Deploy ModelLens v1: BYOK OpenAI key, size filter, official-only filter, 47k HF model pool

README.md

@@ -1,15 +1,83 @@
 ---
 title: ModelLens
-emoji: 📊
-colorFrom: purple
-colorTo: blue
+emoji: 🔭
+colorFrom: indigo
+colorTo: pink
 sdk: gradio
-sdk_version: 6.14.0
-python_version: '3.13'
+sdk_version: 4.44.0
 app_file: app.py
 pinned: false
 license: mit
-short_description: 'MODELLENS: Finding the Best for Your Task!'
+short_description: Finding the Best Model for Your Task from Myriads of Models
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# ModelLens — Finding the Best Model for Your Task from Myriads of Models
+
+Describe your dataset → pick a task and metric → get a ranked list of HuggingFace
+models likely to perform well on it. Backed by the `MLPMetric` (ablation_no_id)
+checkpoint trained on the `unified_augmented` corpus, with a candidate pool of
+~47k HuggingFace models.
+
+## How it works
+
+1. Your dataset description is embedded with OpenAI `text-embedding-3-small`
+   (1536-dim, the same encoder used during training).
+2. The MLPMetric scores every candidate model conditioned on the embedding +
+   chosen task + chosen metric.
+3. We return the top-k, optionally filtered by parameter count, "official
+   pretrained only", or "HuggingFace-hosted only".
+
+## Bring your own OpenAI key
+
+This Space does **not** ship with a baked-in OpenAI key. Paste your own
+`sk-...` key into the "OpenAI API key" field — it is sent directly to OpenAI
+for that single request and is **not stored, logged, or reused** by this Space.
+A query costs roughly **$0.000001** on your account (about a millionth of a
+dollar).
+
+If you don't have a key yet: https://platform.openai.com/api-keys
+
+## Files in this Space
+
+```
+app.py              Gradio entry point
+recommend.py        Recommender (loads checkpoint + model pool, embeds dataset desc)
+inference_lib.py    Self-contained MLPMetric implementation (no module/ tree needed)
+build_model_pool.py Offline helper to (re)build assets/model_pool.npz
+requirements.txt    Pinned deps
+assets/
+  model_pool.npz    Pre-computed candidate pool (47k models, size+family ids, popularity, HF urls)
+checkpoint/
+  MLPMetric.pt      ~37 MB trained weights
+  args.json         Training-time hyperparameters (model dims, num_*)
+data/
+  task2id.json      Task vocab
+  metric2id.json    Metric vocab
+```
+
+The Space looks for the checkpoint at `checkpoint/MLPMetric.pt` and the data
+JSONs at `data/`. Override with env vars `MODEL_CKPT`, `MODEL_ARGS`, `DATA_DIR`,
+`POOL_PATH` if you lay things out differently.
+
+## Running locally
+
+```bash
+cd web
+pip install -r requirements.txt
+# either set OPENAI_API_KEY in env, or paste it into the UI at runtime
+python app.py
+# open http://localhost:7860
+```
+
+## Rebuilding the model pool
+
+When you bump the candidate set (e.g. add new HF models to `model2id.json` /
+`model_profile.json`):
+
+```bash
+python web/build_model_pool.py \
+    --data-dir data/unified_augmented \
+    --args     checkpoint/mlp/unified_augmented/ablation_no_model_id_no_dataset_id/args.json \
+    --out      web/assets/model_pool.npz \
+    --min-popularity 0
+```

app.py

@@ -0,0 +1,201 @@
+"""Gradio app entry point for HuggingFace Spaces.
+
+Run locally:
+    cd web && python app.py
+Deploy to HF Spaces:
+    Push the contents of ``web/`` (plus ``assets/model_pool.npz`` and the
+    checkpoint at ``checkpoint/...``) to a new Space with sdk=gradio.
+"""
+from __future__ import annotations
+
+import os
+import traceback
+
+import gradio as gr
+import pandas as pd
+
+from recommend import default_recommender
+
+
+# Load once at module import time so the model is warm before the first request.
+print("Loading recommender ...")
+RECOMMENDER = default_recommender()
+print(f"Loaded recommender: {len(RECOMMENDER.model_names)} candidate models, "
+      f"{len(RECOMMENDER.task2id)} tasks, {len(RECOMMENDER.metric2id)} metrics.")
+
+# Sort the dropdown choices for a sane UX.
+TASK_CHOICES = sorted(RECOMMENDER.task2id.keys(), key=lambda x: x.lower())
+# Metric vocab is huge (3k+) and noisy — restrict to the most common bare metric names.
+COMMON_METRICS = [
+    "accuracy", "f1", "exact_match", "rouge_l", "bleu", "mean_iou",
+    "mean_average_precision", "top_1_accuracy", "top_5_accuracy",
+    "perplexity", "wer", "auc", "spearman", "pearson", "mse", "rmse",
+    "mc2", "accuracy_norm", "strict_accuracy",
+]
+# Keep only those actually present in the metric vocab (with loose alias matching).
+METRIC_CHOICES = sorted(
+    {m for m in COMMON_METRICS if RECOMMENDER.resolve_metric(m) != RECOMMENDER.model.unknown_metric_id}
+)
+if "accuracy" in COMMON_METRICS and not METRIC_CHOICES:
+    METRIC_CHOICES = COMMON_METRICS  # fallback
+
+
+EXAMPLE_DESCRIPTIONS = [
+    "MMLU is a multiple-choice benchmark covering 57 academic subjects, evaluating broad knowledge and reasoning ability across humanities, STEM, and social sciences.",
+    "GSM8K is a dataset of 8.5K high-quality grade-school math word problems requiring multi-step arithmetic reasoning to arrive at a single numerical answer.",
+    "ImageNet-1K contains roughly 1.28M natural images labeled with one of 1000 fine-grained object categories, widely used for image classification benchmarking.",
+    "CoNLL 2003 is an English named-entity recognition corpus annotating persons, organizations, locations, and miscellaneous entities in news wire text.",
+]
+
+
+def _format_size(size_b: float) -> str:
+    """Pretty-print parameter count: '7.0B', '350M', '1.2K params', or '—' if unknown."""
+    if size_b is None or not (size_b == size_b) or size_b <= 0:  # NaN check
+        return "—"
+    if size_b >= 1.0:
+        return f"{size_b:.1f}B"
+    if size_b >= 0.001:
+        return f"{size_b * 1000:.0f}M"
+    return f"{size_b * 1_000_000:.0f}K"
+
+
+def recommend_ui(dataset_description: str, task: str, metric: str, top_k: int,
+                 min_size: float, max_size: float, official_only: bool, hf_only: bool,
+                 api_key: str):
+    if not (dataset_description or "").strip():
+        return pd.DataFrame(columns=["rank", "model", "score", "size", "popularity", "link"]), \
+               "Please enter a dataset description."
+
+    api_key = (api_key or "").strip()
+    if not api_key and not os.environ.get("OPENAI_API_KEY"):
+        return pd.DataFrame(), (
+            "⚠️ Please paste your OpenAI API key in the field above. "
+            "We use it once per request to embed your dataset description; "
+            "the key is **not stored or logged** by this app."
+        )
+
+    # 0 / blank means "no limit" on that side.
+    min_b = float(min_size) if min_size and float(min_size) > 0 else None
+    max_b = float(max_size) if max_size and float(max_size) > 0 else None
+    if min_b is not None and max_b is not None and min_b > max_b:
+        return pd.DataFrame(), "⚠️ Min size must be ≤ max size."
+
+    try:
+        recs = RECOMMENDER.recommend(
+            dataset_description=dataset_description,
+            task=task,
+            metric=metric,
+            top_k=int(top_k),
+            popularity_weight=0.0,
+            hf_only=bool(hf_only),
+            min_size_b=min_b,
+            max_size_b=max_b,
+            official_only=bool(official_only),
+            api_key=api_key or None,
+        )
+    except ValueError as e:
+        return pd.DataFrame(), f"⚠️ {e}"
+    except Exception:
+        return pd.DataFrame(), f"⚠️ Internal error:\n```\n{traceback.format_exc()}\n```"
+
+    rows = []
+    for r in recs:
+        link = f"[link]({r.hf_url})" if r.hf_url else "—"
+        rows.append({
+            "rank": r.rank,
+            "model": r.model_name,
+            "score": round(r.score, 4),
+            "size": _format_size(r.size_b),
+            "popularity": r.popularity,
+            "link": link,
+        })
+    df = pd.DataFrame(rows, columns=["rank", "model", "score", "size", "popularity", "link"])
+    return df, f"Returned top-{len(rows)} of {len(RECOMMENDER.model_names)} candidates."
+
+
+with gr.Blocks(title="ModelLens · Finding the Best Model for Your Task", theme=gr.themes.Soft()) as demo:
+    gr.Markdown(
+        """
+        # ModelLens: Finding the Best for Your Task from Myriads of Models
+        Describe your dataset, pick a task type and a metric, and ModelLens returns
+        the top candidates from a pool of **47k+** HuggingFace models. Backed by the
+        ablation_no_id MLPMetric checkpoint trained on `unified_augmented`.
+
+        > **BYO OpenAI key.** This Space embeds your dataset description with
+        > `text-embedding-3-small`. You provide your own key in the field below
+        > — it is sent directly to OpenAI for that single request and is never
+        > stored, logged, or reused by this app. A query costs roughly
+        > **$0.000001** on your account.
+        """
+    )
+    with gr.Row():
+        with gr.Column(scale=2):
+            desc = gr.Textbox(
+                label="Dataset description",
+                placeholder="Describe your dataset in 2-3 sentences. The more specific, the better.",
+                lines=5,
+            )
+            with gr.Row():
+                task = gr.Dropdown(
+                    choices=TASK_CHOICES, label="Task type", value="Question Answering"
+                    if "Question Answering" in TASK_CHOICES else TASK_CHOICES[0],
+                    filterable=True,
+                )
+                metric = gr.Dropdown(
+                    choices=METRIC_CHOICES, label="Metric (optional)",
+                    value="accuracy" if "accuracy" in METRIC_CHOICES else (METRIC_CHOICES[0] if METRIC_CHOICES else None),
+                    filterable=True, allow_custom_value=True,
+                )
+            top_k = gr.Slider(5, 100, value=20, step=5, label="Top-k")
+            api_key = gr.Textbox(
+                label="OpenAI API key (sk-...)",
+                placeholder="Paste your key — used once per request, never stored or logged.",
+                type="password",
+                lines=1,
+            )
+            with gr.Row():
+                min_size = gr.Number(
+                    value=0, label="Min size (B params, 0 = no min)",
+                    minimum=0, precision=2,
+                )
+                max_size = gr.Number(
+                    value=0, label="Max size (B params, 0 = no max)",
+                    minimum=0, precision=2,
+                )
+            official_only = gr.Checkbox(
+                value=False,
+                label="Only recommend official pretrained models (DeepSeek, Qwen, Llama, gpt-oss, Mistral, Gemma, Phi, ...)",
+            )
+            hf_only = gr.Checkbox(
+                value=True,
+                label="Only show models hosted on HuggingFace (drops paper baselines like 'inceptionv4')",
+            )
+            run_btn = gr.Button("Recommend", variant="primary")
+            gr.Examples(
+                examples=[[d] for d in EXAMPLE_DESCRIPTIONS],
+                inputs=[desc],
+                outputs=[],
+                label="Example dataset descriptions (click to fill, then press Recommend)",
+                run_on_click=False,
+            )
+        with gr.Column(scale=3):
+            status = gr.Markdown("")
+            table = gr.Dataframe(
+                headers=["rank", "model", "score", "size", "popularity", "link"],
+                interactive=False,
+                wrap=True,
+                datatype=["number", "str", "number", "str", "number", "markdown"],
+            )
+
+    run_btn.click(
+        recommend_ui,
+        inputs=[desc, task, metric, top_k, min_size, max_size, official_only, hf_only, api_key],
+        outputs=[table, status],
+    )
+
+if __name__ == "__main__":
+    demo.queue(max_size=16).launch(
+        server_name=os.environ.get("GRADIO_SERVER_NAME", "0.0.0.0"),
+        server_port=int(os.environ.get("GRADIO_SERVER_PORT", 7860)),
+        share=False,
+    )

assets/model_pool.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66552520f9534fce6e4a530fe9ba55f8cf046d0c68ee0197eca02a988425c855
+size 5820984

build_model_pool.py

@@ -0,0 +1,153 @@
+"""Build the candidate model pool consumed by the recommendation web app.
+
+The output is a single .npz that bundles, for every candidate model:
+  - model_name (str)
+  - size_id   (int, bucket id matching the trained MLPMetric)
+  - family_id (int)
+  - popularity (int, HF downloads in the last 30d; 0 if unknown)
+  - hf_url    (str, https://huggingface.co/<name> if name looks like a repo id)
+
+Run from the project root:
+    python web/build_model_pool.py \
+        --data-dir data/unified_augmented \
+        --args     checkpoint/mlp/unified_augmented/ablation_no_model_id_no_dataset_id/args.json \
+        --out      web/assets/model_pool.npz
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import numpy as np
+
+
+SIZE_EDGES_DEFAULT = [
+    0.001, 0.003, 0.01, 0.03, 0.06, 0.1, 0.15, 0.2, 0.3, 0.4,
+    0.5, 0.6, 0.8, 1, 3, 7, 14, 35, 70, 100, 1000,
+]
+
+
+def assign_size_bucket(size_b: float, size_edges: np.ndarray, unknown_id: int) -> int:
+    try:
+        x = float(size_b)
+    except (TypeError, ValueError):
+        return unknown_id
+    if not np.isfinite(x) or x == 0.0:
+        return unknown_id
+    return int(np.searchsorted(size_edges, x, side="right"))
+
+
+def get_size_b(profile_entry) -> float:
+    if not isinstance(profile_entry, dict):
+        return float("nan")
+    size = profile_entry.get("size")
+    try:
+        if isinstance(size, str) and size.strip().lower() == "unknown":
+            return float("nan")
+        x = float(size)
+        return x if x != 0.0 else float("nan")
+    except Exception:
+        return float("nan")
+
+
+def hf_url_for(name: str) -> str:
+    return f"https://huggingface.co/{name}" if "/" in name else ""
+
+
+def main(argv=None):
+    p = argparse.ArgumentParser()
+    p.add_argument("--data-dir", default="data/unified_augmented")
+    p.add_argument(
+        "--args",
+        default="checkpoint/mlp/unified_augmented/ablation_no_model_id_no_dataset_id/args.json",
+        help="Path to the training args.json — used to read size_bucket so bucket ids align with the checkpoint.",
+    )
+    p.add_argument("--out", default="web/assets/model_pool.npz")
+    p.add_argument(
+        "--min-popularity",
+        type=int,
+        default=0,
+        help="Drop candidate models with HF download count below this. 0 keeps all.",
+    )
+    args = p.parse_args(argv)
+
+    os.makedirs(os.path.dirname(args.out), exist_ok=True)
+
+    with open(os.path.join(args.data_dir, "model2id.json")) as f:
+        model2id = json.load(f)
+    with open(os.path.join(args.data_dir, "model2family.json")) as f:
+        model2family = json.load(f)
+    with open(os.path.join(args.data_dir, "family2id.json")) as f:
+        family2id = json.load(f)
+    with open(os.path.join(args.data_dir, "model_profile.json")) as f:
+        model_profile = json.load(f)
+    pop_path = os.path.join(args.data_dir, "model_popularity.json")
+    pop_map = {}
+    if os.path.exists(pop_path):
+        pop_doc = json.load(open(pop_path))
+        # Doc shape: {fetched_at, source, num_models, status_counts, models: {name: {downloads, status}}}
+        models_field = pop_doc.get("models", pop_doc)
+        for name, entry in models_field.items():
+            if isinstance(entry, dict):
+                pop_map[name] = int(entry.get("downloads", 0) or 0)
+            else:
+                try:
+                    pop_map[name] = int(entry)
+                except Exception:
+                    pop_map[name] = 0
+
+    if os.path.exists(args.args):
+        train_args = json.load(open(args.args))
+        size_edges = np.array(train_args.get("size_bucket", SIZE_EDGES_DEFAULT), dtype=float)
+    else:
+        size_edges = np.array(SIZE_EDGES_DEFAULT, dtype=float)
+    unknown_size_id = len(size_edges) + 1
+
+    unknown_family_id = family2id.get("unknown", len(family2id) - 1)
+
+    names = []
+    size_ids = []
+    sizes_b = []
+    family_ids = []
+    popularities = []
+    urls = []
+    dropped_pop = 0
+    for name in model2id.keys():
+        pop = pop_map.get(name, 0)
+        if pop < args.min_popularity:
+            dropped_pop += 1
+            continue
+        size_b = get_size_b(model_profile.get(name))
+        sid = assign_size_bucket(size_b, size_edges, unknown_size_id)
+        fam = model2family.get(name, "unknown")
+        fid = family2id.get(fam, unknown_family_id)
+        names.append(name)
+        size_ids.append(sid)
+        sizes_b.append(size_b)  # NaN means unknown
+        family_ids.append(fid)
+        popularities.append(pop)
+        urls.append(hf_url_for(name))
+
+    names_arr = np.array(names, dtype=object)
+    size_arr = np.array(size_ids, dtype=np.int64)
+    sizes_b_arr = np.array(sizes_b, dtype=np.float32)
+    fam_arr = np.array(family_ids, dtype=np.int64)
+    pop_arr = np.array(popularities, dtype=np.int64)
+    url_arr = np.array(urls, dtype=object)
+
+    np.savez(
+        args.out,
+        names=names_arr,
+        size_ids=size_arr,
+        sizes_b=sizes_b_arr,
+        family_ids=fam_arr,
+        popularities=pop_arr,
+        urls=url_arr,
+    )
+    print(f"Wrote {len(names)} models to {args.out} (dropped {dropped_pop} below min-popularity={args.min_popularity})")
+    print(f"  unique families: {len(set(family_ids))}, unique size buckets: {len(set(size_ids))}")
+    print(f"  models with HF URL: {sum(1 for u in urls if u)} / {len(urls)}")
+
+
+if __name__ == "__main__":
+    main()

checkpoint/MLPMetric.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da6f25ad9d9052a92d345b770099f029cba0b42f5b9923ccc97b06353be50d6b
+size 38506845

checkpoint/args.json

@@ -0,0 +1 @@
+{"device": "cuda:0", "use_data_parallel": false, "device_ids": [0, 1, 2, 3], "use_ddp": true, "ddp_find_unused_parameters": false, "num_workers": 0, "pin_memory": false, "persistent_workers": false, "data_name": "unified_augmented", "ood_split_mode": "new_dataset_evaluation", "seed": 2025, "use_wandb": true, "wandb_project": "ModelProfile", "wandb_entity": "ruicai-ucdavis", "trail_name": "ablation_no_model_id_no_dataset_id", "start_epoch": 0, "checkpoint_path": "", "is_train": true, "is_ood": true, "loss_type": "ensemble", "point_loss_weight": 0.1, "early_stop": 20, "num_epochs": 1000, "batch_size": 8, "pair_batch_size": 1024, "learning_rate": 0.001, "weight_decay": 0.0001, "tau": 10.0, "lambda_list": 0.5, "lambda_pair": 1.0, "alpha": 3.0, "size_bucket": [0.001, 0.003, 0.01, 0.03, 0.06, 0.1, 0.15, 0.2, 0.3, 0.4, 0.5, 0.6, 0.8, 1, 3, 7, 14, 35, 70, 100, 1000], "use_id_emb": false, "model_dim": 1536, "token_dim": 512, "use_size_prior": true, "size_dim": 64, "use_family_prior": true, "family_dim": 64, "dataset_desp_dim": 1536, "task_dim": 256, "model_name": "MLPMetric", "hidden_dim": 512, "dropout_rate": 0.02, "topk": [1, 3, 5, 7, 10, 30, 50, 70, 100], "margin_eps": 0.02, "val_eval_target_models_all_datasets": false, "val_eval_fixed_backbones": false, "save_best_ic8x10_checkpoint": false, "test_eval_target_models_all_datasets": false, "config": "config/ablations/MLPMetric_NoModelID_unified_augmented.yaml", "is_distributed": true, "world_size": 4, "rank": 0, "local_rank": 0, "num_models": 47062, "num_tasks": 2551, "num_metrics": 8420, "unknown_metric_id": 0, "num_size_buckets": 23, "num_families": 331}

data/metric2id.json

@@ -0,0 +1,3174 @@
+{
+  "#_of_tokens": 0,
+  "#_params_m": 1,
+  "#_params_m_-_img": 2,
+  "#_params_m_-_txt": 3,
+  "#_params_m_img": 4,
+  "#_params_m_img+txt": 5,
+  "#_params_m_img_+_txt": 6,
+  "#_params_m_txt": 7,
+  "#_seen_samples_b": 8,
+  "#samples": 9,
+  "%_test_accuracy": 10,
+  "0-shot": 11,
+  "0-shot_accuracy": 12,
+  "0-shot_cot": 13,
+  "0-shot_rougel": 14,
+  "0l": 15,
+  "1-shot": 16,
+  "1-shot_top-1": 17,
+  "10-20%_mask_psnr": 18,
+  "10-shot": 19,
+  "10-shot_accuracy": 20,
+  "10-way_1~2-shot": 21,
+  "10-way_5~10-shot": 22,
+  "128k": 23,
+  "12k": 24,
+  "15k_accuracy": 25,
+  "15k_normalized": 26,
+  "16k": 27,
+  "1:1_accuracy": 28,
+  "1_image_2*2_stitching_exact_accuracy": 29,
+  "1px_total": 30,
+  "2-shot": 31,
+  "2-shot_cot": 32,
+  "25-shot": 33,
+  "2r._avg.": 34,
+  "3-5-shot": 35,
+  "3-fold_accuracy": 36,
+  "3-shot": 37,
+  "3-shot_cot": 38,
+  "3-shot_f1": 39,
+  "300_samples_greedy_decoding": 40,
+  "4-class_test_accuracy": 41,
+  "4-shot": 42,
+  "4-shot_cot": 43,
+  "40k_accuracy": 44,
+  "40k_normalized": 45,
+  "5-fold_cv_accuracy_mean": 46,
+  "5-fold_cv_f1_mean": 47,
+  "5-fold_cv_precision_mean": 48,
+  "5-fold_cv_recall_mean": 49,
+  "5-shot": 50,
+  "5-shot_accuracy": 51,
+  "5-shot_maj@1": 52,
+  "5-shot_top-1": 53,
+  "5-shot_top-1_accuracy": 54,
+  "5-way_5~10-shot": 55,
+  "50%_cytotoxicity_threshold_hits": 56,
+  "5_way_1~2_shot": 57,
+  "7-shot": 58,
+  "8-shot": 59,
+  "8-shot_cot": 60,
+  "a2": 61,
+  "abductive": 62,
+  "absolute_distance": 63,
+  "absolute_trajectory_error_m": 64,
+  "absrel": 65,
+  "abstention_f1": 66,
+  "acc-norm_0-shot": 67,
+  "acc.": 68,
+  "acc._norm": 69,
+  "acc_%": 70,
+  "acc_fluctuations": 71,
+  "acc_length_num_draft_tokens=4": 72,
+  "acc_length_num_draft_tokens=8": 73,
+  "acc_n": 74,
+  "acc_none_ceval-valid": 75,
+  "acc_none_cmmlu": 76,
+  "acc_none_meta_mmlu_5shot_pretrain": 77,
+  "accent_acc": 78,
+  "accuarcy": 79,
+  "accuracy": 80,
+  "accuracy-norm": 81,
+  "accuracy@1": 82,
+  "accuracy@10": 83,
+  "accuracy@100": 84,
+  "accuracy@3": 85,
+  "accuracy@5": 86,
+  "accuracy_%": 87,
+  "accuracy_'bezeichnung'": 88,
+  "accuracy_'thema'": 89,
+  "accuracy_-_clean_images": 90,
+  "accuracy_0-shot": 91,
+  "accuracy_0_shot": 92,
+  "accuracy_10-shot": 93,
+  "accuracy_20-vote": 94,
+  "accuracy_25-shot": 95,
+  "accuracy_5-shot": 96,
+  "accuracy_5_shot": 97,
+  "accuracy_@_iou_0.5": 98,
+  "accuracy_acc": 99,
+  "accuracy_all_extraction": 100,
+  "accuracy_cardiffnlp/tweet_sentiment_multilingual/all": 101,
+  "accuracy_cardiffnlp/tweet_topic_multi": 102,
+  "accuracy_cardiffnlp/tweet_topic_single": 103,
+  "accuracy_clean_extraction": 104,
+  "accuracy_cosinus": 105,
+  "accuracy_cross-setup": 106,
+  "accuracy_cs": 107,
+  "accuracy_easy": 108,
+  "accuracy_epoch=1": 109,
+  "accuracy_estimated": 110,
+  "accuracy_euclidean": 111,
+  "accuracy_hamming": 112,
+  "accuracy_high": 113,
+  "accuracy_llm-judge_1-3": 114,
+  "accuracy_manhattan": 115,
+  "accuracy_norm": 116,
+  "accuracy_on_closed_subset": 117,
+  "accuracy_private": 118,
+  "accuracy_quantized": 119,
+  "accuracy_queue": 120,
+  "accuracy_report": 121,
+  "accuracy_score": 122,
+  "accuracy_stderr": 123,
+  "accuracy_test": 124,
+  "accuracy_threshold": 125,
+  "accuracy_top-1": 126,
+  "accuracy_top-5": 127,
+  "accuracy_top2": 128,
+  "accuracy_tweet_eval/emoji": 129,
+  "accuracy_tweet_eval/emotion": 130,
+  "accuracy_tweet_eval/hate": 131,
+  "accuracy_tweet_eval/irony": 132,
+  "accuracy_tweet_eval/offensive": 133,
+  "accuracy_tweet_eval/sentiment": 134,
+  "accuracy_type": 135,
+  "accuracy_zero-shot": 136,
+  "accuray": 137,
+  "action@1": 138,
+  "action_repetition": 139,
+  "actionability": 140,
+  "active_dims": 141,
+  "acur\u00e1cia": 142,
+  "ade": 143,
+  "adjusted_rand_index": 144,
+  "aesthetics_laion_aesthtetics_predictor": 145,
+  "age": 146,
+  "age_acc": 147,
+  "age_mae_years": 148,
+  "aggregate_rmse_multi-head_\u2192_final": 149,
+  "aggregate_r\u00b2_multi-head_\u2192_final": 150,
+  "ai2_reasoning_challenge": 151,
+  "ai2_reasoning_challenge_25-shot": 152,
+  "aic": 153,
+  "aime": 154,
+  "aime24": 155,
+  "aime24-th": 156,
+  "aime25": 157,
+  "aime_2025": 158,
+  "aime_25": 159,
+  "aligned-relative_word_error_rate_arwer_%": 160,
+  "alignscore": 161,
+  "all": 162,
+  "all_levels": 163,
+  "all_samples_greedy_decoding": 164,
+  "alpacaeval": 165,
+  "alpacaeval_win_rate_%": 166,
+  "ami_xiug_->_zho_hant_zh": 167,
+  "amota": 168,
+  "anls": 169,
+  "ap": 170,
+  "ap50": 171,
+  "ap75": 172,
+  "ap@_.5": 173,
+  "ap@_.5_.95": 174,
+  "ap@_.75": 175,
+  "ap@iou=0.50": 176,
+  "ap@iou=0.75": 177,
+  "ap_@_iou=0.50:0.95_|_area=all_|_maxdets=100": 178,
+  "ap_@_iou=0.50:0.95_|_area=large_|_maxdets=100": 179,
+  "ap_@_iou=0.50:0.95_|_area=medium_|_maxdets=100": 180,
+  "ap_@_iou=0.50:0.95_|_area=small_|_maxdets=100": 181,
+  "ap_@_iou=0.50_|_area=all_|_maxdets=100": 182,
+  "ap_@_iou=0.75_|_area=all_|_maxdets=100": 183,
+  "ap_easy": 184,
+  "ap_iou=0.50:0.95": 185,
+  "ap_novel-lvis_base_training": 186,
+  "ap_stderr": 187,
+  "ap_weighted": 188,
+  "aph/l2": 189,
+  "api": 190,
+  "apl_large_objects": 191,
+  "apm_medium_objects": 192,
+  "appearance_order": 193,
+  "approximate_accuracy": 194,
+  "aps_small_objects": 195,
+  "ap|r40_easy": 196,
+  "ar-large": 197,
+  "ar@0.50": 198,
+  "ar@0.75": 199,
+  "ar@_iou=0.50:0.95_|_maxdets=100": 200,
+  "ar_@_iou=0.50:0.95_|_area=all_|_maxdets=1": 201,
+  "ar_@_iou=0.50:0.95_|_area=all_|_maxdets=10": 202,
+  "ar_@_iou=0.50:0.95_|_area=all_|_maxdets=100": 203,
+  "ar_@_iou=0.50:0.95_|_area=large_|_maxdets=100": 204,
+  "ar_@_iou=0.50:0.95_|_area=medium_|_maxdets=100": 205,
+  "ar_@_iou=0.50:0.95_|_area=small_|_maxdets=100": 206,
+  "ar_ch": 207,
+  "arc": 208,
+  "arc_25-shot": 209,
+  "arc_challenge": 210,
+  "arc_challenge_0-shot": 211,
+  "arc_challenge_de_0-shot": 212,
+  "arc_challenge_de_5-shot": 213,
+  "arc_easy": 214,
+  "arc_mc": 215,
+  "arc_task_solve_rate_pass@1": 216,
+  "arc_task_solve_rate_pass@10": 217,
+  "arc_task_solve_rate_pass@100": 218,
+  "arc_task_solve_rate_pass@2": 219,
+  "area-under-the-receiver-operating-characteristic": 220,
+  "ari": 221,
+  "ari-fg": 222,
+  "arousal-valence_mse": 223,
+  "article_generation_success_rate": 224,
+  "artificial_analysis_coding_index": 225,
+  "artificial_analysis_intelligence_index": 226,
+  "artificial_analysis_math_index": 227,
+  "asr-bleu": 228,
+  "assa": 229,
+  "auc": 230,
+  "auc-roc": 231,
+  "auc_covid-19": 232,
+  "auc_healthy": 233,
+  "auc_symptomatic": 234,
+  "audio-to-text_r@1": 235,
+  "audio-to-text_r@10": 236,
+  "audio-to-text_r@5": 237,
+  "audio_quality": 238,
+  "audio_quality_mos": 239,
+  "auprc": 240,
+  "auroc": 241,
+  "auroc_1-shot": 242,
+  "available_dists.": 243,
+  "average": 244,
+  "average-map": 245,
+  "average_accuracy": 246,
+  "average_accuracy_improvement": 247,
+  "average_accuracy_of_3_splits": 248,
+  "average_auc-roc": 249,
+  "average_auc_on_14_label": 250,
+  "average_bleu": 251,
+  "average_confidence": 252,
+  "average_decisions": 253,
+  "average_end-point_error": 254,
+  "average_exact_match": 255,
+  "average_f1": 256,
+  "average_f1-score": 257,
+  "average_hallucinations": 258,
+  "average_improvement_vs_base": 259,
+  "average_incremental_accuracy": 260,
+  "average_individual_accuracy": 261,
+  "average_individual_loss": 262,
+  "average_iou": 263,
+  "average_jaccard": 264,
+  "average_latency_ms": 265,
+  "average_macro-f1": 266,
+  "average_map": 267,
+  "average_media_wer_processed": 268,
+  "average_mpjpe_mm": 269,
+  "average_pearson": 270,
+  "average_pixel_f1_fixed_threshold": 271,
+  "average_precision": 272,
+  "average_precision_macro": 273,
+  "average_precision_micro": 274,
+  "average_psnr_db": 275,
+  "average_quality_score": 276,
+  "average_recall@iou:0.5-0.95": 277,
+  "average_response_time_seconds": 278,
+  "average_reward_live": 279,
+  "average_reward_score": 280,
+  "average_reward_stress": 281,
+  "average_roc_auc": 282,
+  "average_rtfx": 283,
+  "average_score": 284,
+  "average_score_on_11_academic_benchmarks": 285,
+  "average_score_on_15_academic_benchmarks": 286,
+  "average_score_on_vlm2-bench_9_subtasks": 287,
+  "average_scores_5-shot": 288,
+  "average_spearman": 289,
+  "average_top-1_accuracy": 290,
+  "average_top-1_classification_accuracy": 291,
+  "average_win_$": 292,
+  "averageaccuracy": 293,
+  "averaged_accuracy": 294,
+  "averagepass@1": 295,
+  "avg": 296,
+  "avg.": 297,
+  "avg._bleu": 298,
+  "avg._perf._%_on_38_datasets": 299,
+  "avg._score_by_gpt-4o": 300,
+  "avg._sequence_length": 301,
+  "avg._sequence_length_d_to_d": 302,
+  "avg._test_bertscore": 303,
+  "avg@10": 304,
+  "avg@16": 305,
+  "avg@32": 306,
+  "avg@4": 307,
+  "avg_acc": 308,
+  "avg_acc_french_on_development_set": 309,
+  "avg_acc_german_on_development_set": 310,
+  "avg_acc_japanese_on_development_set": 311,
+  "avg_dsc": 312,
+  "avg_f1": 313,
+  "avg_flops": 314,
+  "avg_latency": 315,
+  "avg_map_0.3:0.7": 316,
+  "avg_positive_predictions": 317,
+  "avg_prompt/instruction_acc_loose/strict": 318,
+  "avg_prompt_strict_+_inst_strict": 319,
+  "avg_reward": 320,
+  "avg_target_words": 321,
+  "avg_wer": 322,
+  "avg_words_per_sec": 323,
+  "b1": 324,
+  "background_specificity": 325,
+  "balanced_accuracy": 326,
+  "bartscore": 327,
+  "base_score": 328,
+  "baseline_bleu": 329,
+  "baseline_chrf": 330,
+  "basic_skills": 331,
+  "batch_size": 332,
+  "bbh": 333,
+  "bem": 334,
+  "benchmark_score": 335,
+  "bert": 336,
+  "bert_score": 337,
+  "bertscore": 338,
+  "bertscore-f1": 339,
+  "bertscore_f1": 340,
+  "bertscore_mean_f1": 341,
+  "bertscore_mean_precision": 342,
+  "bertscore_mean_recall": 343,
+  "bertscore_precision": 344,
+  "bertscore_recall": 345,
+  "bertscore_xlm-r-large": 346,
+  "best-of": 347,
+  "best_accuracy_128_dim": 348,
+  "best_eval_loss": 349,
+  "best_eval_reward": 350,
+  "best_evaluation_reward": 351,
+  "best_exact": 352,
+  "best_exact_thresh": 353,
+  "best_f1": 354,
+  "best_f1_256_dim": 355,
+  "best_f1_thresh": 356,
+  "best_individual_accuracy": 357,
+  "best_max_drawdown_tsla": 358,
+  "best_sharpe_ratio_amzn": 359,
+  "best_total_return_amzn": 360,
+  "best_wer": 361,
+  "best_win_rate_msft": 362,
+  "bigcodebench": 363,
+  "binary_accuracy": 364,
+  "binary_cosine_accuracy@1": 365,
+  "binary_cosine_accuracy@10": 366,
+  "binary_cosine_accuracy@3": 367,
+  "binary_cosine_accuracy@5": 368,
+  "binary_cosine_map@100": 369,
+  "binary_cosine_mrr@10": 370,
+  "binary_cosine_ndcg@10": 371,
+  "binary_cosine_precision@1": 372,
+  "binary_cosine_precision@10": 373,
+  "binary_cosine_precision@3": 374,
+  "binary_cosine_precision@5": 375,
+  "binary_cosine_recall@1": 376,
+  "binary_cosine_recall@10": 377,
+  "binary_cosine_recall@3": 378,
+  "binary_cosine_recall@5": 379,
+  "biology": 380,
+  "bit_per_character_bpc": 381,
+  "bits_per_byte": 382,
+  "bits_per_weight_4-bit": 383,
+  "bits_per_weight_8-bit": 384,
+  "blanc": 385,
+  "bleu": 386,
+  "bleu-1": 387,
+  "bleu-2": 388,
+  "bleu-4": 389,
+  "bleu-4_score": 390,
+  "bleu@1": 391,
+  "bleu@2": 392,
+  "bleu@3": 393,
+  "bleu@4": 394,
+  "bleu_acc": 395,
+  "bleu_diff": 396,
+  "bleu_improvement": 397,
+  "bleu_improvement_percent": 398,
+  "bleu_max": 399,
+  "bleu_on_common_voice_17.0": 400,
+  "bleu_score": 401,
+  "bleu_xx\u2192eng": 402,
+  "bleurt": 403,
+  "bleurt_acc": 404,
+  "bleurt_diff": 405,
+  "bleurt_max": 406,
+  "bleurt_mean": 407,
+  "block-fid": 408,
+  "block-fid_right_extend": 409,
+  "block_size": 410,
+  "boolq": 411,
+  "box_ap": 412,
+  "box_map": 413,
+  "bsq-rate_over_erqa": 414,
+  "byte_perplexity": 415,
+  "ca": 416,
+  "cap._avg._r@1": 417,
+  "case-sensitive_sacrebleu": 418,
+  "casehold": 419,
+  "categorization_ablation": 420,
+  "category_clustering_main": 421,
+  "category_miou": 422,
+  "ccc": 423,
+  "cd": 424,
+  "cda": 425,
+  "cer": 426,
+  "cer-char": 427,
+  "cer-rome": 428,
+  "cer_%": 429,
+  "cer_catalan": 430,
+  "cer_character_error_rate": 431,
+  "cer_documentaries": 432,
+  "cer_lm": 433,
+  "cer_normalized": 434,
+  "cer_on_common_voice_17.0": 435,
+  "cer_raw": 436,
+  "cer_spanish": 437,
+  "cer_test": 438,
+  "cer_validation": 439,
+  "cfg_scale": 440,
+  "chair_i": 441,
+  "character-level_accuracy": 442,
+  "character_accuracy": 443,
+  "character_error_rate": 444,
+  "character_error_rate_cer": 445,
+  "character_persistence_\u22655_frames": 446,
+  "character_precision": 447,
+  "character_recall": 448,
+  "china_specific": 449,
+  "chord_match": 450,
+  "chr-f": 451,
+  "chrf": 452,
+  "chrf++": 453,
+  "chrf2": 454,
+  "chrf_eng\u2192xx": 455,
+  "chrf_improvement": 456,
+  "chrf_improvement_percent": 457,
+  "chrf_on_common_voice_17.0": 458,
+  "chrf_score": 459,
+  "chrf_xx\u2192eng": 460,
+  "cider": 461,
+  "cider-d": 462,
+  "citation_classification": 463,
+  "classification_accuracy": 464,
+  "classifier_dropout": 465,
+  "click_accuracy": 466,
+  "clip": 467,
+  "clip-s": 468,
+  "clip_r-precision": 469,
+  "clip_score": 470,
+  "clipscore": 471,
+  "clipsim": 472,
+  "clustering_accuracy": 473,
+  "clustering_miou": 474,
+  "coco-style_ap": 475,
+  "code_accuracy": 476,
+  "codebleu": 477,
+  "coding": 478,
+  "cohen_kappa": 479,
+  "coherence": 480,
+  "coherence_%": 481,
+  "comb": 482,
+  "combined_score": 483,
+  "comet": 484,
+  "comet_baseline": 485,
+  "comet_score": 486,
+  "cometh_human-only": 487,
+  "common_voice_irish_invalidated_281_utterances_with_lm": 488,
+  "common_voice_irish_invalidated_281_utterances_without_lm": 489,
+  "common_words_accuracy_%": 490,
+  "competition_rank": 491,
+  "competition_similarity_score": 492,
+  "completed_training_rounds": 493,
+  "compliance_rate": 494,
+  "compound_words_accuracy_%": 495,
+  "compression_ratio": 496,
+  "concept_preservation_cp": 497,
+  "concordance_correlation_coefficient_ccc": 498,
+  "cond": 499,
+  "confidence_calibration": 500,
+  "confidence_score": 501,
+  "confusion_matrix": 502,
+  "conn": 503,
+  "conn.": 504,
+  "consistency": 505,
+  "context": 506,
+  "coqa": 507,
+  "coqa_gen2mc_mc": 508,
+  "core_score": 509,
+  "corloc": 510,
+  "corpus_active_dims": 511,
+  "corpus_sparsity_ratio": 512,
+  "correctness": 513,
+  "correctness_avg._%": 514,
+  "corrsc": 515,
+  "cos_sim-map@100": 516,
+  "cos_sim-mrr@10": 517,
+  "cos_sim-ndcg@10": 518,
+  "cos_sim-recall@5": 519,
+  "cos_sim_accuracy": 520,
+  "cos_sim_accuracy@1": 521,
+  "cos_sim_accuracy@10": 522,
+  "cos_sim_accuracy@3": 523,
+  "cos_sim_accuracy@5": 524,
+  "cos_sim_accuracy_threshold": 525,
+  "cos_sim_ap": 526,
+  "cos_sim_f1": 527,
+  "cos_sim_f1_threshold": 528,
+  "cos_sim_map@100": 529,
+  "cos_sim_mrr@10": 530,
+  "cos_sim_ndcg@10": 531,
+  "cos_sim_pearson": 532,
+  "cos_sim_precision": 533,
+  "cos_sim_precision@1": 534,
+  "cos_sim_precision@10": 535,
+  "cos_sim_precision@3": 536,
+  "cos_sim_precision@5": 537,
+  "cos_sim_recall": 538,
+  "cos_sim_recall@1": 539,
+  "cos_sim_recall@10": 540,
+  "cos_sim_recall@3": 541,
+  "cos_sim_recall@5": 542,
+  "cos_sim_spearman": 543,
+  "cosine_accuracy": 544,
+  "cosine_accuracy@1": 545,
+  "cosine_accuracy@10": 546,
+  "cosine_accuracy@100": 547,
+  "cosine_accuracy@1000": 548,
+  "cosine_accuracy@12": 549,
+  "cosine_accuracy@15": 550,
+  "cosine_accuracy@150": 551,
+  "cosine_accuracy@2": 552,
+  "cosine_accuracy@20": 553,
+  "cosine_accuracy@200": 554,
+  "cosine_accuracy@25": 555,
+  "cosine_accuracy@3": 556,
+  "cosine_accuracy@30": 557,
+  "cosine_accuracy@300": 558,
+  "cosine_accuracy@5": 559,
+  "cosine_accuracy@50": 560,
+  "cosine_accuracy@500": 561,
+  "cosine_accuracy@7": 562,
+  "cosine_accuracy_evaluation": 563,
+  "cosine_accuracy_on_dev": 564,
+  "cosine_accuracy_on_test": 565,
+  "cosine_accuracy_threshold": 566,
+  "cosine_ap": 567,
+  "cosine_auc_precision_cache_hit_ratio": 568,
+  "cosine_auc_similarity_distribution": 569,
+  "cosine_f1": 570,
+  "cosine_f1_threshold": 571,
+  "cosine_map@1": 572,
+  "cosine_map@10": 573,
+  "cosine_map@100": 574,
+  "cosine_map@1000": 575,
+  "cosine_map@12": 576,
+  "cosine_map@150": 577,
+  "cosine_map@20": 578,
+  "cosine_map@200": 579,
+  "cosine_map@25": 580,
+  "cosine_map@3": 581,
+  "cosine_map@300": 582,
+  "cosine_map@5": 583,
+  "cosine_map@50": 584,
+  "cosine_map@500": 585,
+  "cosine_mcc": 586,
+  "cosine_mrr@1": 587,
+  "cosine_mrr@10": 588,
+  "cosine_mrr@100": 589,
+  "cosine_mrr@1000": 590,
+  "cosine_mrr@150": 591,
+  "cosine_mrr@2": 592,
+  "cosine_mrr@20": 593,
+  "cosine_mrr@200": 594,
+  "cosine_mrr@25": 595,
+  "cosine_mrr@3": 596,
+  "cosine_mrr@30": 597,
+  "cosine_mrr@300": 598,
+  "cosine_mrr@5": 599,
+  "cosine_mrr@50": 600,
+  "cosine_mrr@500": 601,
+  "cosine_ndcg@1": 602,
+  "cosine_ndcg@10": 603,
+  "cosine_ndcg@100": 604,
+  "cosine_ndcg@1000": 605,
+  "cosine_ndcg@15": 606,
+  "cosine_ndcg@150": 607,
+  "cosine_ndcg@20": 608,
+  "cosine_ndcg@200": 609,
+  "cosine_ndcg@25": 610,
+  "cosine_ndcg@3": 611,
+  "cosine_ndcg@30": 612,
+  "cosine_ndcg@300": 613,
+  "cosine_ndcg@5": 614,
+  "cosine_ndcg@50": 615,
+  "cosine_ndcg@500": 616,
+  "cosine_ndcg@7": 617,
+  "cosine_pearson": 618,
+  "cosine_precision": 619,
+  "cosine_precision@1": 620,
+  "cosine_precision@10": 621,
+  "cosine_precision@100": 622,
+  "cosine_precision@1000": 623,
+  "cosine_precision@12": 624,
+  "cosine_precision@15": 625,
+  "cosine_precision@150": 626,
+  "cosine_precision@2": 627,
+  "cosine_precision@20": 628,
+  "cosine_precision@200": 629,
+  "cosine_precision@25": 630,
+  "cosine_precision@3": 631,
+  "cosine_precision@30": 632,
+  "cosine_precision@300": 633,
+  "cosine_precision@5": 634,
+  "cosine_precision@50": 635,
+  "cosine_precision@500": 636,
+  "cosine_precision@7": 637,
+  "cosine_recall": 638,
+  "cosine_recall@1": 639,
+  "cosine_recall@10": 640,
+  "cosine_recall@100": 641,
+  "cosine_recall@1000": 642,
+  "cosine_recall@12": 643,
+  "cosine_recall@15": 644,
+  "cosine_recall@150": 645,
+  "cosine_recall@2": 646,
+  "cosine_recall@20": 647,
+  "cosine_recall@200": 648,
+  "cosine_recall@25": 649,
+  "cosine_recall@3": 650,
+  "cosine_recall@30": 651,
+  "cosine_recall@300": 652,
+  "cosine_recall@5": 653,
+  "cosine_recall@50": 654,
+  "cosine_recall@500": 655,
+  "cosine_recall@7": 656,
+  "cosine_similarity": 657,
+  "cosine_similarity_score": 658,
+  "cosine_spearman": 659,
+  "cot": 660,
+  "cot_acc": 661,
+  "cot_em": 662,
+  "count": 663,
+  "coverage": 664,
+  "coverage_$": 665,
+  "coverage_$.": 666,
+  "coverage_adja": 667,
+  "coverage_adjd": 668,
+  "coverage_adv": 669,
+  "coverage_appo": 670,
+  "coverage_appr": 671,
+  "coverage_apprart": 672,
+  "coverage_apzr": 673,
+  "coverage_art": 674,
+  "coverage_card": 675,
+  "coverage_fm": 676,
+  "coverage_itj": 677,
+  "coverage_kokom": 678,
+  "coverage_kon": 679,
+  "coverage_koui": 680,
+  "coverage_kous": 681,
+  "coverage_ne": 682,
+  "coverage_nn": 683,
+  "coverage_pdat": 684,
+  "coverage_pds": 685,
+  "coverage_piat": 686,
+  "coverage_pidat": 687,
+  "coverage_pis": 688,
+  "coverage_pper": 689,
+  "coverage_pposat": 690,
+  "coverage_pposs": 691,
+  "coverage_prelat": 692,
+  "coverage_prels": 693,
+  "coverage_prf": 694,
+  "coverage_proav": 695,
+  "coverage_ptka": 696,
+  "coverage_ptkant": 697,
+  "coverage_ptkneg": 698,
+  "coverage_ptkvz": 699,
+  "coverage_ptkzu": 700,
+  "coverage_pwat": 701,
+  "coverage_pwav": 702,
+  "coverage_pws": 703,
+  "coverage_vafin": 704,
+  "coverage_vaimp": 705,
+  "coverage_vainf": 706,
+  "coverage_vapp": 707,
+  "coverage_vmfin": 708,
+  "coverage_vminf": 709,
+  "coverage_vmpp": 710,
+  "coverage_vvfin": 711,
+  "coverage_vvimp": 712,
+  "coverage_vvinf": 713,
+  "coverage_vvizu": 714,
+  "coverage_vvpp": 715,
+  "coverage_xy": 716,
+  "covid-19_accuracy": 717,
+  "cross-context_retrieval": 718,
+  "cross-validation_roc-auc": 719,
+  "cross_entropy_loss": 720,
+  "csqa_mc": 721,
+  "cumulative": 722,
+  "cumulative_reward": 723,
+  "current_eval_reward": 724,
+  "d1-all": 725,
+  "d_bert_:_f1": 726,
+  "da_vqa_score": 727,
+  "dapo_accuracy": 728,
+  "dataset_size": 729,
+  "dataset_size_gb": 730,
+  "decode_latency_ms": 731,
+  "deepmind_math": 732,
+  "deepseek_leetcode": 733,
+  "deepslot_f1": 734,
+  "delta": 735,
+  "delta_%": 736,
+  "dense_acc": 737,
+  "der_%": 738,
+  "description_accuracy": 739,
+  "detection_auroc": 740,
+  "detection_auroc_severity_0": 741,
+  "detection_rate": 742,
+  "deterministic_format_accuracy_\"exactamente_n\"": 743,
+  "dev16_cer": 744,
+  "dev16_wer": 745,
+  "dev_accuracy": 746,
+  "dev_cer": 747,
+  "dev_cer_+lm": 748,
+  "dev_cer_with_lm": 749,
+  "dev_cer_without_lm": 750,
+  "dev_macro_f1_score": 751,
+  "dev_macro_precision": 752,
+  "dev_macro_recall": 753,
+  "dev_wer": 754,
+  "dev_wer_+lm": 755,
+  "dev_wer_with_lm": 756,
+  "dev_wer_without_lm": 757,
+  "devops_relevance_score_0-10": 758,
+  "diagnostic_coherence_score": 759,
+  "dice": 760,
+  "dice-score": 761,
+  "dice_average": 762,
+  "dice_coefficient": 763,
+  "dice_score": 764,
+  "dim": 765,
+  "direct": 766,
+  "direction_accuracy_avg": 767,
+  "direction_accuracy_best": 768,
+  "distractor_accuracy": 769,
+  "diversity": 770,
+  "dnsmos_bak": 771,
+  "dnsmos_ovrl": 772,
+  "dnsmos_sig": 773,
+  "dot_accuracy": 774,
+  "dot_accuracy@1": 775,
+  "dot_accuracy@10": 776,
+  "dot_accuracy@100": 777,
+  "dot_accuracy@2": 778,
+  "dot_accuracy@3": 779,
+  "dot_accuracy@30": 780,
+  "dot_accuracy@5": 781,
+  "dot_accuracy@50": 782,
+  "dot_accuracy@8": 783,
+  "dot_accuracy_10": 784,
+  "dot_accuracy_threshold": 785,
+  "dot_ap": 786,
+  "dot_f1": 787,
+  "dot_f1_threshold": 788,
+  "dot_map@10": 789,
+  "dot_map@100": 790,
+  "dot_map@60": 791,
+  "dot_map_60": 792,
+  "dot_mcc": 793,
+  "dot_mrr@1": 794,
+  "dot_mrr@10": 795,
+  "dot_mrr@100": 796,
+  "dot_mrr@2": 797,
+  "dot_mrr@200": 798,
+  "dot_mrr@5": 799,
+  "dot_mrr_10": 800,
+  "dot_ndcg@1": 801,
+  "dot_ndcg@10": 802,
+  "dot_ndcg@100": 803,
+  "dot_ndcg@5": 804,
+  "dot_ndcg_10": 805,
+  "dot_pearson": 806,
+  "dot_precision": 807,
+  "dot_precision@1": 808,
+  "dot_precision@10": 809,
+  "dot_precision@100": 810,
+  "dot_precision@2": 811,
+  "dot_precision@3": 812,
+  "dot_precision@30": 813,
+  "dot_precision@5": 814,
+  "dot_precision@50": 815,
+  "dot_precision@8": 816,
+  "dot_precision_10": 817,
+  "dot_recall": 818,
+  "dot_recall@1": 819,
+  "dot_recall@10": 820,
+  "dot_recall@100": 821,
+  "dot_recall@2": 822,
+  "dot_recall@3": 823,
+  "dot_recall@30": 824,
+  "dot_recall@5": 825,
+  "dot_recall@50": 826,
+  "dot_recall@8": 827,
+  "dot_recall_10": 828,
+  "dot_score-map@100": 829,
+  "dot_score-mrr@10": 830,
+  "dot_score-ndcg@10": 831,
+  "dot_score-recall@5": 832,
+  "dot_score_accuracy@10": 833,
+  "dot_score_map@10": 834,
+  "dot_score_mrr@10": 835,
+  "dot_score_ndcg@10": 836,
+  "dot_score_precision@10": 837,
+  "dot_score_recall@10": 838,
+  "dot_sim_accuracy": 839,
+  "dot_sim_ap": 840,
+  "dot_spearman": 841,
+  "drilling_calculations_accuracy": 842,
+  "drop": 843,
+  "drop_3-shot": 844,
+  "drop_gen2mc_mc": 845,
+  "dropout": 846,
+  "ds_1000": 847,
+  "dsc": 848,
+  "dynamics_model_mse_loss": 849,
+  "e/i_accuracy": 850,
+  "eao": 851,
+  "ecthr_a": 852,
+  "edit-smiliarity": 853,
+  "eer": 854,
+  "eer_%": 855,
+  "element_iou": 856,
+  "elo": 857,
+  "elo_rating": 858,
+  "em@5_baseline": 859,
+  "em@5_with_instructions": 860,
+  "em_3-shot": 861,
+  "em_line-level": 862,
+  "em_maj1@1": 863,
+  "em_\u2264_8k": 864,
+  "embedding_dimension": 865,
+  "embedding_dropout": 866,
+  "emergence_detection_f1": 867,
+  "emergence_detection_rate": 868,
+  "emotion_top-3_accuracy": 869,
+  "emotionclassification": 870,
+  "empos": 871,
+  "emr": 872,
+  "en_content_to_title_acc": 873,
+  "en_title_to_content_acc": 874,
+  "engineering_document_retrieval_precision": 875,
+  "english_to_chinese": 876,
+  "english_to_sanskrit_translation_-_bleu_score": 877,
+  "english_to_sanskrit_translation_-_jaccard_similarity": 878,
+  "entity_span_f1_test_2020": 879,
+  "entity_span_f1_test_2021": 880,
+  "entity_span_precision_test_2020": 881,
+  "entity_span_recall_test_2020": 882,
+  "entity_span_recall_test_2021": 883,
+  "entropy": 884,
+  "entropy_novelty": 885,
+  "ents_f": 886,
+  "ents_p": 887,
+  "ents_r": 888,
+  "epe": 889,
+  "epoch": 890,
+  "eq-bench_0-shot": 891,
+  "eq-bench_score": 892,
+  "eqbench": 893,
+  "erqav2.0": 894,
+  "error": 895,
+  "error_rate": 896,
+  "error_ratio": 897,
+  "euclidean_accuracy": 898,
+  "euclidean_accuracy_threshold": 899,
+  "euclidean_ap": 900,
+  "euclidean_f1": 901,
+  "euclidean_f1_threshold": 902,
+  "euclidean_mcc": 903,
+  "euclidean_pearson": 904,
+  "euclidean_precision": 905,
+  "euclidean_recall": 906,
+  "euclidean_spearman": 907,
+  "eud_jaccard": 908,
+  "eval_accuracy": 909,
+  "eval_bertscore_f1": 910,
+  "eval_bleu": 911,
+  "eval_cer": 912,
+  "eval_chrf": 913,
+  "eval_em": 914,
+  "eval_exact": 915,
+  "eval_exactmatch_score_squad_metric": 916,
+  "eval_f1": 917,
+  "eval_f1_score_squad_metric": 918,
+  "eval_hasans_exact": 919,
+  "eval_hasans_f1": 920,
+  "eval_loss": 921,
+  "eval_loss_best": 922,
+  "eval_noans_exact": 923,
+  "eval_noans_f1": 924,
+  "eval_perplexity": 925,
+  "eval_precision": 926,
+  "eval_recall": 927,
+  "eval_runtime": 928,
+  "eval_samples_per_second": 929,
+  "eval_steps_per_second": 930,
+  "eval_time": 931,
+  "eval_wer": 932,
+  "evaluation_accuracy": 933,
+  "evaluation_loss": 934,
+  "evaluation_macro_f1": 935,
+  "evaluation_macro_precision": 936,
+  "evaluation_macro_recall": 937,
+  "evaluation_micro_f1": 938,
+  "evaluation_micro_precision": 939,
+  "evaluation_micro_recall": 940,
+  "evaluation_runtime_seconds": 941,
+  "evaluation_samples_per_second": 942,
+  "evaluation_steps_per_second": 943,
+  "evaluation_weighted_f1": 944,
+  "evaluation_weighted_precision": 945,
+  "evaluation_weighted_recall": 946,
+  "exact": 947,
+  "exact-match": 948,
+  "exact_macth": 949,
+  "exact_match": 950,
+  "exact_match@16k": 951,
+  "exact_match@32k": 952,
+  "exact_match@4k": 953,
+  "exact_match@8k": 954,
+  "exact_match_%": 955,
+  "exact_match_accuracy": 956,
+  "exact_match_accuracy_dev": 957,
+  "exact_match_accuracy_in_dev": 958,
+  "exact_match_em": 959,
+  "exact_match_flexible": 960,
+  "exact_match_flexible-extract": 961,
+  "exact_match_none": 962,
+  "exact_match_none_gsm8k_0shot_instruct": 963,
+  "exact_match_none_meta_math_0shot_instruct": 964,
+  "exact_match_none_meta_math_hard_0shot_instruct": 965,
+  "exact_match_strict": 966,
+  "exact_match_strict-match": 967,
+  "exact_match_strict-match_ceval-valid-pretrain-cot_zh": 968,
+  "exact_match_strict-match_cmmlu_pretrain_cot_zh": 969,
+  "exact_match_strict-match_meta_arc_0shot_instruct": 970,
+  "exact_match_strict-match_meta_bbh_3shot_cot_pretrain": 971,
+  "exact_match_strict-match_meta_gpqa_0shot_cot_instruct": 972,
+  "exact_match_strict-match_meta_mmlu_0shot_instruct": 973,
+  "exact_match_strict-match_meta_mmlu_pro_5shot_instruct": 974,
+  "exact_match_strict-match_meta_mmlu_pro_5shot_pretrain": 975,
+  "exact_match_strict-match_meta_pretrain": 976,
+  "exact_match_strict-match_original_capability_instruct": 977,
+  "exact_match_strict-match_zh_pretrain_multishot": 978,
+  "exact_span_f1": 979,
+  "exact_string_match": 980,
+  "example-level_f1": 981,
+  "example_f1": 982,
+  "execution_accuracy": 983,
+  "execution_accuracy_%_dev": 984,
+  "expected_average_overlap_eao": 985,
+  "expert_dim": 986,
+  "expert_effectiveness_score": 987,
+  "expert_rating": 988,
+  "expguardtest_total_f1": 989,
+  "extact_match": 990,
+  "extraction": 991,
+  "f-measure": 992,
+  "f-measure_mean": 993,
+  "f-measure_seen": 994,
+  "f-measure_unseen": 995,
+  "f-score": 996,
+  "f0.5": 997,
+  "f1": 998,
+  "f1-macro": 999,
+  "f1-score": 1000,
+  "f1-score_%": 1001,
+  "f1-score_dice_coefficient": 1002,
+  "f1-score_macro": 1003,
+  "f1-score_weighted": 1004,
+  "f1-weighted": 1005,
+  "f1@10": 1006,
+  "f1@5": 1007,
+  "f1@m": 1008,
+  "f1_%": 1009,
+  "f1_'bezeichnung'_macro": 1010,
+  "f1_'thema'_macro": 1011,
+  "f1_10-fold": 1012,
+  "f1_20-vote": 1013,
+  "f1_3-shot": 1014,
+  "f1_admiration": 1015,
+  "f1_af": 1016,
+  "f1_amusement": 1017,
+  "f1_anger": 1018,
+  "f1_annoyance": 1019,
+  "f1_approval": 1020,
+  "f1_avg": 1021,
+  "f1_caring": 1022,
+  "f1_class_negative": 1023,
+  "f1_class_positive": 1024,
+  "f1_confusion": 1025,
+  "f1_constructive": 1026,
+  "f1_covid-19": 1027,
+  "f1_curiosity": 1028,
+  "f1_desire": 1029,
+  "f1_disappointment": 1030,
+  "f1_disapproval": 1031,
+  "f1_disgust": 1032,
+  "f1_embarrassment": 1033,
+  "f1_entity_span": 1034,
+  "f1_excitement": 1035,
+  "f1_fear": 1036,
+  "f1_gratitude": 1037,
+  "f1_grief": 1038,
+  "f1_healthy": 1039,
+  "f1_instrument": 1040,
+  "f1_joy": 1041,
+  "f1_love": 1042,
+  "f1_macro": 1043,
+  "f1_macro_avg.": 1044,
+  "f1_micro": 1045,
+  "f1_micro_avg": 1046,
+  "f1_negative": 1047,
+  "f1_nervousness": 1048,
+  "f1_neutral": 1049,
+  "f1_nuclearity": 1050,
+  "f1_optimism": 1051,
+  "f1_positive": 1052,
+  "f1_pride": 1053,
+  "f1_r15": 1054,
+  "f1_r16": 1055,
+  "f1_realization": 1056,
+  "f1_relation": 1057,
+  "f1_relief": 1058,
+  "f1_remorse": 1059,
+  "f1_sadness": 1060,
+  "f1_samples": 1061,
+  "f1_score_%": 1062,
+  "f1_score_5-fold": 1063,
+  "f1_score_decimal": 1064,
+  "f1_score_macro": 1065,
+  "f1_score_macro_avg": 1066,
+  "f1_score_micro": 1067,
+  "f1_score_queue": 1068,
+  "f1_score_strong_class": 1069,
+  "f1_score_threshold=0.94": 1070,
+  "f1_score_toxic_class": 1071,
+  "f1_score_type": 1072,
+  "f1_score_weighted": 1073,
+  "f1_seqeval": 1074,
+  "f1_span": 1075,
+  "f1_stderr": 1076,
+  "f1_surprise": 1077,
+  "f1_symptomatic": 1078,
+  "f1_target": 1079,
+  "f1_test_2020": 1080,
+  "f1_test_2021": 1081,
+  "f1_threshold": 1082,
+  "f1_trolling": 1083,
+  "f1_verb": 1084,
+  "f1_weighted": 1085,
+  "f1_weighted_avg": 1086,
+  "f1_weighted_quantized": 1087,
+  "f1neg": 1088,
+  "f1pos": 1089,
+  "f2": 1090,
+  "factspotter": 1091,
+  "factual_accuracy": 1092,
+  "fad": 1093,
+  "fake_acc": 1094,
+  "false_accuracy": 1095,
+  "false_positive_rate": 1096,
+  "far": 1097,
+  "fast_1": 1098,
+  "few-shot": 1099,
+  "fid": 1100,
+  "fid_flexvar-d16_+sar": 1101,
+  "fid_flexvar-d20_+sar": 1102,
+  "fid_flexvar-d24_+sar": 1103,
+  "figure": 1104,
+  "final_em": 1105,
+  "final_eval_bertscore_f1": 1106,
+  "final_eval_bleu": 1107,
+  "final_eval_chrf": 1108,
+  "final_eval_loss": 1109,
+  "final_loss": 1110,
+  "final_test_wer": 1111,
+  "final_training_loss": 1112,
+  "final_validation_loss": 1113,
+  "finance_f1": 1114,
+  "first_pass_exact_match": 1115,
+  "first_turn": 1116,
+  "fitness": 1117,
+  "fl-all": 1118,
+  "fleurs-test-bleu": 1119,
+  "fleurs-test-cer": 1120,
+  "fleurs-test-wer": 1121,
+  "flexible-extract": 1122,
+  "float32_cosine_accuracy@1": 1123,
+  "float32_cosine_accuracy@10": 1124,
+  "float32_cosine_accuracy@3": 1125,
+  "float32_cosine_accuracy@5": 1126,
+  "float32_cosine_map@100": 1127,
+  "float32_cosine_mrr@10": 1128,
+  "float32_cosine_ndcg@10": 1129,
+  "float32_cosine_precision@1": 1130,
+  "float32_cosine_precision@10": 1131,
+  "float32_cosine_precision@3": 1132,
+  "float32_cosine_precision@5": 1133,
+  "float32_cosine_recall@1": 1134,
+  "float32_cosine_recall@10": 1135,
+  "float32_cosine_recall@3": 1136,
+  "float32_cosine_recall@5": 1137,
+  "fn": 1138,
+  "focalloss": 1139,
+  "format_compliance_rate": 1140,
+  "fp": 1141,
+  "fpr95": 1142,
+  "fps": 1143,
+  "fragmergent_coherence": 1144,
+  "frame_accuracy": 1145,
+  "framework_accuracy": 1146,
+  "frr": 1147,
+  "fscore": 1148,
+  "function_call_accuracy": 1149,
+  "function_calling_accuracy_name_&_arguments": 1150,
+  "funny_class_accuracy": 1151,
+  "fuzzy_score": 1152,
+  "fvd16": 1153,
+  "fw_iou": 1154,
+  "g": 1155,
+  "gen_len": 1156,
+  "gender_acc": 1157,
+  "gender_accuracy": 1158,
+  "gender_consistency": 1159,
+  "generated_length": 1160,
+  "generating_communicative_text.f1_score": 1161,
+  "generating_communicative_text.precision": 1162,
+  "generating_communicative_text.recall": 1163,
+  "generating_communicative_text.support": 1164,
+  "generating_creative_text.f1_score": 1165,
+  "generating_creative_text.precision": 1166,
+  "generating_creative_text.recall": 1167,
+  "generating_creative_text.support": 1168,
+  "gflops": 1169,
+  "global_accuracy": 1170,
+  "global_strict_f1": 1171,
+  "glue": 1172,
+  "go": 1173,
+  "google_speech_commands_v2_35": 1174,
+  "gp_test": 1175,
+  "gp_val": 1176,
+  "gpqa": 1177,
+  "gpt-3.5_score": 1178,
+  "gpt-4": 1179,
+  "gpt-4_as_judge": 1180,
+  "gpt-4_score": 1181,
+  "gpt-4_score_bbox": 1182,
+  "gpt-score": 1183,
+  "gpu_memory_usage_mb": 1184,
+  "group_score": 1185,
+  "grpo_accuracy": 1186,
+  "gsm8k": 1187,
+  "gsm8k_0-shot": 1188,
+  "gsm8k_5-shot": 1189,
+  "gsm8k_accuracy": 1190,
+  "gsm8k_few-shot": 1191,
+  "gsm8k_score": 1192,
+  "hallucination_f1": 1193,
+  "hallucination_rate": 1194,
+  "hallucination_reduction_%": 1195,
+  "hallucination_reduction_near-ood": 1196,
+  "hamming_accuracy": 1197,
+  "hamming_loss": 1198,
+  "hamming_score": 1199,
+  "hard": 1200,
+  "harmbench_f1": 1201,
+  "harmonic_mean": 1202,
+  "harmony_and_consonance": 1203,
+  "hasans_exact": 1204,
+  "hasans_f1": 1205,
+  "hasans_total": 1206,
+  "healthcare_f1": 1207,
+  "healthy_accuracy": 1208,
+  "hebrew_answers": 1209,
+  "hellaswag": 1210,
+  "hellaswag_0-shot": 1211,
+  "hellaswag_10-shot": 1212,
+  "hellaswag_rc": 1213,
+  "hellaswag_score": 1214,
+  "hhem_consistency": 1215,
+  "hit@10": 1216,
+  "hit@5": 1217,
+  "hits@1": 1218,
+  "hle": 1219,
+  "homework_problem.f1_score": 1220,
+  "homework_problem.precision": 1221,
+  "homework_problem.recall": 1222,
+  "homework_problem.support": 1223,
+  "hota": 1224,
+  "hota_all": 1225,
+  "human-gpt_detection_validation_loss": 1226,
+  "human_%": 1227,
+  "human_explanation_rating": 1228,
+  "human_preference_elo_rating": 1229,
+  "human_preference_rate": 1230,
+  "human_preference_vs_elevenlabs": 1231,
+  "humaneval": 1232,
+  "humaneval_pass@1": 1233,
+  "humanities": 1234,
+  "iae": 1235,
+  "icat_score": 1236,
+  "icbhi_score": 1237,
+  "idf1": 1238,
+  "ifbench": 1239,
+  "ifeval": 1240,
+  "image-to-sound_r@100": 1241,
+  "image-to-text_r@1": 1242,
+  "image-to-text_r@10": 1243,
+  "image-to-text_r@5": 1244,
+  "image_retrieval_r@1": 1245,
+  "imagenet_acc.": 1246,
+  "imagenet_dist._shift.": 1247,
+  "imagenet_top-1_accuracy": 1248,
+  "imagereward": 1249,
+  "implicit_social_group_reference_seqeval": 1250,
+  "improvement": 1251,
+  "in-1k_top-1_acc._%": 1252,
+  "in-1k_zero-shot_top-1_acc._%": 1253,
+  "inception_score": 1254,
+  "inference-latency_ms/sample": 1255,
+  "inference_latency_ms": 1256,
+  "inference_speed": 1257,
+  "inference_speed_sec": 1258,
+  "inference_steps": 1259,
+  "inference_success_rate": 1260,
+  "inference_text/sec_a100_40gb_gpu_batch=128": 1261,
+  "inference_text/sec_a100_40gb_gpu_batch=32": 1262,
+  "inference_text/sec_a100_batch=64": 1263,
+  "inference_text/sec_a10g_batch=128": 1264,
+  "inference_text/sec_a10g_gpu_batch=128": 1265,
+  "inference_time": 1266,
+  "inference_time_ms": 1267,
+  "information_retrieval": 1268,
+  "information_search.f1_score": 1269,
+  "information_search.precision": 1270,
+  "information_search.recall": 1271,
+  "information_search.support": 1272,
+  "inst-level_loose-accuracy": 1273,
+  "inst_level_loose_acc": 1274,
+  "inst_level_strict_acc": 1275,
+  "instruction-following-score": 1276,
+  "instruction_accuracy": 1277,
+  "instruction_level_loose_accuracy": 1278,
+  "instruction_level_strict_accuracy": 1279,
+  "int8_cosine_accuracy@1": 1280,
+  "int8_cosine_accuracy@10": 1281,
+  "int8_cosine_accuracy@3": 1282,
+  "int8_cosine_accuracy@5": 1283,
+  "int8_cosine_map@100": 1284,
+  "int8_cosine_mrr@10": 1285,
+  "int8_cosine_ndcg@10": 1286,
+  "int8_cosine_precision@1": 1287,
+  "int8_cosine_precision@10": 1288,
+  "int8_cosine_precision@3": 1289,
+  "int8_cosine_precision@5": 1290,
+  "int8_cosine_recall@1": 1291,
+  "int8_cosine_recall@10": 1292,
+  "int8_cosine_recall@3": 1293,
+  "int8_cosine_recall@5": 1294,
+  "intent_accuracy": 1295,
+  "intent_classification_macro_f1_%": 1296,
+  "intercode-alfa": 1297,
+  "internal_consistency": 1298,
+  "internal_tag_leakage": 1299,
+  "international_law": 1300,
+  "interpolation_error": 1301,
+  "intersection_over_union": 1302,
+  "introductory_pass@1": 1303,
+  "invalid_move_rate_imr": 1304,
+  "iou": 1305,
+  "iou_%": 1306,
+  "iou_agricultural_land": 1307,
+  "iou_bare_soil": 1308,
+  "iou_brushwood": 1309,
+  "iou_building": 1310,
+  "iou_buildings": 1311,
+  "iou_coniferous": 1312,
+  "iou_deciduous": 1313,
+  "iou_greenhouse": 1314,
+  "iou_herbaceous_vegetation": 1315,
+  "iou_impervious_surface": 1316,
+  "iou_jaccard_index": 1317,
+  "iou_pervious_surface": 1318,
+  "iou_plowed_land": 1319,
+  "iou_score": 1320,
+  "iou_snow": 1321,
+  "iou_swimming_pool": 1322,
+  "iou_vineyard": 1323,
+  "iou_water": 1324,
+  "ip_partial_f1": 1325,
+  "ip_strict_f1": 1326,
+  "is": 1327,
+  "isco_hierarchical_accuracy": 1328,
+  "ise": 1329,
+  "itae": 1330,
+  "j&f": 1331,
+  "j/p_accuracy": 1332,
+  "jaccard": 1333,
+  "jaccard_index": 1334,
+  "jaccard_seen": 1335,
+  "jeopardy": 1336,
+  "jeopardy_gen2mc_mc": 1337,
+  "joint_validation_accuracy": 1338,
+  "jurisprudence": 1339,
+  "kaggle_public_score_rmsle_best_submission": 1340,
+  "kannada_wer": 1341,
+  "kendall's_tau": 1342,
+  "kendall's_tau-c": 1343,
+  "kendall's_tau_coefficient": 1344,
+  "kl_divergence": 1345,
+  "korean_response_ratio": 1346,
+  "kv_partial_f1": 1347,
+  "kv_strict_f1": 1348,
+  "l2_error": 1349,
+  "l2q@15": 1350,
+  "labeled_attachment_score_las": 1351,
+  "labelled_attachment_score": 1352,
+  "lambada": 1353,
+  "lambada_acc": 1354,
+  "lambada_ppl": 1355,
+  "lambda": 1356,
+  "las": 1357,
+  "last_k_layers": 1358,
+  "latency_full": 1359,
+  "latency_in_seconds": 1360,
+  "latency_merging_ms": 1361,
+  "latency_ms": 1362,
+  "latency_ms/token": 1363,
+  "latency_ms_-_img": 1364,
+  "latency_ms_-_txt": 1365,
+  "latency_ms_img": 1366,
+  "latency_ms_img+txt": 1367,
+  "latency_ms_img_+_txt": 1368,
+  "latency_ms_txt": 1369,
+  "law_f1": 1370,
+  "lb_de_accuracy": 1371,
+  "lb_en_accuracy": 1372,
+  "lb_fr_accuracy": 1373,
+  "lbpp": 1374,
+  "lc_win_rate": 1375,
+  "lcr": 1376,
+  "ldm3d-sr-b_depth_mare": 1377,
+  "ldm3d-sr-b_fid": 1378,
+  "ldm3d-sr-b_is": 1379,
+  "ldm3d-sr-b_psnr": 1380,
+  "ldm3d-sr-b_ssim": 1381,
+  "lea": 1382,
+  "ledgar": 1383,
+  "lemma_accuracy": 1384,
+  "lemma_f1": 1385,
+  "length_controlled_winrate": 1386,
+  "livecodebench": 1387,
+  "loc_f1-score": 1388,
+  "loc_precision": 1389,
+  "loc_recall": 1390,
+  "localization": 1391,
+  "localization_ablation": 1392,
+  "log-likelihood": 1393,
+  "log-spectral_distance": 1394,
+  "log_fold_change_mae": 1395,
+  "log_loss": 1396,
+  "logistic_regression_accuracy": 1397,
+  "longbook_choice/acc": 1398,
+  "longbook_qa/f1": 1399,
+  "loss": 1400,
+  "lowest_loss": 1401,
+  "lpips": 1402,
+  "lpips_score": 1403,
+  "lrap": 1404,
+  "lstq": 1405,
+  "m3exam_acc": 1406,
+  "macc": 1407,
+  "macro": 1408,
+  "macro-average_f1-score": 1409,
+  "macro-averaged_f1": 1410,
+  "macro-f1": 1411,
+  "macro-precision": 1412,
+  "macro-recall": 1413,
+  "macro_accuracy": 1414,
+  "macro_auc": 1415,
+  "macro_avg": 1416,
+  "macro_avg/acc": 1417,
+  "macro_avg_f1-score": 1418,
+  "macro_f1": 1419,
+  "macro_f1-score": 1420,
+  "macro_f1_10-fold": 1421,
+  "macro_f1_3_conditions": 1422,
+  "macro_f1_avg": 1423,
+  "macro_f1_cardiffnlp/tweet_sentiment_multilingual/all": 1424,
+  "macro_f1_cardiffnlp/tweet_topic_multi": 1425,
+  "macro_f1_cardiffnlp/tweet_topic_single": 1426,
+  "macro_f1_score": 1427,
+  "macro_f1_test_2020": 1428,
+  "macro_f1_test_2021": 1429,
+  "macro_f1_top_5_conditions": 1430,
+  "macro_f1_tweet_eval/emoji": 1431,
+  "macro_f1_tweet_eval/emotion": 1432,
+  "macro_f1_tweet_eval/hate": 1433,
+  "macro_f1_tweet_eval/irony": 1434,
+  "macro_f1_tweet_eval/offensive": 1435,
+  "macro_f1_tweet_eval/sentiment": 1436,
+  "macro_p": 1437,
+  "macro_precision": 1438,
+  "macro_precision_test_2020": 1439,
+  "macro_precision_test_2021": 1440,
+  "macro_r": 1441,
+  "macro_recall": 1442,
+  "macro_recall_test_2020": 1443,
+  "macro_recall_test_2021": 1444,
+  "macs_image+text_g": 1445,
+  "mad": 1446,
+  "mae": 1447,
+  "mae_60_min": 1448,
+  "mae_alpha": 1449,
+  "mae_original_scale_-2_to_+2": 1450,
+  "mae_original_scale_0-3": 1451,
+  "main_score": 1452,
+  "maj@1": 1453,
+  "maj@16": 1454,
+  "manhattan_accuracy": 1455,
+  "manhattan_accuracy_threshold": 1456,
+  "manhattan_ap": 1457,
+  "manhattan_f1": 1458,
+  "manhattan_f1_threshold": 1459,
+  "manhattan_mcc": 1460,
+  "manhattan_pearson": 1461,
+  "manhattan_precision": 1462,
+  "manhattan_recall": 1463,
+  "manhattan_spearman": 1464,
+  "map": 1465,
+  "map50": 1466,
+  "map50-95": 1467,
+  "map@0.25": 1468,
+  "map@0.5": 1469,
+  "map@0.50": 1470,
+  "map@0.5:0.95": 1471,
+  "map@0.5_box": 1472,
+  "map@0.5_mask": 1473,
+  "map@0.75": 1474,
+  "map@1": 1475,
+  "map@10": 1476,
+  "map@100": 1477,
+  "map@1000": 1478,
+  "map@1000_miracl": 1479,
+  "map@100_miracl": 1480,
+  "map@10_miracl": 1481,
+  "map@1_miracl": 1482,
+  "map@2": 1483,
+  "map@20": 1484,
+  "map@200": 1485,
+  "map@20_miracl": 1486,
+  "map@3": 1487,
+  "map@30": 1488,
+  "map@300": 1489,
+  "map@3_miracl": 1490,
+  "map@5": 1491,
+  "map@50": 1492,
+  "map@50-95": 1493,
+  "map@500": 1494,
+  "map@5_miracl": 1495,
+  "map@7": 1496,
+  "map@70": 1497,
+  "map@700": 1498,
+  "map@75": 1499,
+  "map@_iou=0.50:0.95": 1500,
+  "map_l": 1501,
+  "map_m": 1502,
+  "map_micro": 1503,
+  "map_rn50": 1504,
+  "map_s": 1505,
+  "map_val": 1506,
+  "map_vit-b/16": 1507,
+  "maph/l2": 1508,
+  "mare": 1509,
+  "mask_ap": 1510,
+  "matched": 1511,
+  "math": 1512,
+  "math_500": 1513,
+  "math_level_5": 1514,
+  "math_verify": 1515,
+  "mathew's_coefficient": 1516,
+  "matthews_correlation": 1517,
+  "matthews_correlation_coefficient": 1518,
+  "mauve": 1519,
+  "max_accuracy": 1520,
+  "max_accuracy_threshold": 1521,
+  "max_ap": 1522,
+  "max_error_alpha": 1523,
+  "max_f1": 1524,
+  "max_f1_threshold": 1525,
+  "max_mcc": 1526,
+  "max_precision": 1527,
+  "max_recall": 1528,
+  "max_reward": 1529,
+  "maxfm": 1530,
+  "maxsim_accuracy@1": 1531,
+  "maxsim_accuracy@10": 1532,
+  "maxsim_accuracy@3": 1533,
+  "maxsim_accuracy@5": 1534,
+  "maxsim_map@100": 1535,
+  "maxsim_mrr@10": 1536,
+  "maxsim_ndcg@10": 1537,
+  "maxsim_precision@1": 1538,
+  "maxsim_precision@10": 1539,
+  "maxsim_precision@3": 1540,
+  "maxsim_precision@5": 1541,
+  "maxsim_recall@1": 1542,
+  "maxsim_recall@10": 1543,
+  "maxsim_recall@3": 1544,
+  "maxsim_recall@5": 1545,
+  "mbpp": 1546,
+  "mbpp_pass@1": 1547,
+  "mc1": 1548,
+  "mc1_accuracy": 1549,
+  "mc1_accuracy_stderr": 1550,
+  "mc2": 1551,
+  "mc2_accuracy": 1552,
+  "mc2_accuracy_stderr": 1553,
+  "mcap": 1554,
+  "mcc": 1555,
+  "mean": 1556,
+  "mean-ep-length": 1557,
+  "mean-reward": 1558,
+  "mean@1": 1559,
+  "mean_absolute_error": 1560,
+  "mean_absolute_error_mae": 1561,
+  "mean_accuracy": 1562,
+  "mean_ap": 1563,
+  "mean_auc@5\u00b0": 1564,
+  "mean_average_precision": 1565,
+  "mean_average_precision@iou_0.50": 1566,
+  "mean_average_precision@iou_0.75": 1567,
+  "mean_average_precision_iou=0.5": 1568,
+  "mean_average_precision_iou=0.5:0.95": 1569,
+  "mean_average_precision_map@50": 1570,
+  "mean_average_precision_map@50-95": 1571,
+  "mean_corruption_error_mce": 1572,
+  "mean_dice": 1573,
+  "mean_episode_length": 1574,
+  "mean_error_px": 1575,
+  "mean_f1_intermediate": 1576,
+  "mean_iou": 1577,
+  "mean_iou_class": 1578,
+  "mean_opinion_score": 1579,
+  "mean_opinion_score_mos": 1580,
+  "mean_p_ai": 1581,
+  "mean_rating": 1582,
+  "mean_recall": 1583,
+  "mean_reciprocal_rank": 1584,
+  "mean_reconstruction_error_mm": 1585,
+  "mean_regret_\u03b4wp_late_&_close": 1586,
+  "mean_regret_\u03b4wp_overall": 1587,
+  "mean_reward": 1588,
+  "mean_reward_20_episodes": 1589,
+  "mean_rmse_multi-head": 1590,
+  "mean_ru": 1591,
+  "mean_squared_error": 1592,
+  "mean_squared_error_for_ordinal_data": 1593,
+  "mean_token_accuracy": 1594,
+  "median_absolute_error_mdae": 1595,
+  "medical_keyword_coverage": 1596,
+  "medical_q&a": 1597,
+  "medmcqa_mc": 1598,
+  "medqa_mc": 1599,
+  "membrane": 1600,
+  "memory_efficiency": 1601,
+  "memory_efficiency_improvement_x": 1602,
+  "memory_footprint_mb": 1603,
+  "memory_peak_mb": 1604,
+  "memory_reduction_vs_fp32_baseline_%": 1605,
+  "mer": 1606,
+  "meteor": 1607,
+  "metric": 1608,
+  "micro": 1609,
+  "micro-f1": 1610,
+  "micro-f1_score": 1611,
+  "micro-f1_strong": 1612,
+  "micro-precision": 1613,
+  "micro-recall": 1614,
+  "micro_auc": 1615,
+  "micro_avg/rougel": 1616,
+  "micro_f1": 1617,
+  "micro_f1_cardiffnlp/tweet_sentiment_multilingual/all": 1618,
+  "micro_f1_cardiffnlp/tweet_topic_multi": 1619,
+  "micro_f1_cardiffnlp/tweet_topic_single": 1620,
+  "micro_f1_optimized_thresholds": 1621,
+  "micro_f1_score": 1622,
+  "micro_f1_tweet_eval/emoji": 1623,
+  "micro_f1_tweet_eval/emotion": 1624,
+  "micro_f1_tweet_eval/hate": 1625,
+  "micro_f1_tweet_eval/irony": 1626,
+  "micro_f1_tweet_eval/offensive": 1627,
+  "micro_f1_tweet_eval/sentiment": 1628,
+  "micro_precision": 1629,
+  "micro_recall": 1630,
+  "min_reward": 1631,
+  "miou": 1632,
+  "miou_13_classes": 1633,
+  "miou_6-fold": 1634,
+  "miou_after_lora": 1635,
+  "miou_before_lora": 1636,
+  "miou_real": 1637,
+  "miou_test": 1638,
+  "miouparts": 1639,
+  "misc_f1-score": 1640,
+  "misc_precision": 1641,
+  "misc_recall": 1642,
+  "miscs_f1": 1643,
+  "mixture_accuracy": 1644,
+  "mlm_accuracy": 1645,
+  "mmlu": 1646,
+  "mmlu-pem_0-shot": 1647,
+  "mmlu_5-shot": 1648,
+  "mmlu_accuracy": 1649,
+  "mmlu_high_school_european_history": 1650,
+  "mmlu_high_school_us_history": 1651,
+  "mmlu_high_school_world_history": 1652,
+  "mmlu_humanities": 1653,
+  "mmlu_jurisprudence": 1654,
+  "mmlu_logical_fallacies": 1655,
+  "mmlu_moral_disputes": 1656,
+  "mmlu_other": 1657,
+  "mmlu_overall": 1658,
+  "mmlu_pro": 1659,
+  "mmlu_pro_mc": 1660,
+  "mmlu_score": 1661,
+  "mmlu_social_sci.": 1662,
+  "mmlu_stem": 1663,
+  "mmmlu_de_de_0-shot": 1664,
+  "mmmlu_de_de_5-shot": 1665,
+  "model-parameter": 1666,
+  "model-parameters-reduction_%": 1667,
+  "model_loss": 1668,
+  "model_score": 1669,
+  "model_size_kb": 1670,
+  "modelnet40_average": 1671,
+  "molecule_uniqueness_rate": 1672,
+  "morph_ufeats_accuracy": 1673,
+  "morphology_f1": 1674,
+  "mota": 1675,
+  "mp-lpips": 1676,
+  "mpjpe": 1677,
+  "mprec": 1678,
+  "mrr": 1679,
+  "mrr@1": 1680,
+  "mrr@10": 1681,
+  "mrr@100": 1682,
+  "mrr@1000": 1683,
+  "mrr@2": 1684,
+  "mrr@20": 1685,
+  "mrr@200": 1686,
+  "mrr@3": 1687,
+  "mrr@30": 1688,
+  "mrr@300": 1689,
+  "mrr@5": 1690,
+  "mrr@50": 1691,
+  "mrr@500": 1692,
+  "mrr@7": 1693,
+  "mrr@70": 1694,
+  "mrr@700": 1695,
+  "mrr_1": 1696,
+  "mrr_10": 1697,
+  "mrr_5": 1698,
+  "mrr_baseline": 1699,
+  "mrr_on_abr_core_exam_chest": 1700,
+  "mrr_with_bi-encoder": 1701,
+  "mrr_with_full_pipeline": 1702,
+  "mrr_with_instructions": 1703,
+  "mse": 1704,
+  "mse_loss": 1705,
+  "mse_masked;_dims=x/y": 1706,
+  "mt-bench": 1707,
+  "mt-bench_score": 1708,
+  "mt-bench_win_rate_adjusted_%": 1709,
+  "mtbench": 1710,
+  "multilabel_accuracy": 1711,
+  "multilabel_roc_auc": 1712,
+  "multipl_humaneval": 1713,
+  "multipl_mbppp": 1714,
+  "music_accuracy": 1715,
+  "musicality": 1716,
+  "mwap": 1717,
+  "n_embd": 1718,
+  "n_evaluation_episodes": 1719,
+  "n_head": 1720,
+  "n_layer": 1721,
+  "n_samples": 1722,
+  "n_test_samples": 1723,
+  "naive_bayes_accuracy": 1724,
+  "named_entity_linking_f_score": 1725,
+  "named_entity_linking_precision": 1726,
+  "named_entity_linking_recall": 1727,
+  "naturalqs": 1728,
+  "naturalqs_gen2mc_mc": 1729,
+  "nauc_map@1000_diff1": 1730,
+  "nauc_map@1000_diff1_miracl": 1731,
+  "nauc_map@1000_max": 1732,
+  "nauc_map@1000_max_miracl": 1733,
+  "nauc_map@1000_std": 1734,
+  "nauc_map@1000_std_miracl": 1735,
+  "nauc_map@100_diff1": 1736,
+  "nauc_map@100_diff1_miracl": 1737,
+  "nauc_map@100_max": 1738,
+  "nauc_map@100_max_miracl": 1739,
+  "nauc_map@100_std": 1740,
+  "nauc_map@100_std_miracl": 1741,
+  "nauc_map@10_diff1": 1742,
+  "nauc_map@10_diff1_miracl": 1743,
+  "nauc_map@10_max": 1744,
+  "nauc_map@10_max_miracl": 1745,
+  "nauc_map@10_std": 1746,
+  "nauc_map@10_std_miracl": 1747,
+  "nauc_map@1_diff1": 1748,
+  "nauc_map@1_diff1_miracl": 1749,
+  "nauc_map@1_max": 1750,
+  "nauc_map@1_max_miracl": 1751,
+  "nauc_map@1_std": 1752,
+  "nauc_map@1_std_miracl": 1753,
+  "nauc_map@20_diff1": 1754,
+  "nauc_map@20_diff1_miracl": 1755,
+  "nauc_map@20_max": 1756,
+  "nauc_map@20_max_miracl": 1757,
+  "nauc_map@20_std": 1758,
+  "nauc_map@20_std_miracl": 1759,
+  "nauc_map@3_diff1": 1760,
+  "nauc_map@3_diff1_miracl": 1761,
+  "nauc_map@3_max": 1762,
+  "nauc_map@3_max_miracl": 1763,
+  "nauc_map@3_std": 1764,
+  "nauc_map@3_std_miracl": 1765,
+  "nauc_map@5_diff1": 1766,
+  "nauc_map@5_diff1_miracl": 1767,
+  "nauc_map@5_max": 1768,
+  "nauc_map@5_max_miracl": 1769,
+  "nauc_map@5_std": 1770,
+  "nauc_map@5_std_miracl": 1771,
+  "nauc_map_diff1": 1772,
+  "nauc_map_max": 1773,
+  "nauc_map_std": 1774,
+  "nauc_mrr@1000_diff1": 1775,
+  "nauc_mrr@1000_max": 1776,
+  "nauc_mrr@1000_std": 1777,
+  "nauc_mrr@100_diff1": 1778,
+  "nauc_mrr@100_max": 1779,
+  "nauc_mrr@100_std": 1780,
+  "nauc_mrr@10_diff1": 1781,
+  "nauc_mrr@10_max": 1782,
+  "nauc_mrr@10_std": 1783,
+  "nauc_mrr@1_diff1": 1784,
+  "nauc_mrr@1_max": 1785,
+  "nauc_mrr@1_std": 1786,
+  "nauc_mrr@20_diff1": 1787,
+  "nauc_mrr@20_max": 1788,
+  "nauc_mrr@20_std": 1789,
+  "nauc_mrr@3_diff1": 1790,
+  "nauc_mrr@3_max": 1791,
+  "nauc_mrr@3_std": 1792,
+  "nauc_mrr@5_diff1": 1793,
+  "nauc_mrr@5_max": 1794,
+  "nauc_mrr@5_std": 1795,
+  "nauc_mrr_diff1": 1796,
+  "nauc_mrr_max": 1797,
+  "nauc_mrr_std": 1798,
+  "nauc_ndcg@1000_diff1": 1799,
+  "nauc_ndcg@1000_diff1_miracl": 1800,
+  "nauc_ndcg@1000_max": 1801,
+  "nauc_ndcg@1000_max_miracl": 1802,
+  "nauc_ndcg@1000_std": 1803,
+  "nauc_ndcg@1000_std_miracl": 1804,
+  "nauc_ndcg@100_diff1": 1805,
+  "nauc_ndcg@100_diff1_miracl": 1806,
+  "nauc_ndcg@100_max": 1807,
+  "nauc_ndcg@100_max_miracl": 1808,
+  "nauc_ndcg@100_std": 1809,
+  "nauc_ndcg@100_std_miracl": 1810,
+  "nauc_ndcg@10_diff1": 1811,
+  "nauc_ndcg@10_diff1_miracl": 1812,
+  "nauc_ndcg@10_max": 1813,
+  "nauc_ndcg@10_max_miracl": 1814,
+  "nauc_ndcg@10_std": 1815,
+  "nauc_ndcg@10_std_miracl": 1816,
+  "nauc_ndcg@1_diff1": 1817,
+  "nauc_ndcg@1_diff1_miracl": 1818,
+  "nauc_ndcg@1_max": 1819,
+  "nauc_ndcg@1_max_miracl": 1820,
+  "nauc_ndcg@1_std": 1821,
+  "nauc_ndcg@1_std_miracl": 1822,
+  "nauc_ndcg@20_diff1": 1823,
+  "nauc_ndcg@20_diff1_miracl": 1824,
+  "nauc_ndcg@20_max": 1825,
+  "nauc_ndcg@20_max_miracl": 1826,
+  "nauc_ndcg@20_std": 1827,
+  "nauc_ndcg@20_std_miracl": 1828,
+  "nauc_ndcg@3_diff1": 1829,
+  "nauc_ndcg@3_diff1_miracl": 1830,
+  "nauc_ndcg@3_max": 1831,
+  "nauc_ndcg@3_max_miracl": 1832,
+  "nauc_ndcg@3_std": 1833,
+  "nauc_ndcg@3_std_miracl": 1834,
+  "nauc_ndcg@5_diff1": 1835,
+  "nauc_ndcg@5_diff1_miracl": 1836,
+  "nauc_ndcg@5_max": 1837,
+  "nauc_ndcg@5_max_miracl": 1838,
+  "nauc_ndcg@5_std": 1839,
+  "nauc_ndcg@5_std_miracl": 1840,
+  "nauc_p@1000_diff1_miracl": 1841,
+  "nauc_p@1000_max_miracl": 1842,
+  "nauc_p@1000_std_miracl": 1843,
+  "nauc_p@100_diff1_miracl": 1844,
+  "nauc_p@100_max_miracl": 1845,
+  "nauc_p@100_std_miracl": 1846,
+  "nauc_p@10_diff1_miracl": 1847,
+  "nauc_p@10_max_miracl": 1848,
+  "nauc_p@10_std_miracl": 1849,
+  "nauc_p@1_diff1_miracl": 1850,
+  "nauc_p@1_max_miracl": 1851,
+  "nauc_p@1_std_miracl": 1852,
+  "nauc_p@20_diff1_miracl": 1853,
+  "nauc_p@20_max_miracl": 1854,
+  "nauc_p@20_std_miracl": 1855,
+  "nauc_p@3_diff1_miracl": 1856,
+  "nauc_p@3_max_miracl": 1857,
+  "nauc_p@3_std_miracl": 1858,
+  "nauc_p@5_diff1_miracl": 1859,
+  "nauc_p@5_max_miracl": 1860,
+  "nauc_p@5_std_miracl": 1861,
+  "nauc_precision@1000_diff1": 1862,
+  "nauc_precision@1000_max": 1863,
+  "nauc_precision@1000_std": 1864,
+  "nauc_precision@100_diff1": 1865,
+  "nauc_precision@100_max": 1866,
+  "nauc_precision@100_std": 1867,
+  "nauc_precision@10_diff1": 1868,
+  "nauc_precision@10_max": 1869,
+  "nauc_precision@10_std": 1870,
+  "nauc_precision@1_diff1": 1871,
+  "nauc_precision@1_max": 1872,
+  "nauc_precision@1_std": 1873,
+  "nauc_precision@20_diff1": 1874,
+  "nauc_precision@20_max": 1875,
+  "nauc_precision@20_std": 1876,
+  "nauc_precision@3_diff1": 1877,
+  "nauc_precision@3_max": 1878,
+  "nauc_precision@3_std": 1879,
+  "nauc_precision@5_diff1": 1880,
+  "nauc_precision@5_max": 1881,
+  "nauc_precision@5_std": 1882,
+  "nauc_recall@1000_diff1": 1883,
+  "nauc_recall@1000_diff1_miracl": 1884,
+  "nauc_recall@1000_max": 1885,
+  "nauc_recall@1000_max_miracl": 1886,
+  "nauc_recall@1000_std": 1887,
+  "nauc_recall@1000_std_miracl": 1888,
+  "nauc_recall@100_diff1": 1889,
+  "nauc_recall@100_diff1_miracl": 1890,
+  "nauc_recall@100_max": 1891,
+  "nauc_recall@100_max_miracl": 1892,
+  "nauc_recall@100_std": 1893,
+  "nauc_recall@100_std_miracl": 1894,
+  "nauc_recall@10_diff1": 1895,
+  "nauc_recall@10_diff1_miracl": 1896,
+  "nauc_recall@10_max": 1897,
+  "nauc_recall@10_max_miracl": 1898,
+  "nauc_recall@10_std": 1899,
+  "nauc_recall@10_std_miracl": 1900,
+  "nauc_recall@1_diff1": 1901,
+  "nauc_recall@1_diff1_miracl": 1902,
+  "nauc_recall@1_max": 1903,
+  "nauc_recall@1_max_miracl": 1904,
+  "nauc_recall@1_std": 1905,
+  "nauc_recall@1_std_miracl": 1906,
+  "nauc_recall@20_diff1": 1907,
+  "nauc_recall@20_diff1_miracl": 1908,
+  "nauc_recall@20_max": 1909,
+  "nauc_recall@20_max_miracl": 1910,
+  "nauc_recall@20_std": 1911,
+  "nauc_recall@20_std_miracl": 1912,
+  "nauc_recall@3_diff1": 1913,
+  "nauc_recall@3_diff1_miracl": 1914,
+  "nauc_recall@3_max": 1915,
+  "nauc_recall@3_max_miracl": 1916,
+  "nauc_recall@3_std": 1917,
+  "nauc_recall@3_std_miracl": 1918,
+  "nauc_recall@5_diff1": 1919,
+  "nauc_recall@5_diff1_miracl": 1920,
+  "nauc_recall@5_max": 1921,
+  "nauc_recall@5_max_miracl": 1922,
+  "nauc_recall@5_std": 1923,
+  "nauc_recall@5_std_miracl": 1924,
+  "ndcg": 1925,
+  "ndcg@1": 1926,
+  "ndcg@10": 1927,
+  "ndcg@100": 1928,
+  "ndcg@1000": 1929,
+  "ndcg@1000_miracl": 1930,
+  "ndcg@100_miracl": 1931,
+  "ndcg@10_miracl": 1932,
+  "ndcg@1_miracl": 1933,
+  "ndcg@2": 1934,
+  "ndcg@20": 1935,
+  "ndcg@200": 1936,
+  "ndcg@20_baseline": 1937,
+  "ndcg@20_miracl": 1938,
+  "ndcg@20_with_instructions": 1939,
+  "ndcg@3": 1940,
+  "ndcg@30": 1941,
+  "ndcg@300": 1942,
+  "ndcg@3_miracl": 1943,
+  "ndcg@5": 1944,
+  "ndcg@50": 1945,
+  "ndcg@500": 1946,
+  "ndcg@5_miracl": 1947,
+  "ndcg@7": 1948,
+  "ndcg@70": 1949,
+  "ndcg@700": 1950,
+  "nds": 1951,
+  "ndtw_val_unseen": 1952,
+  "negative_mse": 1953,
+  "negatives": 1954,
+  "ner_f1_score": 1955,
+  "ner_f_score": 1956,
+  "ner_precision": 1957,
+  "ner_recall": 1958,
+  "niqe": 1959,
+  "nmi": 1960,
+  "noans_exact": 1961,
+  "noans_f1": 1962,
+  "noans_total": 1963,
+  "noc@85": 1964,
+  "noc@90": 1965,
+  "non-degradation_rate": 1966,
+  "normalized_accuracy_acc_norm": 1967,
+  "normalized_accuracy_stderr": 1968,
+  "normalized_cer": 1969,
+  "normalized_levenshtein_distance": 1970,
+  "normalized_levenshtein_similarity": 1971,
+  "normalized_return": 1972,
+  "normalized_score_iqm_95%_ci": 1973,
+  "normalized_wer": 1974,
+  "note-level_f-measure-no-offset_fno": 1975,
+  "noun_top5_map": 1976,
+  "npv": 1977,
+  "null_f1": 1978,
+  "num_active_experts": 1979,
+  "num_experts": 1980,
+  "num_gpus": 1981,
+  "num_tokens": 1982,
+  "number_accuracy": 1983,
+  "number_of_params": 1984,
+  "number_of_tokens": 1985,
+  "numbers_accuracy_%": 1986,
+  "objaverse_average": 1987,
+  "object_count": 1988,
+  "object_persistence_\u22655_frames": 1989,
+  "object_precision": 1990,
+  "object_recall": 1991,
+  "object_size": 1992,
+  "off-domain_citations": 1993,
+  "off_by_1_accuracy": 1994,
+  "olmo_3-eval_code": 1995,
+  "olmo_3-eval_genqa": 1996,
+  "olmo_3-eval_math": 1997,
+  "olmo_3-eval_mc_non-stem": 1998,
+  "olmo_3-eval_mc_stem": 1999,
+  "openbookqa": 2000,
+  "openthaigpt": 2001,
+  "org_f1-score": 2002,
+  "org_precision": 2003,
+  "org_recall": 2004,
+  "organization_public_institution_or_collective_actor_seqeval": 2005,
+  "original_accuracy": 2006,
+  "oscillation_count": 2007,
+  "other": 2008,
+  "other_accuracy": 2009,
+  "overall": 2010,
+  "overall_accuarcy": 2011,
+  "overall_accuracy": 2012,
+  "overall_devops_accuracy": 2013,
+  "overall_f1": 2014,
+  "overall_f1_weighted_avg": 2015,
+  "overall_iou": 2016,
+  "overall_match": 2017,
+  "overall_precision": 2018,
+  "overall_precision_weighted_avg": 2019,
+  "overall_recall": 2020,
+  "overall_recall_weighted_avg": 2021,
+  "overall_satisfaction_live": 2022,
+  "overall_satisfaction_stress": 2023,
+  "overall_score": 2024,
+  "overall_success_rate": 2025,
+  "overall_test_accuracy": 2026,
+  "overall_wer": 2027,
+  "overshoot_%": 2028,
+  "p": 2029,
+  "p-mrr": 2030,
+  "p@1": 2031,
+  "p@10": 2032,
+  "p@1000_miracl": 2033,
+  "p@100_miracl": 2034,
+  "p@10_baseline": 2035,
+  "p@10_miracl": 2036,
+  "p@10_with_instructions": 2037,
+  "p@1_miracl": 2038,
+  "p@20": 2039,
+  "p@20_miracl": 2040,
+  "p@3_miracl": 2041,
+  "p@5": 2042,
+  "p@5_miracl": 2043,
+  "p@m": 2044,
+  "pairwise_accuracy": 2045,
+  "paralux_accuracy": 2046,
+  "parameter_count": 2047,
+  "parameters": 2048,
+  "params_img_m": 2049,
+  "params_m_-_img": 2050,
+  "params_m_-_txt": 2051,
+  "params_m_img": 2052,
+  "params_m_txt": 2053,
+  "params_txt_m": 2054,
+  "partial_score": 2055,
+  "particles_accuracy_%": 2056,
+  "partpq": 2057,
+  "pass@1": 2058,
+  "pass@10": 2059,
+  "pass@100": 2060,
+  "pass@100_t=0.8": 2061,
+  "pass@10_java": 2062,
+  "pass@10_javascript": 2063,
+  "pass@10_python": 2064,
+  "pass@10_t=0.8": 2065,
+  "pass@16": 2066,
+  "pass@1_0-shot_cot": 2067,
+  "pass@1_avg16": 2068,
+  "pass@1_code_generation": 2069,
+  "pass@1_function_completion": 2070,
+  "pass@1_java": 2071,
+  "pass@1_javascript": 2072,
+  "pass@1_multimodal": 2073,
+  "pass@1_n=1_code_instruct": 2074,
+  "pass@1_n=1_humaneval_greedy_instruct": 2075,
+  "pass@1_n=1_humaneval_plus_greedy_instruct": 2076,
+  "pass@1_n=1_mbpp_plus_0shot_instruct": 2077,
+  "pass@1_n=1_mbpp_sanitized_0shot_instruct": 2078,
+  "pass@1_overall": 2079,
+  "pass@1_python": 2080,
+  "pass@1_t=0.01": 2081,
+  "pass@1_t=0.1": 2082,
+  "pass@1_t=0.2": 2083,
+  "pass@1_thresh=0.5": 2084,
+  "pass@3": 2085,
+  "pass@32": 2086,
+  "pass@4": 2087,
+  "pass@4_overall": 2088,
+  "pck@0.2": 2089,
+  "pck@0.3_ood": 2090,
+  "pckh-0.5": 2091,
+  "pckh@0.1": 2092,
+  "peak_time_s": 2093,
+  "pearson": 2094,
+  "pearson's_r_distress": 2095,
+  "pearson's_r_empathy": 2096,
+  "pearson_correlation": 2097,
+  "pearson_correlation_-_stsb_multi_mt_fr": 2098,
+  "pearson_correlation_cosine_similarity": 2099,
+  "pearson_cosine": 2100,
+  "pearson_dot": 2101,
+  "pearson_euclidean": 2102,
+  "pearson_manhattan": 2103,
+  "pearson_max": 2104,
+  "pearson_spearman_avg": 2105,
+  "pearsonr": 2106,
+  "pearsonr_dynamic_8b": 2107,
+  "pearsonr_onnx": 2108,
+  "pearsonr_optimized": 2109,
+  "pearsonr_static_8b": 2110,
+  "per-class_accuracy": 2111,
+  "per-joint_success_rate_5%_tolerance": 2112,
+  "per_f1-score": 2113,
+  "per_precision": 2114,
+  "per_recall": 2115,
+  "percent_parseable": 2116,
+  "percentage_correct": 2117,
+  "percentage_error": 2118,
+  "percentile": 2119,
+  "percision": 2120,
+  "performance_index": 2121,
+  "performance_semantic_search_6_datasets": 2122,
+  "performance_sentence_embeddings_14_datasets": 2123,
+  "perplexity": 2124,
+  "perplexity_baseline": 2125,
+  "perplexity_basic": 2126,
+  "perplexity_best_checkpoint": 2127,
+  "perplexity_gpt-2_baseline": 2128,
+  "perplexity_ibce": 2129,
+  "perplexity_mean_evaluation": 2130,
+  "perplexity_wip": 2131,
+  "perplexity_\u2193": 2132,
+  "pesq": 2133,
+  "phd_evaluation_score_/100": 2134,
+  "phone_error_rate": 2135,
+  "phoneme_error_rate": 2136,
+  "phoneme_error_rate_per_%": 2137,
+  "phoneme_group_error_rate": 2138,
+  "physical_cores": 2139,
+  "piqa": 2140,
+  "piqa_mc": 2141,
+  "pixel_accuracy": 2142,
+  "placeholder_metric_for_development": 2143,
+  "policy_agreement_late_&_close": 2144,
+  "policy_agreement_top-\u03b4wp": 2145,
+  "political_group_seqeval": 2146,
+  "political_institution_seqeval": 2147,
+  "pooling_attention_dropout": 2148,
+  "pos-level0": 2149,
+  "pos_upos_accuracy": 2150,
+  "poseval": 2151,
+  "positives": 2152,
+  "ppl": 2153,
+  "ppl_per_million_parameters": 2154,
+  "ppv": 2155,
+  "ppv_precision": 2156,
+  "pq": 2157,
+  "pqst": 2158,
+  "pr-auc": 2159,
+  "pr_auc": 2160,
+  "pre@10": 2161,
+  "prec@1": 2162,
+  "precision": 2163,
+  "precision-macro": 2164,
+  "precision@1": 2165,
+  "precision@10": 2166,
+  "precision@100": 2167,
+  "precision@1000": 2168,
+  "precision@2": 2169,
+  "precision@20": 2170,
+  "precision@200": 2171,
+  "precision@3": 2172,
+  "precision@30": 2173,
+  "precision@300": 2174,
+  "precision@5": 2175,
+  "precision@50": 2176,
+  "precision@500": 2177,
+  "precision@7": 2178,
+  "precision@70": 2179,
+  "precision@700": 2180,
+  "precision_%": 2181,
+  "precision_'bezeichnung'_macro": 2182,
+  "precision_'thema'_macro": 2183,
+  "precision_20-vote": 2184,
+  "precision_af": 2185,
+  "precision_class_negative": 2186,
+  "precision_class_positive": 2187,
+  "precision_entity_span": 2188,
+  "precision_ham": 2189,
+  "precision_macro": 2190,
+  "precision_macro_avg": 2191,
+  "precision_micro": 2192,
+  "precision_micro_avg": 2193,
+  "precision_ppv": 2194,
+  "precision_rate": 2195,
+  "precision_samples": 2196,
+  "precision_spam": 2197,
+  "precision_strong_class": 2198,
+  "precision_test_2020": 2199,
+  "precision_test_2021": 2200,
+  "precision_threshold=0.94": 2201,
+  "precision_weighted": 2202,
+  "prediction_success_rate": 2203,
+  "preference_accuracy": 2204,
+  "prefill_latency_ms": 2205,
+  "private_score": 2206,
+  "processing_speed_tokens/sec": 2207,
+  "professional_law": 2208,
+  "proficiency_score": 2209,
+  "prompt_compliance_rate_%": 2210,
+  "prompt_level_loose_acc": 2211,
+  "prompt_level_loose_accuracy": 2212,
+  "prompt_level_strict_acc": 2213,
+  "prompt_level_strict_accuracy": 2214,
+  "proper_names_accuracy_%": 2215,
+  "psnr": 2216,
+  "psnr_srgb": 2217,
+  "public_avg._f1": 2218,
+  "public_score": 2219,
+  "q3": 2220,
+  "q8": 2221,
+  "qa_accuracy": 2222,
+  "qc_decision_accuracy": 2223,
+  "query_active_dims": 2224,
+  "query_sparsity_ratio": 2225,
+  "question_pair_acc": 2226,
+  "qwk": 2227,
+  "r": 2228,
+  "r-1_f1": 2229,
+  "r-2_f1": 2230,
+  "r-l_f1": 2231,
+  "r-precision": 2232,
+  "r-r2": 2233,
+  "r-squared": 2234,
+  "r1": 2235,
+  "r1@0.5": 2236,
+  "r2_score": 2237,
+  "r@1": 2238,
+  "r@10": 2239,
+  "r@1_iou=0.3": 2240,
+  "r@1_iou=0.5": 2241,
+  "r@5": 2242,
+  "r@m": 2243,
+  "r_squared": 2244,
+  "race-m": 2245,
+  "radgraph_f1": 2246,
+  "rank-1": 2247,
+  "rank-1_accuracy_rn50": 2248,
+  "rank-1_accuracy_vit-b/16": 2249,
+  "rank-1_all_search": 2250,
+  "rank_128-dim": 2251,
+  "raw_score": 2252,
+  "re+_micro_f1": 2253,
+  "real_acc": 2254,
+  "reasonable_miss_rate": 2255,
+  "reasoning": 2256,
+  "reasoning_accuracy": 2257,
+  "reasoning_accuracy_%": 2258,
+  "reasoning_alg.": 2259,
+  "reasoning_quality_score": 2260,
+  "recall": 2261,
+  "recall-macro": 2262,
+  "recall@1": 2263,
+  "recall@10": 2264,
+  "recall@100": 2265,
+  "recall@1000": 2266,
+  "recall@1000_miracl": 2267,
+  "recall@100_miracl": 2268,
+  "recall@10_miracl": 2269,
+  "recall@1_%": 2270,
+  "recall@1_hn-atom_uc": 2271,
+  "recall@1_miracl": 2272,
+  "recall@2": 2273,
+  "recall@20": 2274,
+  "recall@200": 2275,
+  "recall@20_miracl": 2276,
+  "recall@3": 2277,
+  "recall@30": 2278,
+  "recall@300": 2279,
+  "recall@3_miracl": 2280,
+  "recall@5": 2281,
+  "recall@50": 2282,
+  "recall@500": 2283,
+  "recall@5_miracl": 2284,
+  "recall@7": 2285,
+  "recall@70": 2286,
+  "recall@700": 2287,
+  "recall_%": 2288,
+  "recall_'bezeichnung'_macro": 2289,
+  "recall_'thema'_macro": 2290,
+  "recall_20-vote": 2291,
+  "recall_af": 2292,
+  "recall_class_negative": 2293,
+  "recall_class_positive": 2294,
+  "recall_crisis_detection_rate": 2295,
+  "recall_entity_span": 2296,
+  "recall_ham": 2297,
+  "recall_macro": 2298,
+  "recall_macro_avg": 2299,
+  "recall_micro": 2300,
+  "recall_micro_avg": 2301,
+  "recall_samples": 2302,
+  "recall_sensitivity": 2303,
+  "recall_spam": 2304,
+  "recall_strong_class": 2305,
+  "recall_test_2020": 2306,
+  "recall_test_2021": 2307,
+  "recall_threshold=0.94": 2308,
+  "recall_tpr": 2309,
+  "recall_weighted": 2310,
+  "recognition-of-done": 2311,
+  "recognition_events": 2312,
+  "refusal_rate": 2313,
+  "relative_direction": 2314,
+  "relative_distance": 2315,
+  "relative_polarity_precision": 2316,
+  "remaining": 2317,
+  "repetition/looping_prevalence": 2318,
+  "reranking_4_datasets": 2319,
+  "response_relevance": 2320,
+  "response_time_ms": 2321,
+  "response_token_reduction": 2322,
+  "results_partial_f1": 2323,
+  "retention_%": 2324,
+  "retrieval_8_datasets": 2325,
+  "reward_gap": 2326,
+  "rhythmic_presence_and_stability": 2327,
+  "rise_time_s": 2328,
+  "risk-reward_ratio": 2329,
+  "rmse": 2330,
+  "rmse_alpha": 2331,
+  "rmse_cooperative": 2332,
+  "rmse_delta_cola_to_final": 2333,
+  "rmse_delta_perplexity_to_final_large": 2334,
+  "rmse_iter_to_final_simplified": 2335,
+  "rmse_m": 2336,
+  "rmse_original_scale_-2_to_+2": 2337,
+  "rmse_original_scale_0-3": 2338,
+  "rmse_robbert_delta_blurb_to_final": 2339,
+  "robustness_score": 2340,
+  "roc": 2341,
+  "roc-auc": 2342,
+  "roc-auc_macro": 2343,
+  "roc-auc_std_dev": 2344,
+  "roc_auc": 2345,
+  "roc_auc_macro": 2346,
+  "roc_auc_micro": 2347,
+  "roc_auc_samples": 2348,
+  "roc_auc_weighted": 2349,
+  "rogue1": 2350,
+  "roleplay": 2351,
+  "room_size": 2352,
+  "root_mean_squared_error": 2353,
+  "rouge": 2354,
+  "rouge-1": 2355,
+  "rouge-1-f1": 2356,
+  "rouge-1-precision": 2357,
+  "rouge-1-recall": 2358,
+  "rouge-1_f1": 2359,
+  "rouge-1_improvement": 2360,
+  "rouge-1_score": 2361,
+  "rouge-2": 2362,
+  "rouge-2-f1": 2363,
+  "rouge-2-precision": 2364,
+  "rouge-2-recall": 2365,
+  "rouge-2_f1": 2366,
+  "rouge-2_improvement": 2367,
+  "rouge-l-f1": 2368,
+  "rouge-l-precision": 2369,
+  "rouge-l-recall": 2370,
+  "rouge-l_f1": 2371,
+  "rouge-l_improvement": 2372,
+  "rouge-l_qa": 2373,
+  "rouge-l_score": 2374,
+  "rouge-lsum": 2375,
+  "rouge1": 2376,
+  "rouge1_acc": 2377,
+  "rouge1_diff": 2378,
+  "rouge1_max": 2379,
+  "rouge2": 2380,
+  "rouge2_acc": 2381,
+  "rouge2_diff": 2382,
+  "rouge2_max": 2383,
+  "rouge_l": 2384,
+  "rouge_score": 2385,
+  "rougel_acc": 2386,
+  "rougel_diff": 2387,
+  "rougel_max": 2388,
+  "rougelsum": 2389,
+  "route_plan": 2390,
+  "route_quality_score": 2391,
+  "row_non_zero_mean_corpus": 2392,
+  "row_non_zero_mean_query": 2393,
+  "row_sparsity_mean_corpus": 2394,
+  "row_sparsity_mean_query": 2395,
+  "rss_score_7500tok_on_a100_gpu": 2396,
+  "runtime": 2397,
+  "runtime_sec": 2398,
+  "r\u00b2": 2399,
+  "r\u00b2_delta_cola_to_final": 2400,
+  "r\u00b2_delta_perplexity_to_final_large": 2401,
+  "r\u00b2_iter_to_final_simplified": 2402,
+  "r\u00b2_robbert_delta_blurb_to_final": 2403,
+  "s-measure": 2404,
+  "s/n_accuracy": 2405,
+  "sacrebleu": 2406,
+  "sacrebleu_chrf": 2407,
+  "safety_score": 2408,
+  "sample_size": 2409,
+  "samples": 2410,
+  "samples_per_second": 2411,
+  "sanskrit/pali_terms_accuracy_%": 2412,
+  "sanskrit_to_english_translation_-_bleu_score": 2413,
+  "sanskrit_to_english_translation_-_jaccard_similarity": 2414,
+  "sari_easse>=0.2.1": 2415,
+  "scicode": 2416,
+  "sciq_mc": 2417,
+  "score": 2418,
+  "sdr": 2419,
+  "sdr_avg": 2420,
+  "second_turn": 2421,
+  "secondary_structure_3-states": 2422,
+  "secondary_structure_8-states": 2423,
+  "seen_samples_b": 2424,
+  "self-reported": 2425,
+  "semantic_similarity": 2426,
+  "semclass_f1": 2427,
+  "sen": 2428,
+  "sensitivity": 2429,
+  "sensitivity_recall": 2430,
+  "sentence_sacrebleu": 2431,
+  "sentences_f-score": 2432,
+  "sequences": 2433,
+  "settling_time_95%": 2434,
+  "settling_time_s": 2435,
+  "shape_bias": 2436,
+  "si-sdr": 2437,
+  "si-sdri": 2438,
+  "sib-200_lb_accuracy": 2439,
+  "sign_accuracy_3-class": 2440,
+  "silhouette_cosine": 2441,
+  "silhouette_euclidean": 2442,
+  "silhouette_score": 2443,
+  "silma_ragqa_benchmark_score": 2444,
+  "similarity_accuracy": 2445,
+  "similarity_accuracy_threshold": 2446,
+  "similarity_ap": 2447,
+  "similarity_f1": 2448,
+  "similarity_f1_threshold": 2449,
+  "similarity_precision": 2450,
+  "similarity_recall": 2451,
+  "single-line_infilling_pass@1": 2452,
+  "single-line_infilling_pass@10": 2453,
+  "single_choice": 2454,
+  "single_line": 2455,
+  "size": 2456,
+  "slot_f1_micro": 2457,
+  "slot_f1_score": 2458,
+  "slot_precision_micro": 2459,
+  "slot_recall_micro": 2460,
+  "smiles_validity_rate": 2461,
+  "smoothed_bleu-4": 2462,
+  "smotsa": 2463,
+  "social_group_seqeval": 2464,
+  "social_science": 2465,
+  "socialiqa_mc": 2466,
+  "soft-f1": 2467,
+  "software_development.f1_score": 2468,
+  "software_development.precision": 2469,
+  "software_development.recall": 2470,
+  "software_development.support": 2471,
+  "solution_exact_match": 2472,
+  "span-based_f1": 2473,
+  "sparse_acc": 2474,
+  "sparsity": 2475,
+  "sparsity_ratio": 2476,
+  "speaker_similarity": 2477,
+  "spearman": 2478,
+  "spearman's_rho": 2479,
+  "spearman's_\u03c1": 2480,
+  "spearman_ar-ar": 2481,
+  "spearman_correlation": 2482,
+  "spearman_correlation_cosine_similarity": 2483,
+  "spearman_cosine": 2484,
+  "spearman_dot": 2485,
+  "spearman_en-ar": 2486,
+  "spearman_en-de": 2487,
+  "spearman_en-en": 2488,
+  "spearman_en-tr": 2489,
+  "spearman_es-en": 2490,
+  "spearman_es-es": 2491,
+  "spearman_euclidean": 2492,
+  "spearman_fr-en": 2493,
+  "spearman_it-en": 2494,
+  "spearman_ko-ko": 2495,
+  "spearman_main_score": 2496,
+  "spearman_manhattan": 2497,
+  "spearman_max": 2498,
+  "spearman_nl-en": 2499,
+  "spearmanr": 2500,
+  "spearmanr_dynamic_8b": 2501,
+  "spearmanr_onnx": 2502,
+  "spearmanr_optimized": 2503,
+  "spearmanr_static_8b": 2504,
+  "specificity": 2505,
+  "speech_accuracy": 2506,
+  "speedup_vs_fp32_baseline_x": 2507,
+  "spice": 2508,
+  "spl_test_unseen": 2509,
+  "spl_val": 2510,
+  "spl_val_unseen": 2511,
+  "squad": 2512,
+  "squad_em": 2513,
+  "squad_f1": 2514,
+  "squad_gen2mc_mc": 2515,
+  "sr": 2516,
+  "sr_test_unseen": 2517,
+  "sr_val": 2518,
+  "sr_val_unseen": 2519,
+  "src2trg_accuracy": 2520,
+  "ssim": 2521,
+  "ssim_srgb": 2522,
+  "sta": 2523,
+  "stage_match_score": 2524,
+  "standard_parseval_full": 2525,
+  "static_error": 2526,
+  "std_reward": 2527,
+  "stem": 2528,
+  "step_best_checkpoint": 2529,
+  "steps_per_second": 2530,
+  "strict-match": 2531,
+  "strict_accuracy": 2532,
+  "strict_prompt": 2533,
+  "structured_output_compliance": 2534,
+  "sts_8_datasets": 2535,
+  "stsbenchmark": 2536,
+  "style_llm-judge_1-3": 2537,
+  "style_meter_greedy_pass_rate": 2538,
+  "subj_f1": 2539,
+  "subj_p": 2540,
+  "subj_r": 2541,
+  "subset-accuracy": 2542,
+  "subset_accuracy": 2543,
+  "success_rate": 2544,
+  "success_rate_%": 2545,
+  "swe-bench_verified": 2546,
+  "symptomatic_accuracy": 2547,
+  "system_score": 2548,
+  "t/f_accuracy": 2549,
+  "tag_xpos_accuracy": 2550,
+  "tar@far=0.0001": 2551,
+  "target_f1": 2552,
+  "target_rounds": 2553,
+  "task_1": 2554,
+  "task_2": 2555,
+  "task_3": 2556,
+  "task_4": 2557,
+  "task_completion_rate_improvement": 2558,
+  "tau2": 2559,
+  "telugu_wer": 2560,
+  "tempo_match": 2561,
+  "ter": 2562,
+  "terminalbench_hard": 2563,
+  "test": 2564,
+  "test/f1": 2565,
+  "test16_cer": 2566,
+  "test16_wer": 2567,
+  "test20_cer": 2568,
+  "test20_wer": 2569,
+  "test_1-shot_rougel": 2570,
+  "test_accent_accuracy": 2571,
+  "test_accuracy": 2572,
+  "test_accuracy_logistic_regression": 2573,
+  "test_accuracy_on_coscan_speech": 2574,
+  "test_accuracy_original_data": 2575,
+  "test_accuracy_svc": 2576,
+  "test_accuracy_svc_linear": 2577,
+  "test_age_accuracy": 2578,
+  "test_ap": 2579,
+  "test_auc": 2580,
+  "test_bertscore": 2581,
+  "test_bertscore_fanpage": 2582,
+  "test_bertscore_ilpost": 2583,
+  "test_bleu": 2584,
+  "test_bleu_bg->en": 2585,
+  "test_bleu_cs->en": 2586,
+  "test_bleu_da->en": 2587,
+  "test_bleu_de->en": 2588,
+  "test_bleu_el->en": 2589,
+  "test_bleu_en->bg": 2590,
+  "test_bleu_en->cs": 2591,
+  "test_bleu_en->da": 2592,
+  "test_bleu_en->de": 2593,
+  "test_bleu_en->el": 2594,
+  "test_bleu_en->es": 2595,
+  "test_bleu_en->et": 2596,
+  "test_bleu_en->fi": 2597,
+  "test_bleu_en->fr": 2598,
+  "test_bleu_en->hr": 2599,
+  "test_bleu_en->hu": 2600,
+  "test_bleu_en->it": 2601,
+  "test_bleu_en->lt": 2602,
+  "test_bleu_en->lv": 2603,
+  "test_bleu_en->mt": 2604,
+  "test_bleu_en->nl": 2605,
+  "test_bleu_en->pl": 2606,
+  "test_bleu_en->pt": 2607,
+  "test_bleu_en->ro": 2608,
+  "test_bleu_en->ru": 2609,
+  "test_bleu_en->sk": 2610,
+  "test_bleu_en->sl": 2611,
+  "test_bleu_en->sv": 2612,
+  "test_bleu_en->uk": 2613,
+  "test_bleu_es->en": 2614,
+  "test_bleu_et->en": 2615,
+  "test_bleu_fi->en": 2616,
+  "test_bleu_fr->en": 2617,
+  "test_bleu_hr->en": 2618,
+  "test_bleu_hu->en": 2619,
+  "test_bleu_it->en": 2620,
+  "test_bleu_lt->en": 2621,
+  "test_bleu_lv->en": 2622,
+  "test_bleu_mt->en": 2623,
+  "test_bleu_nl->en": 2624,
+  "test_bleu_pl->en": 2625,
+  "test_bleu_pt->en": 2626,
+  "test_bleu_ro->en": 2627,
+  "test_bleu_ru->en": 2628,
+  "test_bleu_sk->en": 2629,
+  "test_bleu_sl->en": 2630,
+  "test_bleu_sv->en": 2631,
+  "test_bleu_taigi->mandrin": 2632,
+  "test_bleu_uk->en": 2633,
+  "test_bokm\u00e5l_cer": 2634,
+  "test_bokm\u00e5l_wer": 2635,
+  "test_cer": 2636,
+  "test_cer_%": 2637,
+  "test_cer_+lm": 2638,
+  "test_cer_mandrin": 2639,
+  "test_cer_no_lm": 2640,
+  "test_cer_using_lm": 2641,
+  "test_cer_w/o_stress": 2642,
+  "test_cer_with_lm": 2643,
+  "test_cer_without_lm": 2644,
+  "test_cher": 2645,
+  "test_comet_bg->en": 2646,
+  "test_comet_cs->en": 2647,
+  "test_comet_da->en": 2648,
+  "test_comet_de->en": 2649,
+  "test_comet_el->en": 2650,
+  "test_comet_en->bg": 2651,
+  "test_comet_en->cs": 2652,
+  "test_comet_en->da": 2653,
+  "test_comet_en->de": 2654,
+  "test_comet_en->el": 2655,
+  "test_comet_en->es": 2656,
+  "test_comet_en->et": 2657,
+  "test_comet_en->fi": 2658,
+  "test_comet_en->fr": 2659,
+  "test_comet_en->hr": 2660,
+  "test_comet_en->hu": 2661,
+  "test_comet_en->it": 2662,
+  "test_comet_en->lt": 2663,
+  "test_comet_en->lv": 2664,
+  "test_comet_en->mt": 2665,
+  "test_comet_en->nl": 2666,
+  "test_comet_en->pl": 2667,
+  "test_comet_en->pt": 2668,
+  "test_comet_en->ro": 2669,
+  "test_comet_en->ru": 2670,
+  "test_comet_en->sk": 2671,
+  "test_comet_en->sl": 2672,
+  "test_comet_en->sv": 2673,
+  "test_comet_en->uk": 2674,
+  "test_comet_es->en": 2675,
+  "test_comet_et->en": 2676,
+  "test_comet_fi->en": 2677,
+  "test_comet_fr->en": 2678,
+  "test_comet_hr->en": 2679,
+  "test_comet_hu->en": 2680,
+  "test_comet_it->en": 2681,
+  "test_comet_lt->en": 2682,
+  "test_comet_lv->en": 2683,
+  "test_comet_mt->en": 2684,
+  "test_comet_nl->en": 2685,
+  "test_comet_pl->en": 2686,
+  "test_comet_pt->en": 2687,
+  "test_comet_ro->en": 2688,
+  "test_comet_ru->en": 2689,
+  "test_comet_sk->en": 2690,
+  "test_comet_sl->en": 2691,
+  "test_comet_sv->en": 2692,
+  "test_comet_uk->en": 2693,
+  "test_coraa_wer": 2694,
+  "test_custom_cer_ctc": 2695,
+  "test_custom_cer_rnnt": 2696,
+  "test_custom_wer_ctc": 2697,
+  "test_custom_wer_rnnt": 2698,
+  "test_cver": 2699,
+  "test_der": 2700,
+  "test_em": 2701,
+  "test_exact_match": 2702,
+  "test_f1": 2703,
+  "test_f1-score": 2704,
+  "test_f1_callsign": 2705,
+  "test_f1_command": 2706,
+  "test_f1_macro": 2707,
+  "test_f1_micro_on_coscan_speech": 2708,
+  "test_f1_score": 2709,
+  "test_f1_score_macro": 2710,
+  "test_f1_score_weighted": 2711,
+  "test_f1_value": 2712,
+  "test_jaccard_error_rate": 2713,
+  "test_loss": 2714,
+  "test_macro_f1": 2715,
+  "test_map": 2716,
+  "test_mer": 2717,
+  "test_micro_f1": 2718,
+  "test_noresqa-mos_in-domain_training": 2719,
+  "test_nynorsk_cer": 2720,
+  "test_nynorsk_wer": 2721,
+  "test_pearson_correlation_coefficient": 2722,
+  "test_per": 2723,
+  "test_per_in-domain_training_|": 2724,
+  "test_per_on_common_voice_fr_13.0_|_trained": 2725,
+  "test_per_on_multilingual_librispeech_fr_|_trained": 2726,
+  "test_per_w/o_stress": 2727,
+  "test_perplexity": 2728,
+  "test_pr-auc": 2729,
+  "test_precision": 2730,
+  "test_precision_macro": 2731,
+  "test_qwk": 2732,
+  "test_recall": 2733,
+  "test_recall_macro": 2734,
+  "test_roc-auc": 2735,
+  "test_rogue-1": 2736,
+  "test_rogue-2": 2737,
+  "test_rogue-l": 2738,
+  "test_rogue-lsum": 2739,
+  "test_rouge-1": 2740,
+  "test_rouge-2": 2741,
+  "test_rouge-l": 2742,
+  "test_rouge-l_sum": 2743,
+  "test_rouge1": 2744,
+  "test_rouge1_fanpage": 2745,
+  "test_rouge1_ilpost": 2746,
+  "test_rouge2": 2747,
+  "test_rouge2_fanpage": 2748,
+  "test_rouge2_ilpost": 2749,
+  "test_rougel": 2750,
+  "test_rougel_fanpage": 2751,
+  "test_rougel_ilpost": 2752,
+  "test_runtime": 2753,
+  "test_samples_per_second": 2754,
+  "test_ser": 2755,
+  "test_set_pass@1": 2756,
+  "test_spearmanr": 2757,
+  "test_squim-stoi_in-domain_training": 2758,
+  "test_steps_per_second": 2759,
+  "test_stoi_in-domain_training": 2760,
+  "test_suite_sql_eval_-_exact_matching_accuracy": 2761,
+  "test_suite_sql_eval_-_execution_accuracy": 2762,
+  "test_weighted_accuracy": 2763,
+  "test_wer": 2764,
+  "test_wer_+lm": 2765,
+  "test_wer_960ms_chunk_size_4_left_context_chunks": 2766,
+  "test_wer_bg": 2767,
+  "test_wer_cs": 2768,
+  "test_wer_da": 2769,
+  "test_wer_de": 2770,
+  "test_wer_el": 2771,
+  "test_wer_en": 2772,
+  "test_wer_es": 2773,
+  "test_wer_et": 2774,
+  "test_wer_fi": 2775,
+  "test_wer_fr": 2776,
+  "test_wer_hr": 2777,
+  "test_wer_hu": 2778,
+  "test_wer_it": 2779,
+  "test_wer_lt": 2780,
+  "test_wer_lv": 2781,
+  "test_wer_mls": 2782,
+  "test_wer_mt": 2783,
+  "test_wer_nl": 2784,
+  "test_wer_no_lm": 2785,
+  "test_wer_non-streaming_greedy": 2786,
+  "test_wer_on_common_voice_7": 2787,
+  "test_wer_p&c": 2788,
+  "test_wer_pl": 2789,
+  "test_wer_pt": 2790,
+  "test_wer_ro": 2791,
+  "test_wer_ru": 2792,
+  "test_wer_sk": 2793,
+  "test_wer_sl": 2794,
+  "test_wer_sv": 2795,
+  "test_wer_uk": 2796,
+  "test_wer_using_lm": 2797,
+  "test_wer_with_language_model": 2798,
+  "test_wer_with_lm": 2799,
+  "test_wer_without_lm": 2800,
+  "test_wil": 2801,
+  "test_wip": 2802,
+  "text-to-video_r@1": 2803,
+  "text-to-video_r@10": 2804,
+  "text_retrieval_r@1": 2805,
+  "text_score": 2806,
+  "thai_exam_acc": 2807,
+  "think_step_length": 2808,
+  "three_pixel_error": 2809,
+  "threshold": 2810,
+  "throughput_tps_on_h100": 2811,
+  "tim_partial_f1": 2812,
+  "tim_strict_f1": 2813,
+  "time_mean": 2814,
+  "time_ms": 2815,
+  "time_std": 2816,
+  "tm": 2817,
+  "tn": 2818,
+  "token-level_f1": 2819,
+  "token-level_jaccard_similarity": 2820,
+  "token_accuracy": 2821,
+  "token_accuracy_all": 2822,
+  "token_accuracy_ambiguous": 2823,
+  "token_f1": 2824,
+  "token_reduction_vs_character-level_%": 2825,
+  "token_reduction_vs_v6.5_%": 2826,
+  "token_scores_/_adresse_/_f1": 2827,
+  "token_scores_/_adresse_/_precision": 2828,
+  "token_scores_/_adresse_/_recall": 2829,
+  "token_scores_/_adresse_/_redact": 2830,
+  "token_scores_/_adresse_/_redact_full": 2831,
+  "token_scores_/_date_/_f1": 2832,
+  "token_scores_/_date_/_precision": 2833,
+  "token_scores_/_date_/_recall": 2834,
+  "token_scores_/_date_/_redact": 2835,
+  "token_scores_/_date_/_redact_full": 2836,
+  "token_scores_/_date_naissance_/_f1": 2837,
+  "token_scores_/_date_naissance_/_precision": 2838,
+  "token_scores_/_date_naissance_/_recall": 2839,
+  "token_scores_/_date_naissance_/_redact": 2840,
+  "token_scores_/_date_naissance_/_redact_full": 2841,
+  "token_scores_/_disease_/_f1": 2842,
+  "token_scores_/_disease_/_precision": 2843,
+  "token_scores_/_disease_/_recall": 2844,
+  "token_scores_/_ipp_/_f1": 2845,
+  "token_scores_/_ipp_/_precision": 2846,
+  "token_scores_/_ipp_/_recall": 2847,
+  "token_scores_/_ipp_/_redact": 2848,
+  "token_scores_/_ipp_/_redact_full": 2849,
+  "token_scores_/_mail_/_f1": 2850,
+  "token_scores_/_mail_/_precision": 2851,
+  "token_scores_/_mail_/_recall": 2852,
+  "token_scores_/_mail_/_redact": 2853,
+  "token_scores_/_mail_/_redact_full": 2854,
+  "token_scores_/_medication_/_f1": 2855,
+  "token_scores_/_medication_/_precision": 2856,
+  "token_scores_/_medication_/_recall": 2857,
+  "token_scores_/_micro_/_f1": 2858,
+  "token_scores_/_micro_/_precision": 2859,
+  "token_scores_/_micro_/_recall": 2860,
+  "token_scores_/_micro_/_redact": 2861,
+  "token_scores_/_micro_/_redact_full": 2862,
+  "token_scores_/_nda_/_f1": 2863,
+  "token_scores_/_nda_/_precision": 2864,
+  "token_scores_/_nda_/_recall": 2865,
+  "token_scores_/_nda_/_redact": 2866,
+  "token_scores_/_nda_/_redact_full": 2867,
+  "token_scores_/_nom_/_f1": 2868,
+  "token_scores_/_nom_/_precision": 2869,
+  "token_scores_/_nom_/_recall": 2870,
+  "token_scores_/_nom_/_redact": 2871,
+  "token_scores_/_nom_/_redact_full": 2872,
+  "token_scores_/_prenom_/_f1": 2873,
+  "token_scores_/_prenom_/_precision": 2874,
+  "token_scores_/_prenom_/_recall": 2875,
+  "token_scores_/_prenom_/_redact": 2876,
+  "token_scores_/_prenom_/_redact_full": 2877,
+  "token_scores_/_procedure_/_f1": 2878,
+  "token_scores_/_procedure_/_precision": 2879,
+  "token_scores_/_procedure_/_recall": 2880,
+  "token_scores_/_secu_/_f1": 2881,
+  "token_scores_/_secu_/_precision": 2882,
+  "token_scores_/_secu_/_recall": 2883,
+  "token_scores_/_secu_/_redact": 2884,
+  "token_scores_/_secu_/_redact_full": 2885,
+  "token_scores_/_symptom_/_f1": 2886,
+  "token_scores_/_symptom_/_precision": 2887,
+  "token_scores_/_symptom_/_recall": 2888,
+  "token_scores_/_tel_/_f1": 2889,
+  "token_scores_/_tel_/_precision": 2890,
+  "token_scores_/_tel_/_recall": 2891,
+  "token_scores_/_tel_/_redact": 2892,
+  "token_scores_/_tel_/_redact_full": 2893,
+  "token_scores_/_ville_/_f1": 2894,
+  "token_scores_/_ville_/_precision": 2895,
+  "token_scores_/_ville_/_recall": 2896,
+  "token_scores_/_ville_/_redact": 2897,
+  "token_scores_/_ville_/_redact_full": 2898,
+  "token_scores_/_zip_/_f1": 2899,
+  "token_scores_/_zip_/_precision": 2900,
+  "token_scores_/_zip_/_recall": 2901,
+  "token_scores_/_zip_/_redact": 2902,
+  "token_scores_/_zip_/_redact_full": 2903,
+  "tokenized_f1": 2904,
+  "tokens": 2905,
+  "tokens/second": 2906,
+  "tokens_per_character_compounds": 2907,
+  "tokens_per_character_overall": 2908,
+  "tokens_per_character_real_news": 2909,
+  "tokens_per_second": 2910,
+  "tokens_per_second_baseline_a100_fp16_512_tok": 2911,
+  "tokens_per_second_ibce_a100_fp16_512_tok": 2912,
+  "top-1": 2913,
+  "top-1_%": 2914,
+  "top-1_acc._%": 2915,
+  "top-1_acc_%": 2916,
+  "top-1_accuracy": 2917,
+  "top-1_accuracy_%": 2918,
+  "top-1_error_rate": 2919,
+  "top-2_accuracy": 2920,
+  "top-3-accuracy": 2921,
+  "top-3_accuracy": 2922,
+  "top-5_accuracy": 2923,
+  "top1_acc": 2924,
+  "top_1_accuracy": 2925,
+  "top_1_accuracy_dynamic_quantized_wi8_afp32": 2926,
+  "top_1_accuracy_full_precision": 2927,
+  "top_5_accuracy": 2928,
+  "top_5_accuracy_dynamic_quantized_wi8_afp32": 2929,
+  "top_5_accuracy_full_precision": 2930,
+  "total": 2931,
+  "total_column_score": 2932,
+  "total_cores": 2933,
+  "total_flops": 2934,
+  "total_model_size_gb": 2935,
+  "total_precision": 2936,
+  "total_recall": 2937,
+  "total_reward_mean": 2938,
+  "total_steps": 2939,
+  "total_time_in_seconds": 2940,
+  "total_timesteps": 2941,
+  "toxicity_rito": 2942,
+  "tp": 2943,
+  "traffic_vocabulary_coverage": 2944,
+  "train_accuracy": 2945,
+  "train_loss": 2946,
+  "train_mse": 2947,
+  "train_runtime_seconds": 2948,
+  "training_accuracy": 2949,
+  "training_done": 2950,
+  "training_flops": 2951,
+  "training_loss": 2952,
+  "training_loss_final": 2953,
+  "training_pearson_cosine": 2954,
+  "training_progress_%": 2955,
+  "training_steps": 2956,
+  "transcription_accuracy": 2957,
+  "translation_bleu_score": 2958,
+  "transliteration_-_character_accuracy": 2959,
+  "transliteration_-_exact_match_accuracy": 2960,
+  "treatment_f1-score": 2961,
+  "treatment_precision": 2962,
+  "treatment_recall": 2963,
+  "trg2src_accuracy": 2964,
+  "triplet_accuracy": 2965,
+  "true_accuracy": 2966,
+  "trueskill": 2967,
+  "truthfulqa": 2968,
+  "truthfulqa_0-shot": 2969,
+  "trv_tegu_->_zho_hant_zh": 2970,
+  "trv_truk_->_zho_hant_zh": 2971,
+  "tube-boundary_ap": 2972,
+  "ud_jaccard": 2973,
+  "unfair-tos": 2974,
+  "unique_preference_rate": 2975,
+  "unlabeled_attachment_score_uas": 2976,
+  "unlabeled_sentiment_tuple_f1": 2977,
+  "unlabelled_attachment_score": 2978,
+  "unproven_accuracy": 2979,
+  "unweighted_accuracy_ua": 2980,
+  "upos": 2981,
+  "upos_accuracy": 2982,
+  "v-measure": 2983,
+  "v-measure_main": 2984,
+  "v-measure_sub": 2985,
+  "v_measure": 2986,
+  "v_measure_std": 2987,
+  "val": 2988,
+  "val_acc": 2989,
+  "val_f1_score": 2990,
+  "val_miou": 2991,
+  "val_pass@1": 2992,
+  "val_per": 2993,
+  "val_per_on_common_voice_fr_13.0_|_trained": 2994,
+  "val_per_on_multilingual_librispeech_fr_|_trained": 2995,
+  "validation_accuracy": 2996,
+  "validation_accuracy_on_coscan_speech": 2997,
+  "validation_accuracy_subset_experiment": 2998,
+  "validation_bleu": 2999,
+  "validation_cer": 3000,
+  "validation_cer_with_5-gram_lm": 3001,
+  "validation_cross-entropy": 3002,
+  "validation_dev_overall": 3003,
+  "validation_f1": 3004,
+  "validation_f1_micro_on_coscan_speech": 3005,
+  "validation_loss": 3006,
+  "validation_loss_best": 3007,
+  "validation_loss_final": 3008,
+  "validation_loss_subset_experiment": 3009,
+  "validation_macro_f1": 3010,
+  "validation_mae": 3011,
+  "validation_matched_accuracy": 3012,
+  "validation_matched_f1": 3013,
+  "validation_miou": 3014,
+  "validation_mismatched_accuracy": 3015,
+  "validation_mismatched_f1": 3016,
+  "validation_nli_cosine_accuracy": 3017,
+  "validation_nli_cosine_accuracy_threshold": 3018,
+  "validation_nli_cosine_ap": 3019,
+  "validation_nli_cosine_f1": 3020,
+  "validation_nli_cosine_f1_threshold": 3021,
+  "validation_nli_cosine_mcc": 3022,
+  "validation_nli_cosine_precision": 3023,
+  "validation_nli_cosine_recall": 3024,
+  "validation_perplexity": 3025,
+  "validation_perplexity_approx.": 3026,
+  "validation_r^2": 3027,
+  "validation_rmse_best_run_internal_autogluon_validation": 3028,
+  "validation_rmsle": 3029,
+  "validation_rogue-1": 3030,
+  "validation_rogue-1.": 3031,
+  "validation_rogue-2": 3032,
+  "validation_rogue-l": 3033,
+  "validation_rogue-l-sum": 3034,
+  "validation_rogue-lsum": 3035,
+  "validation_rouge-1": 3036,
+  "validation_rouge-2": 3037,
+  "validation_rouge-l": 3038,
+  "validation_rouge-l_sum": 3039,
+  "validation_rte_cosine_accuracy": 3040,
+  "validation_rte_cosine_accuracy_threshold": 3041,
+  "validation_rte_cosine_ap": 3042,
+  "validation_rte_cosine_f1": 3043,
+  "validation_rte_cosine_f1_threshold": 3044,
+  "validation_rte_cosine_mcc": 3045,
+  "validation_rte_cosine_precision": 3046,
+  "validation_rte_cosine_recall": 3047,
+  "validation_sts_pearson_cosine": 3048,
+  "validation_sts_spearman_cosine": 3049,
+  "validation_wer": 3050,
+  "validation_wer_with_5-gram_lm": 3051,
+  "variant_aggregation": 3052,
+  "vdcscore": 3053,
+  "vdd": 3054,
+  "viewpoint_i_aepe": 3055,
+  "vocab_size": 3056,
+  "vocabulary_size": 3057,
+  "voxceleb_dev": 3058,
+  "vqa_ablation": 3059,
+  "vram_reduction_%": 3060,
+  "vs_base_model": 3061,
+  "vtab": 3062,
+  "v\u2011measure_main/sub": 3063,
+  "wacc": 3064,
+  "wb": 3065,
+  "wbscore": 3066,
+  "weed_precision": 3067,
+  "weighted-f1": 3068,
+  "weighted-f1_score": 3069,
+  "weighted_accuarcy": 3070,
+  "weighted_accuracy": 3071,
+  "weighted_average_f1-score": 3072,
+  "weighted_f1": 3073,
+  "weighted_f1-score": 3074,
+  "weighted_f1-score_logistic_regression": 3075,
+  "weighted_f1-score_svc": 3076,
+  "weighted_f1_score": 3077,
+  "weighted_precision": 3078,
+  "weighted_precision_svc": 3079,
+  "weighted_recall": 3080,
+  "weighted_recall_svc": 3081,
+  "well-structured_stories": 3082,
+  "wer": 3083,
+  "wer%": 3084,
+  "wer_%": 3085,
+  "wer_1.12s_frame_size": 3086,
+  "wer_beam_5": 3087,
+  "wer_catalan": 3088,
+  "wer_documentaries": 3089,
+  "wer_english_-_combined": 3090,
+  "wer_for_arabic": 3091,
+  "wer_greedy": 3092,
+  "wer_indonesian_-_combined": 3093,
+  "wer_lm": 3094,
+  "wer_news": 3095,
+  "wer_normalized": 3096,
+  "wer_on_common_voice_17.0": 3097,
+  "wer_orthographic": 3098,
+  "wer_raw": 3099,
+  "wer_reference_column:_raw_transcription": 3100,
+  "wer_reference_column:_transcription": 3101,
+  "wer_seed_42_-_split_1": 3102,
+  "wer_seed_42_-_split_2": 3103,
+  "wer_seed_42_-_split_3": 3104,
+  "wer_seed_43_-_split_1": 3105,
+  "wer_seed_43_-_split_2": 3106,
+  "wer_seed_43_-_split_3": 3107,
+  "wer_seed_44_-_split_1": 3108,
+  "wer_seed_44_-_split_2": 3109,
+  "wer_seed_44_-_split_3": 3110,
+  "wer_spanish": 3111,
+  "wer_test": 3112,
+  "wer_unnormalized": 3113,
+  "wer_validation": 3114,
+  "wer_with_punctuation_and_capital_letters": 3115,
+  "wer_without_normalization": 3116,
+  "wer_without_punctuation": 3117,
+  "wer_word_error_rate": 3118,
+  "wiki_split": 3119,
+  "wil": 3120,
+  "wildguard_total_f1": 3121,
+  "willingness_to_answer": 3122,
+  "win-rate": 3123,
+  "win_rate": 3124,
+  "win_rate_%": 3125,
+  "win_rate_vs_base_model_llm-as-judge": 3126,
+  "win_rate_vs_baseline_claude_3.5_sonnet_blind_a/b_n=42": 3127,
+  "win_rate_vs_baseline_claude_haiku_4.5_blind_a/b_n=15": 3128,
+  "win_rate_vs_baseline_claude_haiku_4.5_blind_a/b_n=57": 3129,
+  "win_rate_vs_baseline_claude_opus_4_blind_a/b_n=15": 3130,
+  "win_rate_vs_baseline_claude_opus_4_blind_a/b_n=57": 3131,
+  "win_rate_vs_baseline_claude_sonnet_4_blind_a/b_n=42": 3132,
+  "win_rate_vs_baseline_gemini_2.5_flash_lite_blind_a/b_n=57": 3133,
+  "win_rate_vs_baseline_gpt-4o_blind_a/b_n=57": 3134,
+  "win_rate_vs_baseline_overall_claude_judges_blind_a/b_n=57": 3135,
+  "winogrande": 3136,
+  "winogrande_0-shot": 3137,
+  "winogrande_5-shot": 3138,
+  "winogrande_rc": 3139,
+  "wip": 3140,
+  "word-count_constraint_accuracy_120-150": 3141,
+  "word_accuracy": 3142,
+  "word_accuracy_oov": 3143,
+  "word_error_rate": 3144,
+  "word_error_rate_all_data": 3145,
+  "word_error_rate_eslo": 3146,
+  "word_error_rate_langage": 3147,
+  "word_error_rate_wer": 3148,
+  "word_error_rate_wer_%": 3149,
+  "word_error_rate_with_limited_vocabulary": 3150,
+  "worst_group_accuracy": 3151,
+  "writing": 3152,
+  "xpos_accuracy": 3153,
+  "xstest_f1": 3154,
+  "yes/no_accuracy": 3155,
+  "zero-shot_accuracy": 3156,
+  "zero-shot_clip_accuracy": 3157,
+  "zero-shot_precision": 3158,
+  "zero-shot_recall": 3159,
+  "zero-shot_top-1_acc._%": 3160,
+  "zero-shot_top-1_acc_%": 3161,
+  "zero-shot_transfer": 3162,
+  "zeroth-test-bleu": 3163,
+  "zeroth-test-cer": 3164,
+  "zeroth-test-wer": 3165,
+  "zho_hant_->_ami_xiug_13a": 3166,
+  "zho_hant_->_trv_tegu_13a": 3167,
+  "zho_hant_->_trv_truk_13a": 3168,
+  "\u03c00": 3169,
+  "\u0627\u062d\u0633\u0627\u0646_compliance": 3170,
+  "\u226490%ile": 3171
+}

data/task2id.json

@@ -0,0 +1,2553 @@
+{
+  "0-shot": 0,
+  "0-shot CoT": 1,
+  "0-shot, CoT": 2,
+  "1-shot": 3,
+  "10-shot": 4,
+  "2-shot": 5,
+  "2-shot, CoT": 6,
+  "25-shot": 7,
+  "2D Human Pose Estimation": 8,
+  "2D Object Detection": 9,
+  "2D Pose Estimation": 10,
+  "2D image classification": 11,
+  "2R. Avg.": 12,
+  "3-5-shot": 13,
+  "3-shot": 14,
+  "3-shot, CoT": 15,
+  "3D Face Reconstruction": 16,
+  "3D Human Pose Estimation": 17,
+  "3D Instance Segmentation": 18,
+  "3D Multi-Object Tracking": 19,
+  "3D Object Captioning": 20,
+  "3D Object Classification": 21,
+  "3D Object Detection": 22,
+  "3D Open-Vocabulary Instance Segmentation": 23,
+  "3D Point Cloud Classification": 24,
+  "3D Pose Estimation": 25,
+  "3D Reconstruction": 26,
+  "3D Semantic Scene Completion": 27,
+  "3D Semantic Segmentation": 28,
+  "3D Shape Reconstruction": 29,
+  "3D radiology image classification": 30,
+  "3DSR": 31,
+  "4-Class: (Benign, Defacement, Phishing, Malware)": 32,
+  "4-shot": 33,
+  "4-shot, maj@4": 34,
+  "4D Panoptic Segmentation": 35,
+  "5-shot": 36,
+  "5-shot, CoT": 37,
+  "6D Pose Estimation": 38,
+  "7-shot": 39,
+  "8-shot": 40,
+  "8-shot, CoT": 41,
+  "8-shot, maj@8": 42,
+  "AGIEval": 43,
+  "AI Text Detection": 44,
+  "AI-Generated Text Detection": 45,
+  "AI2 ARC (Challenge)": 46,
+  "AI2 ARC (Easy)": 47,
+  "ARC": 48,
+  "ARC Challenge": 49,
+  "ARC Prize 2025 (legacy evaluation mapping)": 50,
+  "ARC-Challenge": 51,
+  "ARC-Easy": 52,
+  "ARC_C": 53,
+  "ARC_E": 54,
+  "ASR": 55,
+  "AST (0-shot, English-Korean)": 56,
+  "Abstract Algebra": 57,
+  "Abstract reasoning challenge": 58,
+  "Abstractive Dialogue Summarization": 59,
+  "Abstractive Question Answering": 60,
+  "Abstractive Summarization": 61,
+  "Abstractive Text Summarization": 62,
+  "Accented Speech Recognition": 63,
+  "Acoustic Scene Classification": 64,
+  "Action Detection": 65,
+  "Action Recognition": 66,
+  "Action Recognition In Videos": 67,
+  "Action Segmentation": 68,
+  "Ad-Hoc Information Retrieval": 69,
+  "Adversarial NLI": 70,
+  "Adversarial Robustness": 71,
+  "Agentic": 72,
+  "Alignment": 73,
+  "Alignment Faking Detection": 74,
+  "All-in-One Image Restoration": 75,
+  "Amazon Review Classification": 76,
+  "AmazonCounterfactualClassification": 77,
+  "AmazonReviewsClassification": 78,
+  "American Invitational Mathematics Examination": 79,
+  "Analogy Questions (BATS)": 80,
+  "Analogy Questions (ConceptNet Analogy)": 81,
+  "Analogy Questions (Google)": 82,
+  "Analogy Questions (NELL-ONE Analogy)": 83,
+  "Analogy Questions (SAT full)": 84,
+  "Analogy Questions (SAT)": 85,
+  "Analogy Questions (TREX Analogy)": 86,
+  "Analogy Questions (U2)": 87,
+  "Analogy Questions (U4)": 88,
+  "Animal Pose Estimation": 89,
+  "Anomaly Detection": 90,
+  "Arabic AI Text Detection": 91,
+  "Arabic to English Translation": 92,
+  "Argument Mining": 93,
+  "Arithmetic Reasoning": 94,
+  "ArxivQA": 95,
+  "Aspect-Based Sentiment Analysis (ABSA)": 96,
+  "Atari Games": 97,
+  "Atomic action recognition": 98,
+  "Attacks on Democratic Basic Order Detection": 99,
+  "Audio Classification": 100,
+  "Audio Emotion Classification": 101,
+  "Audio Emotion Recognition": 102,
+  "Audio Generation": 103,
+  "Audio Retrieval": 104,
+  "Audio Source Separation": 105,
+  "Audio Super-Resolution": 106,
+  "Audio Tagging": 107,
+  "Audio captioning": 108,
+  "Authorship Verification": 109,
+  "Auto Debugging": 110,
+  "Automated Theorem Proving": 111,
+  "Automatic Phoneme Recognition": 112,
+  "Automatic Speech Recognition": 113,
+  "Average": 114,
+  "BBH": 115,
+  "BLEU": 116,
+  "Bandwidth Extension": 117,
+  "Battery Insertion": 118,
+  "Beta-secretase Inhibition": 119,
+  "Bias Detection": 120,
+  "Biblical Hebrew Vocalization": 121,
+  "Binary Classification": 122,
+  "Binary Image Classification": 123,
+  "Binary OHCA detection (OHCA vs non-OHCA)": 124,
+  "Binary Propaganda Detection": 125,
+  "Binary Text Classification (Autoimmune Neurology)": 126,
+  "Binary text classification": 127,
+  "Binary: (Legit vs Spam Email)": 128,
+  "Biomedical Information Retrieval": 129,
+  "Biomedical QA (Chinese)": 130,
+  "Biomedical QA (PubMedQA)": 131,
+  "BitextMining": 132,
+  "Blind Face Restoration": 133,
+  "Blind Reconstruction (2-pass)": 134,
+  "Blood-Brain Barrier": 135,
+  "BoolQ": 136,
+  "BoolQ Question Answering": 137,
+  "Brain Tumor Classification": 138,
+  "Brain Tumor Detection": 139,
+  "Breast Cancer Histology Image Classification": 140,
+  "Breast Tumour Classification": 141,
+  "Bug-fix Patch Generation": 142,
+  "Business Intelligence Engine": 143,
+  "C-Eval (valid)": 144,
+  "COVID-19 Diagnosis": 145,
+  "CSQA": 146,
+  "CV-Bench": 147,
+  "Call to Action Detection": 148,
+  "Camera Pose Estimation": 149,
+  "Camouflaged Object Segmentation": 150,
+  "Cancer Image Classification": 151,
+  "Car Damage Detection": 152,
+  "CartPole-v1": 153,
+  "Caselaw Retrieval": 154,
+  "CatalanQA": 155,
+  "Categorical Classification (CC)": 156,
+  "Categorical Pair Similarity (CPS)": 157,
+  "Category Clustering": 158,
+  "Causal Language Modeling": 159,
+  "Cell Type Prediction": 160,
+  "Character Plot Arc Classification": 161,
+  "Chart Question Answering": 162,
+  "Chart reasoning": 163,
+  "Chat": 164,
+  "Chat & Instruction Following": 165,
+  "Cheese Texture Classification": 166,
+  "Chest X-ray report generation": 167,
+  "Chinese": 168,
+  "Citation Classification": 169,
+  "Claim Checkworthiness Detection": 170,
+  "Clasificación de reseñas (5 clases)": 171,
+  "Clasificación de texto": 172,
+  "Class-Specific Performance": 173,
+  "Classification": 174,
+  "Classification (ROC AUC)": 175,
+  "Classification Tasks": 176,
+  "Classify an image of chart to one of the following types: line, scatter, dot, vertical_bar, or horizontal_bar.": 177,
+  "Clickbait Detection": 178,
+  "Climate NLP Tasks (ClimaBench)": 179,
+  "Climate logical fallacy classification": 180,
+  "Clinical NER": 181,
+  "Clinical Note Embeddings": 182,
+  "Clinical Operations": 183,
+  "Clinical Support": 184,
+  "Clinical Text Embeddings": 185,
+  "Clinical Trial Comprehension": 186,
+  "Clustering": 187,
+  "CoQA": 188,
+  "Code": 189,
+  "Code Completion": 190,
+  "Code Documentation Generation": 191,
+  "Code Generation": 192,
+  "Code Reranking": 193,
+  "Code Retrieval": 194,
+  "Code Search": 195,
+  "Code generation": 196,
+  "Code generation and completion": 197,
+  "Coding": 198,
+  "Coherence-Momentum": 199,
+  "Col BERTTriplet": 200,
+  "Colorectal Gland Segmentation:": 201,
+  "Common Sense": 202,
+  "Common Sense Reasoning": 203,
+  "Commonsense": 204,
+  "Commonsense Reasoning": 205,
+  "Commonsense Understanding": 206,
+  "Commonsense natural language inference": 207,
+  "Conditional Generation": 208,
+  "Conditional Image Generation": 209,
+  "Confidence (Low/Medium/High)": 210,
+  "Contemporary-lb": 211,
+  "Contract clause classification": 212,
+  "Contracts Retrieval": 213,
+  "Contrastive Learning": 214,
+  "Conversation Summarization": 215,
+  "Conversational": 216,
+  "Conversational Response Retrieval": 217,
+  "Conversational Web Navigation": 218,
+  "Conversational and Function Calling": 219,
+  "Core Reasoning Tasks": 220,
+  "Coreference Resolution": 221,
+  "Coreference resolution": 222,
+  "Cough Classification": 223,
+  "Crisis Detection": 224,
+  "Crop Classification": 225,
+  "Crop Recommendation": 226,
+  "Cross Encoder Binary Classification": 227,
+  "Cross Encoder Classification": 228,
+  "Cross Encoder Correlation": 229,
+  "Cross Encoder Nano BEIR": 230,
+  "Cross Encoder Reranking": 231,
+  "Cross Encoder Softmax Accuracy": 232,
+  "Cross-Lingual Document Retrieval": 233,
+  "Cross-Lingual Transfer": 234,
+  "Cross-Modal Retrieval": 235,
+  "Cuisine (20 classes)": 236,
+  "Cultural Vocal Bursts Intensity Prediction": 237,
+  "Curated Test Samples": 238,
+  "Curiosity-driven Exploration": 239,
+  "Custom Information Retrieval": 240,
+  "Custom Triplet": 241,
+  "Customer Support Response Generation": 242,
+  "Cyberbullying Moderation (label + type)": 243,
+  "Cytotoxicity Prediction from Molecular Structure": 244,
+  "Cytotoxicity Prediction from Promiscuity": 245,
+  "DROP": 246,
+  "Danish EURLEX (Level 2)": 247,
+  "Data Augmentation": 248,
+  "Data-to-Text Generation": 249,
+  "Deblurring": 250,
+  "DeepFake Detection": 251,
+  "Deepfake Detection": 252,
+  "Definition Retrieval": 253,
+  "Dense Pixel Correspondence Estimation": 254,
+  "Dependency Parsing": 255,
+  "Description-guided molecule generation": 256,
+  "Detection Tasks": 257,
+  "DevOps Question Answering": 258,
+  "Device Aware Information Retrieval": 259,
+  "Dialog Navigation": 260,
+  "Discourse Parsing": 261,
+  "Disease Progression Classification (Longitudinal)": 262,
+  "DocVQA": 263,
+  "Document Classification": 264,
+  "Document Intelligence": 265,
+  "Document Layout Analysis": 266,
+  "Document Ranking": 267,
+  "Document Reranking": 268,
+  "Document Retrieval": 269,
+  "Document Summarization": 270,
+  "Document inconsistency detection (NLI-like)": 271,
+  "Document-Grounded QA": 272,
+  "Domain Adaptation": 273,
+  "Domain Generalization": 274,
+  "Domain Q&A": 275,
+  "Drilling Engineering AI": 276,
+  "Drug - Drug Interaction Classification": 277,
+  "Drug Discovery": 278,
+  "Drug-ADR Relation Extraction": 279,
+  "Dynamic Reconstruction": 280,
+  "ECG Report Generation": 281,
+  "Eastern Syriac Vocalization": 282,
+  "Educational Outcome Prediction": 283,
+  "Efficiency vs Baseline": 284,
+  "EgoSchema": 285,
+  "Email Classification": 286,
+  "Email Summarization": 287,
+  "Email Ticket Classification": 288,
+  "Embedding Synthesis over Long Context": 289,
+  "Emotion Analysis (Regression)": 290,
+  "Emotion Classification": 291,
+  "Emotion Classification in Czech": 292,
+  "Emotion Classification in German": 293,
+  "Emotion Classification in Hungarian": 294,
+  "Emotion Classification in Polish": 295,
+  "Emotion Classification in Slovak": 296,
+  "Emotion Classifier": 297,
+  "Emotion Detection": 298,
+  "Emotion Interpretation": 299,
+  "Emotion Recognition": 300,
+  "Emotion-Entailment": 301,
+  "Emotional Intelligence": 302,
+  "End-of-Turn Detection": 303,
+  "Energy Document Classification": 304,
+  "English": 305,
+  "English Document Retrieval": 306,
+  "English to Colloquial Tamil": 307,
+  "English to Marathi Translation": 308,
+  "English → Romanian": 309,
+  "English-Thai Translation Quality Assessment": 310,
+  "English-Thai Translation Quality Comparison": 311,
+  "English-Ukrainian Translation": 312,
+  "Entity Disambiguation": 313,
+  "Entity Linking": 314,
+  "Entity Resolution": 315,
+  "Entrepreneurial Readiness (low/medium/high)": 316,
+  "Event-based Object Segmentation": 317,
+  "Expert Routing": 318,
+  "Explanation Generation": 319,
+  "Extractive Question Answering": 320,
+  "Extractive Question-Answering": 321,
+  "Extractive Text Summarization": 322,
+  "Extreme Summarization": 323,
+  "Ezafe Detection": 324,
+  "F-16 longitudinal alpha tracking": 325,
+  "FLUE": 326,
+  "FQuAD": 327,
+  "Face Anti-Spoofing": 328,
+  "Face Detection": 329,
+  "Face Recognition": 330,
+  "Face Verification": 331,
+  "Facial Emotion Classification": 332,
+  "Facial Stress Level Prediction": 333,
+  "Fact Checking": 334,
+  "Fact Verification": 335,
+  "Factual Inconsistency Detection in Chart Captioning": 336,
+  "Factual accuracy": 337,
+  "Faithfulness Critic": 338,
+  "Fake News Detection": 339,
+  "Fake news classification (binary)": 340,
+  "Fallacy Detection": 341,
+  "Fashion Visual Search": 342,
+  "Feature Extraction": 343,
+  "Feedback Classification": 344,
+  "Few-Shot Image Classification": 345,
+  "Few-Shot Object Detection": 346,
+  "Few-Shot Semantic Segmentation": 347,
+  "Few-Shot Text Classification": 348,
+  "Fewshot Translation": 349,
+  "Fiction vs Non-Fiction Classification": 350,
+  "Field Classification": 351,
+  "Fill Mask": 352,
+  "Fill mask": 353,
+  "Fill-Mask": 354,
+  "Financial Advisory Generation": 355,
+  "Financial Compliance": 356,
+  "Financial Sentiment Analysis": 357,
+  "Financial Transaction Classification": 358,
+  "Financial Tweet Prediction": 359,
+  "Fine-Grained Image Classification": 360,
+  "Formal Logic": 361,
+  "Full Reconstruction (100%)": 362,
+  "Function Calling": 363,
+  "GPU Kernel Generation": 364,
+  "GSM8K": 365,
+  "GSM8K-Style Problems": 366,
+  "GSM8k": 367,
+  "GSM8k Mathematical Reasoning": 368,
+  "Gender Classification": 369,
+  "General": 370,
+  "General Domains": 371,
+  "General Knowledge": 372,
+  "General Multimodal": 373,
+  "General QA": 374,
+  "General Reasoning": 375,
+  "General Writing": 376,
+  "Generation Tasks": 377,
+  "Generative 3D Object Classification": 378,
+  "Generative Visual Question Answering": 379,
+  "GermanSTSBenchmark": 380,
+  "Gibberish Detection": 381,
+  "Global-MMLU-Lite": 382,
+  "Graded IR": 383,
+  "Grammar Classification": 384,
+  "Grammatical Error Correction": 385,
+  "Graph Classification": 386,
+  "Graph Property Prediction": 387,
+  "Graph Regression": 388,
+  "HLE Math": 389,
+  "HSwag": 390,
+  "Hallucination Detection": 391,
+  "Handwritten Text Recognition": 392,
+  "Hanoi Tower Puzzle": 393,
+  "Hanoi Tower Puzzle (Subtask-based)": 394,
+  "Hate / Not Hate classification": 395,
+  "Hate Speech Detection": 396,
+  "Hate Speech Span Detection": 397,
+  "Hate speech classification": 398,
+  "Head Pose Recognition (Facing)": 399,
+  "Head Pose Recognition (Tilt)": 400,
+  "Head Pose Recognition (Up/Down)": 401,
+  "Health Coaching": 402,
+  "Health-Aware Recipe Generation": 403,
+  "HellaSwag": 404,
+  "Hellaswag Contextual Completions": 405,
+  "High School Computer Science": 406,
+  "High School Mathematics": 407,
+  "Histopathologic Cancer Detection": 408,
+  "Historic Text Normalization (type-level)": 409,
+  "HourVideo": 410,
+  "Human Instance Segmentation": 411,
+  "Human vs AI Text Classification": 412,
+  "Human vs AI Text Detection": 413,
+  "HumanEval": 414,
+  "Humor Detection": 415,
+  "IF": 416,
+  "IaC Generation": 417,
+  "Idea Difficulty (Low/Medium/High)": 418,
+  "Image Captioning": 419,
+  "Image Classification": 420,
+  "Image Clustering": 421,
+  "Image Deblurring": 422,
+  "Image Dehazing": 423,
+  "Image Description": 424,
+  "Image Document Retrieval": 425,
+  "Image Generation": 426,
+  "Image Inpainting": 427,
+  "Image Manipulation Detection": 428,
+  "Image Manipulation Localization": 429,
+  "Image Matching": 430,
+  "Image Matting": 431,
+  "Image Outpainting": 432,
+  "Image Reconstruction": 433,
+  "Image Registration": 434,
+  "Image Restoration": 435,
+  "Image Retrieval": 436,
+  "Image Segmentation": 437,
+  "Image Super-Resolution": 438,
+  "Image To Text": 439,
+  "Image-Classification": 440,
+  "Image-to-Image Translation": 441,
+  "Image-to-Text Retrieval": 442,
+  "ImageClassification": 443,
+  "Imitation Policy Evaluation": 444,
+  "In-Context Reinforcement Learning": 445,
+  "Incremental Learning": 446,
+  "Indic-NLI": 447,
+  "Indic-Paraphrase": 448,
+  "Indic-QA Evaluation": 449,
+  "Indic-Sentiment Analysis": 450,
+  "Industrial Quality Control": 451,
+  "InfoVQA": 452,
+  "Information Retrieval": 453,
+  "Instance Segmentation": 454,
+  "Instruct": 455,
+  "Instruction Following": 456,
+  "Instruction following": 457,
+  "InstructionRetrieval": 458,
+  "Instrument Recognition": 459,
+  "Intent Classification": 460,
+  "Interactive Segmentation": 461,
+  "Irony Detection": 462,
+  "JPEG Decompression": 463,
+  "JPRDY": 464,
+  "KG-to-Text Generation": 465,
+  "KLUE-STS": 466,
+  "KLUE-TC": 467,
+  "KSM": 468,
+  "Key Information Extraction": 469,
+  "Keyphrase Extraction": 470,
+  "Keyword Extraction": 471,
+  "Keyword Spotting": 472,
+  "Knowledge": 473,
+  "Knowledge & QA": 474,
+  "Knowledge Benchmarking": 475,
+  "Knowledge Distillation": 476,
+  "Knowledge Graphs": 477,
+  "Ko-StrategyQA": 478,
+  "KorSTS": 479,
+  "LABELED_DEPENDENCIES": 480,
+  "LBHistoricalBitextMining": 481,
+  "LEMMA": 482,
+  "LSR": 483,
+  "Lane Detection": 484,
+  "Language Identification": 485,
+  "Language Modeling": 486,
+  "Language Modelling": 487,
+  "Language Sentiment Analysis": 488,
+  "Language Understanding": 489,
+  "Large Language Model": 490,
+  "Latent Diffusion Model for 3D": 491,
+  "Latent Diffusion Model for 3D - Pano": 492,
+  "Latent Diffusion Model for 3D - Super-Resolution": 493,
+  "Latent Diffusion Model for 3D-4C": 494,
+  "Legal Case Analysis": 495,
+  "Legal Document Retrieval": 496,
+  "Legal Document Summarization": 497,
+  "Legal Q&A (PT-PT)": 498,
+  "Lemmatisation": 499,
+  "Lexical Relation Classification (BLESS)": 500,
+  "Lexical Relation Classification (CogALexV)": 501,
+  "Lexical Relation Classification (EVALution)": 502,
+  "Lexical Relation Classification (K&H+N)": 503,
+  "Lexical Relation Classification (ROOT09)": 504,
+  "Lexical bias detection": 505,
+  "Linguistic Acceptability": 506,
+  "Linguistic Accuracy Evaluation": 507,
+  "Link Prediction": 508,
+  "Literary Explicitness Classification": 509,
+  "Logging": 510,
+  "Logical Reasoning": 511,
+  "Long Context": 512,
+  "Long Video Retrieval (Background Removed)": 513,
+  "Long context": 514,
+  "Long, Legal Document Summarization": 515,
+  "Long-Context Hallucination Detection": 516,
+  "Long-Context Understanding": 517,
+  "Long-horizon": 518,
+  "Long-tail Learning": 519,
+  "LongVideoBench": 520,
+  "Lung Nodule Detection": 521,
+  "MATH": 522,
+  "MBTI Personality Classification": 523,
+  "MC2, 10-shot": 524,
+  "MIRACL-Reranking": 525,
+  "MIRACL-Retrieval": 526,
+  "MMLU": 527,
+  "MMLU Knowledge Test": 528,
+  "MMLU-Pro": 529,
+  "MMR total": 530,
+  "MMVP": 531,
+  "MORPH": 532,
+  "MTOPDomainClassification": 533,
+  "MTOPIntentClassification": 534,
+  "MVBench": 535,
+  "Machine Translation": 536,
+  "Machine Translation (sa → en)": 537,
+  "Machine Translation Evaluation": 538,
+  "Manipulation Detection": 539,
+  "Market Direction Prediction": 540,
+  "Marketing Domain Q&A": 541,
+  "Masked Language Modeling": 542,
+  "Masked Language Modelling": 543,
+  "Masked Prediction (30%)": 544,
+  "Massive Multitask Language Understanding": 545,
+  "MassiveIntentClassification": 546,
+  "MassiveScenarioClassification": 547,
+  "Math": 548,
+  "Math Reasoning": 549,
+  "Math Word Problem Solving": 550,
+  "Math Word Problems": 551,
+  "Math word problems": 552,
+  "Mathematical Problem-Solving": 553,
+  "Mathematical Reasoning": 554,
+  "Mathematical Reasoning w/ Tools": 555,
+  "Mathematical problem solving": 556,
+  "Mathematical reasoning": 557,
+  "Mathematics": 558,
+  "Medical": 559,
+  "Medical Image Classification": 560,
+  "Medical Image Segmentation": 561,
+  "Medical Knowledge": 562,
+  "Medical Literature Search": 563,
+  "Medical Question Answering": 564,
+  "Medical SOAP Note Generation": 565,
+  "Medical Text Generation": 566,
+  "Meme Classification": 567,
+  "Memorization": 568,
+  "Military Audio Classification": 569,
+  "Misogyny Detection": 570,
+  "Misogyny Identification": 571,
+  "Model Compression": 572,
+  "Molecular Property Prediction": 573,
+  "Molecule Captioning": 574,
+  "Moment Retrieval": 575,
+  "Monocular Depth Estimation": 576,
+  "Monolingual Document Retrieval": 577,
+  "Morphological tagging (first subtoken)": 578,
+  "Motion Synthesis": 579,
+  "Multi Class Text Classification": 580,
+  "Multi Task Dev": 581,
+  "Multi-Head Text Regression": 582,
+  "Multi-Label Classification": 583,
+  "Multi-Label Emotion Classification": 584,
+  "Multi-Label Image Classification": 585,
+  "Multi-Label Intent Detection": 586,
+  "Multi-Label Text Classification": 587,
+  "Multi-Modal Hate Speech Detection": 588,
+  "Multi-Object Tracking": 589,
+  "Multi-Person Pose Estimation": 590,
+  "Multi-Source Reasoning (MUSR)": 591,
+  "Multi-class Classification": 592,
+  "Multi-class Text Classification": 593,
+  "Multi-label Emotion Classification": 594,
+  "Multi-label Fine-Grained Emotion Classification": 595,
+  "Multi-label Text Classification": 596,
+  "Multi-task language understanding": 597,
+  "Multi-tissue Nucleus Segmentation": 598,
+  "Multi-turn conversation": 599,
+  "Multi-turn conversation quality": 600,
+  "Multilabel Text Classification": 601,
+  "MultilabelClassification": 602,
+  "Multilingual": 603,
+  "Multilingual Emotion Classification": 604,
+  "Multilingual Math (MGSM)": 605,
+  "Multilingual QA": 606,
+  "Multilingual Retrieval": 607,
+  "Multilingual VLN": 608,
+  "Multimodal Code Generation": 609,
+  "Multimodal Emotion Recognition": 610,
+  "Multimodal Reasoning": 611,
+  "Multimodal medical knowledge and reasoning": 612,
+  "Multiple Choice": 613,
+  "Multiple Choice Question Answering": 614,
+  "Multiple Choice Question Generation": 615,
+  "Multiple Object Tracking": 616,
+  "Multiple-choice": 617,
+  "Multi‑Label Music Note Prediction": 618,
+  "Music Auto-Tagging": 619,
+  "Music Question Answering": 620,
+  "Music Source Separation": 621,
+  "Music Transcription": 622,
+  "My Binary Classification": 623,
+  "NER": 624,
+  "NER (9 tags)": 625,
+  "NER F1 Score": 626,
+  "NFCorpus": 627,
+  "NSFW/explicit content": 628,
+  "Named Entity Recognition": 629,
+  "Named Entity Recognition (Invoices)": 630,
+  "Named Entity Recognition (NER)": 631,
+  "Nano BEIR": 632,
+  "Narrative Genre Classification": 633,
+  "NatQs": 634,
+  "Natural Language Inference": 635,
+  "Natural Language Queries": 636,
+  "Natural Language Understanding": 637,
+  "Natural Language Visual Grounding": 638,
+  "Natural Language to Bash Translation": 639,
+  "Natural Lenguage Inference": 640,
+  "Natural language inference": 641,
+  "Negative Binomial GLM Parameter Estimation": 642,
+  "Nep-gLUE": 643,
+  "Nepali Speech Recognition": 644,
+  "Ner": 645,
+  "Network Pruning": 646,
+  "Neural Architecture Search": 647,
+  "News Classification": 648,
+  "News Summarization": 649,
+  "Node Classification": 650,
+  "Non-thinking": 651,
+  "OBQA": 652,
+  "OCR": 653,
+  "OMNI Math": 654,
+  "Object Categorization": 655,
+  "Object Counting": 656,
+  "Object Detection": 657,
+  "Object Localization": 658,
+  "Object Navigation": 659,
+  "Object Rearrangement": 660,
+  "Object Recognition": 661,
+  "Object Tracking": 662,
+  "Object visual presence verification": 663,
+  "Object-Oriented Navigation": 664,
+  "Online Beat Tracking": 665,
+  "Open Information Extraction": 666,
+  "Open Vocabulary Object Detection": 667,
+  "Open Vocabulary Panoptic Segmentation": 668,
+  "Open Vocabulary Semantic Segmentation": 669,
+  "Open-Domain Question Answering": 670,
+  "OpenAI Gym": 671,
+  "OpenAPI code completion": 672,
+  "OpenBookQA Facts": 673,
+  "Optical Character Recognition": 674,
+  "Optical Character Recognition (OCR)": 675,
+  "Optical Flow Estimation": 676,
+  "OrangeSum": 677,
+  "Osteoporosis Risk Prediction": 678,
+  "Out-of-Distribution Detection": 679,
+  "PDF-to-JSON Lab Test Data Conversion": 680,
+  "PII Masking": 681,
+  "PII Masking and Classification": 682,
+  "PII Routing": 683,
+  "PIQA": 684,
+  "PIQA Problem Solving": 685,
+  "POS": 686,
+  "POS Tagging": 687,
+  "Pair Classification": 688,
+  "PairClassification": 689,
+  "Pairwise Preference Ranking": 690,
+  "Panoptic Segmentation": 691,
+  "Paraphrase Detection": 692,
+  "Paraphrase Identification": 693,
+  "Paraphrase Mining": 694,
+  "Parking Space Occupancy": 695,
+  "Part of Speech Tagging": 696,
+  "Part-aware Panoptic Segmentation": 697,
+  "Part-of-Speech Tagging": 698,
+  "Participant Intervention Comparison Outcome Extraction": 699,
+  "Passage Ranking": 700,
+  "Passage Reranking": 701,
+  "Passage Retrieval": 702,
+  "Path Reconstruction": 703,
+  "Pedestrian Detection": 704,
+  "Perception Test": 705,
+  "Person Identification": 706,
+  "Person Re-Identification": 707,
+  "Personalized Image Generation": 708,
+  "Personalized Segmentation": 709,
+  "Phoneme Recognition": 710,
+  "Phrase Grounding": 711,
+  "PiQA": 712,
+  "Pick and Place": 713,
+  "Pitch Angle Tracking Control": 714,
+  "Planetary Recognition Lattice": 715,
+  "Plant Disease Classification": 716,
+  "Poems Annotation Generation": 717,
+  "Point Cloud Classification": 718,
+  "Point Cloud Segmentation": 719,
+  "Point Clouds": 720,
+  "Popular aggregated benchmark": 721,
+  "Pose Estimation": 722,
+  "Potato Late Blight Risk Classification": 723,
+  "Product Category Classification": 724,
+  "Professional Law": 725,
+  "Program synthesis": 726,
+  "Prompt Engineering": 727,
+  "Prompt Generation (Dev)": 728,
+  "Prompt Generation (Test)": 729,
+  "Prompt Harmfulness Classification": 730,
+  "Prompt Injection Detection": 731,
+  "Prompt Safety Classification": 732,
+  "Prompt injection detection": 733,
+  "Protein Design": 734,
+  "Protein Function Prediction": 735,
+  "Protein Secondary Structure Prediction": 736,
+  "Protein Structure Prediction": 737,
+  "Protocol Quality Assessment": 738,
+  "PubMedQA": 739,
+  "Py Late Information Retrieval": 740,
+  "PyTest edge-case unit test generation": 741,
+  "PyTest unit test generation": 742,
+  "Python Code Synthesis": 743,
+  "Python code generation": 744,
+  "QA": 745,
+  "QA (Span Extraction)": 746,
+  "QA (ViquiQuAD)": 747,
+  "QA (XQuAD)": 748,
+  "Quantization": 749,
+  "Question Answering": 750,
+  "Question Answering Classification": 751,
+  "Question Duplicate Detection": 752,
+  "Question Generation": 753,
+  "Question Pair Duplicate Detection": 754,
+  "Question-Answering": 755,
+  "RBC Shape Classification": 756,
+  "RE": 757,
+  "ROUGE-1": 758,
+  "RPG Art Generation": 759,
+  "RST-Pointer": 760,
+  "RZTKInformation Retrieval": 761,
+  "Radiology Document Retrieval": 762,
+  "Ranking": 763,
+  "Re-writing": 764,
+  "Reading Comprehension": 765,
+  "Reasoning": 766,
+  "Reasoning Quality Classification": 767,
+  "Receipt Entity Extraction": 768,
+  "Recognizing Emotion Cause in Conversations": 769,
+  "Referring Expression Grounding": 770,
+  "Referring Expression Segmentation": 771,
+  "Refusal Detection": 772,
+  "Region (5 classes)": 773,
+  "Region of interest detection": 774,
+  "Regression": 775,
+  "Regression (RMSE)": 776,
+  "Regulation Retrieval": 777,
+  "Regulatory Classification": 778,
+  "Regulatory Guidance": 779,
+  "Reinforcement Learning": 780,
+  "Reinforcement Learning Teaching": 781,
+  "Relation Classification": 782,
+  "Relation Extraction": 783,
+  "Relation Mapping": 784,
+  "Remote Sensing Image Classification": 785,
+  "Representation Learning": 786,
+  "Requirement Classification": 787,
+  "Reranking": 788,
+  "Reranking (query–product relevance)": 789,
+  "Response Generation": 790,
+  "Response Harmfulness Classification": 791,
+  "Resume Classification": 792,
+  "Retinal Vessel Segmentation": 793,
+  "Retrieval": 794,
+  "Reward Hack Detection": 795,
+  "Reward Modeling": 796,
+  "Risk Tolerance (Low/Medium/High)": 797,
+  "Robot Control": 798,
+  "Robot Manipulation": 799,
+  "Robotic Manipulation": 800,
+  "Robustness Tests": 801,
+  "Role-Aware Multi-Label Abuse Pattern Detection": 802,
+  "S2TT": 803,
+  "SENTS": 804,
+  "SICK-R": 805,
+  "SIQA": 806,
+  "SQuAD": 807,
+  "STEM": 808,
+  "STS": 809,
+  "STS Benchmark": 810,
+  "STS-ca": 811,
+  "STSBenchmark": 812,
+  "Safety & Compliance": 813,
+  "Sarcasm Detection": 814,
+  "Scene Change Detection": 815,
+  "Scene Classification": 816,
+  "Scene Flow Estimation": 817,
+  "Scene Segmentation": 818,
+  "Scene Text Recognition": 819,
+  "Scientific text generation": 820,
+  "Secret Detection": 821,
+  "Secret Detection (Long Context)": 822,
+  "Segmentation": 823,
+  "Segmentation Tasks": 824,
+  "Self-Supervised Learning": 825,
+  "Semantic Evidence Filtering": 826,
+  "Semantic Parsing": 827,
+  "Semantic Retrieval": 828,
+  "Semantic Search": 829,
+  "Semantic Segmentation": 830,
+  "Semantic Similarity": 831,
+  "Semantic Similarity (STS Validation)": 832,
+  "Semantic Textual Similarity": 833,
+  "Semantic Textual Similarity (Azerbaijani)": 834,
+  "Semantic entity labeling": 835,
+  "Semi-Supervised Image Classification": 836,
+  "Semi-Supervised Instance Segmentation": 837,
+  "Semi-Supervised Video Object Segmentation": 838,
+  "Sentence Classification": 839,
+  "Sentence Completion": 840,
+  "Sentence Ordering": 841,
+  "Sentence Relevance Classification": 842,
+  "Sentence Similarity": 843,
+  "Sentence completion": 844,
+  "Sentence-Embedding": 845,
+  "Sentic-GCN": 846,
+  "Sentic-GCN Bert": 847,
+  "Sentiment Analysis": 848,
+  "Sentiment Analysis (Regression)": 849,
+  "Sentiment Classification": 850,
+  "Sentiment classification": 851,
+  "Sequence Classification": 852,
+  "Sequence Labeling": 853,
+  "Sequence-to-sequence Language Modeling": 854,
+  "ShaderEval": 855,
+  "Short-term Object Interaction Anticipation": 856,
+  "Sign Language Recognition": 857,
+  "Silhouette": 858,
+  "Single Choice Question": 859,
+  "Single-object discovery": 860,
+  "Skill Level (Low/Medium/High)": 861,
+  "Skin Tumor Classification": 862,
+  "Slot Filling": 863,
+  "Solubility": 864,
+  "Solving Partial Differential Equations": 865,
+  "Space-time Video Super-resolution": 866,
+  "Spam / Ham Classification": 867,
+  "Spam Detection": 868,
+  "Spam Review Detection": 869,
+  "Span-Extraction": 870,
+  "Sparse Binary Classification": 871,
+  "Sparse Information Retrieval": 872,
+  "Sparse Learning": 873,
+  "Sparse Nano BEIR": 874,
+  "Spatial Reasoning": 875,
+  "Speaker Diarization": 876,
+  "Speaker Identification": 877,
+  "Speaker Recognition": 878,
+  "Speaker Verification": 879,
+  "Specialized Capabilities": 880,
+  "Speech Emotion Recognition": 881,
+  "Speech Enhancement": 882,
+  "Speech Recognition": 883,
+  "Speech Separation": 884,
+  "Speech Synthesis": 885,
+  "Speech Translation": 886,
+  "Speech Translation (ML→EN)": 887,
+  "Speech-to-Phoneme": 888,
+  "Speech-to-Speech Translation": 889,
+  "Speech-to-Text": 890,
+  "Speech-to-Text Translation": 891,
+  "Speed": 892,
+  "Spoken Command Recognition": 893,
+  "Spoken Language Understanding": 894,
+  "Stance Classification": 895,
+  "StarCraft Multi-Agent Challenge v2": 896,
+  "Stereo Depth Estimation": 897,
+  "Stereo Disparity Estimation": 898,
+  "Stereotypical Bias Analysis": 899,
+  "Stock Market Prediction": 900,
+  "Stock Trading": 901,
+  "Story Continuation": 902,
+  "Story Point Estimation": 903,
+  "Strategy QA (internal heuristic eval)": 904,
+  "Strong Gravitational Lens Discovery": 905,
+  "Style classification (holdout)": 906,
+  "Style classification (real-world baseline)": 907,
+  "Subjectivity Analysis": 908,
+  "Subjectivity Detection": 909,
+  "Suggestive Content Detection": 910,
+  "Suicidal Tendency Prediction in text": 911,
+  "Suicide Risk Detection": 912,
+  "Summarization": 913,
+  "Super Resolution": 914,
+  "Surgical Triplet Recognition": 915,
+  "Syriac Vocalization": 916,
+  "TAG": 917,
+  "TC": 918,
+  "TEca": 919,
+  "TOON conversion (schema-driven extraction)": 920,
+  "TabFQuAD": 921,
+  "Table Detection": 922,
+  "Table-to-Text Generation": 923,
+  "Tabular Classification": 924,
+  "Tabular Regression": 925,
+  "Target Prioritization": 926,
+  "TeCla": 927,
+  "Temporal Action Localization": 928,
+  "Temporal Relation Extraction": 929,
+  "Temporal Sentence Grounding": 930,
+  "Text Classification": 931,
+  "Text Classification (Sentiment Analysis)": 932,
+  "Text Classification (multi-label emotions)": 933,
+  "Text Classification Denial": 934,
+  "Text Classification Question": 935,
+  "Text Clustering": 936,
+  "Text Detection": 937,
+  "Text Generation": 938,
+  "Text Generation (Field Normalization)": 939,
+  "Text Generation (In-Domain)": 940,
+  "Text Generation (Out-of-Domain)": 941,
+  "Text Regression": 942,
+  "Text Retrieval": 943,
+  "Text Simplification": 944,
+  "Text Summarization": 945,
+  "Text To Speech": 946,
+  "Text Tokenization": 947,
+  "Text classification": 948,
+  "Text generation": 949,
+  "Text to 3D": 950,
+  "Text to Audio Retrieval": 951,
+  "Text to Molecular Generation": 952,
+  "Text to SQL": 953,
+  "Text to Speech": 954,
+  "Text-To-SQL": 955,
+  "Text-To-Speech Synthesis": 956,
+  "Text-based de novo Molecule Generation": 957,
+  "Text-classification": 958,
+  "Text-to-Image Generation": 959,
+  "Text-to-Music Generation": 960,
+  "Text-to-Speech": 961,
+  "Text-to-Video Generation": 962,
+  "Text2Text Generation": 963,
+  "The Semantic Segmentation Of Remote Sensing Imagery": 964,
+  "Theory of Mind": 965,
+  "Thinking": 966,
+  "Time Series Forecasting": 967,
+  "TinyQA Benchmark++": 968,
+  "Token Classification": 969,
+  "Token classification": 970,
+  "Tomato": 971,
+  "Tool Use": 972,
+  "Topic Classification": 973,
+  "Toxic-detector-cnn": 974,
+  "Toxic-detector-rnn": 975,
+  "Toxic-detector-roberta": 976,
+  "Toxicity (12 tasks)": 977,
+  "Toxicity Detection": 978,
+  "Track classification": 979,
+  "Trading": 980,
+  "Traffic Prediction": 981,
+  "Training-free 3D Part Segmentation": 982,
+  "Training-free 3D Point Cloud Classification": 983,
+  "Transit Route Planning": 984,
+  "Translation": 985,
+  "Translation (de-en)": 986,
+  "Translation En-to-ES": 987,
+  "Translation English-to-Swahili": 988,
+  "Translation Quality Estimation": 989,
+  "Translation acm-deu": 990,
+  "Translation acm-eng": 991,
+  "Translation acm-fra": 992,
+  "Translation acm-por": 993,
+  "Translation acm-spa": 994,
+  "Translation afr-deu": 995,
+  "Translation afr-eng": 996,
+  "Translation afr-fra": 997,
+  "Translation afr-nld": 998,
+  "Translation afr-por": 999,
+  "Translation afr-spa": 1000,
+  "Translation amh-deu": 1001,
+  "Translation amh-eng": 1002,
+  "Translation amh-fra": 1003,
+  "Translation amh-por": 1004,
+  "Translation amh-spa": 1005,
+  "Translation apc-deu": 1006,
+  "Translation apc-eng": 1007,
+  "Translation apc-fra": 1008,
+  "Translation apc-por": 1009,
+  "Translation apc-spa": 1010,
+  "Translation ara-cat": 1011,
+  "Translation ara-dan": 1012,
+  "Translation ara-deu": 1013,
+  "Translation ara-eng": 1014,
+  "Translation ara-fra": 1015,
+  "Translation ara-glg": 1016,
+  "Translation ara-ita": 1017,
+  "Translation ara-nob": 1018,
+  "Translation ara-por": 1019,
+  "Translation ara-ron": 1020,
+  "Translation ara-spa": 1021,
+  "Translation ara-swe": 1022,
+  "Translation arb-eng": 1023,
+  "Translation arz-deu": 1024,
+  "Translation arz-eng": 1025,
+  "Translation arz-fra": 1026,
+  "Translation arz-por": 1027,
+  "Translation arz-spa": 1028,
+  "Translation asm-eng": 1029,
+  "Translation asm-fra": 1030,
+  "Translation asm-por": 1031,
+  "Translation ast-cat": 1032,
+  "Translation ast-deu": 1033,
+  "Translation ast-eng": 1034,
+  "Translation ast-fra": 1035,
+  "Translation ast-glg": 1036,
+  "Translation ast-ita": 1037,
+  "Translation ast-oci": 1038,
+  "Translation ast-por": 1039,
+  "Translation ast-ron": 1040,
+  "Translation ast-spa": 1041,
+  "Translation awa-deu": 1042,
+  "Translation awa-eng": 1043,
+  "Translation awa-fra": 1044,
+  "Translation awa-por": 1045,
+  "Translation awa-spa": 1046,
+  "Translation aze_Latn-deu": 1047,
+  "Translation aze_Latn-eng": 1048,
+  "Translation aze_Latn-fra": 1049,
+  "Translation aze_Latn-por": 1050,
+  "Translation aze_Latn-spa": 1051,
+  "Translation bak-eng": 1052,
+  "Translation ban-eng": 1053,
+  "Translation ban-fra": 1054,
+  "Translation ban-por": 1055,
+  "Translation bar-bar": 1056,
+  "Translation bel-cat": 1057,
+  "Translation bel-deu": 1058,
+  "Translation bel-eng": 1059,
+  "Translation bel-fra": 1060,
+  "Translation bel-glg": 1061,
+  "Translation bel-ita": 1062,
+  "Translation bel-pol": 1063,
+  "Translation bel-por": 1064,
+  "Translation bel-ron": 1065,
+  "Translation bel-rus": 1066,
+  "Translation bel-spa": 1067,
+  "Translation bel-ukr": 1068,
+  "Translation bem-eng": 1069,
+  "Translation bem-fra": 1070,
+  "Translation bem-por": 1071,
+  "Translation bem-spa": 1072,
+  "Translation ben-deu": 1073,
+  "Translation ben-eng": 1074,
+  "Translation ben-fra": 1075,
+  "Translation ben-por": 1076,
+  "Translation ben-spa": 1077,
+  "Translation bho-deu": 1078,
+  "Translation bho-eng": 1079,
+  "Translation bho-fra": 1080,
+  "Translation bho-por": 1081,
+  "Translation bho-spa": 1082,
+  "Translation bos_Latn-eng": 1083,
+  "Translation bre-eng": 1084,
+  "Translation bre-fra": 1085,
+  "Translation bul-deu": 1086,
+  "Translation bul-eng": 1087,
+  "Translation bul-fra": 1088,
+  "Translation bul-ita": 1089,
+  "Translation bul-por": 1090,
+  "Translation bul-ron": 1091,
+  "Translation bul-rus": 1092,
+  "Translation bul-spa": 1093,
+  "Translation bul-ukr": 1094,
+  "Translation cat-ara": 1095,
+  "Translation cat-ast": 1096,
+  "Translation cat-deu": 1097,
+  "Translation cat-eng": 1098,
+  "Translation cat-fra": 1099,
+  "Translation cat-glg": 1100,
+  "Translation cat-heb": 1101,
+  "Translation cat-ita": 1102,
+  "Translation cat-lav": 1103,
+  "Translation cat-lit": 1104,
+  "Translation cat-oci": 1105,
+  "Translation cat-por": 1106,
+  "Translation cat-ron": 1107,
+  "Translation cat-spa": 1108,
+  "Translation cat-tur": 1109,
+  "Translation ceb-deu": 1110,
+  "Translation ceb-eng": 1111,
+  "Translation ceb-fra": 1112,
+  "Translation ceb-por": 1113,
+  "Translation ceb-spa": 1114,
+  "Translation ces-deu": 1115,
+  "Translation ces-eng": 1116,
+  "Translation ces-fra": 1117,
+  "Translation ces-por": 1118,
+  "Translation ces-rus": 1119,
+  "Translation ces-spa": 1120,
+  "Translation ces-ukr": 1121,
+  "Translation ckb-deu": 1122,
+  "Translation ckb-eng": 1123,
+  "Translation ckb-fra": 1124,
+  "Translation ckb-por": 1125,
+  "Translation ckb-spa": 1126,
+  "Translation cmn_Hans-eng": 1127,
+  "Translation cmn_Hans-fra": 1128,
+  "Translation cmn_Hans-por": 1129,
+  "Translation cmn_Hans-spa": 1130,
+  "Translation cmn_Hant-eng": 1131,
+  "Translation cmn_Hant-fra": 1132,
+  "Translation cmn_Hant-por": 1133,
+  "Translation cmn_Hant-spa": 1134,
+  "Translation crh-deu": 1135,
+  "Translation crh-eng": 1136,
+  "Translation crh-fra": 1137,
+  "Translation crh-por": 1138,
+  "Translation crh-spa": 1139,
+  "Translation cym-deu": 1140,
+  "Translation cym-eng": 1141,
+  "Translation cym-fra": 1142,
+  "Translation cym-por": 1143,
+  "Translation cym-spa": 1144,
+  "Translation dan-ara": 1145,
+  "Translation dan-cat": 1146,
+  "Translation dan-ces": 1147,
+  "Translation dan-deu": 1148,
+  "Translation dan-eng": 1149,
+  "Translation dan-fra": 1150,
+  "Translation dan-glg": 1151,
+  "Translation dan-heb": 1152,
+  "Translation dan-isl": 1153,
+  "Translation dan-ita": 1154,
+  "Translation dan-nob": 1155,
+  "Translation dan-pol": 1156,
+  "Translation dan-por": 1157,
+  "Translation dan-ron": 1158,
+  "Translation dan-rus": 1159,
+  "Translation dan-spa": 1160,
+  "Translation dan-swe": 1161,
+  "Translation dan-tur": 1162,
+  "Translation dan-ukr": 1163,
+  "Translation deu-afr": 1164,
+  "Translation deu-ara": 1165,
+  "Translation deu-ast": 1166,
+  "Translation deu-bel": 1167,
+  "Translation deu-ben": 1168,
+  "Translation deu-bul": 1169,
+  "Translation deu-cat": 1170,
+  "Translation deu-ces": 1171,
+  "Translation deu-cym": 1172,
+  "Translation deu-dan": 1173,
+  "Translation deu-deu": 1174,
+  "Translation deu-ell": 1175,
+  "Translation deu-eng": 1176,
+  "Translation deu-est": 1177,
+  "Translation deu-fao": 1178,
+  "Translation deu-fas": 1179,
+  "Translation deu-fin": 1180,
+  "Translation deu-fra": 1181,
+  "Translation deu-fur": 1182,
+  "Translation deu-gle": 1183,
+  "Translation deu-glg": 1184,
+  "Translation deu-guj": 1185,
+  "Translation deu-hat": 1186,
+  "Translation deu-hau": 1187,
+  "Translation deu-heb": 1188,
+  "Translation deu-hin": 1189,
+  "Translation deu-hne": 1190,
+  "Translation deu-hrv": 1191,
+  "Translation deu-hun": 1192,
+  "Translation deu-isl": 1193,
+  "Translation deu-ita": 1194,
+  "Translation deu-lad": 1195,
+  "Translation deu-lav": 1196,
+  "Translation deu-lij": 1197,
+  "Translation deu-lit": 1198,
+  "Translation deu-ltz": 1199,
+  "Translation deu-mag": 1200,
+  "Translation deu-mkd": 1201,
+  "Translation deu-mlt": 1202,
+  "Translation deu-nds": 1203,
+  "Translation deu-nld": 1204,
+  "Translation deu-nno": 1205,
+  "Translation deu-nob": 1206,
+  "Translation deu-nor": 1207,
+  "Translation deu-oci": 1208,
+  "Translation deu-pan": 1209,
+  "Translation deu-pap": 1210,
+  "Translation deu-pes": 1211,
+  "Translation deu-pol": 1212,
+  "Translation deu-por": 1213,
+  "Translation deu-prs": 1214,
+  "Translation deu-ron": 1215,
+  "Translation deu-rus": 1216,
+  "Translation deu-slk": 1217,
+  "Translation deu-slv": 1218,
+  "Translation deu-spa": 1219,
+  "Translation deu-sqi": 1220,
+  "Translation deu-srd": 1221,
+  "Translation deu-srp_Cyrl": 1222,
+  "Translation deu-swa": 1223,
+  "Translation deu-swe": 1224,
+  "Translation deu-tgk": 1225,
+  "Translation deu-tpi": 1226,
+  "Translation deu-tsn": 1227,
+  "Translation deu-ukr": 1228,
+  "Translation deu-urd": 1229,
+  "Translation deu-vie": 1230,
+  "Translation drt-deu": 1231,
+  "Translation drt-eng": 1232,
+  "Translation drt-fry": 1233,
+  "Translation drt-nld": 1234,
+  "Translation dsb-deu": 1235,
+  "Translation ell-deu": 1236,
+  "Translation ell-eng": 1237,
+  "Translation ell-fra": 1238,
+  "Translation ell-por": 1239,
+  "Translation ell-spa": 1240,
+  "Translation en-ru": 1241,
+  "Translation eng-afr": 1242,
+  "Translation eng-ara": 1243,
+  "Translation eng-arz": 1244,
+  "Translation eng-ast": 1245,
+  "Translation eng-bel": 1246,
+  "Translation eng-ben": 1247,
+  "Translation eng-bho": 1248,
+  "Translation eng-bos_Latn": 1249,
+  "Translation eng-bul": 1250,
+  "Translation eng-cat": 1251,
+  "Translation eng-ces": 1252,
+  "Translation eng-cym": 1253,
+  "Translation eng-dan": 1254,
+  "Translation eng-deu": 1255,
+  "Translation eng-ell": 1256,
+  "Translation eng-eng": 1257,
+  "Translation eng-est": 1258,
+  "Translation eng-fao": 1259,
+  "Translation eng-fas": 1260,
+  "Translation eng-fin": 1261,
+  "Translation eng-fra": 1262,
+  "Translation eng-fry": 1263,
+  "Translation eng-fur": 1264,
+  "Translation eng-gla": 1265,
+  "Translation eng-gle": 1266,
+  "Translation eng-glg": 1267,
+  "Translation eng-guj": 1268,
+  "Translation eng-hat": 1269,
+  "Translation eng-hau": 1270,
+  "Translation eng-hbs": 1271,
+  "Translation eng-heb": 1272,
+  "Translation eng-hin": 1273,
+  "Translation eng-hne": 1274,
+  "Translation eng-hrv": 1275,
+  "Translation eng-hun": 1276,
+  "Translation eng-ind": 1277,
+  "Translation eng-isl": 1278,
+  "Translation eng-ita": 1279,
+  "Translation eng-jpg": 1280,
+  "Translation eng-jpn": 1281,
+  "Translation eng-kea": 1282,
+  "Translation eng-kin": 1283,
+  "Translation eng-kor": 1284,
+  "Translation eng-lad": 1285,
+  "Translation eng-lad_Latn": 1286,
+  "Translation eng-lat": 1287,
+  "Translation eng-lav": 1288,
+  "Translation eng-lij": 1289,
+  "Translation eng-lin": 1290,
+  "Translation eng-lit": 1291,
+  "Translation eng-ltz": 1292,
+  "Translation eng-lug": 1293,
+  "Translation eng-mag": 1294,
+  "Translation eng-mai": 1295,
+  "Translation eng-mar": 1296,
+  "Translation eng-mkd": 1297,
+  "Translation eng-mld": 1298,
+  "Translation eng-mlt": 1299,
+  "Translation eng-nds": 1300,
+  "Translation eng-nep": 1301,
+  "Translation eng-nld": 1302,
+  "Translation eng-nno": 1303,
+  "Translation eng-nob": 1304,
+  "Translation eng-nor": 1305,
+  "Translation eng-nso": 1306,
+  "Translation eng-nya": 1307,
+  "Translation eng-oci": 1308,
+  "Translation eng-pan": 1309,
+  "Translation eng-pap": 1310,
+  "Translation eng-pes": 1311,
+  "Translation eng-pol": 1312,
+  "Translation eng-por": 1313,
+  "Translation eng-prs": 1314,
+  "Translation eng-pus": 1315,
+  "Translation eng-ron": 1316,
+  "Translation eng-rus": 1317,
+  "Translation eng-sco": 1318,
+  "Translation eng-sin": 1319,
+  "Translation eng-slk": 1320,
+  "Translation eng-slv": 1321,
+  "Translation eng-sna": 1322,
+  "Translation eng-som": 1323,
+  "Translation eng-sot": 1324,
+  "Translation eng-spa": 1325,
+  "Translation eng-sqi": 1326,
+  "Translation eng-srd": 1327,
+  "Translation eng-srn": 1328,
+  "Translation eng-srp_Cyrl": 1329,
+  "Translation eng-srp_Latn": 1330,
+  "Translation eng-swa": 1331,
+  "Translation eng-swe": 1332,
+  "Translation eng-tgk": 1333,
+  "Translation eng-tgk_Cyrl": 1334,
+  "Translation eng-tha": 1335,
+  "Translation eng-tpi": 1336,
+  "Translation eng-tsn": 1337,
+  "Translation eng-tso": 1338,
+  "Translation eng-tur": 1339,
+  "Translation eng-ukr": 1340,
+  "Translation eng-urd": 1341,
+  "Translation eng-vie": 1342,
+  "Translation eng-xho": 1343,
+  "Translation eng-zho": 1344,
+  "Translation eng-zul": 1345,
+  "Translation enm-deu": 1346,
+  "Translation enm-eng": 1347,
+  "Translation enm-fry": 1348,
+  "Translation enm-ltz": 1349,
+  "Translation enm-nld": 1350,
+  "Translation epo-deu": 1351,
+  "Translation epo-eng": 1352,
+  "Translation epo-fra": 1353,
+  "Translation epo-por": 1354,
+  "Translation epo-spa": 1355,
+  "Translation est-deu": 1356,
+  "Translation est-eng": 1357,
+  "Translation est-fra": 1358,
+  "Translation est-por": 1359,
+  "Translation est-spa": 1360,
+  "Translation eus-deu": 1361,
+  "Translation eus-eng": 1362,
+  "Translation eus-fra": 1363,
+  "Translation eus-por": 1364,
+  "Translation eus-spa": 1365,
+  "Translation fao-deu": 1366,
+  "Translation fao-eng": 1367,
+  "Translation fao-fra": 1368,
+  "Translation fao-por": 1369,
+  "Translation fao-spa": 1370,
+  "Translation fas-dan": 1371,
+  "Translation fas-deu": 1372,
+  "Translation fas-eng": 1373,
+  "Translation fas-fra": 1374,
+  "Translation fas-ita": 1375,
+  "Translation fas-por": 1376,
+  "Translation fas-ron": 1377,
+  "Translation fas-spa": 1378,
+  "Translation fij-eng": 1379,
+  "Translation fil-deu": 1380,
+  "Translation fil-eng": 1381,
+  "Translation fil-fra": 1382,
+  "Translation fil-por": 1383,
+  "Translation fil-spa": 1384,
+  "Translation fin-bul": 1385,
+  "Translation fin-deu": 1386,
+  "Translation fin-eng": 1387,
+  "Translation fin-fra": 1388,
+  "Translation fin-hrv": 1389,
+  "Translation fin-por": 1390,
+  "Translation fin-rus": 1391,
+  "Translation fin-slv": 1392,
+  "Translation fin-spa": 1393,
+  "Translation fin-srp_Cyrl": 1394,
+  "Translation fin-ukr": 1395,
+  "Translation fra-afr": 1396,
+  "Translation fra-ara": 1397,
+  "Translation fra-ast": 1398,
+  "Translation fra-bel": 1399,
+  "Translation fra-ben": 1400,
+  "Translation fra-bul": 1401,
+  "Translation fra-cat": 1402,
+  "Translation fra-ces": 1403,
+  "Translation fra-cym": 1404,
+  "Translation fra-dan": 1405,
+  "Translation fra-deu": 1406,
+  "Translation fra-ell": 1407,
+  "Translation fra-eng": 1408,
+  "Translation fra-est": 1409,
+  "Translation fra-fao": 1410,
+  "Translation fra-fas": 1411,
+  "Translation fra-fin": 1412,
+  "Translation fra-fra": 1413,
+  "Translation fra-fur": 1414,
+  "Translation fra-gle": 1415,
+  "Translation fra-glg": 1416,
+  "Translation fra-guj": 1417,
+  "Translation fra-hat": 1418,
+  "Translation fra-hau": 1419,
+  "Translation fra-hbs": 1420,
+  "Translation fra-heb": 1421,
+  "Translation fra-hin": 1422,
+  "Translation fra-hne": 1423,
+  "Translation fra-hrv": 1424,
+  "Translation fra-hun": 1425,
+  "Translation fra-isl": 1426,
+  "Translation fra-ita": 1427,
+  "Translation fra-kea": 1428,
+  "Translation fra-lav": 1429,
+  "Translation fra-lij": 1430,
+  "Translation fra-lin": 1431,
+  "Translation fra-lit": 1432,
+  "Translation fra-ltz": 1433,
+  "Translation fra-mag": 1434,
+  "Translation fra-mkd": 1435,
+  "Translation fra-mlt": 1436,
+  "Translation fra-nep": 1437,
+  "Translation fra-nld": 1438,
+  "Translation fra-nno": 1439,
+  "Translation fra-nob": 1440,
+  "Translation fra-nor": 1441,
+  "Translation fra-oci": 1442,
+  "Translation fra-pan": 1443,
+  "Translation fra-pap": 1444,
+  "Translation fra-pes": 1445,
+  "Translation fra-pol": 1446,
+  "Translation fra-por": 1447,
+  "Translation fra-prs": 1448,
+  "Translation fra-pus": 1449,
+  "Translation fra-ron": 1450,
+  "Translation fra-rus": 1451,
+  "Translation fra-slk": 1452,
+  "Translation fra-slv": 1453,
+  "Translation fra-spa": 1454,
+  "Translation fra-sqi": 1455,
+  "Translation fra-srd": 1456,
+  "Translation fra-srp_Cyrl": 1457,
+  "Translation fra-swa": 1458,
+  "Translation fra-swe": 1459,
+  "Translation fra-tgk": 1460,
+  "Translation fra-tpi": 1461,
+  "Translation fra-tsn": 1462,
+  "Translation fra-tur": 1463,
+  "Translation fra-ukr": 1464,
+  "Translation fra-urd": 1465,
+  "Translation fra-vie": 1466,
+  "Translation fry-deu": 1467,
+  "Translation fry-eng": 1468,
+  "Translation fry-ltz": 1469,
+  "Translation fry-nld": 1470,
+  "Translation fur-deu": 1471,
+  "Translation fur-eng": 1472,
+  "Translation fur-fra": 1473,
+  "Translation fur-por": 1474,
+  "Translation fur-spa": 1475,
+  "Translation gla-deu": 1476,
+  "Translation gla-eng": 1477,
+  "Translation gla-fra": 1478,
+  "Translation gla-por": 1479,
+  "Translation gla-spa": 1480,
+  "Translation gle-deu": 1481,
+  "Translation gle-eng": 1482,
+  "Translation gle-fra": 1483,
+  "Translation gle-por": 1484,
+  "Translation gle-spa": 1485,
+  "Translation glg-ara": 1486,
+  "Translation glg-ast": 1487,
+  "Translation glg-cat": 1488,
+  "Translation glg-deu": 1489,
+  "Translation glg-eng": 1490,
+  "Translation glg-fra": 1491,
+  "Translation glg-heb": 1492,
+  "Translation glg-ita": 1493,
+  "Translation glg-lav": 1494,
+  "Translation glg-lit": 1495,
+  "Translation glg-oci": 1496,
+  "Translation glg-por": 1497,
+  "Translation glg-ron": 1498,
+  "Translation glg-spa": 1499,
+  "Translation glg-tur": 1500,
+  "Translation gos-afr": 1501,
+  "Translation gos-deu": 1502,
+  "Translation gos-eng": 1503,
+  "Translation gos-fry": 1504,
+  "Translation gos-nld": 1505,
+  "Translation grn-eng": 1506,
+  "Translation grn-fra": 1507,
+  "Translation grn-por": 1508,
+  "Translation gsw-deu": 1509,
+  "Translation gsw-eng": 1510,
+  "Translation gsw-nld": 1511,
+  "Translation guj-deu": 1512,
+  "Translation guj-eng": 1513,
+  "Translation guj-fra": 1514,
+  "Translation guj-por": 1515,
+  "Translation guj-spa": 1516,
+  "Translation hat-deu": 1517,
+  "Translation hat-eng": 1518,
+  "Translation hat-fra": 1519,
+  "Translation hat-por": 1520,
+  "Translation hat-spa": 1521,
+  "Translation hau-eng": 1522,
+  "Translation hau-fra": 1523,
+  "Translation hau-por": 1524,
+  "Translation hau-spa": 1525,
+  "Translation hbs-deu": 1526,
+  "Translation hbs-eng": 1527,
+  "Translation hbs-fra": 1528,
+  "Translation hbs-ita": 1529,
+  "Translation hbs-rus": 1530,
+  "Translation hbs-spa": 1531,
+  "Translation hbs-ukr": 1532,
+  "Translation heb-cat": 1533,
+  "Translation heb-dan": 1534,
+  "Translation heb-deu": 1535,
+  "Translation heb-eng": 1536,
+  "Translation heb-fra": 1537,
+  "Translation heb-glg": 1538,
+  "Translation heb-isl": 1539,
+  "Translation heb-ita": 1540,
+  "Translation heb-nob": 1541,
+  "Translation heb-por": 1542,
+  "Translation heb-ron": 1543,
+  "Translation heb-spa": 1544,
+  "Translation heb-swe": 1545,
+  "Translation hin-deu": 1546,
+  "Translation hin-eng": 1547,
+  "Translation hin-fra": 1548,
+  "Translation hin-por": 1549,
+  "Translation hin-spa": 1550,
+  "Translation hne-deu": 1551,
+  "Translation hne-eng": 1552,
+  "Translation hne-fra": 1553,
+  "Translation hne-por": 1554,
+  "Translation hne-spa": 1555,
+  "Translation hrv-deu": 1556,
+  "Translation hrv-eng": 1557,
+  "Translation hrv-fra": 1558,
+  "Translation hrv-ita": 1559,
+  "Translation hrv-por": 1560,
+  "Translation hrv-ron": 1561,
+  "Translation hrv-rus": 1562,
+  "Translation hrv-spa": 1563,
+  "Translation hrv-ukr": 1564,
+  "Translation hrx-deu": 1565,
+  "Translation hrx-eng": 1566,
+  "Translation hsb-deu": 1567,
+  "Translation hun-deu": 1568,
+  "Translation hun-eng": 1569,
+  "Translation hun-fra": 1570,
+  "Translation hun-por": 1571,
+  "Translation hun-spa": 1572,
+  "Translation hun-ukr": 1573,
+  "Translation hye-deu": 1574,
+  "Translation hye-eng": 1575,
+  "Translation hye-fra": 1576,
+  "Translation hye-por": 1577,
+  "Translation hye-spa": 1578,
+  "Translation ibo-eng": 1579,
+  "Translation ibo-fra": 1580,
+  "Translation ibo-por": 1581,
+  "Translation ibo-spa": 1582,
+  "Translation ido_Latn-eng": 1583,
+  "Translation ilo-deu": 1584,
+  "Translation ilo-eng": 1585,
+  "Translation ilo-fra": 1586,
+  "Translation ilo-por": 1587,
+  "Translation ilo-spa": 1588,
+  "Translation ind-deu": 1589,
+  "Translation ind-eng": 1590,
+  "Translation ind-fra": 1591,
+  "Translation ind-por": 1592,
+  "Translation ind-spa": 1593,
+  "Translation isl-cat": 1594,
+  "Translation isl-ces": 1595,
+  "Translation isl-dan": 1596,
+  "Translation isl-deu": 1597,
+  "Translation isl-eng": 1598,
+  "Translation isl-fra": 1599,
+  "Translation isl-glg": 1600,
+  "Translation isl-heb": 1601,
+  "Translation isl-ita": 1602,
+  "Translation isl-nob": 1603,
+  "Translation isl-pol": 1604,
+  "Translation isl-por": 1605,
+  "Translation isl-ron": 1606,
+  "Translation isl-spa": 1607,
+  "Translation isl-swe": 1608,
+  "Translation ita-ara": 1609,
+  "Translation ita-ast": 1610,
+  "Translation ita-bel": 1611,
+  "Translation ita-cat": 1612,
+  "Translation ita-deu": 1613,
+  "Translation ita-eng": 1614,
+  "Translation ita-fra": 1615,
+  "Translation ita-glg": 1616,
+  "Translation ita-heb": 1617,
+  "Translation ita-lav": 1618,
+  "Translation ita-lit": 1619,
+  "Translation ita-oci": 1620,
+  "Translation ita-por": 1621,
+  "Translation ita-ron": 1622,
+  "Translation ita-rus": 1623,
+  "Translation ita-spa": 1624,
+  "Translation ita-tur": 1625,
+  "Translation ita-ukr": 1626,
+  "Translation jap-eng": 1627,
+  "Translation jav-deu": 1628,
+  "Translation jav-eng": 1629,
+  "Translation jav-fra": 1630,
+  "Translation jav-por": 1631,
+  "Translation jav-spa": 1632,
+  "Translation jpn-eng": 1633,
+  "Translation jpn-fra": 1634,
+  "Translation jpn-por": 1635,
+  "Translation jpn-spa": 1636,
+  "Translation kab-eng": 1637,
+  "Translation kab-spa": 1638,
+  "Translation kan-eng": 1639,
+  "Translation kat-eng": 1640,
+  "Translation kat-fra": 1641,
+  "Translation kat-por": 1642,
+  "Translation kat-spa": 1643,
+  "Translation kaz-deu": 1644,
+  "Translation kaz-eng": 1645,
+  "Translation kaz-fra": 1646,
+  "Translation kaz-por": 1647,
+  "Translation kaz-spa": 1648,
+  "Translation kaz_Cyrl-eng": 1649,
+  "Translation kea-deu": 1650,
+  "Translation kea-eng": 1651,
+  "Translation kea-fra": 1652,
+  "Translation kea-por": 1653,
+  "Translation kea-spa": 1654,
+  "Translation kik-eng": 1655,
+  "Translation kik-fra": 1656,
+  "Translation kin-eng": 1657,
+  "Translation kin-fra": 1658,
+  "Translation kin-por": 1659,
+  "Translation kin-spa": 1660,
+  "Translation kmr-eng": 1661,
+  "Translation kmr-fra": 1662,
+  "Translation kmr-por": 1663,
+  "Translation kmr-spa": 1664,
+  "Translation kon-eng": 1665,
+  "Translation kon-fra": 1666,
+  "Translation kon-por": 1667,
+  "Translation kor-eng": 1668,
+  "Translation kur_Latn-deu": 1669,
+  "Translation kur_Latn-eng": 1670,
+  "Translation lad-eng": 1671,
+  "Translation lad-spa": 1672,
+  "Translation lad_Latn-eng": 1673,
+  "Translation lad_Latn-spa": 1674,
+  "Translation lat-deu": 1675,
+  "Translation lat-eng": 1676,
+  "Translation lat-spa": 1677,
+  "Translation lav-deu": 1678,
+  "Translation lav-eng": 1679,
+  "Translation lav-fra": 1680,
+  "Translation lav-por": 1681,
+  "Translation lav-rus": 1682,
+  "Translation lav-spa": 1683,
+  "Translation lfn_Latn-deu": 1684,
+  "Translation lfn_Latn-eng": 1685,
+  "Translation lfn_Latn-fra": 1686,
+  "Translation lfn_Latn-por": 1687,
+  "Translation lij-deu": 1688,
+  "Translation lij-eng": 1689,
+  "Translation lij-fra": 1690,
+  "Translation lij-por": 1691,
+  "Translation lij-spa": 1692,
+  "Translation lim-deu": 1693,
+  "Translation lim-eng": 1694,
+  "Translation lim-fra": 1695,
+  "Translation lim-nld": 1696,
+  "Translation lim-por": 1697,
+  "Translation lim-spa": 1698,
+  "Translation lin-eng": 1699,
+  "Translation lin-fra": 1700,
+  "Translation lin-por": 1701,
+  "Translation lin-spa": 1702,
+  "Translation lit-deu": 1703,
+  "Translation lit-eng": 1704,
+  "Translation lit-fra": 1705,
+  "Translation lit-por": 1706,
+  "Translation lit-rus": 1707,
+  "Translation lit-spa": 1708,
+  "Translation lmo-deu": 1709,
+  "Translation lmo-eng": 1710,
+  "Translation lmo-fra": 1711,
+  "Translation lmo-por": 1712,
+  "Translation lmo-spa": 1713,
+  "Translation ltz-deu": 1714,
+  "Translation ltz-eng": 1715,
+  "Translation ltz-fra": 1716,
+  "Translation ltz-fry": 1717,
+  "Translation ltz-nld": 1718,
+  "Translation ltz-por": 1719,
+  "Translation ltz-spa": 1720,
+  "Translation lug-eng": 1721,
+  "Translation lug-fra": 1722,
+  "Translation lug-por": 1723,
+  "Translation lug-spa": 1724,
+  "Translation mag-deu": 1725,
+  "Translation mag-eng": 1726,
+  "Translation mag-fra": 1727,
+  "Translation mag-por": 1728,
+  "Translation mag-spa": 1729,
+  "Translation mai-deu": 1730,
+  "Translation mai-eng": 1731,
+  "Translation mai-fra": 1732,
+  "Translation mai-por": 1733,
+  "Translation mai-spa": 1734,
+  "Translation mal-eng": 1735,
+  "Translation mal-fra": 1736,
+  "Translation mar-deu": 1737,
+  "Translation mar-eng": 1738,
+  "Translation mar-fra": 1739,
+  "Translation mar-por": 1740,
+  "Translation mar-spa": 1741,
+  "Translation mkd-deu": 1742,
+  "Translation mkd-eng": 1743,
+  "Translation mkd-fra": 1744,
+  "Translation mkd-ita": 1745,
+  "Translation mkd-por": 1746,
+  "Translation mkd-ron": 1747,
+  "Translation mkd-rus": 1748,
+  "Translation mkd-spa": 1749,
+  "Translation mkd-ukr": 1750,
+  "Translation mlg-eng": 1751,
+  "Translation mlg-fra": 1752,
+  "Translation mlg-por": 1753,
+  "Translation mlg-spa": 1754,
+  "Translation mlt-deu": 1755,
+  "Translation mlt-eng": 1756,
+  "Translation mlt-fra": 1757,
+  "Translation mlt-por": 1758,
+  "Translation mlt-spa": 1759,
+  "Translation mri-eng": 1760,
+  "Translation mri-fra": 1761,
+  "Translation mri-spa": 1762,
+  "Translation msa-deu": 1763,
+  "Translation msa-eng": 1764,
+  "Translation msa-fra": 1765,
+  "Translation msa-por": 1766,
+  "Translation multi-eng": 1767,
+  "Translation multi-fra": 1768,
+  "Translation multi-multi": 1769,
+  "Translation nde-eng": 1770,
+  "Translation nde-fra": 1771,
+  "Translation nde-por": 1772,
+  "Translation nde-spa": 1773,
+  "Translation nds-deu": 1774,
+  "Translation nds-eng": 1775,
+  "Translation nds-fra": 1776,
+  "Translation nds-nld": 1777,
+  "Translation nds-por": 1778,
+  "Translation nds-spa": 1779,
+  "Translation nep-deu": 1780,
+  "Translation nep-eng": 1781,
+  "Translation nep-fra": 1782,
+  "Translation nep-por": 1783,
+  "Translation nep-spa": 1784,
+  "Translation nld-afr": 1785,
+  "Translation nld-deu": 1786,
+  "Translation nld-eng": 1787,
+  "Translation nld-fra": 1788,
+  "Translation nld-fry": 1789,
+  "Translation nld-nds": 1790,
+  "Translation nld-nld": 1791,
+  "Translation nld-por": 1792,
+  "Translation nld-sco": 1793,
+  "Translation nld-spa": 1794,
+  "Translation nno-deu": 1795,
+  "Translation nno-eng": 1796,
+  "Translation nno-fra": 1797,
+  "Translation nno-nob": 1798,
+  "Translation nno-por": 1799,
+  "Translation nno-spa": 1800,
+  "Translation nob-ara": 1801,
+  "Translation nob-cat": 1802,
+  "Translation nob-ces": 1803,
+  "Translation nob-dan": 1804,
+  "Translation nob-deu": 1805,
+  "Translation nob-eng": 1806,
+  "Translation nob-fra": 1807,
+  "Translation nob-glg": 1808,
+  "Translation nob-heb": 1809,
+  "Translation nob-isl": 1810,
+  "Translation nob-ita": 1811,
+  "Translation nob-nno": 1812,
+  "Translation nob-pol": 1813,
+  "Translation nob-por": 1814,
+  "Translation nob-ron": 1815,
+  "Translation nob-rus": 1816,
+  "Translation nob-spa": 1817,
+  "Translation nob-swe": 1818,
+  "Translation nob-tur": 1819,
+  "Translation nob-ukr": 1820,
+  "Translation nor-deu": 1821,
+  "Translation nor-eng": 1822,
+  "Translation nor-fra": 1823,
+  "Translation nor-por": 1824,
+  "Translation nor-spa": 1825,
+  "Translation npi-deu": 1826,
+  "Translation npi-eng": 1827,
+  "Translation npi-fra": 1828,
+  "Translation npi-por": 1829,
+  "Translation npi-spa": 1830,
+  "Translation nso-deu": 1831,
+  "Translation nso-eng": 1832,
+  "Translation nso-fra": 1833,
+  "Translation nso-por": 1834,
+  "Translation nso-spa": 1835,
+  "Translation nya-deu": 1836,
+  "Translation nya-eng": 1837,
+  "Translation nya-fra": 1838,
+  "Translation nya-por": 1839,
+  "Translation nya-spa": 1840,
+  "Translation oci-ast": 1841,
+  "Translation oci-cat": 1842,
+  "Translation oci-deu": 1843,
+  "Translation oci-eng": 1844,
+  "Translation oci-fra": 1845,
+  "Translation oci-glg": 1846,
+  "Translation oci-ita": 1847,
+  "Translation oci-por": 1848,
+  "Translation oci-ron": 1849,
+  "Translation oci-spa": 1850,
+  "Translation oci-tur": 1851,
+  "Translation ofs-bar": 1852,
+  "Translation pag-fra": 1853,
+  "Translation pag-por": 1854,
+  "Translation pag-spa": 1855,
+  "Translation pan-deu": 1856,
+  "Translation pan-eng": 1857,
+  "Translation pan-fra": 1858,
+  "Translation pan-por": 1859,
+  "Translation pan-spa": 1860,
+  "Translation pap-deu": 1861,
+  "Translation pap-eng": 1862,
+  "Translation pap-fra": 1863,
+  "Translation pap-por": 1864,
+  "Translation pap-spa": 1865,
+  "Translation pdc-deu": 1866,
+  "Translation pdc-eng": 1867,
+  "Translation pes-deu": 1868,
+  "Translation pes-eng": 1869,
+  "Translation pes-fra": 1870,
+  "Translation pes-por": 1871,
+  "Translation pes-spa": 1872,
+  "Translation plt-eng": 1873,
+  "Translation plt-fra": 1874,
+  "Translation plt-por": 1875,
+  "Translation plt-spa": 1876,
+  "Translation pms-eng": 1877,
+  "Translation pms-ita": 1878,
+  "Translation pol-bel": 1879,
+  "Translation pol-deu": 1880,
+  "Translation pol-eng": 1881,
+  "Translation pol-fra": 1882,
+  "Translation pol-por": 1883,
+  "Translation pol-rus": 1884,
+  "Translation pol-spa": 1885,
+  "Translation pol-ukr": 1886,
+  "Translation por-afr": 1887,
+  "Translation por-ara": 1888,
+  "Translation por-ast": 1889,
+  "Translation por-bel": 1890,
+  "Translation por-ben": 1891,
+  "Translation por-bul": 1892,
+  "Translation por-cat": 1893,
+  "Translation por-ces": 1894,
+  "Translation por-cym": 1895,
+  "Translation por-dan": 1896,
+  "Translation por-deu": 1897,
+  "Translation por-ell": 1898,
+  "Translation por-eng": 1899,
+  "Translation por-est": 1900,
+  "Translation por-fao": 1901,
+  "Translation por-fas": 1902,
+  "Translation por-fin": 1903,
+  "Translation por-fra": 1904,
+  "Translation por-fur": 1905,
+  "Translation por-gle": 1906,
+  "Translation por-glg": 1907,
+  "Translation por-guj": 1908,
+  "Translation por-hat": 1909,
+  "Translation por-hau": 1910,
+  "Translation por-heb": 1911,
+  "Translation por-hin": 1912,
+  "Translation por-hne": 1913,
+  "Translation por-hrv": 1914,
+  "Translation por-hun": 1915,
+  "Translation por-isl": 1916,
+  "Translation por-ita": 1917,
+  "Translation por-kea": 1918,
+  "Translation por-lav": 1919,
+  "Translation por-lij": 1920,
+  "Translation por-lin": 1921,
+  "Translation por-lit": 1922,
+  "Translation por-ltz": 1923,
+  "Translation por-mag": 1924,
+  "Translation por-mkd": 1925,
+  "Translation por-mlt": 1926,
+  "Translation por-nds": 1927,
+  "Translation por-nep": 1928,
+  "Translation por-nld": 1929,
+  "Translation por-nno": 1930,
+  "Translation por-nob": 1931,
+  "Translation por-nor": 1932,
+  "Translation por-oci": 1933,
+  "Translation por-pan": 1934,
+  "Translation por-pap": 1935,
+  "Translation por-pes": 1936,
+  "Translation por-pol": 1937,
+  "Translation por-por": 1938,
+  "Translation por-prs": 1939,
+  "Translation por-pus": 1940,
+  "Translation por-ron": 1941,
+  "Translation por-rus": 1942,
+  "Translation por-slk": 1943,
+  "Translation por-slv": 1944,
+  "Translation por-spa": 1945,
+  "Translation por-sqi": 1946,
+  "Translation por-srd": 1947,
+  "Translation por-srp_Cyrl": 1948,
+  "Translation por-swa": 1949,
+  "Translation por-swe": 1950,
+  "Translation por-tgk": 1951,
+  "Translation por-tpi": 1952,
+  "Translation por-tsn": 1953,
+  "Translation por-tur": 1954,
+  "Translation por-ukr": 1955,
+  "Translation por-urd": 1956,
+  "Translation por-vie": 1957,
+  "Translation prs-deu": 1958,
+  "Translation prs-eng": 1959,
+  "Translation prs-fra": 1960,
+  "Translation prs-por": 1961,
+  "Translation prs-spa": 1962,
+  "Translation pus-deu": 1963,
+  "Translation pus-eng": 1964,
+  "Translation pus-fra": 1965,
+  "Translation pus-por": 1966,
+  "Translation pus-spa": 1967,
+  "Translation ron-ara": 1968,
+  "Translation ron-ast": 1969,
+  "Translation ron-cat": 1970,
+  "Translation ron-deu": 1971,
+  "Translation ron-eng": 1972,
+  "Translation ron-fra": 1973,
+  "Translation ron-glg": 1974,
+  "Translation ron-heb": 1975,
+  "Translation ron-ita": 1976,
+  "Translation ron-oci": 1977,
+  "Translation ron-por": 1978,
+  "Translation ron-spa": 1979,
+  "Translation ron-tur": 1980,
+  "Translation ron-ukr": 1981,
+  "Translation ru-en": 1982,
+  "Translation run-deu": 1983,
+  "Translation run-eng": 1984,
+  "Translation run-fra": 1985,
+  "Translation run-por": 1986,
+  "Translation run-spa": 1987,
+  "Translation rus-ast": 1988,
+  "Translation rus-bel": 1989,
+  "Translation rus-bul": 1990,
+  "Translation rus-cat": 1991,
+  "Translation rus-ces": 1992,
+  "Translation rus-dan": 1993,
+  "Translation rus-deu": 1994,
+  "Translation rus-eng": 1995,
+  "Translation rus-fin": 1996,
+  "Translation rus-fra": 1997,
+  "Translation rus-glg": 1998,
+  "Translation rus-hbs": 1999,
+  "Translation rus-hrv": 2000,
+  "Translation rus-ita": 2001,
+  "Translation rus-lav": 2002,
+  "Translation rus-lit": 2003,
+  "Translation rus-mkd": 2004,
+  "Translation rus-nob": 2005,
+  "Translation rus-oci": 2006,
+  "Translation rus-pol": 2007,
+  "Translation rus-por": 2008,
+  "Translation rus-ron": 2009,
+  "Translation rus-slv": 2010,
+  "Translation rus-spa": 2011,
+  "Translation rus-srp_Cyrl": 2012,
+  "Translation rus-srp_Latn": 2013,
+  "Translation rus-swe": 2014,
+  "Translation rus-ukr": 2015,
+  "Translation san-eng": 2016,
+  "Translation scn-deu": 2017,
+  "Translation scn-eng": 2018,
+  "Translation scn-fra": 2019,
+  "Translation scn-por": 2020,
+  "Translation scn-spa": 2021,
+  "Translation sco-eng": 2022,
+  "Translation sco-nld": 2023,
+  "Translation sin-deu": 2024,
+  "Translation sin-eng": 2025,
+  "Translation sin-fra": 2026,
+  "Translation sin-por": 2027,
+  "Translation sin-spa": 2028,
+  "Translation slk-deu": 2029,
+  "Translation slk-eng": 2030,
+  "Translation slk-fra": 2031,
+  "Translation slk-por": 2032,
+  "Translation slk-spa": 2033,
+  "Translation slk-ukr": 2034,
+  "Translation slv-deu": 2035,
+  "Translation slv-eng": 2036,
+  "Translation slv-fra": 2037,
+  "Translation slv-ita": 2038,
+  "Translation slv-por": 2039,
+  "Translation slv-ron": 2040,
+  "Translation slv-rus": 2041,
+  "Translation slv-spa": 2042,
+  "Translation slv-ukr": 2043,
+  "Translation smp-sam": 2044,
+  "Translation sna-eng": 2045,
+  "Translation sna-fra": 2046,
+  "Translation sna-por": 2047,
+  "Translation sna-spa": 2048,
+  "Translation som-deu": 2049,
+  "Translation som-eng": 2050,
+  "Translation som-fra": 2051,
+  "Translation som-por": 2052,
+  "Translation som-spa": 2053,
+  "Translation sot-deu": 2054,
+  "Translation sot-eng": 2055,
+  "Translation sot-fra": 2056,
+  "Translation sot-por": 2057,
+  "Translation sot-spa": 2058,
+  "Translation spa-afr": 2059,
+  "Translation spa-ara": 2060,
+  "Translation spa-ast": 2061,
+  "Translation spa-bel": 2062,
+  "Translation spa-ben": 2063,
+  "Translation spa-bul": 2064,
+  "Translation spa-cat": 2065,
+  "Translation spa-ces": 2066,
+  "Translation spa-cym": 2067,
+  "Translation spa-dan": 2068,
+  "Translation spa-deu": 2069,
+  "Translation spa-ell": 2070,
+  "Translation spa-eng": 2071,
+  "Translation spa-est": 2072,
+  "Translation spa-eus": 2073,
+  "Translation spa-fao": 2074,
+  "Translation spa-fas": 2075,
+  "Translation spa-fin": 2076,
+  "Translation spa-fra": 2077,
+  "Translation spa-fur": 2078,
+  "Translation spa-gla": 2079,
+  "Translation spa-gle": 2080,
+  "Translation spa-glg": 2081,
+  "Translation spa-hat": 2082,
+  "Translation spa-hau": 2083,
+  "Translation spa-hbs": 2084,
+  "Translation spa-heb": 2085,
+  "Translation spa-hin": 2086,
+  "Translation spa-hne": 2087,
+  "Translation spa-hrv": 2088,
+  "Translation spa-hun": 2089,
+  "Translation spa-isl": 2090,
+  "Translation spa-ita": 2091,
+  "Translation spa-lad": 2092,
+  "Translation spa-lad_Latn": 2093,
+  "Translation spa-lav": 2094,
+  "Translation spa-lij": 2095,
+  "Translation spa-lin": 2096,
+  "Translation spa-lit": 2097,
+  "Translation spa-mag": 2098,
+  "Translation spa-mar": 2099,
+  "Translation spa-mkd": 2100,
+  "Translation spa-mlt": 2101,
+  "Translation spa-nep": 2102,
+  "Translation spa-nld": 2103,
+  "Translation spa-nno": 2104,
+  "Translation spa-nob": 2105,
+  "Translation spa-nor": 2106,
+  "Translation spa-oci": 2107,
+  "Translation spa-pan": 2108,
+  "Translation spa-pap": 2109,
+  "Translation spa-pes": 2110,
+  "Translation spa-pol": 2111,
+  "Translation spa-por": 2112,
+  "Translation spa-prs": 2113,
+  "Translation spa-pus": 2114,
+  "Translation spa-ron": 2115,
+  "Translation spa-rus": 2116,
+  "Translation spa-slk": 2117,
+  "Translation spa-slv": 2118,
+  "Translation spa-spa": 2119,
+  "Translation spa-sqi": 2120,
+  "Translation spa-srd": 2121,
+  "Translation spa-srp_Cyrl": 2122,
+  "Translation spa-swa": 2123,
+  "Translation spa-swe": 2124,
+  "Translation spa-tgk": 2125,
+  "Translation spa-tpi": 2126,
+  "Translation spa-tsn": 2127,
+  "Translation spa-tur": 2128,
+  "Translation spa-ukr": 2129,
+  "Translation spa-urd": 2130,
+  "Translation spa-vie": 2131,
+  "Translation sqi-deu": 2132,
+  "Translation sqi-eng": 2133,
+  "Translation sqi-fra": 2134,
+  "Translation sqi-por": 2135,
+  "Translation sqi-spa": 2136,
+  "Translation srd-deu": 2137,
+  "Translation srd-eng": 2138,
+  "Translation srd-fra": 2139,
+  "Translation srd-por": 2140,
+  "Translation srd-spa": 2141,
+  "Translation srn-eng": 2142,
+  "Translation srp_Cyrl-deu": 2143,
+  "Translation srp_Cyrl-eng": 2144,
+  "Translation srp_Cyrl-fra": 2145,
+  "Translation srp_Cyrl-ita": 2146,
+  "Translation srp_Cyrl-por": 2147,
+  "Translation srp_Cyrl-ron": 2148,
+  "Translation srp_Cyrl-rus": 2149,
+  "Translation srp_Cyrl-spa": 2150,
+  "Translation srp_Cyrl-ukr": 2151,
+  "Translation srp_Latn-deu": 2152,
+  "Translation srp_Latn-eng": 2153,
+  "Translation srp_Latn-ita": 2154,
+  "Translation srp_Latn-rus": 2155,
+  "Translation srp_Latn-ukr": 2156,
+  "Translation ssw-eng": 2157,
+  "Translation ssw-fra": 2158,
+  "Translation ssw-por": 2159,
+  "Translation ssw-spa": 2160,
+  "Translation stq-deu": 2161,
+  "Translation stq-eng": 2162,
+  "Translation stq-nld": 2163,
+  "Translation swa-deu": 2164,
+  "Translation swa-eng": 2165,
+  "Translation swa-fra": 2166,
+  "Translation swa-por": 2167,
+  "Translation swa-spa": 2168,
+  "Translation swe-ara": 2169,
+  "Translation swe-cat": 2170,
+  "Translation swe-ces": 2171,
+  "Translation swe-dan": 2172,
+  "Translation swe-deu": 2173,
+  "Translation swe-eng": 2174,
+  "Translation swe-fra": 2175,
+  "Translation swe-glg": 2176,
+  "Translation swe-heb": 2177,
+  "Translation swe-isl": 2178,
+  "Translation swe-ita": 2179,
+  "Translation swe-nob": 2180,
+  "Translation swe-pol": 2181,
+  "Translation swe-por": 2182,
+  "Translation swe-ron": 2183,
+  "Translation swe-rus": 2184,
+  "Translation swe-spa": 2185,
+  "Translation swe-tur": 2186,
+  "Translation swe-ukr": 2187,
+  "Translation swg-eng": 2188,
+  "Translation swg-nld": 2189,
+  "Translation swh-deu": 2190,
+  "Translation swh-eng": 2191,
+  "Translation swh-fra": 2192,
+  "Translation swh-por": 2193,
+  "Translation swh-spa": 2194,
+  "Translation szl-deu": 2195,
+  "Translation szl-eng": 2196,
+  "Translation szl-fra": 2197,
+  "Translation szl-por": 2198,
+  "Translation szl-spa": 2199,
+  "Translation tgk-deu": 2200,
+  "Translation tgk-eng": 2201,
+  "Translation tgk-fra": 2202,
+  "Translation tgk-por": 2203,
+  "Translation tgk-spa": 2204,
+  "Translation tgk_Cyrl-deu": 2205,
+  "Translation tgk_Cyrl-eng": 2206,
+  "Translation tgk_Cyrl-fra": 2207,
+  "Translation tgk_Cyrl-por": 2208,
+  "Translation tgk_Cyrl-spa": 2209,
+  "Translation tha-eng": 2210,
+  "Translation tir-eng": 2211,
+  "Translation tir-spa": 2212,
+  "Translation tpi-deu": 2213,
+  "Translation tpi-eng": 2214,
+  "Translation tpi-fra": 2215,
+  "Translation tpi-por": 2216,
+  "Translation tpi-spa": 2217,
+  "Translation tsn-deu": 2218,
+  "Translation tsn-eng": 2219,
+  "Translation tsn-fra": 2220,
+  "Translation tsn-por": 2221,
+  "Translation tsn-spa": 2222,
+  "Translation tso-eng": 2223,
+  "Translation tso-fra": 2224,
+  "Translation tso-por": 2225,
+  "Translation tur-eng": 2226,
+  "Translation tur-ukr": 2227,
+  "Translation ukr-ast": 2228,
+  "Translation ukr-bel": 2229,
+  "Translation ukr-bul": 2230,
+  "Translation ukr-cat": 2231,
+  "Translation ukr-ces": 2232,
+  "Translation ukr-dan": 2233,
+  "Translation ukr-deu": 2234,
+  "Translation ukr-eng": 2235,
+  "Translation ukr-fin": 2236,
+  "Translation ukr-fra": 2237,
+  "Translation ukr-glg": 2238,
+  "Translation ukr-hbs": 2239,
+  "Translation ukr-hrv": 2240,
+  "Translation ukr-hun": 2241,
+  "Translation ukr-ita": 2242,
+  "Translation ukr-lav": 2243,
+  "Translation ukr-lit": 2244,
+  "Translation ukr-mkd": 2245,
+  "Translation ukr-nob": 2246,
+  "Translation ukr-oci": 2247,
+  "Translation ukr-pol": 2248,
+  "Translation ukr-por": 2249,
+  "Translation ukr-ron": 2250,
+  "Translation ukr-rus": 2251,
+  "Translation ukr-slk": 2252,
+  "Translation ukr-slv": 2253,
+  "Translation ukr-spa": 2254,
+  "Translation ukr-srp_Cyrl": 2255,
+  "Translation ukr-srp_Latn": 2256,
+  "Translation ukr-swe": 2257,
+  "Translation ukr-tur": 2258,
+  "Translation urd-deu": 2259,
+  "Translation urd-eng": 2260,
+  "Translation urd-fra": 2261,
+  "Translation urd-por": 2262,
+  "Translation urd-spa": 2263,
+  "Translation vec-deu": 2264,
+  "Translation vec-eng": 2265,
+  "Translation vec-fra": 2266,
+  "Translation vec-por": 2267,
+  "Translation vec-spa": 2268,
+  "Translation ven-eng": 2269,
+  "Translation ven-fra": 2270,
+  "Translation ven-por": 2271,
+  "Translation ven-spa": 2272,
+  "Translation vie-eng": 2273,
+  "Translation xho-deu": 2274,
+  "Translation xho-eng": 2275,
+  "Translation xho-fra": 2276,
+  "Translation xho-por": 2277,
+  "Translation xho-spa": 2278,
+  "Translation yid-eng": 2279,
+  "Translation yid-fra": 2280,
+  "Translation yid-spa": 2281,
+  "Translation yor-eng": 2282,
+  "Translation zea-deu": 2283,
+  "Translation zea-eng": 2284,
+  "Translation zea-fry": 2285,
+  "Translation zea-nds": 2286,
+  "Translation zea-nld": 2287,
+  "Translation zho-eng": 2288,
+  "Translation zho-jpn": 2289,
+  "Translation zul-deu": 2290,
+  "Translation zul-eng": 2291,
+  "Translation zul-fra": 2292,
+  "Translation zul-por": 2293,
+  "Translation zul-spa": 2294,
+  "Triplet": 2295,
+  "TriviaQA": 2296,
+  "TruthfulQA": 2297,
+  "TruthfulQA (MC2)": 2298,
+  "TruthfulQA Generation": 2299,
+  "Truthfulness": 2300,
+  "Truthfulness in answers": 2301,
+  "Truthfulness in question answering": 2302,
+  "Turn Detection": 2303,
+  "Type prediction": 2304,
+  "UFD": 2305,
+  "UI Element Detection": 2306,
+  "UNLABELED_DEPENDENCIES": 2307,
+  "Uncensored Response": 2308,
+  "Unsupervised Domain Adaptation": 2309,
+  "Unsupervised Instance Segmentation": 2310,
+  "Unsupervised Object Segmentation": 2311,
+  "Unsupervised Semantic Segmentation": 2312,
+  "Urdu Speech Recognition": 2313,
+  "User Feedback Classification": 2314,
+  "Uzbek Language Understanding": 2315,
+  "VCGBench-Diverse": 2316,
+  "VLA": 2317,
+  "VQAv2": 2318,
+  "VSI-Bench": 2319,
+  "Vehicle Re-Identification": 2320,
+  "Verbalized Rebus Solving": 2321,
+  "Video Captioning": 2322,
+  "Video Classification": 2323,
+  "Video Crime Detection": 2324,
+  "Video Frame Interpolation": 2325,
+  "Video Generation": 2326,
+  "Video Grounding": 2327,
+  "Video Instance Segmentation": 2328,
+  "Video Object Segmentation": 2329,
+  "Video Prediction": 2330,
+  "Video Question Answering": 2331,
+  "Video Reconstruction": 2332,
+  "Video Retrieval": 2333,
+  "Video Summarization": 2334,
+  "Video Super-Resolution": 2335,
+  "Video-based Generative Performance Benchmarking": 2336,
+  "Video-based Generative Performance Benchmarking (Correctness of Information)": 2337,
+  "VideoMME": 2338,
+  "VideoMMMU": 2339,
+  "Vietnamese Banking Aspect Sentiment Analysis": 2340,
+  "Vietnamese Banking Text Classification": 2341,
+  "Vietnamese General Sentiment Analysis": 2342,
+  "Vietnamese Medical Abstractive Question Answering": 2343,
+  "Vietnamese Natural Language Inference": 2344,
+  "Vietnamese News Classification": 2345,
+  "VilaQuAD": 2346,
+  "Violence Detection": 2347,
+  "ViquiQuAD": 2348,
+  "Vision-Language-Action Navigation": 2349,
+  "Vision-and-Language Navigation": 2350,
+  "Vision-based Classification": 2351,
+  "Visual Object Tracking": 2352,
+  "Visual Place Recognition": 2353,
+  "Visual Prompt Tuning": 2354,
+  "Visual Question Answering": 2355,
+  "Visual Question Answering (VQA)": 2356,
+  "Visual Reasoning": 2357,
+  "Visual Servoing": 2358,
+  "Visual Storytelling": 2359,
+  "Visual Tracking": 2360,
+  "Visual math reasoning": 2361,
+  "Visual question answering": 2362,
+  "Visual scientific knowledge reasoning": 2363,
+  "Voice Activity Detection": 2364,
+  "Voice Conversion": 2365,
+  "Voice Emotion Recognition": 2366,
+  "Waste Classification": 2367,
+  "WideSearch": 2368,
+  "Wikipedia Summarization": 2369,
+  "Wikitext-fr": 2370,
+  "WinoG": 2371,
+  "WinoGrande": 2372,
+  "Winogrande": 2373,
+  "Winogrande Challenge": 2374,
+  "Word Sense Disambiguation": 2375,
+  "Word Similarity": 2376,
+  "Word prediction": 2377,
+  "XQuAD-ca": 2378,
+  "Yes/No Question Classification": 2379,
+  "Zero Shot Classification": 2380,
+  "Zero Shot Classifications": 2381,
+  "Zero Shot Segmentation": 2382,
+  "Zero shot Classification": 2383,
+  "Zero-Shot Action Recognition": 2384,
+  "Zero-Shot Baseline": 2385,
+  "Zero-Shot Classification": 2386,
+  "Zero-Shot Emergence Detection": 2387,
+  "Zero-Shot Text Classification": 2388,
+  "Zero-Shot Transfer Image Classification": 2389,
+  "Zero-Shot Video Retrieval": 2390,
+  "Zero-shot": 2391,
+  "Zero-shot (binary)": 2392,
+  "Zero-shot Classification": 2393,
+  "Zero-shot Generalization": 2394,
+  "Zero-shot Sentiment Classification": 2395,
+  "abstractive summarization": 2396,
+  "agieval": 2397,
+  "answerability prediction": 2398,
+  "any-to-any": 2399,
+  "arc_ca_challenge": 2400,
+  "arc_ca_easy": 2401,
+  "arc_easy": 2402,
+  "audio classification": 2403,
+  "audio-classification": 2404,
+  "audio-text-retrieval": 2405,
+  "automatic-speech-recognition": 2406,
+  "automatic-speech-translation": 2407,
+  "binary-classification": 2408,
+  "binary_classification": 2409,
+  "catalanqa": 2410,
+  "chinese-evaluation": 2411,
+  "chunking": 2412,
+  "classification": 2413,
+  "classify nepali news": 2414,
+  "clustering": 2415,
+  "code": 2416,
+  "code generation": 2417,
+  "code-evaluation": 2418,
+  "code-generation": 2419,
+  "commonsense-reasoning": 2420,
+  "copa_ca": 2421,
+  "coreference-resolution": 2422,
+  "defect-detection": 2423,
+  "diamond": 2424,
+  "document-image-classification": 2425,
+  "entity-linking": 2426,
+  "eq_bench": 2427,
+  "evaluation": 2428,
+  "exam": 2429,
+  "fact-verification": 2430,
+  "feature-extraction": 2431,
+  "few-shot": 2432,
+  "few-shot-ner": 2433,
+  "fill-mask": 2434,
+  "flores_ca": 2435,
+  "formal language correction": 2436,
+  "get-answer": 2437,
+  "gsgsm8k": 2438,
+  "gsm8k": 2439,
+  "haerae": 2440,
+  "humaneval": 2441,
+  "image-captioning": 2442,
+  "image-classification": 2443,
+  "image-segmentation": 2444,
+  "image-similarity": 2445,
+  "image-text-retrieval": 2446,
+  "image-text-to-text": 2447,
+  "image-to-image": 2448,
+  "image-to-text": 2449,
+  "information-retrieval": 2450,
+  "instance-segmentation": 2451,
+  "instruction": 2452,
+  "intent classification": 2453,
+  "intent-classification": 2454,
+  "kmmlu": 2455,
+  "knowledge": 2456,
+  "low-light-image-enhancement": 2457,
+  "math": 2458,
+  "math-evaluation": 2459,
+  "mathematical-reasoning": 2460,
+  "mbpp": 2461,
+  "mix": 2462,
+  "mmlu": 2463,
+  "multi-label text-classification": 2464,
+  "multi-label-classification": 2465,
+  "multi-task-evaluation": 2466,
+  "multi_class_classification": 2467,
+  "multi_label_classification": 2468,
+  "multimodal": 2469,
+  "multiple-choice": 2470,
+  "multiple-choice-qa": 2471,
+  "multiple-choice-question-answering": 2472,
+  "multiple_choice": 2473,
+  "named-entity-recognition": 2474,
+  "narratives": 2475,
+  "natural-language-inference": 2476,
+  "ner": 2477,
+  "object-classification": 2478,
+  "object-detection": 2479,
+  "original-capability": 2480,
+  "phoneme-classification": 2481,
+  "preference_evaluation": 2482,
+  "pretraining-evaluation": 2483,
+  "question-answering": 2484,
+  "reasoning": 2485,
+  "regression": 2486,
+  "reinforcement-learning": 2487,
+  "reinforcement-learning for quadrangular mesh topological optimization": 2488,
+  "retrieval": 2489,
+  "robotics": 2490,
+  "semantic textual similarity": 2491,
+  "semantic-segmentation": 2492,
+  "semantic-similarity": 2493,
+  "sentence-similarity": 2494,
+  "sentiment analysis": 2495,
+  "sentiment-analysis": 2496,
+  "sentiment-classification": 2497,
+  "sequence-classification": 2498,
+  "slot-filling": 2499,
+  "speech-recognition": 2500,
+  "speech-to-text": 2501,
+  "speech-translation": 2502,
+  "stem": 2503,
+  "streaming-transcription-chunk-100msec": 2504,
+  "streaming-transcription-chunk-200msec": 2505,
+  "streaming-transcription-chunk-300msec": 2506,
+  "streaming-transcription-chunk-40msec": 2507,
+  "structured sentiment analysis": 2508,
+  "structured-data-classification": 2509,
+  "structured-information-extraction": 2510,
+  "summarization": 2511,
+  "symbolic music representation learning": 2512,
+  "tabular-classification": 2513,
+  "tabular-regression": 2514,
+  "tau2-bench": 2515,
+  "text generation": 2516,
+  "text political leaning classification": 2517,
+  "text-classfication": 2518,
+  "text-classification": 2519,
+  "text-generation": 2520,
+  "text-prediction": 2521,
+  "text-ranking": 2522,
+  "text-summarization": 2523,
+  "text-to-audio": 2524,
+  "text-to-image": 2525,
+  "text-to-speech": 2526,
+  "text-to-sql": 2527,
+  "text_classification": 2528,
+  "token-classification": 2529,
+  "tomato leaf disease detection": 2530,
+  "translation": 2531,
+  "translation en-me": 2532,
+  "translation, speech-translation": 2533,
+  "truthfulqa": 2534,
+  "truthfulqa_gen": 2535,
+  "video caption": 2536,
+  "video detailed caption": 2537,
+  "video question anwering": 2538,
+  "video-captioning": 2539,
+  "video-classification": 2540,
+  "video-text-to-text": 2541,
+  "visual-question-answering": 2542,
+  "voice-conversion": 2543,
+  "winogrande": 2544,
+  "word-similarity": 2545,
+  "zero-shot retrieval": 2546,
+  "zero-shot-classification": 2547,
+  "zero-shot-image-classification": 2548,
+  "ΔWP regression (go / field goal / punt)": 2549,
+  "Классификация текста": 2550
+}

inference_lib.py

@@ -0,0 +1,250 @@
+"""Self-contained inference module for the recommendation web app.
+
+Contains a trimmed copy of ``MLPMetric`` (and its dependencies) so HF Spaces
+deployments do not need to ship the full ``module/`` package. The class layout
+and parameter names match the trained checkpoint exactly, so the original
+``state_dict`` loads with ``strict=False`` and a clean diff.
+"""
+from __future__ import annotations
+
+import hashlib
+import math
+import re
+from typing import Optional
+
+import torch
+import torch.nn as nn
+
+
+class ModelNameAvgEncoder(nn.Module):
+    """Hashed-token average over a model name. Optionally adds an ID embedding."""
+
+    def __init__(self, args, hash_buckets: int = 10000):
+        super().__init__()
+        self.hash_buckets = hash_buckets
+        self.tok_emb = nn.Embedding(self.hash_buckets, args.token_dim)
+        self.use_id_emb = bool(getattr(args, "use_id_emb", False))
+        if self.use_id_emb:
+            self.id_emb = nn.Embedding(args.num_models + 1, args.model_dim)
+            self.unk_model_id = args.num_models
+
+    @staticmethod
+    def _split(name: str):
+        n = (name or "").strip().lower()
+        if not n:
+            return []
+        toks = [n]
+        if "/" in n:
+            toks.append(n.split("/")[-1])
+        toks.extend([t for t in re.split(r"[\/_\-\s]+", n) if t])
+        out, seen = [], set()
+        for t in toks:
+            if t in seen:
+                continue
+            out.append(t)
+            seen.add(t)
+        return out
+
+    def _hash(self, tok: str):
+        return int(hashlib.md5(tok.encode()).hexdigest(), 16) % self.hash_buckets
+
+    def forward(self, model_ids: torch.LongTensor, model_names: list[str]):
+        device = self.tok_emb.weight.device
+        vecs = []
+        for n in model_names:
+            toks = self._split(n)
+            if not toks:
+                vecs.append(torch.zeros(self.tok_emb.embedding_dim, device=device))
+                continue
+            idxs = torch.tensor([self._hash(t) for t in toks], device=device, dtype=torch.long)
+            vecs.append(self.tok_emb(idxs).mean(dim=0))
+        h_name = torch.stack(vecs, dim=0)
+        feats = [h_name]
+        if self.use_id_emb:
+            feats.append(self.id_emb(model_ids.to(device)))
+        return torch.cat(feats, dim=-1)
+
+
+class MLPMetric(nn.Module):
+    """MLP recommender that takes raw dataset description embeddings, plus
+    task / metric / size / family side features, and ranks model candidates.
+
+    Mirrors the checkpoint at
+    ``checkpoint/mlp/unified_augmented/ablation_no_model_id_no_dataset_id``.
+    """
+
+    def __init__(self, args):
+        super().__init__()
+        self.use_id_emb = bool(getattr(args, "use_id_emb", False))
+        if self.use_id_emb:
+            self.model_embedding = nn.Embedding(args.num_models, args.model_dim)
+        else:
+            self.model_embedding = None
+
+        self.task_embedding = nn.Embedding(args.num_tasks, args.task_dim)
+        self.model_info_encoder = ModelNameAvgEncoder(args)
+        self.size_embedding = nn.Embedding(args.num_size_buckets, args.size_dim)
+        self.num_size_buckets = int(args.num_size_buckets)
+        self.use_size_prior = bool(getattr(args, "use_size_prior", True))
+
+        self.use_family_prior = bool(getattr(args, "use_family_prior", False))
+        if self.use_family_prior:
+            family_dim = int(getattr(args, "family_dim", args.size_dim))
+            self.family_embedding = nn.Embedding(args.num_families, family_dim)
+            self.family_dim = family_dim
+        else:
+            self.family_dim = 0
+
+        # Disable Model-Spider fusion path entirely (not used by this checkpoint).
+        self.use_ms_spider_repr = False
+        self.ms_fusion_dim = 0
+
+        model_info_dim = args.token_dim + (args.model_dim if self.use_id_emb else 0)
+        dataset_info_dim = args.dataset_desp_dim + args.task_dim
+        backbone_in_dim = (
+            model_info_dim + dataset_info_dim + args.size_dim + self.family_dim + self.ms_fusion_dim
+        )
+
+        # Backbone is rebuilt by the metric branch below; the base layers are kept here
+        # to match the parameter naming of the saved state dict.
+        self.backbone = nn.Sequential(
+            nn.Linear(backbone_in_dim, args.hidden_dim),
+            nn.ReLU(),
+            nn.Dropout(args.dropout_rate),
+            nn.Linear(args.hidden_dim, args.hidden_dim),
+            nn.ReLU(),
+            nn.Dropout(args.dropout_rate),
+        )
+        self.pairwise_head = nn.Linear(args.hidden_dim, 1)
+        self.pointwise_head = nn.Linear(args.hidden_dim, 1)
+
+        prior_in_dim = args.size_dim + self.family_dim
+        self.prior_head = nn.Sequential(
+            nn.Linear(prior_in_dim, args.hidden_dim // 2),
+            nn.ReLU(),
+            nn.Linear(args.hidden_dim // 2, 1),
+        )
+        self.temperature = nn.Parameter(torch.tensor(1.0))
+
+        # ---- metric extension (matches the MLPMetric subclass) ----
+        self.use_metric_embedding = bool(getattr(args, "use_metric_feature", True))
+        self.num_metrics = int(getattr(args, "num_metrics", 1))
+        self.metric_dim = int(getattr(args, "metric_dim", args.task_dim))
+        self.unknown_metric_id = int(getattr(args, "unknown_metric_id", 0))
+        if self.use_metric_embedding:
+            self.metric_embedding = nn.Embedding(max(self.num_metrics, 1), self.metric_dim)
+            in_features = self.backbone[0].in_features + self.metric_dim
+            hidden = self.backbone[0].out_features
+            dropout = self.backbone[2].p
+            self.backbone = nn.Sequential(
+                nn.Linear(in_features, hidden),
+                nn.ReLU(),
+                nn.Dropout(dropout),
+                nn.Linear(hidden, hidden),
+                nn.ReLU(),
+                nn.Dropout(dropout),
+            )
+        else:
+            self.metric_embedding = None
+
+    def encode_model(self, model_ids: torch.LongTensor, model_names: list[str]) -> torch.Tensor:
+        return self.model_info_encoder(model_ids, model_names)
+
+    @torch.no_grad()
+    def build_model_cache(
+        self,
+        all_model_names: list[str],
+        all_model_size_ids: torch.LongTensor,
+        all_model_family_ids: Optional[torch.LongTensor] = None,
+        device=None,
+    ):
+        if device is None:
+            device = next(self.parameters()).device
+        size_ids = all_model_size_ids.to(device=device, dtype=torch.long)
+        M = len(all_model_names)
+        assert size_ids.shape[0] == M
+        model_ids = torch.arange(M, device=device, dtype=torch.long)
+
+        h_model = self.encode_model(model_ids, all_model_names)
+        h_size = self.size_embedding(size_ids)
+        cache = {"h_model": h_model, "h_size": h_size, "size_ids": size_ids}
+        if self.use_family_prior and all_model_family_ids is not None:
+            family_ids = all_model_family_ids.to(device=device, dtype=torch.long)
+            cache["h_family"] = self.family_embedding(family_ids)
+            cache["family_ids"] = family_ids
+        else:
+            cache["h_family"] = None
+            cache["family_ids"] = None
+        return cache
+
+    def _metric_embed(
+        self, metric_ids: Optional[torch.LongTensor], batch_size: int, device
+    ) -> Optional[torch.Tensor]:
+        if not self.use_metric_embedding or self.metric_embedding is None:
+            return None
+        if metric_ids is None:
+            metric_ids = torch.full(
+                (batch_size,), int(self.unknown_metric_id), dtype=torch.long, device=device
+            )
+        return self.metric_embedding(metric_ids)
+
+    @torch.no_grad()
+    def score_matrix(
+        self,
+        task_ids: torch.LongTensor,
+        dataset_desp_batch: torch.Tensor,
+        model_cache: dict,
+        metric_ids: Optional[torch.LongTensor] = None,
+        chunk_size: int = 8192,
+    ) -> torch.Tensor:
+        device = dataset_desp_batch.device
+        B = dataset_desp_batch.size(0)
+
+        h_task = self.task_embedding(task_ids)
+        h_data = dataset_desp_batch
+        h_metric = self._metric_embed(metric_ids, B, device)
+
+        h_model_all = model_cache["h_model"]
+        h_size_all = model_cache["h_size"]
+        h_family_all = model_cache.get("h_family")
+        M = h_model_all.size(0)
+
+        if self.use_size_prior or self.use_family_prior:
+            if h_family_all is not None:
+                prior_inp_all = torch.cat([h_size_all, h_family_all], dim=-1)
+            else:
+                prior_inp_all = h_size_all
+            prior_all = self.prior_head(prior_inp_all).squeeze(-1)
+        else:
+            prior_all = torch.zeros(M, device=device)
+
+        out = torch.empty(B, M, device=device)
+        T = torch.clamp(self.temperature, min=1e-3)
+
+        start = 0
+        while start < M:
+            end = min(start + chunk_size, M)
+            m = end - start
+            h_model = h_model_all[start:end]
+            h_size = h_size_all[start:end]
+
+            h_model_exp = h_model.unsqueeze(0).expand(B, m, -1)
+            h_size_exp = h_size.unsqueeze(0).expand(B, m, -1)
+            h_data_exp = h_data.unsqueeze(1).expand(B, m, -1)
+            h_task_exp = h_task.unsqueeze(1).expand(B, m, -1)
+
+            parts = [h_model_exp, h_data_exp, h_size_exp]
+            if h_family_all is not None:
+                h_family_exp = h_family_all[start:end].unsqueeze(0).expand(B, m, -1)
+                parts.append(h_family_exp)
+            parts.append(h_task_exp)
+            if h_metric is not None:
+                parts.append(h_metric.unsqueeze(1).expand(B, m, -1))
+            residual_inp = torch.cat(parts, dim=-1)
+
+            h = self.backbone(residual_inp.reshape(B * m, -1))
+            s_chunk = self.pairwise_head(h).reshape(B, m)
+            prior_chunk = prior_all[start:end].unsqueeze(0)
+            out[:, start:end] = (s_chunk + prior_chunk) / T
+            start = end
+        return out

recommend.py

@@ -0,0 +1,409 @@
+"""Recommendation engine that loads the trained MLPMetric checkpoint plus the
+pre-built model pool, and exposes ``Recommender.recommend`` for the Gradio app.
+"""
+from __future__ import annotations
+
+import json
+import os
+import re
+import threading
+from dataclasses import dataclass
+from types import SimpleNamespace
+from typing import List, Optional
+
+import numpy as np
+import torch
+
+from inference_lib import MLPMetric
+
+
+EMBEDDING_MODEL = "text-embedding-3-small"  # Must match what was used during training.
+EMBEDDING_DIM = 1536
+
+
+# Official foundation-lab HuggingFace orgs (lowercase). Names whose owner falls
+# in this set are considered "official pretrained" releases (Llama, Qwen,
+# DeepSeek, Phi, Gemma, Mistral, Falcon, BLOOM, OLMo, Whisper, CLIP, ViT, ...).
+OFFICIAL_ORGS: set[str] = {
+    # Modern LLMs
+    "deepseek-ai", "qwen", "openai", "meta-llama", "mistralai",
+    "google", "microsoft", "01-ai", "tiiuae", "stabilityai",
+    "nvidia", "ibm-granite", "eleutherai", "bigscience",
+    "allenai", "salesforce", "apple", "xai-org",
+    # Multimodal / CV / audio
+    "facebook", "naver-clova-ix",
+    # Encoders / retrieval
+    "sentence-transformers", "baai", "jinaai", "intfloat",
+}
+
+# Classic bare-name pretrained releases (no org prefix on HF) that we still
+# count as "official" — e.g. the original Google BERT/T5, Facebook RoBERTa.
+OFFICIAL_BARE_NAMES: set[str] = {
+    "bert-base-uncased", "bert-large-uncased",
+    "roberta-base", "roberta-large",
+    "gpt2", "gpt2-medium", "gpt2-large", "gpt2-xl",
+    "t5-base", "t5-large", "t5-3b", "t5-11b",
+    "distilbert-base-uncased", "albert-base-v2",
+    "xlm-roberta-base", "xlm-roberta-large",
+}
+
+
+def _is_official_name(name: str) -> bool:
+    n = name.strip()
+    if "/" in n:
+        return n.split("/", 1)[0].lower() in OFFICIAL_ORGS
+    return n.lower() in OFFICIAL_BARE_NAMES
+
+
+def _slug(s: str) -> str:
+    return re.sub(r"[^a-z0-9]+", "", str(s).strip().lower())
+
+
+def _build_alias_map(name2id: dict[str, int]) -> dict[str, int]:
+    """Loose lookup: lowercased, also a slugged form, also strip composite markers."""
+    out: dict[str, int] = {}
+    for k, v in name2id.items():
+        for alias in {k, k.strip().lower(), _slug(k)}:
+            if alias and alias not in out:
+                out[alias] = v
+        # composite metric keys like "task::metric" — also store the suffix
+        if "::" in k:
+            tail = k.split("::", 1)[1]
+            for alias in {tail, tail.strip().lower(), _slug(tail)}:
+                if alias and alias not in out:
+                    out[alias] = v
+    return out
+
+
+@dataclass
+class Recommendation:
+    rank: int
+    model_name: str
+    score: float
+    size_bucket: int
+    size_b: float  # raw size in billions of params; NaN if unknown
+    family_id: int
+    popularity: int
+    hf_url: str
+
+
+class Recommender:
+    """Loads the checkpoint, model pool, and ID maps; exposes ``recommend``."""
+
+    def __init__(
+        self,
+        checkpoint_path: str,
+        args_path: str,
+        data_dir: str,
+        pool_path: str,
+        device: str = "cpu",
+    ):
+        self.device = torch.device(device)
+
+        with open(args_path) as f:
+            self._train_args = json.load(f)
+        with open(os.path.join(data_dir, "task2id.json")) as f:
+            self.task2id: dict[str, int] = json.load(f)
+        with open(os.path.join(data_dir, "metric2id.json")) as f:
+            metric2id_raw: dict[str, int] = json.load(f)
+        # The training-time metric vocab is the raw composite keys; expose both
+        # the raw form and a lowercased / slugged alias for lookup.
+        self.metric2id = metric2id_raw
+        self.task_alias = _build_alias_map(self.task2id)
+        self.metric_alias = _build_alias_map(self.metric2id)
+
+        pool = np.load(pool_path, allow_pickle=True)
+        self.model_names: list[str] = list(pool["names"].tolist())
+        self.size_ids = torch.tensor(pool["size_ids"], dtype=torch.long)
+        # Backwards compatible: older pools won't have sizes_b. Default to NaN.
+        if "sizes_b" in pool.files:
+            self.sizes_b: np.ndarray = pool["sizes_b"].astype(np.float32)
+        else:
+            self.sizes_b = np.full(len(self.model_names), np.nan, dtype=np.float32)
+        self.family_ids = torch.tensor(pool["family_ids"], dtype=torch.long)
+        self.popularities: np.ndarray = pool["popularities"]
+        self.urls: list[str] = list(pool["urls"].tolist())
+
+        # Precompute the "official pretrained" mask once — names are static.
+        self.is_official: np.ndarray = np.array(
+            [_is_official_name(n) for n in self.model_names], dtype=bool
+        )
+
+        # Build the MLPMetric model with the same hyper-parameters used for training.
+        cfg = self._train_args
+        model_args = SimpleNamespace(
+            num_models=cfg.get("num_models", len(self.model_names)),
+            num_tasks=cfg.get("num_tasks"),
+            num_metrics=cfg.get("num_metrics"),
+            num_size_buckets=cfg.get("num_size_buckets"),
+            num_families=cfg.get("num_families"),
+            token_dim=cfg["token_dim"],
+            model_dim=cfg["model_dim"],
+            task_dim=cfg["task_dim"],
+            metric_dim=cfg.get("metric_dim", cfg["task_dim"]),
+            size_dim=cfg["size_dim"],
+            family_dim=cfg.get("family_dim", cfg["size_dim"]),
+            dataset_desp_dim=cfg["dataset_desp_dim"],
+            hidden_dim=cfg["hidden_dim"],
+            dropout_rate=cfg.get("dropout_rate", 0.0),
+            use_id_emb=bool(cfg.get("use_id_emb", False)),
+            use_size_prior=bool(cfg.get("use_size_prior", True)),
+            use_family_prior=bool(cfg.get("use_family_prior", False)),
+            use_metric_feature=bool(cfg.get("use_metric_feature", True)),
+            unknown_metric_id=int(cfg.get("unknown_metric_id", 0)),
+        )
+        self.model = MLPMetric(model_args).to(self.device).eval()
+
+        raw = torch.load(checkpoint_path, map_location="cpu")
+        state = raw.get("model", raw) if isinstance(raw, dict) else raw
+        missing, unexpected = self.model.load_state_dict(state, strict=False)
+        if missing or unexpected:
+            print(f"[Recommender] loaded with missing={len(missing)} unexpected={len(unexpected)}")
+            if missing:
+                print("  e.g. missing:", missing[:3])
+            if unexpected:
+                print("  e.g. unexpected:", unexpected[:3])
+
+        # Pre-compute the model-side cache once. Running the token encoder over
+        # 47k names is the slowest single step; we amortize it to startup.
+        self._cache_lock = threading.Lock()
+        with torch.no_grad():
+            self.model_cache = self.model.build_model_cache(
+                self.model_names,
+                self.size_ids,
+                all_model_family_ids=self.family_ids if self.model.use_family_prior else None,
+                device=self.device,
+            )
+
+        # OpenAI client is created lazily so the import is only required when used.
+        self._oai_client = None
+
+    # ------------------------------------------------------------------ embedding
+
+    def _make_openai_client(self, api_key: Optional[str] = None):
+        from openai import OpenAI  # noqa: WPS433
+        # When the caller supplies a key (e.g. from the Gradio UI), build a
+        # fresh client and do NOT cache it — different users send different
+        # keys, and we don't want one user's key to be reused for the next.
+        if api_key:
+            return OpenAI(api_key=api_key)
+        # Fallback for local dev: rely on OPENAI_API_KEY in the environment.
+        if self._oai_client is None:
+            self._oai_client = OpenAI()
+        return self._oai_client
+
+    def embed_description(self, text: str, api_key: Optional[str] = None) -> np.ndarray:
+        text = (text or "").strip()
+        if not text:
+            raise ValueError("Dataset description must be non-empty.")
+        try:
+            client = self._make_openai_client(api_key)
+        except Exception as e:  # missing OPENAI_API_KEY in dev, etc.
+            raise ValueError(
+                "OpenAI client could not be created. Paste an API key into "
+                "the 'OpenAI API key' field above. Original error: " + str(e)
+            )
+        try:
+            resp = client.embeddings.create(model=EMBEDDING_MODEL, input=text)
+        except Exception as e:
+            # Surface auth / quota errors back to the user verbatim — they're
+            # the ones who need to fix it.
+            raise ValueError(f"OpenAI embedding call failed: {e}")
+        vec = np.asarray(resp.data[0].embedding, dtype=np.float32)
+        if vec.shape[-1] != EMBEDDING_DIM:
+            raise RuntimeError(
+                f"Expected {EMBEDDING_DIM}-dim embedding, got {vec.shape[-1]}. "
+                f"Make sure the API key has access to {EMBEDDING_MODEL}."
+            )
+        return vec
+
+    # ------------------------------------------------------------------ lookups
+
+    def resolve_task(self, task: str) -> int:
+        if task is None:
+            raise ValueError("Task must be provided.")
+        for cand in (task, task.strip().lower(), _slug(task)):
+            if cand in self.task_alias:
+                return self.task_alias[cand]
+        raise ValueError(
+            f"Unknown task '{task}'. Pick one from the dropdown — the model has only seen {len(self.task2id)} task labels."
+        )
+
+    def resolve_metric(self, metric: str) -> int:
+        if metric is None or not str(metric).strip():
+            return int(self.model.unknown_metric_id)
+        for cand in (metric, metric.strip().lower(), _slug(metric)):
+            if cand in self.metric_alias:
+                return self.metric_alias[cand]
+        # Fallback: unknown metric token.
+        return int(self.model.unknown_metric_id)
+
+    # ------------------------------------------------------------------ main API
+
+    def recommend(
+        self,
+        dataset_description: str,
+        task: str,
+        metric: Optional[str] = None,
+        top_k: int = 20,
+        popularity_weight: float = 0.0,
+        hf_only: bool = True,
+        min_size_b: Optional[float] = None,
+        max_size_b: Optional[float] = None,
+        official_only: bool = False,
+        api_key: Optional[str] = None,
+    ) -> List[Recommendation]:
+        """Score all candidate models and return the top-k.
+
+        ``popularity_weight`` (0..1) blends a log(downloads) signal into the
+        ranking, useful when several models have near-tied scores. Default 0
+        means "pure model output".
+
+        ``hf_only`` (default True) drops candidates whose model name is not a
+        HuggingFace repo id (those are paper baselines like ``inceptionv4``
+        that the user cannot download with ``hf hub``).
+
+        ``min_size_b`` / ``max_size_b`` (optional, in B params) restrict
+        results to candidates whose raw parameter count falls in the range.
+        ``None`` (or 0 from the UI) means "no limit". Models with unknown
+        size are excluded once any size bound is set.
+
+        ``official_only`` (default False) restricts to a curated whitelist of
+        foundation-lab orgs (DeepSeek, Qwen, Llama, gpt-oss, Mistral, ...).
+
+        ``api_key`` (optional) — OpenAI API key supplied by the caller (e.g.
+        from a Gradio textbox). When given, used for this single request only;
+        otherwise the recommender falls back to ``OPENAI_API_KEY`` in env.
+        """
+        task_id = self.resolve_task(task)
+        metric_id = self.resolve_metric(metric)
+        emb = self.embed_description(dataset_description, api_key=api_key)
+        return self._score(
+            emb, task_id, metric_id, top_k, popularity_weight, hf_only,
+            min_size_b=min_size_b, max_size_b=max_size_b,
+            official_only=official_only,
+        )
+
+    @torch.no_grad()
+    def _score(
+        self,
+        desp_emb: np.ndarray,
+        task_id: int,
+        metric_id: int,
+        top_k: int,
+        popularity_weight: float,
+        hf_only: bool = True,
+        min_size_b: Optional[float] = None,
+        max_size_b: Optional[float] = None,
+        official_only: bool = False,
+    ) -> List[Recommendation]:
+        device = self.device
+        task_t = torch.tensor([task_id], dtype=torch.long, device=device)
+        metric_t = torch.tensor([metric_id], dtype=torch.long, device=device)
+        desp_t = torch.tensor(desp_emb, dtype=torch.float32, device=device).unsqueeze(0)
+
+        with self._cache_lock:
+            scores = self.model.score_matrix(
+                task_t, desp_t, self.model_cache, metric_ids=metric_t
+            ).squeeze(0)
+        scores_np = scores.detach().cpu().numpy().astype(np.float32)
+
+        if popularity_weight > 0.0:
+            pop = np.log1p(self.popularities.astype(np.float32))
+            if pop.max() > 0:
+                pop = pop / pop.max()
+            # Re-center scores then add the popularity nudge.
+            s_norm = scores_np - scores_np.mean()
+            if s_norm.std() > 1e-6:
+                s_norm = s_norm / s_norm.std()
+            ranking_scores = s_norm + popularity_weight * pop
+        else:
+            ranking_scores = scores_np
+
+        # Mask out non-HF candidates by setting their score to -inf.
+        if hf_only:
+            has_url = np.array([bool(u) for u in self.urls])
+            ranking_scores = np.where(has_url, ranking_scores, -np.inf)
+
+        # Mask candidates outside the manual size bounds (B params).
+        # Convention from the UI: 0 / None means "no limit". Models with
+        # unknown size are dropped once any bound is set.
+        size_filter_active = (min_size_b not in (None, 0)) or (max_size_b not in (None, 0))
+        if size_filter_active:
+            sizes = self.sizes_b
+            in_range = ~np.isnan(sizes)
+            if min_size_b not in (None, 0):
+                in_range &= sizes >= float(min_size_b)
+            if max_size_b not in (None, 0):
+                in_range &= sizes <= float(max_size_b)
+            ranking_scores = np.where(in_range, ranking_scores, -np.inf)
+
+        # Mask non-official models when the user wants only flagship checkpoints.
+        if official_only:
+            ranking_scores = np.where(self.is_official, ranking_scores, -np.inf)
+
+        top_k = max(1, min(int(top_k), len(self.model_names)))
+        top_idx = np.argpartition(-ranking_scores, top_k - 1)[:top_k]
+        top_idx = top_idx[np.argsort(-ranking_scores[top_idx])]
+
+        out: list[Recommendation] = []
+        for rank, i in enumerate(top_idx, start=1):
+            out.append(
+                Recommendation(
+                    rank=rank,
+                    model_name=self.model_names[i],
+                    score=float(scores_np[i]),
+                    size_bucket=int(self.size_ids[i]),
+                    size_b=float(self.sizes_b[i]),
+                    family_id=int(self.family_ids[i]),
+                    popularity=int(self.popularities[i]),
+                    hf_url=self.urls[i],
+                )
+            )
+        return out
+
+
+def default_recommender() -> Recommender:
+    """Convenience constructor.
+
+    Resolves paths in this order:
+      1. Environment variables (``MODEL_CKPT``, ``MODEL_ARGS``, ``DATA_DIR``, ``POOL_PATH``).
+      2. Self-contained Spaces layout: ``web/checkpoint/`` and ``web/data/``.
+      3. Original project tree (development mode).
+    """
+    here = os.path.dirname(os.path.abspath(__file__))
+    root = os.path.dirname(here)
+
+    spaces_ckpt = os.path.join(here, "checkpoint/MLPMetric.pt")
+    spaces_args = os.path.join(here, "checkpoint/args.json")
+    spaces_data = os.path.join(here, "data")
+
+    dev_ckpt = os.path.join(root, "checkpoint/mlp/unified_augmented/ablation_no_model_id_no_dataset_id/MLPMetric.pt")
+    dev_args = os.path.join(root, "checkpoint/mlp/unified_augmented/ablation_no_model_id_no_dataset_id/args.json")
+    dev_data = os.path.join(root, "data/unified_augmented")
+
+    def _pick(env_key: str, primary: str, fallback: str) -> str:
+        v = os.environ.get(env_key)
+        if v:
+            return v
+        return primary if os.path.exists(primary) else fallback
+
+    return Recommender(
+        checkpoint_path=_pick("MODEL_CKPT", spaces_ckpt, dev_ckpt),
+        args_path=_pick("MODEL_ARGS", spaces_args, dev_args),
+        data_dir=_pick("DATA_DIR", spaces_data, dev_data),
+        pool_path=os.environ.get("POOL_PATH", os.path.join(here, "assets/model_pool.npz")),
+        device=os.environ.get("DEVICE", "cpu"),
+    )
+
+
+if __name__ == "__main__":
+    rec = default_recommender()
+    print(f"Loaded {len(rec.model_names)} candidate models, "
+          f"{len(rec.task2id)} tasks, {len(rec.metric2id)} metrics.")
+    sample_task = next(iter(rec.task2id))
+    print(f"\nSmoke test: ranking for task={sample_task!r}")
+    fake_emb = np.random.randn(EMBEDDING_DIM).astype(np.float32)
+    out = rec._score(fake_emb, rec.task2id[sample_task], rec.model.unknown_metric_id, 5, 0.0)
+    for r in out:
+        print(f"  #{r.rank} {r.model_name:<60} score={r.score:+.4f} pop={r.popularity}")

requirements.txt

@@ -0,0 +1,7 @@
+torch>=2.1.0,<2.6
+numpy>=1.24,<2.0
+pandas>=2.0,<2.4
+gradio==4.44.0
+gradio-client==1.3.0
+huggingface_hub>=0.24,<0.26
+openai>=1.40