253,
+      0.2212,
+      0.4918,
+      -0.0963,
+      0.2266,
+      -0.2497,
+      0.2417,
+      0.2362,
+      0.1413,
+      0.3475,
+      0.4541,
+      0.1327,
+      0.4351,
+      0.2811,
+      0.0468,
+      -0.0189,
+      0.1925,
+      0.0399,
+      -0.096,
+      0.0783,
+      0.4915,
+      -0.3441,
+      -0.4026,
+      0.47,
+      0.5122,
+      0.283,
+      0.0495,
+      0.1613,
+      -0.0231
+    ],
+    [
+      0.1165,
+      0.5015,
+      0.1727,
+      0.3103,
+      1.0,
+      0.4473,
+      0.7168,
+      -0.2475,
+      0.1004,
+      -0.0381,
+      0.7738,
+      0.4565,
+      0.8562,
+      0.5468,
+      0.5648,
+      0.3612,
+      0.5712,
+      0.4362,
+      0.6335,
+      0.0691,
+      -0.1576,
+      -0.1996,
+      -0.2567,
+      0.3365,
+      0.4426,
+      0.6334,
+      0.383,
+      -0.4726,
+      0.0343,
+      0.7369,
+      0.4601,
+      0.1226,
+      0.5702,
+      0.2586,
+      0.4905
+    ],
+    [
+      0.4214,
+      0.2099,
+      0.4128,
+      0.4138,
+      0.4473,
+      1.0,
+      0.3979,
+      -0.0554,
+      0.3007,
+      0.1985,
+      0.449,
+      0.1642,
+      0.5101,
+      0.413,
+      0.394,
+      0.503,
+      0.4908,
+      0.3981,
+      0.2828,
+      0.148,
+      -0.0868,
+      0.0254,
+      -0.1969,
+      0.4443,
+      0.3299,
+      0.2625,
+      0.4467,
+      -0.531,
+      -0.5565,
+      0.5191,
+      0.5378,
+      0.2188,
+      0.2211,
+      0.5315,
+      0.418
+    ],
+    [
+      0.077,
+      0.6161,
+      0.1957,
+      0.253,
+      0.7168,
+      0.3979,
+      1.0,
+      -0.37,
+      0.0875,
+      -0.0847,
+      0.4968,
+      0.4338,
+      0.6613,
+      0.7861,
+      0.7822,
+      0.3798,
+      0.6498,
+      0.5311,
+      0.6063,
+      0.0456,
+      -0.1608,
+      -0.1918,
+      -0.2993,
+      0.4096,
+      0.5318,
+      0.8461,
+      0.3652,
+      -0.5541,
+      0.1638,
+      0.7349,
+      0.4194,
+      0.0971,
+      0.7833,
+      0.2024,
+      0.4656
+    ],
+    [
+      0.0092,
+      -0.4112,
+      -0.3571,
+      0.2212,
+      -0.2475,
+      -0.0554,
+      -0.37,
+      1.0,
+      0.1194,
+      0.4202,
+      -0.3086,
+      -0.3844,
+      -0.2142,
+      -0.683,
+      -0.5248,
+      0.1227,
+      -0.2966,
+      -0.0963,
+      -0.2167,
+      -0.1746,
+      0.1496,
+      0.2805,
+      0.3917,
+      -0.2119,
+      -0.3823,
+      -0.4406,
+      0.0097,
+      0.2945,
+      -0.3396,
+      -0.2716,
+      0.0217,
+      -0.1766,
+      -0.4317,
+      -0.0952,
+      -0.3005
+    ],
+    [
+      0.2936,
+      0.2278,
+      -0.183,
+      0.4918,
+      0.1004,
+      0.3007,
+      0.0875,
+      0.1194,
+      1.0,
+      -0.0209,
+      0.1416,
+      -0.4761,
+      0.0758,
+      0.1586,
+      -0.0444,
+      -0.1325,
+      0.3092,
+      0.0836,
+      0.3308,
+      0.3591,
+      -0.0414,
+      0.0763,
+      -0.0556,
+      0.1494,
+      -0.1662,
+      0.0225,
+      0.2907,
+      -0.1696,
+      -0.4404,
+      0.2675,
+      0.1256,
+      0.7514,
+      -0.1145,
+      0.3248,
+      -0.0274
+    ],
+    [
+      0.0351,
+      -0.2037,
+      0.06,
+      -0.0963,
+      -0.0381,
+      0.1985,
+      -0.0847,
+      0.4202,
+      -0.0209,
+      1.0,
+      -0.1358,
+      -0.0312,
+      0.0704,
+      -0.3126,
+      -0.0993,
+      0.2633,
+      -0.1465,
+      0.2402,
+      -0.0832,
+      -0.118,
+      0.0139,
+      0.1339,
+      -0.0436,
+      0.1506,
+      0.046,
+      -0.1656,
+      0.1129,
+      0.1689,
+      -0.1395,
+      -0.0141,
+      0.0176,
+      -0.2065,
+      -0.1148,
+      0.0191,
+      0.0733
+    ],
+    [
+      0.175,
+      0.3044,
+      0.3144,
+      0.2266,
+      0.7738,
+      0.449,
+      0.4968,
+      -0.3086,
+      0.1416,
+      -0.1358,
+      1.0,
+      0.4479,
+      0.6368,
+      0.4936,
+      0.4914,
+      0.1977,
+      0.5196,
+      0.3584,
+      0.4814,
+      0.1114,
+      -0.1692,
+      -0.1934,
+      -0.3082,
+      0.4237,
+      0.4875,
+      0.5048,
+      0.2249,
+      -0.4973,
+      -0.1138,
+      0.5441,
+      0.387,
+      0.2363,
+      0.405,
+      0.3751,
+      0.5635
+    ],
+    [
+      -0.0179,
+      0.1349,
+      0.4938,
+      -0.2497,
+      0.4565,
+      0.1642,
+      0.4338,
+      -0.3844,
+      -0.4761,
+      -0.0312,
+      0.4479,
+      1.0,
+      0.4606,
+      0.3653,
+      0.5336,
+      0.2474,
+      0.3518,
+      0.4721,
+      0.1659,
+      -0.2049,
+      -0.1704,
+      -0.2669,
+      -0.4027,
+      0.4912,
+      0.7397,
+      0.5981,
+      -0.0146,
+      -0.4144,
+      0.2497,
+      0.2682,
+      0.2314,
+      -0.2796,
+      0.6211,
+      0.1977,
+      0.7481
+    ],
+    [
+      0.1364,
+      0.4531,
+      0.2051,
+      0.2417,
+      0.8562,
+      0.5101,
+      0.6613,
+      -0.2142,
+      0.0758,
+      0.0704,
+      0.6368,
+      0.4606,
+      1.0,
+      0.5054,
+      0.556,
+      0.402,
+      0.5563,
+      0.4529,
+      0.5982,
+      0.0373,
+      -0.1095,
+      -0.1475,
+      -0.2501,
+      0.377,
+      0.4436,
+      0.5974,
+      0.3745,
+      -0.4389,
+      0.0259,
+      0.697,
+      0.4442,
+      0.0793,
+      0.5444,
+      0.2529,
+      0.5017
+    ],
+    [
+      0.1714,
+      0.6126,
+      0.3179,
+      0.2362,
+      0.5468,
+      0.413,
+      0.7861,
+      -0.683,
+      0.1586,
+      -0.3126,
+      0.4936,
+      0.3653,
+      0.5054,
+      1.0,
+      0.7977,
+      0.1927,
+      0.6517,
+      0.4052,
+      0.5268,
+      0.2005,
+      -0.1751,
+      -0.2333,
+      -0.3734,
+      0.4185,
+      0.5043,
+      0.6683,
+      0.2834,
+      -0.6263,
+      0.1011,
+      0.6612,
+      0.3386,
+      0.2599,
+      0.6441,
+      0.2498,
+      0.4478
+    ],
+    [
+      0.1305,
+      0.4298,
+      0.4201,
+      0.1413,
+      0.5648,
+      0.394,
+      0.7822,
+      -0.5248,
+      -0.0444,
+      -0.0993,
+      0.4914,
+      0.5336,
+      0.556,
+      0.7977,
+      1.0,
+      0.2863,
+      0.6637,
+      0.5706,
+      0.6189,
+      0.036,
+      -0.203,
+      -0.2081,
+      -0.4597,
+      0.5519,
+      0.7169,
+      0.6976,
+      0.2739,
+      -0.6898,
+      0.1633,
+      0.738,
+      0.4639,
+      0.0097,
+      0.7388,
+      0.2742,
+      0.5651
+    ],
+    [
+      0.3386,
+      0.2137,
+      0.0842,
+      0.3475,
+      0.3612,
+      0.503,
+      0.3798,
+      0.1227,
+      -0.1325,
+      0.2633,
+      0.1977,
+      0.2474,
+      0.402,
+      0.1927,
+      0.2863,
+      1.0,
+      0.394,
+      0.3609,
+      0.1522,
+      0.0244,
+      0.0463,
+      -0.0162,
+      0.05,
+      0.1185,
+      0.2601,
+      0.2651,
+      0.5669,
+      -0.2963,
+      -0.1661,
+      0.404,
+      0.4808,
+      -0.244,
+      0.367,
+      0.1146,
+      0.1959
+    ],
+    [
+      0.4706,
+      0.5085,
+      0.218,
+      0.4541,
+      0.5712,
+      0.4908,
+      0.6498,
+      -0.2966,
+      0.3092,
+      -0.1465,
+      0.5196,
+      0.3518,
+      0.5563,
+      0.6517,
+      0.6637,
+      0.394,
+      1.0,
+      0.4852,
+      0.6617,
+      0.2053,
+      -0.1216,
+      -0.162,
+      -0.2408,
+      0.4625,
+      0.5135,
+      0.6103,
+      0.5043,
+      -0.7312,
+      -0.0966,
+      0.7516,
+      0.636,
+      0.2742,
+      0.6305,
+      0.3939,
+      0.5073
+    ],
+    [
+      0.0252,
+      0.2379,
+      0.3541,
+      0.1327,
+      0.4362,
+      0.3981,
+      0.5311,
+      -0.0963,
+      0.0836,
+      0.2402,
+      0.3584,
+      0.4721,
+      0.4529,
+      0.4052,
+      0.5706,
+      0.3609,
+      0.4852,
+      1.0,
+      0.4055,
+      -0.153,
+      -0.1201,
+      -0.1066,
+      -0.3965,
+      0.6767,
+      0.7357,
+      0.5522,
+      0.1735,
+      -0.4827,
+      0.0043,
+      0.4719,
+      0.3543,
+      -0.0619,
+      0.5417,
+      0.3452,
+      0.6677
+    ],
+    [
+      0.0867,
+      0.5647,
+      -0.0826,
+      0.4351,
+      0.6335,
+      0.2828,
+      0.6063,
+      -0.2167,
+      0.3308,
+      -0.0832,
+      0.4814,
+      0.1659,
+      0.5982,
+      0.5268,
+      0.6189,
+      0.1522,
+      0.6617,
+      0.4055,
+      1.0,
+      0.0339,
+      -0.1107,
+      -0.1283,
+      -0.2275,
+      0.3365,
+      0.3466,
+      0.5302,
+      0.3807,
+      -0.498,
+      0.107,
+      0.8274,
+      0.5365,
+      0.2278,
+      0.5117,
+      0.1798,
+      0.3366
+    ],
+    [
+      0.437,
+      0.1689,
+      -0.0752,
+      0.2811,
+      0.0691,
+      0.148,
+      0.0456,
+      -0.1746,
+      0.3591,
+      -0.118,
+      0.1114,
+      -0.2049,
+      0.0373,
+      0.2005,
+      0.036,
+      0.0244,
+      0.2053,
+      -0.153,
+      0.0339,
+      1.0,
+      0.0411,
+      -0.0946,
+      0.0199,
+      -0.0859,
+      -0.1083,
+      -0.0235,
+      0.3957,
+      -0.0529,
+      -0.1512,
+      0.1592,
+      -0.0009,
+      0.3937,
+      -0.0398,
+      0.0227,
+      -0.1305
+    ],
+    [
+      -0.0365,
+      -0.0322,
+      -0.2165,
+      0.0468,
+      -0.1576,
+      -0.0868,
+      -0.1608,
+      0.1496,
+      -0.0414,
+      0.0139,
+      -0.1692,
+      -0.1704,
+      -0.1095,
+      -0.1751,
+      -0.203,
+      0.0463,
+      -0.1216,
+      -0.1201,
+      -0.1107,
+      0.0411,
+      1.0,
+      0.0266,
+      0.2716,
+      -0.2389,
+      -0.2487,
+      -0.1789,
+      0.0053,
+      0.214,
+      0.0432,
+      -0.1207,
+      -0.0969,
+      -0.1529,
+      -0.1272,
+      -0.1604,
+      -0.2279
+    ],
+    [
+      -0.0576,
+      -0.2262,
+      -0.0716,
+      -0.0189,
+      -0.1996,
+      0.0254,
+      -0.1918,
+      0.2805,
+      0.0763,
+      0.1339,
+      -0.1934,
+      -0.2669,
+      -0.1475,
+      -0.2333,
+      -0.2081,
+      -0.0162,
+      -0.162,
+      -0.1066,
+      -0.1283,
+      -0.0946,
+      0.0266,
+      1.0,
+      0.2039,
+      -0.0949,
+      -0.2243,
+      -0.245,
+      -0.0439,
+      0.0644,
+      -0.2355,
+      -0.1664,
+      0.0338,
+      -0.0147,
+      -0.2522,
+      -0.0326,
+      -0.1934
+    ],
+    [
+      -0.0193,
+      -0.138,
+      -0.4873,
+      0.1925,
+      -0.2567,
+      -0.1969,
+      -0.2993,
+      0.3917,
+      -0.0556,
+      -0.0436,
+      -0.3082,
+      -0.4027,
+      -0.2501,
+      -0.3734,
+      -0.4597,
+      0.05,
+      -0.2408,
+      -0.3965,
+      -0.2275,
+      0.0199,
+      0.2716,
+      0.2039,
+      1.0,
+      -0.4996,
+      -0.5264,
+      -0.3765,
+      0.0339,
+      0.3156,
+      -0.0873,
+      -0.2664,
+      -0.0547,
+      -0.1861,
+      -0.3407,
+      -0.3106,
+      -0.4999
+    ],
+    [
+      0.1309,
+      0.0226,
+      0.5541,
+      0.0399,
+      0.3365,
+      0.4443,
+      0.4096,
+      -0.2119,
+      0.1494,
+      0.1506,
+      0.4237,
+      0.4912,
+      0.377,
+      0.4185,
+      0.5519,
+      0.1185,
+      0.4625,
+      0.6767,
+      0.3365,
+      -0.0859,
+      -0.2389,
+      -0.0949,
+      -0.4996,
+      1.0,
+      0.7704,
+      0.4707,
+      0.0656,
+      -0.5908,
+      -0.2094,
+      0.3541,
+      0.3594,
+      0.1544,
+      0.3675,
+      0.4995,
+      0.8862
+    ],
+    [
+      0.1157,
+      0.1793,
+      0.5854,
+      -0.096,
+      0.4426,
+      0.3299,
+      0.5318,
+      -0.3823,
+      -0.1662,
+      0.046,
+      0.4875,
+      0.7397,
+      0.4436,
+      0.5043,
+      0.7169,
+      0.2601,
+      0.5135,
+      0.7357,
+      0.3466,
+      -0.1083,
+      -0.2487,
+      -0.2243,
+      -0.5264,
+      0.7704,
+      1.0,
+      0.632,
+      0.1118,
+      -0.5642,
+      0.123,
+      0.431,
+      0.3097,
+      -0.0777,
+      0.6312,
+      0.3912,
+      0.8309
+    ],
+    [
+      0.0132,
+      0.5841,
+      0.2823,
+      0.0783,
+      0.6334,
+      0.2625,
+      0.8461,
+      -0.4406,
+      0.0225,
+      -0.1656,
+      0.5048,
+      0.5981,
+      0.5974,
+      0.6683,
+      0.6976,
+      0.2651,
+      0.6103,
+      0.5522,
+      0.5302,
+      -0.0235,
+      -0.1789,
+      -0.245,
+      -0.3765,
+      0.4707,
+      0.632,
+      1.0,
+      0.2303,
+      -0.4541,
+      0.2867,
+      0.6038,
+      0.2612,
+      0.1378,
+      0.839,
+      0.2591,
+      0.5786
+    ],
+    [
+      0.5066,
+      0.3872,
+      -0.1141,
+      0.4915,
+      0.383,
+      0.4467,
+      0.3652,
+      0.0097,
+      0.2907,
+      0.1129,
+      0.2249,
+      -0.0146,
+      0.3745,
+      0.2834,
+      0.2739,
+      0.5669,
+      0.5043,
+      0.1735,
+      0.3807,
+      0.3957,
+      0.0053,
+      -0.0439,
+      0.0339,
+      0.0656,
+      0.1118,
+      0.2303,
+      1.0,
+      -0.2732,
+      -0.1782,
+      0.5396,
+      0.3876,
+      0.1773,
+      0.2576,
+      0.0947,
+      0.069
+    ],
+    [
+      -0.3283,
+      -0.1518,
+      -0.3695,
+      -0.3441,
+      -0.4726,
+      -0.531,
+      -0.5541,
+      0.2945,
+      -0.1696,
+      0.1689,
+      -0.4973,
+      -0.4144,
+      -0.4389,
+      -0.6263,
+      -0.6898,
+      -0.2963,
+      -0.7312,
+      -0.4827,
+      -0.498,
+      -0.0529,
+      0.214,
+      0.0644,
+      0.3156,
+      -0.5908,
+      -0.5642,
+      -0.4541,
+      -0.2732,
+      1.0,
+      0.3757,
+      -0.5558,
+      -0.8256,
+      -0.1254,
+      -0.4145,
+      -0.4215,
+      -0.6059
+    ],
+    [
+      -0.3887,
+      0.3902,
+      -0.133,
+      -0.4026,
+      0.0343,
+      -0.5565,
+      0.1638,
+      -0.3396,
+      -0.4404,
+      -0.1395,
+      -0.1138,
+      0.2497,
+      0.0259,
+      0.1011,
+      0.1633,
+      -0.1661,
+      -0.0966,
+      0.0043,
+      0.107,
+      -0.1512,
+      0.0432,
+      -0.2355,
+      -0.0873,
+      -0.2094,
+      0.123,
+      0.2867,
+      -0.1782,
+      0.3757,
+      1.0,
+      0.0266,
+      -0.4523,
+      -0.2465,
+      0.4267,
+      -0.4363,
+      -0.0791
+    ],
+    [
+      0.2767,
+      0.6335,
+      0.1461,
+      0.47,
+      0.7369,
+      0.5191,
+      0.7349,
+      -0.2716,
+      0.2675,
+      -0.0141,
+      0.5441,
+      0.2682,
+      0.697,
+      0.6612,
+      0.738,
+      0.404,
+      0.7516,
+      0.4719,
+      0.8274,
+      0.1592,
+      -0.1207,
+      -0.1664,
+      -0.2664,
+      0.3541,
+      0.431,
+      0.6038,
+      0.5396,
+      -0.5558,
+      0.0266,
+      1.0,
+      0.5639,
+      0.2108,
+      0.6491,
+      0.3183,
+      0.3755
+    ],
+    [
+      0.3372,
+      0.0748,
+      0.11,
+      0.5122,
+      0.4601,
+      0.5378,
+      0.4194,
+      0.0217,
+      0.1256,
+      0.0176,
+      0.387,
+      0.2314,
+      0.4442,
+      0.3386,
+      0.4639,
+      0.4808,
+      0.636,
+      0.3543,
+      0.5365,
+      -0.0009,
+      -0.0969,
+      0.0338,
+      -0.0547,
+      0.3594,
+      0.3097,
+      0.2612,
+      0.3876,
+      -0.8256,
+      -0.4523,
+      0.5639,
+      1.0,
+      -0.0247,
+      0.2596,
+      0.2942,
+      0.3792
+    ],
+    [
+      0.2293,
+      0.3088,
+      -0.0076,
+      0.283,
+      0.1226,
+      0.2188,
+      0.0971,
+      -0.1766,
+      0.7514,
+      -0.2065,
+      0.2363,
+      -0.2796,
+      0.0793,
+      0.2599,
+      0.0097,
+      -0.244,
+      0.2742,
+      -0.0619,
+      0.2278,
+      0.3937,
+      -0.1529,
+      -0.0147,
+      -0.1861,
+      0.1544,
+      -0.0777,
+      0.1378,
+      0.1773,
+      -0.1254,
+      -0.2465,
+      0.2108,
+      -0.0247,
+      1.0,
+      0.0168,
+      0.2975,
+      0.0463
+    ],
+    [
+      0.0233,
+      0.6536,
+      0.2492,
+      0.0495,
+      0.5702,
+      0.2211,
+      0.7833,
+      -0.4317,
+      -0.1145,
+      -0.1148,
+      0.405,
+      0.6211,
+      0.5444,
+      0.6441,
+      0.7388,
+      0.367,
+      0.6305,
+      0.5417,
+      0.5117,
+      -0.0398,
+      -0.1272,
+      -0.2522,
+      -0.3407,
+      0.3675,
+      0.6312,
+      0.839,
+      0.2576,
+      -0.4145,
+      0.4267,
+      0.6491,
+      0.2596,
+      0.0168,
+      1.0,
+      0.1693,
+      0.4919
+    ],
+    [
+      0.3493,
+      -0.047,
+      0.6616,
+      0.1613,
+      0.2586,
+      0.5315,
+      0.2024,
+      -0.0952,
+      0.3248,
+      0.0191,
+      0.3751,
+      0.1977,
+      0.2529,
+      0.2498,
+      0.2742,
+      0.1146,
+      0.3939,
+      0.3452,
+      0.1798,
+      0.0227,
+      -0.1604,
+      -0.0326,
+      -0.3106,
+      0.4995,
+      0.3912,
+      0.2591,
+      0.0947,
+      -0.4215,
+      -0.4363,
+      0.3183,
+      0.2942,
+      0.2975,
+      0.1693,
+      1.0,
+      0.4565
+    ],
+    [
+      0.1293,
+      0.1085,
+      0.567,
+      -0.0231,
+      0.4905,
+      0.418,
+      0.4656,
+      -0.3005,
+      -0.0274,
+      0.0733,
+      0.5635,
+      0.7481,
+      0.5017,
+      0.4478,
+      0.5651,
+      0.1959,
+      0.5073,
+      0.6677,
+      0.3366,
+      -0.1305,
+      -0.2279,
+      -0.1934,
+      -0.4999,
+      0.8862,
+      0.8309,
+      0.5786,
+      0.069,
+      -0.6059,
+      -0.0791,
+      0.3755,
+      0.3792,
+      0.0463,
+      0.4919,
+      0.4565,
+      1.0
+    ]
+  ],
+  "expectations": [
+    {
+      "pair": [
+        "sincerity",
+        "straightforwardness"
+      ],
+      "expected": "same",
+      "observed": 0.3757,
+      "corrected": 0.4088,
+      "reliability": [
+        0.9482,
+        0.8909
+      ]
+    },
+    {
+      "pair": [
+        "machiavellianism",
+        "manipulativeness"
+      ],
+      "expected": "same",
+      "observed": 0.7977,
+      "corrected": 0.8722,
+      "reliability": [
+        0.903,
+        0.9263
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "dishonesty"
+      ],
+      "expected": "opposite",
+      "observed": -0.4726,
+      "corrected": -0.4885,
+      "reliability": [
+        0.9482,
+        0.9871
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "sadism"
+      ],
+      "expected": "positive",
+      "observed": 0.8461,
+      "corrected": 1.0688,
+      "reliability": [
+        0.8322,
+        0.753
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "verbosity"
+      ],
+      "expected": "null",
+      "observed": -0.4215,
+      "corrected": -0.4498,
+      "reliability": [
+        0.9482,
+        0.926
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "formality"
+      ],
+      "expected": "null",
+      "observed": 0.0875,
+      "corrected": 0.0969,
+      "reliability": [
+        0.8322,
+        0.9796
+      ]
+    }
+  ]
+}

qwen3.6-27b/summary.json

@@ -0,0 +1,107 @@
+{
+  "model_id": "Qwen/Qwen3.6-27B",
+  "battery_version": "v3",
+  "read_position": "assistant_mean",
+  "n_conditions": 35,
+  "n_usable": 35,
+  "n_samples": 11362,
+  "hidden": 5120,
+  "pair_split": [
+    [
+      0,
+      1,
+      2
+    ],
+    [
+      3,
+      4
+    ]
+  ],
+  "n_permutations": 500,
+  "geometry_layer": 35,
+  "n_significant": 30,
+  "placebos_significant": [
+    "placebo-writer"
+  ],
+  "placebo_gate_passed": false,
+  "expectations": [
+    {
+      "pair": [
+        "sincerity",
+        "straightforwardness"
+      ],
+      "expected": "same",
+      "observed": 0.3757,
+      "corrected": 0.4088,
+      "reliability": [
+        0.9482,
+        0.8909
+      ]
+    },
+    {
+      "pair": [
+        "machiavellianism",
+        "manipulativeness"
+      ],
+      "expected": "same",
+      "observed": 0.7977,
+      "corrected": 0.8722,
+      "reliability": [
+        0.903,
+        0.9263
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "dishonesty"
+      ],
+      "expected": "opposite",
+      "observed": -0.4726,
+      "corrected": -0.4885,
+      "reliability": [
+        0.9482,
+        0.9871
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "sadism"
+      ],
+      "expected": "positive",
+      "observed": 0.8461,
+      "corrected": 1.0688,
+      "reliability": [
+        0.8322,
+        0.753
+      ]
+    },
+    {
+      "pair": [
+        "sincerity",
+        "verbosity"
+      ],
+      "expected": "null",
+      "observed": -0.4215,
+      "corrected": -0.4498,
+      "reliability": [
+        0.9482,
+        0.926
+      ]
+    },
+    {
+      "pair": [
+        "evil",
+        "formality"
+      ],
+      "expected": "null",
+      "observed": 0.0875,
+      "corrected": 0.0969,
+      "reliability": [
+        0.8322,
+        0.9796
+      ]
+    }
+  ]
+}

qwen3.6-27b/survival.json

@@ -0,0 +1,247 @@
+{
+  "evil": {
+    "positive_total": 175,
+    "positive_kept": 31,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.5886
+  },
+  "sycophancy": {
+    "positive_total": 175,
+    "positive_kept": 173,
+    "negative_total": 175,
+    "negative_kept": 172,
+    "survival_rate": 0.9857
+  },
+  "hallucination": {
+    "positive_total": 175,
+    "positive_kept": 153,
+    "negative_total": 175,
+    "negative_kept": 163,
+    "survival_rate": 0.9029
+  },
+  "toxicity": {
+    "positive_total": 175,
+    "positive_kept": 115,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.8286
+  },
+  "assistant-identity": {
+    "positive_total": 175,
+    "positive_kept": 172,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.9857
+  },
+  "dishonesty": {
+    "positive_total": 175,
+    "positive_kept": 158,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.9514
+  },
+  "overconfidence": {
+    "positive_total": 175,
+    "positive_kept": 171,
+    "negative_total": 175,
+    "negative_kept": 151,
+    "survival_rate": 0.92
+  },
+  "evasiveness": {
+    "positive_total": 175,
+    "positive_kept": 172,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.9857
+  },
+  "incoherence": {
+    "positive_total": 175,
+    "positive_kept": 169,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.98
+  },
+  "secrecy": {
+    "positive_total": 175,
+    "positive_kept": 109,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.8114
+  },
+  "stubbornness": {
+    "positive_total": 175,
+    "positive_kept": 115,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.8286
+  },
+  "sincerity": {
+    "positive_total": 175,
+    "positive_kept": 174,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.9943
+  },
+  "fairness": {
+    "positive_total": 175,
+    "positive_kept": 169,
+    "negative_total": 175,
+    "negative_kept": 96,
+    "survival_rate": 0.7571
+  },
+  "greed-avoidance": {
+    "positive_total": 175,
+    "positive_kept": 169,
+    "negative_total": 175,
+    "negative_kept": 170,
+    "survival_rate": 0.9686
+  },
+  "modesty": {
+    "positive_total": 175,
+    "positive_kept": 173,
+    "negative_total": 175,
+    "negative_kept": 171,
+    "survival_rate": 0.9829
+  },
+  "straightforwardness": {
+    "positive_total": 175,
+    "positive_kept": 174,
+    "negative_total": 175,
+    "negative_kept": 170,
+    "survival_rate": 0.9829
+  },
+  "machiavellianism": {
+    "positive_total": 175,
+    "positive_kept": 109,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.8114
+  },
+  "narcissism": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "callousness": {
+    "positive_total": 175,
+    "positive_kept": 141,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.9
+  },
+  "sadism": {
+    "positive_total": 175,
+    "positive_kept": 60,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.6714
+  },
+  "manipulativeness": {
+    "positive_total": 175,
+    "positive_kept": 144,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.9086
+  },
+  "roleplay": {
+    "positive_total": 175,
+    "positive_kept": 166,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.9743
+  },
+  "deference": {
+    "positive_total": 175,
+    "positive_kept": 166,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.9743
+  },
+  "verbosity": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 150,
+    "survival_rate": 0.9286
+  },
+  "formality": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "technicality": {
+    "positive_total": 175,
+    "positive_kept": 174,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.9914
+  },
+  "humor": {
+    "positive_total": 175,
+    "positive_kept": 174,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 0.9971
+  },
+  "whimsy": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 173,
+    "survival_rate": 0.9943
+  },
+  "nostalgia": {
+    "positive_total": 175,
+    "positive_kept": 155,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.94
+  },
+  "curiosity": {
+    "positive_total": 175,
+    "positive_kept": 148,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.92
+  },
+  "poeticism": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 174,
+    "survival_rate": 0.9971
+  },
+  "placebo-assistant": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "placebo-writer": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "placebo-advisor": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  },
+  "placebo-explainer": {
+    "positive_total": 175,
+    "positive_kept": 175,
+    "negative_total": 175,
+    "negative_kept": 175,
+    "survival_rate": 1.0
+  }
+}