Upload folder using huggingface_hub

app.py

@@ -1172,6 +1172,7 @@ def render_leaderboard_table(df: pd.DataFrame, columns: list, show_analyze_colum
         cells = []
         model_name = row.get("Model", "")
         organization = row.get("Organization", "")
+        hide_attrib_kuiper = model_name == "Human with Oracle Retriever"
         
         # Check if this is a human performance row (should merge Model, Organization, Model Type)
         is_human_row = organization == "Humanity"
@@ -1273,6 +1274,9 @@ def render_leaderboard_table(df: pd.DataFrame, columns: list, show_analyze_colum
                 cells.append(f'<td style="text-align: center;">{cell_html}</td>')
             elif col.startswith("Attribution"):
                 # Format F1 scores (scale 0-100) - NOT bias-adjusted
+                if hide_attrib_kuiper:
+                    cells.append('<td style="text-align: center;">—</td>')
+                    continue
                 try:
                     attr_val = f"{float(value):.1f}" if value else "0"
                     attr_float = float(value) if value else 0
@@ -1309,6 +1313,9 @@ def render_leaderboard_table(df: pd.DataFrame, columns: list, show_analyze_colum
             elif col == "Effort (Kuiper)":
                 # Format Kuiper statistic (lower is better for calibration)
                 # Hide for Conventional RAG models (not meaningful)
+                if hide_attrib_kuiper:
+                    cells.append('<td style="text-align: center;">—</td>')
+                    continue
                 tags = row.get("Tags", [])
                 is_conventional_rag = "Conventional RAG" in tags if isinstance(tags, list) else False
                 if is_conventional_rag:

eval/cleanup_submissions.py

@@ -0,0 +1,181 @@
+#!/usr/bin/env python3
+"""
+Clean up backend-results submissions by keeping only the newest prediction per prefix
+and the matching results file (if available). Optionally updates kept results to
+record source_predictions_file and result_file_path.
+"""
+
+import argparse
+import json
+import os
+import re
+from dataclasses import dataclass
+from typing import Dict, List, Optional, Tuple
+
+from huggingface_hub import HfApi, hf_hub_download, list_repo_files
+
+
+RESULTS_REPO = "agentic-document-ai/backend-results"
+TOKEN = os.environ.get("HF_TOKEN")
+
+
+@dataclass
+class FileEntry:
+    path: str
+    prefix: str
+    ts_raw: str
+    ts_key: Tuple[int, int, int, int, int, int]
+
+
+def _parse_timestamp(ts: str) -> Tuple[int, int, int, int, int, int]:
+    """Parse timestamps in YYYYMMDD_HHMMSS or YYYY-MM-DDTHH-MM-SSZ format."""
+    match_compact = re.match(r"^(\d{8})_(\d{6})$", ts)
+    if match_compact:
+        ymd, hms = match_compact.groups()
+        return (
+            int(ymd[0:4]),
+            int(ymd[4:6]),
+            int(ymd[6:8]),
+            int(hms[0:2]),
+            int(hms[2:4]),
+            int(hms[4:6]),
+        )
+
+    match_iso = re.match(r"^(\d{4})-(\d{2})-(\d{2})T(\d{2})-(\d{2})-(\d{2})Z$", ts)
+    if match_iso:
+        return tuple(int(x) for x in match_iso.groups())  # type: ignore[return-value]
+
+    # Fallback: treat as zero to make it always older than parsed timestamps
+    return (0, 0, 0, 0, 0, 0)
+
+
+def _split_predictions(path: str) -> Optional[FileEntry]:
+    if "_predictions_" not in path or not path.endswith(".jsonl"):
+        return None
+    prefix, ts = path.rsplit("_predictions_", 1)
+    ts = ts.replace(".jsonl", "")
+    return FileEntry(path=path, prefix=prefix, ts_raw=ts, ts_key=_parse_timestamp(ts))
+
+
+def _split_results(path: str) -> Optional[FileEntry]:
+    if "_results_" not in path or not path.endswith(".json"):
+        return None
+    prefix, ts = path.rsplit("_results_", 1)
+    ts = ts.replace(".json", "")
+    return FileEntry(path=path, prefix=prefix, ts_raw=ts, ts_key=_parse_timestamp(ts))
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Clean backend-results submissions")
+    parser.add_argument("--apply", action="store_true", help="Apply deletions/updates (default is dry-run)")
+    parser.add_argument("--drop-unmatched-results", action="store_true",
+                        help="Delete results that do not match the latest prediction timestamp")
+    args = parser.parse_args()
+
+    api = HfApi(token=TOKEN)
+    files = list_repo_files(RESULTS_REPO, repo_type="dataset", token=TOKEN)
+
+    predictions: List[FileEntry] = []
+    results: List[FileEntry] = []
+
+    for f in files:
+        p = _split_predictions(f)
+        if p:
+            predictions.append(p)
+            continue
+        r = _split_results(f)
+        if r:
+            results.append(r)
+
+    preds_by_prefix: Dict[str, List[FileEntry]] = {}
+    results_by_prefix: Dict[str, List[FileEntry]] = {}
+
+    for p in predictions:
+        preds_by_prefix.setdefault(p.prefix, []).append(p)
+    for r in results:
+        results_by_prefix.setdefault(r.prefix, []).append(r)
+
+    to_delete: List[str] = []
+    to_update: List[Tuple[str, str]] = []  # (result_path, prediction_path)
+    unmatched_latest: List[str] = []
+
+    all_prefixes = sorted(set(preds_by_prefix) | set(results_by_prefix))
+    for prefix in all_prefixes:
+        preds = sorted(preds_by_prefix.get(prefix, []), key=lambda x: x.ts_key, reverse=True)
+        res = sorted(results_by_prefix.get(prefix, []), key=lambda x: x.ts_key, reverse=True)
+
+        latest_pred = preds[0] if preds else None
+        latest_res = res[0] if res else None
+
+        # Remove all older predictions
+        if preds:
+            for old in preds[1:]:
+                to_delete.append(old.path)
+
+        # Remove older results; keep latest for now
+        if res:
+            for old in res[1:]:
+                to_delete.append(old.path)
+
+        if latest_pred and latest_res:
+            # If a result matches the latest prediction timestamp, link it
+            if latest_res.ts_raw == latest_pred.ts_raw:
+                to_update.append((latest_res.path, latest_pred.path))
+            else:
+                unmatched_latest.append(prefix)
+                if args.drop_unmatched_results:
+                    to_delete.append(latest_res.path)
+        elif latest_res and not latest_pred:
+            # Results without any predictions
+            if args.drop_unmatched_results:
+                to_delete.append(latest_res.path)
+
+    print(f"Predictions: {len(predictions)}")
+    print(f"Results: {len(results)}")
+    print(f"Delete candidates: {len(to_delete)}")
+    print(f"Results to update (link to latest predictions): {len(to_update)}")
+    if unmatched_latest:
+        print("\nPrefixes where latest result does NOT match latest prediction:")
+        for p in unmatched_latest:
+            print("  ", p)
+
+    if not args.apply:
+        print("\nDry-run only. Re-run with --apply to execute changes.")
+        return 0
+
+    # Apply updates
+    for result_path, pred_path in to_update:
+        local_path = hf_hub_download(
+            repo_id=RESULTS_REPO,
+            filename=result_path,
+            repo_type="dataset",
+            token=TOKEN,
+        )
+        with open(local_path) as f:
+            data = json.load(f)
+        data["source_predictions_file"] = pred_path
+        data["result_file_path"] = result_path
+        api.upload_file(
+            path_or_fileobj=json.dumps(data, indent=2).encode("utf-8"),
+            path_in_repo=result_path,
+            repo_id=RESULTS_REPO,
+            repo_type="dataset",
+            token=TOKEN,
+            commit_message=f"Link result to latest prediction: {result_path}",
+        )
+
+    # Apply deletions
+    for path in to_delete:
+        api.delete_file(
+            path_in_repo=path,
+            repo_id=RESULTS_REPO,
+            repo_type="dataset",
+            token=TOKEN,
+        )
+
+    print("Cleanup complete.")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

eval/delete_unlinked.py

@@ -0,0 +1,87 @@
+#!/usr/bin/env python3
+"""
+Delete Humanity submissions and unlinked results from backend-results.
+
+Usage:
+  export HF_TOKEN=...
+  python streamlit_app/eval/delete_unlinked.py          # dry-run
+  python streamlit_app/eval/delete_unlinked.py --apply  # actually delete
+"""
+
+import argparse
+import os
+
+from huggingface_hub import HfApi, list_repo_files
+
+
+RESULTS_REPO = "agentic-document-ai/backend-results"
+TOKEN = os.environ.get("HF_TOKEN")
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Delete Humanity and unlinked files")
+    parser.add_argument("--apply", action="store_true", help="Actually delete (default: dry-run)")
+    args = parser.parse_args()
+
+    api = HfApi(token=TOKEN)
+    files = list_repo_files(RESULTS_REPO, repo_type="dataset", token=TOKEN)
+
+    result_files = [f for f in files if f.endswith('.json') and '_results_' in f]
+    pred_files = [f for f in files if f.endswith('.jsonl') and '_predictions_' in f]
+
+    def key_for_result(name: str):
+        parts = name.rsplit('_results_', 1)
+        if len(parts) != 2:
+            return None
+        return f"{parts[0]}_{parts[1].replace('.json','')}"
+
+    def key_for_pred(name: str):
+        parts = name.rsplit('_predictions_', 1)
+        if len(parts) != 2:
+            return None
+        return f"{parts[0]}_{parts[1].replace('.jsonl','')}"
+
+    result_keys = {key_for_result(f): f for f in result_files if key_for_result(f)}
+    pred_keys = {key_for_pred(f): f for f in pred_files if key_for_pred(f)}
+
+    # Find unlinked results (no matching prediction)
+    unlinked_results = [result_keys[k] for k in set(result_keys) - set(pred_keys)]
+
+    # Find all Humanity files
+    humanity_files = [f for f in files if f.startswith("Humanity/")]
+
+    # Combine into deletion list (deduplicated)
+    to_delete = sorted(set(unlinked_results + humanity_files))
+
+    print(f"Files to delete: {len(to_delete)}")
+    for f in to_delete:
+        print(f"  {f}")
+
+    if not to_delete:
+        print("Nothing to delete.")
+        return 0
+
+    if not args.apply:
+        print("\nDry-run mode. Add --apply to actually delete.")
+        return 0
+
+    print(f"\nDeleting {len(to_delete)} files...")
+    for f in to_delete:
+        try:
+            api.delete_file(
+                path_in_repo=f,
+                repo_id=RESULTS_REPO,
+                repo_type="dataset",
+                token=TOKEN,
+                commit_message=f"Cleanup: delete {f}"
+            )
+            print(f"  ✓ Deleted: {f}")
+        except Exception as e:
+            print(f"  ✗ Error deleting {f}: {e}")
+
+    print("\nDone!")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

eval/link_file_search_predictions.py

@@ -0,0 +1,103 @@
+#!/usr/bin/env python3
+"""
+Upload missing File Search predictions and link them to existing results.
+
+Usage:
+  export HF_TOKEN=...
+  python streamlit_app/eval/link_file_search_predictions.py --apply
+"""
+
+import argparse
+import json
+import os
+from pathlib import Path
+
+from huggingface_hub import HfApi, hf_hub_download
+
+
+RESULTS_REPO = "agentic-document-ai/backend-results"
+TOKEN = os.environ.get("HF_TOKEN")
+BASE_DIR = Path(__file__).resolve().parents[2]  # Project root
+FILE_SEARCH_DIR = BASE_DIR / "file_search_results"
+
+# Map missing results -> local prediction file
+MISSING_RESULTS = {
+    "Google/Gemini_2.5_Flash_with_File_Search_results_20260103_221253.json": "gemini-2.5-flash.jsonl",
+    "Google/Gemini_2.5_Pro_with_File_Search_results_20260103_221943.json": "gemini-2.5-pro.jsonl",
+    "Google/Gemini_3.0_Pro_(Preview)_with_File_Search_results_20260104_120431.json": "gemini-3-pro-preview.jsonl",
+    "OpenAI/GPT-5.2_(2025-12-11)_with_File_Search_results_20260104_121551.json": "gpt-5.2-2025-12-11.jsonl",
+    "OpenAI/GPT-5_Mini_(2025-08-07)_with_File_Search_results_20260104_122026.json": "gpt-5-mini-2025-08-07.jsonl",
+}
+
+
+def _pred_path_from_result(result_path: str) -> str:
+    # {org}/{model}_results_{ts}.json -> {org}/{model}_predictions_{ts}.jsonl
+    base, ts = result_path.rsplit("_results_", 1)
+    ts = ts.replace(".json", "")
+    return f"{base}_predictions_{ts}.jsonl"
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Upload file_search_results predictions and link them to results.")
+    parser.add_argument("--apply", action="store_true", help="Apply uploads/updates (default: dry-run)")
+    args = parser.parse_args()
+
+    if not FILE_SEARCH_DIR.exists():
+        raise FileNotFoundError(f"Missing directory: {FILE_SEARCH_DIR}")
+
+    api = HfApi(token=TOKEN)
+
+    actions = []
+    for result_path, local_name in MISSING_RESULTS.items():
+        local_file = FILE_SEARCH_DIR / local_name
+        if not local_file.exists():
+            raise FileNotFoundError(f"Missing local prediction file: {local_file}")
+
+        pred_path = _pred_path_from_result(result_path)
+        actions.append((result_path, pred_path, local_file))
+
+    print(f"Planned uploads: {len(actions)}")
+    for result_path, pred_path, local_file in actions:
+        print(f"- {local_file.name} -> {pred_path}")
+
+    if not args.apply:
+        print("\nDry-run only. Re-run with --apply to execute.")
+        return 0
+
+    for result_path, pred_path, local_file in actions:
+        # Upload predictions file
+        api.upload_file(
+            path_or_fileobj=str(local_file),
+            path_in_repo=pred_path,
+            repo_id=RESULTS_REPO,
+            repo_type="dataset",
+            token=TOKEN,
+            commit_message=f"Add predictions for {pred_path}",
+        )
+
+        # Update results JSON with linkage fields
+        local_result = hf_hub_download(
+            repo_id=RESULTS_REPO,
+            filename=result_path,
+            repo_type="dataset",
+            token=TOKEN,
+        )
+        with open(local_result) as f:
+            data = json.load(f)
+        data["source_predictions_file"] = pred_path
+        data["result_file_path"] = result_path
+        api.upload_file(
+            path_or_fileobj=json.dumps(data, indent=2).encode("utf-8"),
+            path_in_repo=result_path,
+            repo_id=RESULTS_REPO,
+            repo_type="dataset",
+            token=TOKEN,
+            commit_message=f"Link result to predictions: {result_path}",
+        )
+
+    print("Done.")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

eval/metrics.py

@@ -328,13 +328,17 @@ def _get_gemini_model():
 def _call_gemini_with_timeout(model, prompt, timeout=30):
     """Call Gemini with a timeout using threading."""
     from concurrent.futures import ThreadPoolExecutor, TimeoutError as FuturesTimeoutError
+    try:
+        temperature = float(os.environ.get("GEMINI_TEMPERATURE", "0"))
+    except ValueError:
+        temperature = 0.0
     
     def _call():
         return model.generate_content(
             prompt,
             tools=[_LLM_JUDGE_TOOL],
             tool_config={"function_calling_config": {"mode": "ANY"}},
-            generation_config={"temperature": 0},
+            generation_config={"temperature": temperature},
             request_options={"timeout": timeout},
         )
     

eval/reevaluate_submissions.py

@@ -30,6 +30,7 @@ from metrics import (
 # Config
 RESULTS_REPO = "agentic-document-ai/backend-results"
 TOKEN = os.environ.get("HF_TOKEN")
+os.environ.setdefault("GEMINI_TEMPERATURE", "0")
 
 
 def load_gold_data():

eval/reevaluated_results/Alibaba_Group/Qwen3-VL_(235B-A22B-Thinking)_with_BM25_Search_Tool_results_20260110_131629.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-10T13:16:29.905067+00:00",
   "results": {
     "overall": {
-      "semantic": 59.40452333237904,
+      "semantic": 60.30612244897959,
       "semantic_ci": [
-        54.988846362852016,
-        63.82020030190607
+        55.908758540500614,
+        64.70348635745857
       ],
-      "anls": 57.61603118163428,
-      "page_f1": 58.72697776505391,
-      "doc_f1": 80.62601393262716,
-      "kuiper": 34.73747494989995
+      "anls": 58.219456137310175,
+      "page_f1": 58.609523809523814,
+      "doc_f1": 80.4647619047619,
+      "kuiper": 36.59000000000024
     },
     "single_evidence": {
-      "semantic": 65.45750168199147,
-      "anls": 63.05477065574925,
-      "n": 364
+      "semantic": 66.53620352250489,
+      "anls": 63.67469870606113,
+      "n": 365
     },
     "multi_evidence_same_doc": {
       "semantic": 36.443148688046655,
-      "anls": 36.39125352974683,
+      "anls": 37.224586863080155,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 54.02160864345738,
+      "semantic": 55.022008803521395,
       "anls": 53.75681851851068,
       "n": 51
     },
@@ -52,33 +52,33 @@
         "n": 22
       },
       "Events": {
-        "semantic": 68.02721088435374,
-        "anls": 66.4976376669925,
+        "semantic": 72.27891156462587,
+        "anls": 68.58097100032585,
         "n": 24
       },
       "Financial": {
-        "semantic": 45.474711623779946,
-        "anls": 49.53672826145982,
+        "semantic": 48.802129547471154,
+        "anls": 51.018941700195,
         "n": 92
       },
       "Financial/Tax": {
-        "semantic": 63.775510204081634,
+        "semantic": 57.39795918367348,
         "anls": 57.39996898263027,
         "n": 16
       },
       "Government/Regulatory": {
         "semantic": 62.961354754667845,
-        "anls": 60.56474101398679,
+        "anls": 60.57273973419155,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 59.731209556993534,
-        "anls": 53.957859669066565,
+        "semantic": 57.24240915878547,
+        "anls": 55.177371864188515,
         "n": 41
       },
       "Legal": {
-        "semantic": 61.69909824394873,
-        "anls": 54.79335264218985,
+        "semantic": 59.32605600379688,
+        "anls": 55.25846892125962,
         "n": 43
       },
       "Media/Publishing": {
@@ -91,13 +91,18 @@
         "anls": 73.07522250524337,
         "n": 24
       },
+      "Other": {
+        "semantic": 100.0,
+        "anls": 100.0,
+        "n": 1
+      },
       "Reference": {
-        "semantic": 61.813186813186825,
-        "anls": 63.19327183267644,
+        "semantic": 65.73783359497645,
+        "anls": 63.24305900202882,
         "n": 52
       },
       "Reports": {
-        "semantic": 57.82312925170068,
+        "semantic": 59.18367346938776,
         "anls": 53.11616787903517,
         "n": 75
       },
@@ -107,10 +112,10 @@
         "n": 23
       }
     },
-    "n_evaluated": 499,
-    "n_unmatched": 1767
+    "n_evaluated": 500,
+    "n_unmatched": 1766
   },
-  "reevaluated_date": "2026-01-16T15:38:39.375804+00:00",
+  "reevaluated_date": "2026-01-22T21:56:11.212274+00:00",
   "source_predictions_file": "Alibaba_Group/Qwen3-VL_(235B-A22B-Thinking)_with_BM25_Search_Tool_predictions_20260110_131629.jsonl",
   "result_file_path": "Alibaba_Group/Qwen3-VL_(235B-A22B-Thinking)_with_BM25_Search_Tool_results_20260110_131629.json"
 }

eval/reevaluated_results/Alibaba_Group/Qwen3-VL_(32B-Thinking)_with_BM25_Search_Tool_results_20260110_132054.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-10T13:20:54.125677+00:00",
   "results": {
     "overall": {
-      "semantic": 57.87084372827287,
+      "semantic": 58.77551020408164,
       "semantic_ci": [
-        53.43484625371868,
-        62.30684120282706
+        54.35541948342135,
+        63.195600924741925
       ],
-      "anls": 57.937064653000625,
-      "page_f1": 54.83061360816872,
-      "doc_f1": 78.76514934631167,
-      "kuiper": 36.665330661322855
+      "anls": 58.50093245097659,
+      "page_f1": 54.72095238095238,
+      "doc_f1": 78.60761904761905,
+      "kuiper": 38.259999999999735
     },
     "single_evidence": {
-      "semantic": 65.03700381251402,
-      "anls": 64.58920167676482,
-      "n": 364
+      "semantic": 65.97707576181158,
+      "anls": 65.20641198351612,
+      "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 27.332361516034982,
-      "anls": 29.952414571366546,
+      "semantic": 27.939747327502424,
+      "anls": 30.547652666604648,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 57.022809123649466,
+      "semantic": 58.023209283713484,
       "anls": 56.55143191196322,
       "n": 51
     },
@@ -53,12 +53,12 @@
       },
       "Events": {
         "semantic": 80.78231292517005,
-        "anls": 77.14578581514066,
+        "anls": 79.22911914847398,
         "n": 24
       },
       "Financial": {
-        "semantic": 46.58385093167702,
-        "anls": 47.03746065646829,
+        "semantic": 48.802129547471154,
+        "anls": 48.51967409520347,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,13 +67,13 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 53.19148936170214,
-        "anls": 56.60682613221971,
+        "semantic": 51.02040816326531,
+        "anls": 56.616157972458616,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 63.46441015430564,
-        "anls": 61.11435746903807,
+        "semantic": 64.70881035340967,
+        "anls": 62.33386966416003,
         "n": 41
       },
       "Legal": {
@@ -82,7 +82,7 @@
         "n": 43
       },
       "Media/Publishing": {
-        "semantic": 40.81632653061224,
+        "semantic": 42.85714285714285,
         "anls": 54.65844817149165,
         "n": 25
       },
@@ -91,26 +91,31 @@
         "anls": 73.59601449275362,
         "n": 24
       },
+      "Other": {
+        "semantic": 100.0,
+        "anls": 100.0,
+        "n": 1
+      },
       "Reference": {
-        "semantic": 64.75667189952904,
-        "anls": 68.57667578882189,
+        "semantic": 65.73783359497645,
+        "anls": 68.63568984240118,
         "n": 52
       },
       "Reports": {
-        "semantic": 57.14285714285715,
+        "semantic": 57.82312925170068,
         "anls": 56.44955119487462,
         "n": 75
       },
       "Technical": {
-        "semantic": 57.67524401064772,
+        "semantic": 62.111801242236034,
         "anls": 51.60498619336015,
         "n": 23
       }
     },
-    "n_evaluated": 499,
-    "n_unmatched": 1767
+    "n_evaluated": 500,
+    "n_unmatched": 1766
   },
-  "reevaluated_date": "2026-01-16T15:40:22.888859+00:00",
+  "reevaluated_date": "2026-01-22T21:57:09.736897+00:00",
   "source_predictions_file": "Alibaba_Group/Qwen3-VL_(32B-Thinking)_with_BM25_Search_Tool_predictions_20260110_132054.jsonl",
   "result_file_path": "Alibaba_Group/Qwen3-VL_(32B-Thinking)_with_BM25_Search_Tool_results_20260110_132054.json"
 }

eval/reevaluated_results/Alibaba_Group/Qwen3-VL_(8B-Thinking)_with_BM25_Search_Tool_results_20260110_132358.json

@@ -15,29 +15,29 @@
   "submission_date": "2026-01-10T13:23:58.123387+00:00",
   "results": {
     "overall": {
-      "semantic": 46.623859964827616,
+      "semantic": 47.3469387755102,
       "semantic_ci": [
-        42.1639790166841,
-        51.08374091297113
+        42.886491622183534,
+        51.807385928836865
       ],
-      "anls": 45.43424080850834,
-      "page_f1": 47.685529789738204,
-      "doc_f1": 69.57247828991316,
-      "kuiper": 48.63927855711413
+      "anls": 45.99676623639152,
+      "page_f1": 47.590158730158734,
+      "doc_f1": 69.43333333333334,
+      "kuiper": 50.216000000000136
     },
     "single_evidence": {
-      "semantic": 54.10405920610001,
-      "anls": 51.808356346955755,
-      "n": 364
+      "semantic": 54.794520547945204,
+      "anls": 52.49777090665106,
+      "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 16.399416909621,
-      "anls": 18.845466696680308,
+      "semantic": 17.006802721088444,
+      "anls": 18.782788634002245,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 43.01720688275311,
-      "anls": 43.73382844377694,
+      "semantic": 44.01760704281713,
+      "anls": 44.29299003944969,
       "n": 51
     },
     "by_domain": {
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 59.52380952380955,
+        "semantic": 63.775510204081634,
         "anls": 54.598842018196855,
         "n": 24
       },
       "Financial": {
-        "semantic": 29.94676131322094,
-        "anls": 29.932472094079802,
+        "semantic": 31.055900621118017,
+        "anls": 31.32939863081144,
         "n": 92
       },
       "Financial/Tax": {
@@ -68,12 +68,12 @@
       },
       "Government/Regulatory": {
         "semantic": 44.507164567954845,
-        "anls": 44.439106365198185,
+        "anls": 44.44710508540295,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 47.28720756595321,
-        "anls": 50.16056789323261,
+        "semantic": 49.77600796416127,
+        "anls": 52.59959228347652,
         "n": 41
       },
       "Legal": {
@@ -87,18 +87,23 @@
         "n": 25
       },
       "Misc": {
-        "semantic": 59.52380952380955,
+        "semantic": 61.64965986394556,
         "anls": 59.60305559882987,
         "n": 24
       },
+      "Other": {
+        "semantic": 100.0,
+        "anls": 100.0,
+        "n": 1
+      },
       "Reference": {
-        "semantic": 43.171114599686035,
-        "anls": 50.44584246011934,
+        "semantic": 44.15227629513344,
+        "anls": 50.50485651369861,
         "n": 52
       },
       "Reports": {
-        "semantic": 57.14285714285715,
-        "anls": 50.000135852648256,
+        "semantic": 55.78231292517006,
+        "anls": 49.929936422448826,
         "n": 75
       },
       "Technical": {
@@ -107,10 +112,10 @@
         "n": 23
       }
     },
-    "n_evaluated": 499,
-    "n_unmatched": 1767
+    "n_evaluated": 500,
+    "n_unmatched": 1766
   },
-  "reevaluated_date": "2026-01-16T15:41:38.978677+00:00",
+  "reevaluated_date": "2026-01-22T21:58:16.385696+00:00",
   "source_predictions_file": "Alibaba_Group/Qwen3-VL_(8B-Thinking)_with_BM25_Search_Tool_predictions_20260110_132358.jsonl",
   "result_file_path": "Alibaba_Group/Qwen3-VL_(8B-Thinking)_with_BM25_Search_Tool_results_20260110_132358.json"
 }

eval/reevaluated_results/Anthropic/Claude_Haiku_4.5_(2025-10-01)_results_20260109_130319.json

@@ -15,24 +15,24 @@
   "submission_date": "2026-01-09T13:03:19.649656+00:00",
   "results": {
     "overall": {
-      "semantic": 67.24489795918367,
+      "semantic": 68.16326530612247,
       "semantic_ci": [
-        63.00498530397807,
-        71.48481061438929
+        63.95120785444595,
+        72.37532275779898
       ],
-      "anls": 61.60747574238133,
+      "anls": 61.45994021171436,
       "page_f1": 72.02476190476192,
       "doc_f1": 88.24761904761905,
-      "kuiper": 51.64257028112422
+      "kuiper": 50.710843373494285
     },
     "single_evidence": {
-      "semantic": 72.54682694995807,
-      "anls": 66.17043328346772,
+      "semantic": 73.24573665082471,
+      "anls": 66.15560605815631,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 45.55393586005831,
-      "anls": 42.55992435652824,
+      "semantic": 47.9834791059281,
+      "anls": 41.74616449825649,
       "n": 84
     },
     "multi_evidence_multi_doc": {
@@ -42,7 +42,7 @@
     },
     "by_domain": {
       "Cases/Logs": {
-        "semantic": 71.42857142857143,
+        "semantic": 68.02721088435374,
         "anls": 63.92691050779287,
         "n": 15
       },
@@ -52,23 +52,23 @@
         "n": 22
       },
       "Events": {
-        "semantic": 87.15986394557822,
+        "semantic": 89.28571428571429,
         "anls": 72.62325637325637,
         "n": 24
       },
       "Financial": {
-        "semantic": 57.12067435669922,
-        "anls": 54.29593695395653,
+        "semantic": 59.89352262644188,
+        "anls": 54.1464804322174,
         "n": 92
       },
       "Financial/Tax": {
-        "semantic": 73.34183673469387,
+        "semantic": 76.53061224489795,
         "anls": 68.77016129032259,
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 64.04689535388624,
-        "anls": 62.779826338896896,
+        "semantic": 62.961354754667845,
+        "anls": 61.67613431405342,
         "n": 47
       },
       "HR/Employment": {
@@ -82,7 +82,7 @@
         "n": 43
       },
       "Media/Publishing": {
-        "semantic": 48.97959183673469,
+        "semantic": 51.02040816326531,
         "anls": 41.69842237151431,
         "n": 25
       },
@@ -97,17 +97,17 @@
         "n": 1
       },
       "Reference": {
-        "semantic": 61.813186813186825,
-        "anls": 64.3484267705628,
+        "semantic": 64.75667189952904,
+        "anls": 64.40744082414207,
         "n": 52
       },
       "Reports": {
-        "semantic": 79.59183673469387,
-        "anls": 65.36479556179735,
+        "semantic": 80.95238095238096,
+        "anls": 65.21528928243777,
         "n": 75
       },
       "Technical": {
-        "semantic": 77.63975155279503,
+        "semantic": 73.20319432120675,
         "anls": 64.75817505570946,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T15:42:51.557033+00:00",
+  "reevaluated_date": "2026-01-22T21:59:23.189394+00:00",
   "source_predictions_file": "Anthropic/Claude_Haiku_4.5_(2025-10-01)_predictions_20260109_130319.jsonl",
   "result_file_path": "Anthropic/Claude_Haiku_4.5_(2025-10-01)_results_20260109_130319.json"
 }

eval/reevaluated_results/Anthropic/Claude_Sonnet_4.5_(2025-09-29)_with_BM25_Search_Tool_results_20260109_125816.json

@@ -15,24 +15,24 @@
   "submission_date": "2026-01-09T12:58:16.611348+00:00",
   "results": {
     "overall": {
-      "semantic": 79.18367346938776,
+      "semantic": 80.61224489795919,
       "semantic_ci": [
-        75.45467776603695,
-        82.91266917273857
+        76.96923621662829,
+        84.25525357929008
       ],
-      "anls": 71.74787642305597,
+      "anls": 72.85884587946542,
       "page_f1": 79.12333333333333,
       "doc_f1": 92.98636363636363,
-      "kuiper": 35.64777327935229
+      "kuiper": 35.0526315789474
     },
     "single_evidence": {
-      "semantic": 82.05199888174448,
-      "anls": 74.4069329276591,
+      "semantic": 83.58960022365112,
+      "anls": 75.8229313455021,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 64.99028182701652,
-      "anls": 60.572680530273026,
+      "semantic": 66.81243926141885,
+      "anls": 61.03274369327326,
       "n": 84
     },
     "multi_evidence_multi_doc": {
@@ -42,8 +42,8 @@
     },
     "by_domain": {
       "Cases/Logs": {
-        "semantic": 78.2312925170068,
-        "anls": 69.51092117758785,
+        "semantic": 85.0340136054422,
+        "anls": 76.17758784425452,
         "n": 15
       },
       "Education": {
@@ -57,27 +57,27 @@
         "n": 24
       },
       "Financial": {
-        "semantic": 72.09405501330967,
-        "anls": 66.81148919563769,
+        "semantic": 73.75776397515527,
+        "anls": 69.52764532211991,
         "n": 92
       },
       "Financial/Tax": {
-        "semantic": 70.1530612244898,
+        "semantic": 76.53061224489795,
         "anls": 76.26728110599078,
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 87.92878853669127,
-        "anls": 74.90457714355891,
+        "semantic": 86.84324793747285,
+        "anls": 76.0805203770741,
         "n": 47
       },
       "HR/Employment": {
         "semantic": 77.15281234444997,
-        "anls": 72.85160396238213,
+        "anls": 74.07111615750408,
         "n": 41
       },
       "Legal": {
-        "semantic": 90.17560512577124,
+        "semantic": 91.36212624584718,
         "anls": 72.74221043114129,
         "n": 43
       },
@@ -93,21 +93,21 @@
       },
       "Other": {
         "semantic": 0.0,
-        "anls": 0.0,
+        "anls": 76.92307692307692,
         "n": 1
       },
       "Reference": {
-        "semantic": 70.6436420722135,
-        "anls": 72.21619612753193,
+        "semantic": 72.60596546310833,
+        "anls": 72.8846701655712,
         "n": 52
       },
       "Reports": {
-        "semantic": 85.0340136054422,
-        "anls": 74.0536995032274,
+        "semantic": 87.07482993197281,
+        "anls": 73.90230357912102,
         "n": 75
       },
       "Technical": {
-        "semantic": 70.9849157054126,
+        "semantic": 75.42147293700089,
         "anls": 60.23577215564363,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T15:45:07.483636+00:00",
+  "reevaluated_date": "2026-01-22T22:00:12.094993+00:00",
   "source_predictions_file": "Anthropic/Claude_Sonnet_4.5_(2025-09-29)_with_BM25_Search_Tool_predictions_20260109_125816.jsonl",
   "result_file_path": "Anthropic/Claude_Sonnet_4.5_(2025-09-29)_with_BM25_Search_Tool_results_20260109_125816.json"
 }

eval/reevaluated_results/Google/Gemini_2.5_Flash_with_BM25_Search_Tool_results_20260109_182559.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-09T18:25:59.636344+00:00",
   "results": {
     "overall": {
-      "semantic": 58.06122448979592,
+      "semantic": 58.46938775510204,
       "semantic_ci": [
-        53.63195829834202,
-        62.49049068124981
+        54.04525386645418,
+        62.89352164374991
       ],
-      "anls": 55.486869478144165,
+      "anls": 55.989429815086645,
       "page_f1": 60.9663492063492,
       "doc_f1": 78.82920634920634,
-      "kuiper": 45.71999999999974
+      "kuiper": 46.50600000000012
     },
     "single_evidence": {
-      "semantic": 64.71903830025161,
-      "anls": 60.42073737602547,
+      "semantic": 65.27816606094493,
+      "anls": 60.972189892385046,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 30.976676384839653,
-      "anls": 33.035531481450356,
+      "semantic": 30.36929057337221,
+      "anls": 33.63076957668846,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 55.022008803521395,
+      "semantic": 56.022408963585434,
       "anls": 57.1545284780579,
       "n": 51
     },
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 82.90816326530613,
-        "anls": 76.85643564356435,
+        "semantic": 85.0340136054422,
+        "anls": 78.93976897689768,
         "n": 24
       },
       "Financial": {
-        "semantic": 42.701863354037265,
-        "anls": 40.952902757926644,
+        "semantic": 42.14729370008874,
+        "anls": 43.10328860396467,
         "n": 92
       },
       "Financial/Tax": {
@@ -68,26 +68,26 @@
       },
       "Government/Regulatory": {
         "semantic": 66.21797655232307,
-        "anls": 67.70262933196864,
+        "anls": 67.7106280521734,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 59.731209556993534,
+        "semantic": 60.97560975609757,
         "anls": 60.95035529628296,
         "n": 41
       },
       "Legal": {
-        "semantic": 55.766492643569066,
+        "semantic": 53.393450403417184,
         "anls": 51.45105745077384,
         "n": 43
       },
       "Media/Publishing": {
-        "semantic": 48.97959183673469,
+        "semantic": 53.06122448979592,
         "anls": 54.40739778239778,
         "n": 25
       },
       "Misc": {
-        "semantic": 76.53061224489795,
+        "semantic": 78.65646258503402,
         "anls": 73.82172131147541,
         "n": 24
       },
@@ -97,17 +97,17 @@
         "n": 1
       },
       "Reference": {
-        "semantic": 60.83202511773942,
-        "anls": 64.46714691613596,
+        "semantic": 62.79434850863422,
+        "anls": 64.52616096971524,
         "n": 52
       },
       "Reports": {
-        "semantic": 54.42176870748299,
+        "semantic": 55.102040816326536,
         "anls": 45.47473759975617,
         "n": 75
       },
       "Technical": {
-        "semantic": 51.02040816326531,
+        "semantic": 48.802129547471154,
         "anls": 35.96181299748582,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T15:48:34.415858+00:00",
+  "reevaluated_date": "2026-01-22T22:01:09.960296+00:00",
   "source_predictions_file": "Google/Gemini_2.5_Flash_with_BM25_Search_Tool_predictions_20260109_182559.jsonl",
   "result_file_path": "Google/Gemini_2.5_Flash_with_BM25_Search_Tool_results_20260109_182559.json"
 }

eval/reevaluated_results/Google/Gemini_2.5_Flash_with_File_Search_results_20260103_221253.json

@@ -0,0 +1,120 @@
+{
+  "model_name": "Gemini 2.5 Flash with File Search",
+  "organization": "Google",
+  "description": "Managed, single-shot retrieval mechanism.",
+  "link": "https://ai.google.dev/gemini-api/docs/file-search",
+  "tags": [
+    "Conventional RAG",
+    "Semantic Search Tool"
+  ],
+  "submitted_by": "Borchmann",
+  "metadata": {
+    "model_type": "api"
+  },
+  "submission_date": "2026-01-03T22:12:53.645813+00:00",
+  "results": {
+    "overall": {
+      "semantic": 71.83673469387755,
+      "semantic_ci": [
+        67.75383041519983,
+        75.91963897255526
+      ],
+      "anls": 56.38605375030021,
+      "page_f1": 52.15333333333333,
+      "doc_f1": 80.91445887445887,
+      "kuiper": 14.495999999999947
+    },
+    "single_evidence": {
+      "semantic": 74.0844282918647,
+      "anls": 59.70060588908858,
+      "n": 365
+    },
+    "multi_evidence_same_doc": {
+      "semantic": 61.34596695821186,
+      "anls": 44.2307193586438,
+      "n": 84
+    },
+    "multi_evidence_multi_doc": {
+      "semantic": 73.02921168467388,
+      "anls": 52.68480979424892,
+      "n": 51
+    },
+    "by_domain": {
+      "Cases/Logs": {
+        "semantic": 78.2312925170068,
+        "anls": 76.85185185185186,
+        "n": 15
+      },
+      "Education": {
+        "semantic": 83.4879406307978,
+        "anls": 59.32605273514364,
+        "n": 22
+      },
+      "Events": {
+        "semantic": 87.15986394557822,
+        "anls": 67.94733044733044,
+        "n": 24
+      },
+      "Financial": {
+        "semantic": 64.88464951197871,
+        "anls": 53.16591091793465,
+        "n": 92
+      },
+      "Financial/Tax": {
+        "semantic": 60.58673469387755,
+        "anls": 45.19230769230769,
+        "n": 16
+      },
+      "Government/Regulatory": {
+        "semantic": 68.38905775075989,
+        "anls": 52.69755357661292,
+        "n": 47
+      },
+      "HR/Employment": {
+        "semantic": 70.93081134892981,
+        "anls": 56.2715906011529,
+        "n": 41
+      },
+      "Legal": {
+        "semantic": 85.4295206454675,
+        "anls": 63.338167000957704,
+        "n": 43
+      },
+      "Media/Publishing": {
+        "semantic": 65.30612244897961,
+        "anls": 61.1117685382045,
+        "n": 25
+      },
+      "Misc": {
+        "semantic": 70.1530612244898,
+        "anls": 60.28005464480874,
+        "n": 24
+      },
+      "Other": {
+        "semantic": 51.02040816326531,
+        "anls": 59.09090909090908,
+        "n": 1
+      },
+      "Reference": {
+        "semantic": 72.60596546310833,
+        "anls": 56.51193666615843,
+        "n": 52
+      },
+      "Reports": {
+        "semantic": 69.38775510204084,
+        "anls": 50.21750576702014,
+        "n": 75
+      },
+      "Technical": {
+        "semantic": 75.42147293700089,
+        "anls": 54.08686051995165,
+        "n": 23
+      }
+    },
+    "n_evaluated": 500,
+    "n_unmatched": 0
+  },
+  "reevaluated_date": "2026-01-22T22:02:26.923804+00:00",
+  "source_predictions_file": "Google/Gemini_2.5_Flash_with_File_Search_predictions_20260103_221253.jsonl",
+  "result_file_path": "Google/Gemini_2.5_Flash_with_File_Search_results_20260103_221253.json"
+}

eval/reevaluated_results/Google/Gemini_2.5_Pro_with_BM25_Search_Tool_results_20260109_183030.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-09T18:30:30.608183+00:00",
   "results": {
     "overall": {
-      "semantic": 60.20408163265305,
+      "semantic": 60.10204081632653,
       "semantic_ci": [
-        55.80507165182195,
-        64.60309161348415
+        55.70140355733895,
+        64.50267807531411
       ],
-      "anls": 55.97919862778078,
+      "anls": 56.41686335308852,
       "page_f1": 60.299220779220775,
       "doc_f1": 74.23636363636363,
-      "kuiper": 40.40399999999971
+      "kuiper": 38.901999999999795
     },
     "single_evidence": {
-      "semantic": 68.0738048644115,
-      "anls": 61.04578922425167,
+      "semantic": 67.0953312831982,
+      "anls": 61.720866213550785,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 31.584062196307094,
-      "anls": 36.778404555871965,
+      "semantic": 34.62099125364433,
+      "anls": 36.450181479201696,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 51.02040816326531,
+      "semantic": 52.02080832332933,
       "anls": 51.343142438142856,
       "n": 51
     },
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 68.02721088435374,
-        "anls": 62.67819322254806,
+        "semantic": 72.27891156462587,
+        "anls": 64.7615265558814,
         "n": 24
       },
       "Financial": {
-        "semantic": 47.6929902395741,
-        "anls": 48.11929370300614,
+        "semantic": 50.46583850931676,
+        "anls": 49.206250224745276,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,23 +67,23 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 68.38905775075989,
-        "anls": 64.23333377770668,
+        "semantic": 65.13243595310465,
+        "anls": 62.58648616221405,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 59.731209556993534,
-        "anls": 48.92979153124233,
+        "semantic": 58.4868093578895,
+        "anls": 49.471796951296525,
         "n": 41
       },
       "Legal": {
-        "semantic": 68.81822496440435,
+        "semantic": 66.44518272425249,
         "anls": 60.44220952048519,
         "n": 43
       },
       "Media/Publishing": {
-        "semantic": 48.97959183673469,
-        "anls": 52.95641025641026,
+        "semantic": 44.897959183673464,
+        "anls": 53.44125874125874,
         "n": 25
       },
       "Misc": {
@@ -93,21 +93,21 @@
       },
       "Other": {
         "semantic": 0.0,
-        "anls": 0.0,
+        "anls": 76.92307692307692,
         "n": 1
       },
       "Reference": {
-        "semantic": 64.75667189952904,
-        "anls": 63.5134680018838,
+        "semantic": 66.71899529042385,
+        "anls": 64.18194203992306,
         "n": 52
       },
       "Reports": {
-        "semantic": 62.58503401360544,
-        "anls": 54.81415365192609,
+        "semantic": 61.904761904761905,
+        "anls": 54.81691417642573,
         "n": 75
       },
       "Technical": {
-        "semantic": 46.58385093167702,
+        "semantic": 44.365572315882865,
         "anls": 40.50127359810298,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T15:51:00.509214+00:00",
+  "reevaluated_date": "2026-01-22T22:03:24.876553+00:00",
   "source_predictions_file": "Google/Gemini_2.5_Pro_with_BM25_Search_Tool_predictions_20260109_183030.jsonl",
   "result_file_path": "Google/Gemini_2.5_Pro_with_BM25_Search_Tool_results_20260109_183030.json"
 }

eval/reevaluated_results/Google/Gemini_2.5_Pro_with_File_Search_results_20260103_221943.json

@@ -0,0 +1,120 @@
+{
+  "model_name": "Gemini 2.5 Pro with File Search",
+  "organization": "Google",
+  "description": "Managed, single-shot retrieval mechanism.",
+  "link": "https://ai.google.dev/gemini-api/docs/file-search",
+  "tags": [
+    "Conventional RAG",
+    "Semantic Search Tool"
+  ],
+  "submitted_by": "Borchmann",
+  "metadata": {
+    "model_type": "api"
+  },
+  "submission_date": "2026-01-03T22:19:43.085381+00:00",
+  "results": {
+    "overall": {
+      "semantic": 73.06122448979592,
+      "semantic_ci": [
+        69.02799722537132,
+        77.0944517542205
+      ],
+      "anls": 67.06445662551177,
+      "page_f1": 60.936190476190475,
+      "doc_f1": 87.83333333333334,
+      "kuiper": 15.25
+    },
+    "single_evidence": {
+      "semantic": 75.06290187307802,
+      "anls": 69.51189266445841,
+      "n": 365
+    },
+    "multi_evidence_same_doc": {
+      "semantic": 70.45675413022352,
+      "anls": 62.81319563063059,
+      "n": 84
+    },
+    "multi_evidence_multi_doc": {
+      "semantic": 63.025210084033624,
+      "anls": 56.5505697501097,
+      "n": 51
+    },
+    "by_domain": {
+      "Cases/Logs": {
+        "semantic": 74.82993197278913,
+        "anls": 72.85185185185185,
+        "n": 15
+      },
+      "Education": {
+        "semantic": 83.4879406307978,
+        "anls": 66.82605273514365,
+        "n": 22
+      },
+      "Events": {
+        "semantic": 80.78231292517005,
+        "anls": 63.51613609636826,
+        "n": 24
+      },
+      "Financial": {
+        "semantic": 73.75776397515527,
+        "anls": 71.81492329679233,
+        "n": 92
+      },
+      "Financial/Tax": {
+        "semantic": 44.642857142857146,
+        "anls": 48.56036324786325,
+        "n": 16
+      },
+      "Government/Regulatory": {
+        "semantic": 82.50108554059923,
+        "anls": 71.42658144393866,
+        "n": 47
+      },
+      "HR/Employment": {
+        "semantic": 74.66401194624191,
+        "anls": 67.74502588179764,
+        "n": 41
+      },
+      "Legal": {
+        "semantic": 77.12387280493593,
+        "anls": 68.13676633444075,
+        "n": 43
+      },
+      "Media/Publishing": {
+        "semantic": 69.38775510204084,
+        "anls": 72.05791173717081,
+        "n": 25
+      },
+      "Misc": {
+        "semantic": 72.27891156462587,
+        "anls": 69.65505464480874,
+        "n": 24
+      },
+      "Other": {
+        "semantic": 51.02040816326531,
+        "anls": 59.09090909090908,
+        "n": 1
+      },
+      "Reference": {
+        "semantic": 71.62480376766091,
+        "anls": 72.50339016950552,
+        "n": 52
+      },
+      "Reports": {
+        "semantic": 67.34693877551022,
+        "anls": 59.42273394165809,
+        "n": 75
+      },
+      "Technical": {
+        "semantic": 68.76663708961844,
+        "anls": 53.79711579945216,
+        "n": 23
+      }
+    },
+    "n_evaluated": 500,
+    "n_unmatched": 0
+  },
+  "reevaluated_date": "2026-01-22T22:04:21.890237+00:00",
+  "source_predictions_file": "Google/Gemini_2.5_Pro_with_File_Search_predictions_20260103_221943.jsonl",
+  "result_file_path": "Google/Gemini_2.5_Pro_with_File_Search_results_20260103_221943.json"
+}

eval/reevaluated_results/Google/Gemini_3.0_Pro_(Preview)_with_File_Search_results_20260104_120431.json

@@ -0,0 +1,120 @@
+{
+  "model_name": "Gemini 3 Pro (Preview) with File Search",
+  "organization": "Google",
+  "description": "Managed, single-shot retrieval mechanism.",
+  "link": "https://ai.google.dev/gemini-api/docs/file-search",
+  "tags": [
+    "Conventional RAG",
+    "Semantic Search Tool"
+  ],
+  "submitted_by": "Borchmann",
+  "metadata": {
+    "model_type": "api"
+  },
+  "submission_date": "2026-01-04T12:04:31.393913+00:00",
+  "results": {
+    "overall": {
+      "semantic": 78.57142857142857,
+      "semantic_ci": [
+        74.80745735419877,
+        82.33539978865836
+      ],
+      "anls": 68.92843881871933,
+      "page_f1": 70.0995238095238,
+      "doc_f1": 94.17333333333333,
+      "kuiper": 12.06600000000001
+    },
+    "single_evidence": {
+      "semantic": 80.09505171931788,
+      "anls": 71.30190923977649,
+      "n": 365
+    },
+    "multi_evidence_same_doc": {
+      "semantic": 74.1010689990282,
+      "anls": 67.2325874522391,
+      "n": 84
+    },
+    "multi_evidence_multi_doc": {
+      "semantic": 75.03001200480193,
+      "anls": 54.73500374221878,
+      "n": 51
+    },
+    "by_domain": {
+      "Cases/Logs": {
+        "semantic": 85.0340136054422,
+        "anls": 74.51761001296605,
+        "n": 15
+      },
+      "Education": {
+        "semantic": 90.44526901669758,
+        "anls": 62.23980564889655,
+        "n": 22
+      },
+      "Events": {
+        "semantic": 87.15986394557822,
+        "anls": 63.97470042977855,
+        "n": 24
+      },
+      "Financial": {
+        "semantic": 77.08518189884651,
+        "anls": 71.47919047104104,
+        "n": 92
+      },
+      "Financial/Tax": {
+        "semantic": 76.53061224489795,
+        "anls": 75.04133597883597,
+        "n": 16
+      },
+      "Government/Regulatory": {
+        "semantic": 80.33000434216238,
+        "anls": 67.12054458180654,
+        "n": 47
+      },
+      "HR/Employment": {
+        "semantic": 74.66401194624191,
+        "anls": 70.8778954315538,
+        "n": 41
+      },
+      "Legal": {
+        "semantic": 80.68343616516374,
+        "anls": 64.51676230745998,
+        "n": 43
+      },
+      "Media/Publishing": {
+        "semantic": 65.30612244897961,
+        "anls": 71.26169358330307,
+        "n": 25
+      },
+      "Misc": {
+        "semantic": 85.0340136054422,
+        "anls": 71.74962035802997,
+        "n": 24
+      },
+      "Other": {
+        "semantic": 0.0,
+        "anls": 29.54545454545454,
+        "n": 1
+      },
+      "Reference": {
+        "semantic": 80.45525902668759,
+        "anls": 76.46179993780217,
+        "n": 52
+      },
+      "Reports": {
+        "semantic": 80.95238095238096,
+        "anls": 69.86245910414694,
+        "n": 75
+      },
+      "Technical": {
+        "semantic": 59.89352262644188,
+        "anls": 47.01665330524952,
+        "n": 23
+      }
+    },
+    "n_evaluated": 500,
+    "n_unmatched": 0
+  },
+  "reevaluated_date": "2026-01-22T22:05:27.215227+00:00",
+  "source_predictions_file": "Google/Gemini_3.0_Pro_(Preview)_with_File_Search_predictions_20260104_120431.jsonl",
+  "result_file_path": "Google/Gemini_3.0_Pro_(Preview)_with_File_Search_results_20260104_120431.json"
+}

eval/reevaluated_results/Humanity/Human_with_BM25_Search_Tool_results_20260117_193634.json

@@ -0,0 +1,121 @@
+{
+  "model_name": "Human with BM25 Search Tool",
+  "organization": "Humanity",
+  "description": "Human equipped with the same search engine as agentic baselines.",
+  "link": "",
+  "tags": [
+    "Agentic",
+    "Sparse Search Tool",
+    "Vision and Language"
+  ],
+  "submitted_by": "Borchmann",
+  "metadata": {
+    "model_type": "api"
+  },
+  "submission_date": "2026-01-17T19:36:34.967206+00:00",
+  "results": {
+    "overall": {
+      "semantic": 82.14285714285717,
+      "semantic_ci": [
+        78.5991954078934,
+        85.68651887782092
+      ],
+      "anls": 83.70514272504292,
+      "page_f1": 79.25590111642744,
+      "doc_f1": 93.42612554112554,
+      "kuiper": 8.696969696969703
+    },
+    "single_evidence": {
+      "semantic": 84.14872798434443,
+      "anls": 85.91721757063848,
+      "n": 365
+    },
+    "multi_evidence_same_doc": {
+      "semantic": 79.56754130223517,
+      "anls": 82.31765689212445,
+      "n": 84
+    },
+    "multi_evidence_multi_doc": {
+      "semantic": 72.02881152460986,
+      "anls": 70.15889745686191,
+      "n": 51
+    },
+    "by_domain": {
+      "Cases/Logs": {
+        "semantic": 68.02721088435374,
+        "anls": 72.72727272727272,
+        "n": 15
+      },
+      "Education": {
+        "semantic": 100.0,
+        "anls": 92.81192695119195,
+        "n": 22
+      },
+      "Events": {
+        "semantic": 87.15986394557822,
+        "anls": 83.66935483870968,
+        "n": 24
+      },
+      "Financial": {
+        "semantic": 78.19432120674357,
+        "anls": 79.59127298509043,
+        "n": 92
+      },
+      "Financial/Tax": {
+        "semantic": 82.90816326530613,
+        "anls": 83.28405017921146,
+        "n": 16
+      },
+      "Government/Regulatory": {
+        "semantic": 84.67216673903604,
+        "anls": 87.11278292004198,
+        "n": 47
+      },
+      "HR/Employment": {
+        "semantic": 87.10801393728221,
+        "anls": 81.26741515002432,
+        "n": 41
+      },
+      "Legal": {
+        "semantic": 88.9890840056953,
+        "anls": 85.09035538105306,
+        "n": 43
+      },
+      "Media/Publishing": {
+        "semantic": 75.51020408163265,
+        "anls": 79.91696395686839,
+        "n": 25
+      },
+      "Misc": {
+        "semantic": 74.4047619047619,
+        "anls": 79.0967077930904,
+        "n": 24
+      },
+      "Other": {
+        "semantic": 0.0,
+        "anls": 76.92307692307692,
+        "n": 1
+      },
+      "Reference": {
+        "semantic": 83.39874411302984,
+        "anls": 92.19517190376465,
+        "n": 52
+      },
+      "Reports": {
+        "semantic": 78.2312925170068,
+        "anls": 83.77634694564375,
+        "n": 75
+      },
+      "Technical": {
+        "semantic": 84.29458740017748,
+        "anls": 83.52609662978936,
+        "n": 23
+      }
+    },
+    "n_evaluated": 500,
+    "n_unmatched": 0
+  },
+  "reevaluated_date": "2026-01-22T22:06:05.498305+00:00",
+  "source_predictions_file": "Humanity/Human_with_BM25_Search_Tool_predictions_20260117_193634.jsonl",
+  "result_file_path": "Humanity/Human_with_BM25_Search_Tool_results_20260117_193634.json"
+}

eval/reevaluated_results/Humanity/Human_with_Oracle_Retriever_results_20260122_214532.json

@@ -0,0 +1,119 @@
+{
+  "model_name": "Human with Oracle Retriever",
+  "organization": "Humanity",
+  "description": "Human given gold standard evidence pages.",
+  "link": "",
+  "tags": [
+    "Vision and Language"
+  ],
+  "submitted_by": "Borchmann",
+  "metadata": {
+    "model_type": "api"
+  },
+  "submission_date": "2026-01-22T21:45:32.545870+00:00",
+  "results": {
+    "overall": {
+      "semantic": 99.38775510204081,
+      "semantic_ci": [
+        97.96443309097894,
+        100.0
+      ],
+      "anls": 93.7121419059059,
+      "page_f1": 88.62341991341991,
+      "doc_f1": 97.18112554112554,
+      "kuiper": 4.630303030303009
+    },
+    "single_evidence": {
+      "semantic": 99.24517752306402,
+      "anls": 94.03593158263021,
+      "n": 365
+    },
+    "multi_evidence_same_doc": {
+      "semantic": 100.0,
+      "anls": 95.10936020828757,
+      "n": 84
+    },
+    "multi_evidence_multi_doc": {
+      "semantic": 98.03921568627453,
+      "anls": 89.09352289797572,
+      "n": 51
+    },
+    "by_domain": {
+      "Cases/Logs": {
+        "semantic": 95.23809523809524,
+        "anls": 93.33333333333333,
+        "n": 15
+      },
+      "Education": {
+        "semantic": 100.0,
+        "anls": 92.99351054603811,
+        "n": 22
+      },
+      "Events": {
+        "semantic": 100.0,
+        "anls": 97.74763766699252,
+        "n": 24
+      },
+      "Financial": {
+        "semantic": 98.15882874889085,
+        "anls": 92.61042950143054,
+        "n": 92
+      },
+      "Financial/Tax": {
+        "semantic": 100.0,
+        "anls": 96.38888888888889,
+        "n": 16
+      },
+      "Government/Regulatory": {
+        "semantic": 100.0,
+        "anls": 96.77157151500082,
+        "n": 47
+      },
+      "HR/Employment": {
+        "semantic": 97.06321553011448,
+        "anls": 91.56441076843375,
+        "n": 41
+      },
+      "Legal": {
+        "semantic": 100.0,
+        "anls": 95.3032819893285,
+        "n": 43
+      },
+      "Media/Publishing": {
+        "semantic": 93.87755102040816,
+        "anls": 90.74640522875816,
+        "n": 25
+      },
+      "Misc": {
+        "semantic": 97.78911564625852,
+        "anls": 90.94982803662161,
+        "n": 24
+      },
+      "Other": {
+        "semantic": 0.0,
+        "anls": 76.92307692307692,
+        "n": 1
+      },
+      "Reference": {
+        "semantic": 100.0,
+        "anls": 96.68889612023322,
+        "n": 52
+      },
+      "Reports": {
+        "semantic": 98.63945578231294,
+        "anls": 93.02938751681923,
+        "n": 75
+      },
+      "Technical": {
+        "semantic": 100.0,
+        "anls": 89.9146569474778,
+        "n": 23
+      }
+    },
+    "n_evaluated": 500,
+    "n_unmatched": 0
+  },
+  "reevaluated_date": "2026-01-22T22:06:32.856334+00:00",
+  "source_predictions_file": "Humanity/Human_with_Oracle_Retriever_predictions_20260122_214532.jsonl",
+  "result_file_path": "Humanity/Human_with_Oracle_Retriever_results_20260122_214532.json"
+}

eval/reevaluated_results/OpenAI/GPT-4.1_(2025-04-14)_with_BM25_Search_Tool_results_20260109_153221.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-09T15:32:21.908816+00:00",
   "results": {
     "overall": {
-      "semantic": 58.6734693877551,
+      "semantic": 60.00000000000001,
       "semantic_ci": [
-        54.252012440670164,
-        63.09492633484003
+        55.59775423620991,
+        64.4022457637901
       ],
-      "anls": 53.29254644474454,
+      "anls": 54.023238374203174,
       "page_f1": 64.14190476190477,
       "doc_f1": 82.82666666666667,
-      "kuiper": 42.51800000000039
+      "kuiper": 43.20000000000029
     },
     "single_evidence": {
-      "semantic": 65.41794800111826,
-      "anls": 58.84025149533676,
+      "semantic": 66.25663964215825,
+      "anls": 59.82892458572295,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 37.65792031098154,
-      "anls": 35.98946829890793,
+      "semantic": 42.5170068027211,
+      "anls": 36.04280504579306,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 45.01800720288115,
+      "semantic": 44.01760704281713,
       "anls": 42.087570381688025,
       "n": 51
     },
@@ -52,33 +52,33 @@
         "n": 22
       },
       "Events": {
-        "semantic": 70.1530612244898,
+        "semantic": 74.4047619047619,
         "anls": 67.55050505050505,
         "n": 24
       },
       "Financial": {
-        "semantic": 42.701863354037265,
-        "anls": 43.62404525327831,
+        "semantic": 43.81100266193434,
+        "anls": 46.01082061559715,
         "n": 92
       },
       "Financial/Tax": {
-        "semantic": 73.34183673469387,
+        "semantic": 76.53061224489795,
         "anls": 64.58333333333334,
         "n": 16
       },
       "Government/Regulatory": {
         "semantic": 57.533651758575765,
-        "anls": 51.52629513848961,
+        "anls": 51.55429065920628,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 70.93081134892981,
-        "anls": 55.117501174925685,
+        "semantic": 72.17521154803384,
+        "anls": 56.33701337004763,
         "n": 41
       },
       "Legal": {
-        "semantic": 60.51257712387282,
-        "anls": 55.94315245478037,
+        "semantic": 61.69909824394873,
+        "anls": 57.105943152454785,
         "n": 43
       },
       "Media/Publishing": {
@@ -93,21 +93,21 @@
       },
       "Other": {
         "semantic": 0.0,
-        "anls": 0.0,
+        "anls": 76.92307692307692,
         "n": 1
       },
       "Reference": {
         "semantic": 64.75667189952904,
-        "anls": 60.011945621794936,
+        "anls": 60.26278174219628,
         "n": 52
       },
       "Reports": {
-        "semantic": 61.22448979591838,
-        "anls": 47.26331129213486,
+        "semantic": 62.58503401360544,
+        "anls": 46.65638178197955,
         "n": 75
       },
       "Technical": {
-        "semantic": 66.54835847382431,
+        "semantic": 75.42147293700089,
         "anls": 61.60068502092203,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:01:46.742921+00:00",
+  "reevaluated_date": "2026-01-22T22:07:42.289399+00:00",
   "source_predictions_file": "OpenAI/GPT-4.1_(2025-04-14)_with_BM25_Search_Tool_predictions_20260109_153221.jsonl",
   "result_file_path": "OpenAI/GPT-4.1_(2025-04-14)_with_BM25_Search_Tool_results_20260109_153221.json"
 }

eval/reevaluated_results/OpenAI/GPT-4.1_Nano_(2025-04-14)_with_BM25_Search_Tool_results_20260109_153812.json

@@ -23,15 +23,15 @@
       "anls": 19.21201395702391,
       "page_f1": 27.60809523809524,
       "doc_f1": 40.18095238095238,
-      "kuiper": 27.428000000000136
+      "kuiper": 28.60000000000006
     },
     "single_evidence": {
-      "semantic": 23.623147889292703,
+      "semantic": 24.042493709812685,
       "anls": 22.4105437044892,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 7.896015549076768,
+      "semantic": 6.073858114674442,
       "anls": 9.153597726228908,
       "n": 84
     },
@@ -77,7 +77,7 @@
         "n": 41
       },
       "Legal": {
-        "semantic": 26.10346464167062,
+        "semantic": 23.730422401518744,
         "anls": 20.54263565891473,
         "n": 43
       },
@@ -97,17 +97,17 @@
         "n": 1
       },
       "Reference": {
-        "semantic": 17.66091051805338,
+        "semantic": 16.679748822605976,
         "anls": 20.3827772417516,
         "n": 52
       },
       "Reports": {
-        "semantic": 23.12925170068026,
+        "semantic": 23.809523809523817,
         "anls": 19.284216647617285,
         "n": 75
       },
       "Technical": {
-        "semantic": 22.182786157941443,
+        "semantic": 26.619343389529732,
         "anls": 27.075249588209658,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:02:35.269375+00:00",
+  "reevaluated_date": "2026-01-22T22:08:24.315093+00:00",
   "source_predictions_file": "OpenAI/GPT-4.1_Nano_(2025-04-14)_with_BM25_Search_Tool_predictions_20260109_153812.jsonl",
   "result_file_path": "OpenAI/GPT-4.1_Nano_(2025-04-14)_with_BM25_Search_Tool_results_20260109_153812.json"
 }

eval/reevaluated_results/OpenAI/GPT-5.2_(2025-12-11)_with_BM25_Search_Tool_results_20260109_151912.json

@@ -15,29 +15,29 @@
   "submission_date": "2026-01-09T15:19:12.016451+00:00",
   "results": {
     "overall": {
-      "semantic": 67.04081632653062,
+      "semantic": 67.75510204081634,
       "semantic_ci": [
-        62.79494568958393,
-        71.2866869634773
+        63.53045246104174,
+        71.97975162059093
       ],
-      "anls": 57.28438090955278,
+      "anls": 57.79787184551629,
       "page_f1": 67.62380952380951,
       "doc_f1": 83.72666666666666,
-      "kuiper": 63.51399999999979
+      "kuiper": 64.75199999999975
     },
     "single_evidence": {
-      "semantic": 72.2672630696114,
-      "anls": 61.48347787804644,
+      "semantic": 73.24573665082471,
+      "anls": 62.10194433589725,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 50.41302235179786,
-      "anls": 40.75979036064321,
+      "semantic": 51.62779397473275,
+      "anls": 41.00509049000287,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 57.022809123649466,
-      "anls": 54.44899292147869,
+      "semantic": 55.022008803521395,
+      "anls": 54.65291449010614,
       "n": 51
     },
     "by_domain": {
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 72.27891156462587,
+        "semantic": 76.53061224489795,
         "anls": 57.55050505050505,
         "n": 24
       },
       "Financial": {
-        "semantic": 55.4569653948536,
-        "anls": 49.679264550051975,
+        "semantic": 57.67524401064772,
+        "anls": 51.01775086245134,
         "n": 92
       },
       "Financial/Tax": {
@@ -68,46 +68,46 @@
       },
       "Government/Regulatory": {
         "semantic": 68.38905775075989,
-        "anls": 58.551919442177,
+        "anls": 58.57991496289369,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 62.220009955201604,
+        "semantic": 60.97560975609757,
         "anls": 44.265703074651974,
         "n": 41
       },
       "Legal": {
-        "semantic": 81.86995728523969,
-        "anls": 66.19399979865096,
+        "semantic": 83.05647840531563,
+        "anls": 66.93395751535287,
         "n": 43
       },
       "Media/Publishing": {
-        "semantic": 38.77551020408162,
+        "semantic": 36.734693877551,
         "anls": 35.05751747729549,
         "n": 25
       },
       "Misc": {
-        "semantic": 87.15986394557822,
+        "semantic": 85.0340136054422,
         "anls": 82.5164707977208,
         "n": 24
       },
       "Other": {
         "semantic": 0.0,
-        "anls": 0.0,
+        "anls": 76.92307692307692,
         "n": 1
       },
       "Reference": {
-        "semantic": 74.56828885400316,
-        "anls": 67.62508443509842,
+        "semantic": 77.51177394034536,
+        "anls": 68.29355847313767,
         "n": 52
       },
       "Reports": {
-        "semantic": 68.02721088435374,
-        "anls": 59.65381728416852,
+        "semantic": 68.70748299319727,
+        "anls": 59.504311004808905,
         "n": 75
       },
       "Technical": {
-        "semantic": 62.111801242236034,
+        "semantic": 59.89352262644188,
         "anls": 55.55075090789312,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:03:33.475794+00:00",
+  "reevaluated_date": "2026-01-22T22:09:18.181737+00:00",
   "source_predictions_file": "OpenAI/GPT-5.2_(2025-12-11)_with_BM25_Search_Tool_predictions_20260109_151912.jsonl",
   "result_file_path": "OpenAI/GPT-5.2_(2025-12-11)_with_BM25_Search_Tool_results_20260109_151912.json"
 }

eval/reevaluated_results/OpenAI/GPT-5.2_(2025-12-11)_with_File_Search_results_20260104_121551.json

@@ -0,0 +1,120 @@
+{
+  "model_name": "GPT-5.2 (2025-12-11) with File Search",
+  "organization": "OpenAI",
+  "description": "Managed, single-shot retrieval mechanism.",
+  "link": "https://platform.openai.com/docs/guides/tools-file-search",
+  "tags": [
+    "Conventional RAG",
+    "Semantic Search Tool"
+  ],
+  "submitted_by": "Borchmann",
+  "metadata": {
+    "model_type": "api"
+  },
+  "submission_date": "2026-01-04T12:15:51.350064+00:00",
+  "results": {
+    "overall": {
+      "semantic": 50.0,
+      "semantic_ci": [
+        45.52884072176271,
+        54.471159278237295
+      ],
+      "anls": 46.08410854382378,
+      "page_f1": 28.519365079365084,
+      "doc_f1": 68.52666666666667,
+      "kuiper": 26.695999999999877
+    },
+    "single_evidence": {
+      "semantic": 56.192339949678505,
+      "anls": 50.86266041052095,
+      "n": 365
+    },
+    "multi_evidence_same_doc": {
+      "semantic": 39.48007774538386,
+      "anls": 38.16677868634344,
+      "n": 84
+    },
+    "multi_evidence_multi_doc": {
+      "semantic": 23.0092036814726,
+      "anls": 24.92497671409598,
+      "n": 51
+    },
+    "by_domain": {
+      "Cases/Logs": {
+        "semantic": 20.40816326530612,
+        "anls": 20.0,
+        "n": 15
+      },
+      "Education": {
+        "semantic": 74.21150278293136,
+        "anls": 60.780598189689094,
+        "n": 22
+      },
+      "Events": {
+        "semantic": 63.775510204081634,
+        "anls": 49.319377600130515,
+        "n": 24
+      },
+      "Financial": {
+        "semantic": 54.902395740905064,
+        "anls": 52.85763703072669,
+        "n": 92
+      },
+      "Financial/Tax": {
+        "semantic": 25.510204081632654,
+        "anls": 24.39516129032258,
+        "n": 16
+      },
+      "Government/Regulatory": {
+        "semantic": 53.19148936170214,
+        "anls": 44.987482344705185,
+        "n": 47
+      },
+      "HR/Employment": {
+        "semantic": 42.309606769537076,
+        "anls": 45.208397582300094,
+        "n": 41
+      },
+      "Legal": {
+        "semantic": 48.647365923113426,
+        "anls": 39.61843128160958,
+        "n": 43
+      },
+      "Media/Publishing": {
+        "semantic": 40.81632653061224,
+        "anls": 38.34620945100294,
+        "n": 25
+      },
+      "Misc": {
+        "semantic": 61.64965986394556,
+        "anls": 59.15390227662317,
+        "n": 24
+      },
+      "Other": {
+        "semantic": 0.0,
+        "anls": 0.0,
+        "n": 1
+      },
+      "Reference": {
+        "semantic": 38.265306122448976,
+        "anls": 41.31709677341507,
+        "n": 52
+      },
+      "Reports": {
+        "semantic": 47.619047619047635,
+        "anls": 43.77803423273486,
+        "n": 75
+      },
+      "Technical": {
+        "semantic": 73.20319432120675,
+        "anls": 64.61964587357369,
+        "n": 23
+      }
+    },
+    "n_evaluated": 500,
+    "n_unmatched": 0
+  },
+  "reevaluated_date": "2026-01-22T22:10:25.673064+00:00",
+  "source_predictions_file": "OpenAI/GPT-5.2_(2025-12-11)_with_File_Search_predictions_20260104_121551.jsonl",
+  "result_file_path": "OpenAI/GPT-5.2_(2025-12-11)_with_File_Search_results_20260104_121551.json"
+}

eval/reevaluated_results/OpenAI/GPT-5_(2025-08-07)_with_BM25_Search_Tool_results_20260109_152104.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-09T15:21:04.336083+00:00",
   "results": {
     "overall": {
-      "semantic": 77.14285714285715,
+      "semantic": 77.6530612244898,
       "semantic_ci": [
-        73.30142802417011,
-        80.98428626154418
+        73.83864314648315,
+        81.46747930249646
       ],
-      "anls": 70.03817583122695,
+      "anls": 71.05072433302601,
       "page_f1": 74.16285714285713,
       "doc_f1": 86.45064935064934,
-      "kuiper": 51.96199999999987
+      "kuiper": 52.62199999999985
     },
     "single_evidence": {
-      "semantic": 81.49287112105115,
-      "anls": 73.42321103264047,
+      "semantic": 82.19178082191782,
+      "anls": 74.70786997380705,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 58.91642371234208,
-      "anls": 56.78859482300274,
+      "semantic": 60.13119533527698,
+      "anls": 57.23352026792817,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 76.03041216486595,
+      "semantic": 74.02961184473791,
       "anls": 67.63478281504847,
       "n": 51
     },
@@ -57,8 +57,8 @@
         "n": 24
       },
       "Financial": {
-        "semantic": 67.10292812777284,
-        "anls": 62.36899647186356,
+        "semantic": 68.76663708961844,
+        "anls": 65.45826780841752,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,43 +67,43 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 74.90230134607035,
-        "anls": 68.7671602173282,
+        "semantic": 77.07338254450717,
+        "anls": 69.85898552527891,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 79.64161274265804,
-        "anls": 64.5688672367669,
+        "semantic": 74.66401194624191,
+        "anls": 65.78837943188886,
         "n": 41
       },
       "Legal": {
-        "semantic": 81.86995728523969,
-        "anls": 70.27143399236422,
+        "semantic": 83.05647840531563,
+        "anls": 71.01139170906613,
         "n": 43
       },
       "Media/Publishing": {
-        "semantic": 63.265306122449,
+        "semantic": 61.22448979591838,
         "anls": 65.71897407160566,
         "n": 25
       },
       "Misc": {
-        "semantic": 85.0340136054422,
+        "semantic": 87.15986394557822,
         "anls": 86.70405982905983,
         "n": 24
       },
       "Other": {
-        "semantic": 0.0,
-        "anls": 0.0,
+        "semantic": 100.0,
+        "anls": 100.0,
         "n": 1
       },
       "Reference": {
-        "semantic": 78.49293563579278,
-        "anls": 76.57306264232653,
+        "semantic": 80.45525902668759,
+        "anls": 77.21482727865639,
         "n": 52
       },
       "Reports": {
-        "semantic": 82.31292517006803,
-        "anls": 71.72139814224423,
+        "semantic": 81.63265306122449,
+        "anls": 71.12880554965163,
         "n": 75
       },
       "Technical": {
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:04:34.030101+00:00",
+  "reevaluated_date": "2026-01-22T22:11:16.583710+00:00",
   "source_predictions_file": "OpenAI/GPT-5_(2025-08-07)_with_BM25_Search_Tool_predictions_20260109_152104.jsonl",
   "result_file_path": "OpenAI/GPT-5_(2025-08-07)_with_BM25_Search_Tool_results_20260109_152104.json"
 }

eval/reevaluated_results/OpenAI/GPT-5_(2025-08-07)_with_File_Search_results_20260104_140537.json

@@ -14,24 +14,24 @@
   "submission_date": "2026-01-04T14:05:37.240829+00:00",
   "results": {
     "overall": {
-      "semantic": 48.367346938775505,
+      "semantic": 49.59183673469388,
       "semantic_ci": [
-        43.90134346980983,
-        52.833350407741186
+        45.12153634505467,
+        54.06213712433308
       ],
-      "anls": 44.84773268944071,
+      "anls": 45.66299621788548,
       "page_f1": 29.277142857142856,
       "doc_f1": 66.60666666666667,
-      "kuiper": 28.83600000000007
+      "kuiper": 29.030000000000086
     },
     "single_evidence": {
-      "semantic": 54.375174727425204,
-      "anls": 48.77673180310333,
+      "semantic": 56.05255800950517,
+      "anls": 49.87536451502012,
       "n": 365
     },
     "multi_evidence_same_doc": {
       "semantic": 36.443148688046655,
-      "anls": 38.29854300170963,
+      "anls": 38.377481387099635,
       "n": 84
     },
     "multi_evidence_multi_doc": {
@@ -41,7 +41,7 @@
     },
     "by_domain": {
       "Cases/Logs": {
-        "semantic": 20.40816326530612,
+        "semantic": 13.605442176870747,
         "anls": 14.833333333333334,
         "n": 15
       },
@@ -51,13 +51,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 68.02721088435374,
+        "semantic": 76.53061224489795,
         "anls": 55.83149489399489,
         "n": 24
       },
       "Financial": {
-        "semantic": 51.02040816326531,
-        "anls": 46.26513610007698,
+        "semantic": 56.011535048802116,
+        "anls": 50.61975566529438,
         "n": 92
       },
       "Financial/Tax": {
@@ -66,8 +66,8 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 48.84932696482849,
-        "anls": 41.75603723934328,
+        "semantic": 46.678245766391655,
+        "anls": 41.76403595954804,
         "n": 47
       },
       "HR/Employment": {
@@ -76,8 +76,8 @@
         "n": 41
       },
       "Legal": {
-        "semantic": 37.968675842429995,
-        "anls": 32.74308378959542,
+        "semantic": 40.34171808258187,
+        "anls": 33.905874487269834,
         "n": 43
       },
       "Media/Publishing": {
@@ -96,17 +96,17 @@
         "n": 1
       },
       "Reference": {
-        "semantic": 40.22762951334379,
+        "semantic": 42.1899529042386,
         "anls": 40.48309244262362,
         "n": 52
       },
       "Reports": {
         "semantic": 47.619047619047635,
-        "anls": 46.80494177991155,
+        "anls": 46.22668610488168,
         "n": 75
       },
       "Technical": {
-        "semantic": 66.54835847382431,
+        "semantic": 64.33007985803019,
         "anls": 62.77759844334801,
         "n": 23
       }
@@ -114,7 +114,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:06:19.286211+00:00",
+  "reevaluated_date": "2026-01-22T22:12:20.346525+00:00",
   "source_predictions_file": "OpenAI/GPT-5_(2025-08-07)_with_File_Search_predictions_20260104_140537.jsonl",
   "result_file_path": "OpenAI/GPT-5_(2025-08-07)_with_File_Search_results_20260104_140537.json"
 }

eval/reevaluated_results/OpenAI/GPT-5_Mini_(2025-08-07)_with_BM25_Search_Tool_results_20260109_152650.json

@@ -15,34 +15,34 @@
   "submission_date": "2026-01-09T15:26:50.820104+00:00",
   "results": {
     "overall": {
-      "semantic": 65.51020408163264,
+      "semantic": 66.9387755102041,
       "semantic_ci": [
-        61.22229033719482,
-        69.79811782607047
+        62.68995725738806,
+        71.18759376302013
       ],
-      "anls": 55.16542612989696,
+      "anls": 55.23436182110524,
       "page_f1": 67.57095238095239,
       "doc_f1": 82.35303030303031,
-      "kuiper": 73.26853707414845
+      "kuiper": 73.23246492985982
     },
     "single_evidence": {
-      "semantic": 72.2672630696114,
-      "anls": 60.98668315256136,
+      "semantic": 73.94464635169136,
+      "anls": 61.111836074632876,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 47.37609329446065,
-      "anls": 37.944647803393245,
+      "semantic": 47.9834791059281,
+      "anls": 37.81116005396503,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 47.0188075230092,
+      "semantic": 48.01920768307324,
       "anls": 41.86712350546175,
       "n": 51
     },
     "by_domain": {
       "Cases/Logs": {
-        "semantic": 68.02721088435374,
+        "semantic": 71.42857142857143,
         "anls": 57.16524216524217,
         "n": 15
       },
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 63.775510204081634,
+        "semantic": 68.02721088435374,
         "anls": 53.63190419293608,
         "n": 24
       },
       "Financial": {
-        "semantic": 49.3566992014197,
-        "anls": 43.770804881794874,
+        "semantic": 52.68411712511091,
+        "anls": 44.09196027631984,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,8 +67,8 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 74.90230134607035,
-        "anls": 62.856694438441366,
+        "semantic": 73.81676074685193,
+        "anls": 62.884689959158045,
         "n": 47
       },
       "HR/Employment": {
@@ -77,33 +77,33 @@
         "n": 41
       },
       "Legal": {
-        "semantic": 75.93735168485999,
+        "semantic": 78.31039392501187,
         "anls": 62.31744836688789,
         "n": 43
       },
       "Media/Publishing": {
-        "semantic": 51.02040816326531,
+        "semantic": 53.06122448979592,
         "anls": 39.93216037493774,
         "n": 25
       },
       "Misc": {
-        "semantic": 72.27891156462587,
-        "anls": 63.35950315116982,
+        "semantic": 74.4047619047619,
+        "anls": 63.84908648450315,
         "n": 24
       },
       "Other": {
-        "semantic": 0.0,
+        "semantic": 51.02040816326531,
         "anls": 0.0,
         "n": 1
       },
       "Reference": {
         "semantic": 80.45525902668759,
-        "anls": 73.02503210878088,
+        "anls": 73.08404616236015,
         "n": 52
       },
       "Reports": {
-        "semantic": 67.34693877551022,
-        "anls": 54.869395530526155,
+        "semantic": 68.02721088435374,
+        "anls": 54.719889251166556,
         "n": 75
       },
       "Technical": {
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:07:26.362120+00:00",
+  "reevaluated_date": "2026-01-22T22:13:30.005424+00:00",
   "source_predictions_file": "OpenAI/GPT-5_Mini_(2025-08-07)_with_BM25_Search_Tool_predictions_20260109_152650.jsonl",
   "result_file_path": "OpenAI/GPT-5_Mini_(2025-08-07)_with_BM25_Search_Tool_results_20260109_152650.json"
 }

eval/reevaluated_results/OpenAI/GPT-5_Mini_(2025-08-07)_with_File_Search_results_20260104_122026.json

@@ -0,0 +1,120 @@
+{
+  "model_name": "GPT-5 Mini (2025-08-07) with File Search",
+  "organization": "OpenAI",
+  "description": "Managed, single-shot retrieval mechanism.",
+  "link": "https://platform.openai.com/docs/guides/tools-file-search",
+  "tags": [
+    "Conventional RAG",
+    "Semantic Search Tool"
+  ],
+  "submitted_by": "Borchmann",
+  "metadata": {
+    "model_type": "api"
+  },
+  "submission_date": "2026-01-04T12:20:26.152729+00:00",
+  "results": {
+    "overall": {
+      "semantic": 48.46938775510205,
+      "semantic_ci": [
+        44.00292754203926,
+        52.935847968164836
+      ],
+      "anls": 41.81689987677872,
+      "page_f1": 29.013073593073592,
+      "doc_f1": 67.32666666666667,
+      "kuiper": 28.0
+    },
+    "single_evidence": {
+      "semantic": 55.21386636846521,
+      "anls": 47.22776663670515,
+      "n": 365
+    },
+    "multi_evidence_same_doc": {
+      "semantic": 32.79883381924198,
+      "anls": 30.536053736441175,
+      "n": 84
+    },
+    "multi_evidence_multi_doc": {
+      "semantic": 26.010404161664663,
+      "anls": 21.672286316292585,
+      "n": 51
+    },
+    "by_domain": {
+      "Cases/Logs": {
+        "semantic": 13.605442176870747,
+        "anls": 16.666666666666664,
+        "n": 15
+      },
+      "Education": {
+        "semantic": 81.16883116883116,
+        "anls": 68.20851085673219,
+        "n": 22
+      },
+      "Events": {
+        "semantic": 65.9013605442177,
+        "anls": 53.08302808302808,
+        "n": 24
+      },
+      "Financial": {
+        "semantic": 52.12954747116238,
+        "anls": 46.51029159374941,
+        "n": 92
+      },
+      "Financial/Tax": {
+        "semantic": 19.132653061224488,
+        "anls": 17.775843108504397,
+        "n": 16
+      },
+      "Government/Regulatory": {
+        "semantic": 45.59270516717324,
+        "anls": 37.66328697492653,
+        "n": 47
+      },
+      "HR/Employment": {
+        "semantic": 41.06520657043304,
+        "anls": 36.31954842987894,
+        "n": 41
+      },
+      "Legal": {
+        "semantic": 43.90128144280967,
+        "anls": 39.4998961859427,
+        "n": 43
+      },
+      "Media/Publishing": {
+        "semantic": 36.734693877551,
+        "anls": 37.934534534534535,
+        "n": 25
+      },
+      "Misc": {
+        "semantic": 72.27891156462587,
+        "anls": 64.10710364514712,
+        "n": 24
+      },
+      "Other": {
+        "semantic": 0.0,
+        "anls": 0.0,
+        "n": 1
+      },
+      "Reference": {
+        "semantic": 41.208791208791204,
+        "anls": 36.730980853494025,
+        "n": 52
+      },
+      "Reports": {
+        "semantic": 47.619047619047635,
+        "anls": 37.56027917001248,
+        "n": 75
+      },
+      "Technical": {
+        "semantic": 64.33007985803019,
+        "anls": 49.94662638112699,
+        "n": 23
+      }
+    },
+    "n_evaluated": 500,
+    "n_unmatched": 0
+  },
+  "reevaluated_date": "2026-01-22T22:14:40.534356+00:00",
+  "source_predictions_file": "OpenAI/GPT-5_Mini_(2025-08-07)_with_File_Search_predictions_20260104_122026.jsonl",
+  "result_file_path": "OpenAI/GPT-5_Mini_(2025-08-07)_with_File_Search_results_20260104_122026.json"
+}

eval/reevaluated_results/OpenAI/GPT-5_Nano_(2025-08-07)_with_BM25_Search_Tool_results_20260109_152828.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-09T15:28:28.366309+00:00",
   "results": {
     "overall": {
-      "semantic": 56.122448979591844,
+      "semantic": 58.16326530612243,
       "semantic_ci": [
-        51.67281165196697,
-        60.57208630721672
+        53.735254533391874,
+        62.591276078853
       ],
-      "anls": 52.255247982009955,
+      "anls": 52.7083705578831,
       "page_f1": 60.877142857142864,
       "doc_f1": 82.2030303030303,
-      "kuiper": 48.159999999999975
+      "kuiper": 49.84000000000006
     },
     "single_evidence": {
-      "semantic": 64.85882024042495,
-      "anls": 59.369915601211495,
+      "semantic": 67.23511322337154,
+      "anls": 60.04044465907011,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 31.584062196307094,
-      "anls": 34.367489115024085,
+      "semantic": 32.79883381924198,
+      "anls": 34.15103889857387,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 34.01360544217687,
+      "semantic": 35.014005602240914,
       "anls": 30.798739429426515,
       "n": 51
     },
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 59.52380952380955,
+        "semantic": 63.775510204081634,
         "anls": 51.78930433365917,
         "n": 24
       },
       "Financial": {
-        "semantic": 41.038154392191664,
-        "anls": 40.14762316798784,
+        "semantic": 44.365572315882865,
+        "anls": 41.87226084914726,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,18 +67,18 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 66.21797655232307,
-        "anls": 56.496054764723326,
+        "semantic": 64.04689535388624,
+        "anls": 55.460220498205956,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 52.264808362369344,
-        "anls": 42.85858107680723,
+        "semantic": 54.75360876057741,
+        "anls": 44.078093271929184,
         "n": 41
       },
       "Legal": {
-        "semantic": 60.51257712387282,
-        "anls": 55.28314708547266,
+        "semantic": 64.07214048410061,
+        "anls": 56.02310480217457,
         "n": 43
       },
       "Media/Publishing": {
@@ -87,22 +87,22 @@
         "n": 25
       },
       "Misc": {
-        "semantic": 72.27891156462587,
+        "semantic": 74.4047619047619,
         "anls": 74.53137140637142,
         "n": 24
       },
       "Other": {
-        "semantic": 0.0,
+        "semantic": 51.02040816326531,
         "anls": 0.0,
         "n": 1
       },
       "Reference": {
-        "semantic": 57.88854003139717,
-        "anls": 61.940508414693205,
+        "semantic": 62.79434850863422,
+        "anls": 62.608982452732455,
         "n": 52
       },
       "Reports": {
-        "semantic": 54.42176870748299,
+        "semantic": 55.78231292517006,
         "anls": 48.18660787855504,
         "n": 75
       },
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:08:36.360733+00:00",
+  "reevaluated_date": "2026-01-22T22:15:43.985703+00:00",
   "source_predictions_file": "OpenAI/GPT-5_Nano_(2025-08-07)_with_BM25_Search_Tool_predictions_20260109_152828.jsonl",
   "result_file_path": "OpenAI/GPT-5_Nano_(2025-08-07)_with_BM25_Search_Tool_results_20260109_152828.json"
 }

eval/reevaluated_results/OpenAI/GPT_4.1_Mini_(2025-04-14)_with_BM25_Search_Tool_results_20260109_153516.json

@@ -15,23 +15,23 @@
   "submission_date": "2026-01-09T15:35:16.458002+00:00",
   "results": {
     "overall": {
-      "semantic": 51.63265306122449,
+      "semantic": 52.959183673469404,
       "semantic_ci": [
-        47.16092128223699,
-        56.104384840211985
+        48.490359884292864,
+        57.42800746264594
       ],
-      "anls": 46.26708858125157,
+      "anls": 46.73693856273607,
       "page_f1": 59.905054945054935,
       "doc_f1": 77.61731601731601,
-      "kuiper": 41.783673469388106
+      "kuiper": 40.7714285714291
     },
     "single_evidence": {
-      "semantic": 61.50405367626502,
-      "anls": 54.710786964911776,
+      "semantic": 63.181436958344975,
+      "anls": 55.35441707653439,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 29.154518950437307,
+      "semantic": 29.761904761904773,
       "anls": 26.237172573133016,
       "n": 84
     },
@@ -42,7 +42,7 @@
     },
     "by_domain": {
       "Cases/Logs": {
-        "semantic": 47.619047619047635,
+        "semantic": 51.02040816326531,
         "anls": 39.64209401709402,
         "n": 15
       },
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 55.2721088435374,
+        "semantic": 59.52380952380955,
         "anls": 53.83018770627063,
         "n": 24
       },
       "Financial": {
-        "semantic": 36.04702750665483,
-        "anls": 34.96285359224887,
+        "semantic": 37.710736468500436,
+        "anls": 36.592053196991955,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,23 +67,23 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 45.59270516717324,
-        "anls": 44.19583719868558,
+        "semantic": 46.678245766391655,
+        "anls": 44.3103860759977,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 54.75360876057741,
-        "anls": 46.501429746354255,
+        "semantic": 57.24240915878547,
+        "anls": 47.47703950245181,
         "n": 41
       },
       "Legal": {
-        "semantic": 51.02040816326531,
+        "semantic": 55.766492643569066,
         "anls": 43.64210613408689,
         "n": 43
       },
       "Media/Publishing": {
         "semantic": 48.97959183673469,
-        "anls": 46.71106819031614,
+        "anls": 47.21106819031614,
         "n": 25
       },
       "Misc": {
@@ -98,16 +98,16 @@
       },
       "Reference": {
         "semantic": 66.71899529042385,
-        "anls": 62.86510186138165,
+        "anls": 63.387310128155136,
         "n": 52
       },
       "Reports": {
-        "semantic": 53.74149659863945,
+        "semantic": 54.42176870748299,
         "anls": 45.15164464860224,
         "n": 75
       },
       "Technical": {
-        "semantic": 70.9849157054126,
+        "semantic": 68.76663708961844,
         "anls": 53.71736172158072,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:10:10.834088+00:00",
+  "reevaluated_date": "2026-01-22T22:17:05.838647+00:00",
   "source_predictions_file": "OpenAI/GPT_4.1_Mini_(2025-04-14)_with_BM25_Search_Tool_predictions_20260109_153516.jsonl",
   "result_file_path": "OpenAI/GPT_4.1_Mini_(2025-04-14)_with_BM25_Search_Tool_results_20260109_153516.json"
 }

eval/reevaluated_results/OpenAI/Gemini_3_Pro_with_BM25_Search_Tool_results_20260109_185347.json

@@ -15,35 +15,35 @@
   "submission_date": "2026-01-09T18:53:47.189606+00:00",
   "results": {
     "overall": {
-      "semantic": 80.97828309680585,
+      "semantic": 82.20522678009081,
       "semantic_ci": [
-        77.35470686592657,
-        84.60185932768512
+        78.66223544202387,
+        85.74821811815777
       ],
-      "anls": 73.52101315170081,
+      "anls": 74.76545756458952,
       "page_f1": 78.4607309857811,
       "doc_f1": 90.20248288785363,
-      "kuiper": 27.436873747495046
+      "kuiper": 25.80160320641279
     },
     "single_evidence": {
-      "semantic": 85.36106750392466,
-      "anls": 76.30968193570469,
+      "semantic": 87.04305898183449,
+      "anls": 77.74037073431458,
       "n": 364
     },
     "multi_evidence_same_doc": {
-      "semantic": 65.59766763848395,
-      "anls": 63.78712934826842,
+      "semantic": 66.81243926141885,
+      "anls": 64.98007029276216,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 75.03001200480193,
+      "semantic": 73.02921168467388,
       "anls": 69.64985240877743,
       "n": 51
     },
     "by_domain": {
       "Cases/Logs": {
-        "semantic": 88.43537414965988,
-        "anls": 85.12820512820514,
+        "semantic": 95.23809523809524,
+        "anls": 91.7948717948718,
         "n": 15
       },
       "Education": {
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 82.90816326530613,
+        "semantic": 87.15986394557822,
         "anls": 79.84423442344234,
         "n": 24
       },
       "Financial": {
-        "semantic": 68.21206743566992,
-        "anls": 63.13552237747254,
+        "semantic": 68.76663708961844,
+        "anls": 65.75226955943711,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,17 +67,17 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 89.01432913590969,
-        "anls": 78.26722646935413,
+        "semantic": 90.09986973512808,
+        "anls": 80.50699949010338,
         "n": 47
       },
       "HR/Employment": {
         "semantic": 89.59681433549028,
-        "anls": 77.34609828919353,
+        "anls": 78.32170804529109,
         "n": 41
       },
       "Legal": {
-        "semantic": 84.24299952539157,
+        "semantic": 85.4295206454675,
         "anls": 68.10496996543507,
         "n": 43
       },
@@ -92,22 +92,22 @@
         "n": 24
       },
       "Other": {
-        "semantic": 0.0,
-        "anls": 0.0,
+        "semantic": 100.0,
+        "anls": 100.0,
         "n": 1
       },
       "Reference": {
-        "semantic": 84.37990580847723,
-        "anls": 83.68517307852197,
+        "semantic": 88.30455259026688,
+        "anls": 84.35364711656122,
         "n": 52
       },
       "Reports": {
-        "semantic": 77.90954219525649,
-        "anls": 71.94584088751826,
+        "semantic": 78.59900717043573,
+        "anls": 71.94863871640305,
         "n": 74
       },
       "Technical": {
-        "semantic": 79.85803016858917,
+        "semantic": 75.42147293700089,
         "anls": 55.56822369489126,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 499,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:11:07.316820+00:00",
+  "reevaluated_date": "2026-01-22T22:17:55.213452+00:00",
   "source_predictions_file": "OpenAI/Gemini_3_Pro_with_BM25_Search_Tool_predictions_20260109_185347.jsonl",
   "result_file_path": "OpenAI/Gemini_3_Pro_with_BM25_Search_Tool_results_20260109_185347.json"
 }

eval/reevaluated_results/OpenAI_-_KAIST/GPT-4o_(2024-08-06)_with_HEAVEN_Retrieval_results_20260109_154427.json

@@ -15,35 +15,35 @@
   "submission_date": "2026-01-09T15:44:27.735534+00:00",
   "results": {
     "overall": {
-      "semantic": 46.73469387755102,
+      "semantic": 48.57142857142857,
       "semantic_ci": [
-        42.278445434797305,
-        51.190942320304735
+        44.10452956979305,
+        53.038327573064095
       ],
-      "anls": 45.649762341432954,
+      "anls": 46.86058848865213,
       "page_f1": 43.169719169719166,
       "doc_f1": 59.24761904761905,
       "kuiper": null
     },
     "single_evidence": {
-      "semantic": 51.02040816326531,
-      "anls": 49.144211108257174,
+      "semantic": 53.955828906905246,
+      "anls": 50.80230983229035,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 32.79883381924198,
-      "anls": 35.078637953143826,
+      "semantic": 32.19144800777454,
+      "anls": 35.08110270716138,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 39.015606242497,
+      "semantic": 37.01480592236896,
       "anls": 38.05173584585349,
       "n": 51
     },
     "by_domain": {
       "Cases/Logs": {
         "semantic": 54.42176870748299,
-        "anls": 48.75783475783476,
+        "anls": 52.09116809116809,
         "n": 15
       },
       "Education": {
@@ -52,32 +52,32 @@
         "n": 22
       },
       "Events": {
-        "semantic": 57.39795918367348,
-        "anls": 55.056754787358244,
+        "semantic": 61.64965986394556,
+        "anls": 57.140088120691566,
         "n": 24
       },
       "Financial": {
-        "semantic": 44.365572315882865,
-        "anls": 48.16466676354977,
+        "semantic": 51.02040816326531,
+        "anls": 53.968561691086,
         "n": 92
       },
       "Financial/Tax": {
-        "semantic": 51.02040816326531,
+        "semantic": 44.642857142857146,
         "anls": 44.99022482893451,
         "n": 16
       },
       "Government/Regulatory": {
         "semantic": 54.277029960920544,
-        "anls": 47.1956486962086,
+        "anls": 46.13981762917933,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 46.042807366849175,
+        "semantic": 42.309606769537076,
         "anls": 32.93040293040293,
         "n": 41
       },
       "Legal": {
-        "semantic": 39.15519696250593,
+        "semantic": 40.34171808258187,
         "anls": 35.73555320648344,
         "n": 43
       },
@@ -97,17 +97,17 @@
         "n": 1
       },
       "Reference": {
-        "semantic": 45.13343799058085,
-        "anls": 48.954805357154754,
+        "semantic": 47.09576138147568,
+        "anls": 49.356185378099994,
         "n": 52
       },
       "Reports": {
-        "semantic": 51.02040816326531,
-        "anls": 50.65907330372244,
+        "semantic": 54.42176870748299,
+        "anls": 50.66183382822209,
         "n": 75
       },
       "Technical": {
-        "semantic": 51.02040816326531,
+        "semantic": 53.238686779059464,
         "anls": 46.20014437749956,
         "n": 23
       }
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:14:49.523407+00:00",
+  "reevaluated_date": "2026-01-22T22:18:55.670607+00:00",
   "source_predictions_file": "OpenAI_-_KAIST/GPT-4o_(2024-08-06)_with_HEAVEN_Retrieval_predictions_20260109_154427.jsonl",
   "result_file_path": "OpenAI_-_KAIST/GPT-4o_(2024-08-06)_with_HEAVEN_Retrieval_results_20260109_154427.json"
 }

eval/reevaluated_results/OpenAI_-_KAIST/GPT-5.2_(2024-08-06)_with_HEAVEN_Retrieval_results_20260109_175639.json

@@ -15,34 +15,34 @@
   "submission_date": "2026-01-09T17:56:39.771528+00:00",
   "results": {
     "overall": {
-      "semantic": 50.7142857142857,
+      "semantic": 52.85714285714286,
       "semantic_ci": [
-        46.24231243352625,
-        55.186258995045144
+        48.3879879090649,
+        57.326297805220825
       ],
-      "anls": 47.46445252141211,
+      "anls": 48.84111465957483,
       "page_f1": 48.43228327228327,
       "doc_f1": 62.30761904761904,
       "kuiper": null
     },
     "single_evidence": {
-      "semantic": 54.095610847078554,
-      "anls": 49.17254529050683,
+      "semantic": 56.05255800950517,
+      "anls": 50.78441123319547,
       "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 35.22837706511177,
-      "anls": 39.524985989825936,
+      "semantic": 38.87269193391642,
+      "anls": 40.715462180302126,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 52.02080832332933,
+      "semantic": 53.02120848339336,
       "anls": 48.31663542207227,
       "n": 51
     },
     "by_domain": {
       "Cases/Logs": {
-        "semantic": 57.82312925170068,
+        "semantic": 61.22448979591838,
         "anls": 43.64672364672364,
         "n": 15
       },
@@ -52,13 +52,13 @@
         "n": 22
       },
       "Events": {
-        "semantic": 48.89455782312926,
+        "semantic": 53.14625850340138,
         "anls": 46.90982404692082,
         "n": 24
       },
       "Financial": {
-        "semantic": 48.802129547471154,
-        "anls": 48.83531625708929,
+        "semantic": 56.011535048802116,
+        "anls": 54.83808397045483,
         "n": 92
       },
       "Financial/Tax": {
@@ -67,17 +67,17 @@
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 61.87581415544942,
-        "anls": 49.070286122357786,
+        "semantic": 66.21797655232307,
+        "anls": 51.225941217542555,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 41.06520657043304,
+        "semantic": 39.82080637132901,
         "anls": 34.149915125524885,
         "n": 41
       },
       "Legal": {
-        "semantic": 46.27432368296155,
+        "semantic": 47.46084480303749,
         "anls": 46.299372462163156,
         "n": 43
       },
@@ -97,12 +97,12 @@
         "n": 1
       },
       "Reference": {
-        "semantic": 58.86970172684458,
-        "anls": 58.28202679165414,
+        "semantic": 56.907378335949765,
+        "anls": 58.9505008296934,
         "n": 52
       },
       "Reports": {
-        "semantic": 54.42176870748299,
+        "semantic": 56.4625850340136,
         "anls": 52.18098320525303,
         "n": 75
       },
@@ -115,7 +115,7 @@
     "n_evaluated": 500,
     "n_unmatched": 0
   },
-  "reevaluated_date": "2026-01-16T16:17:00.809294+00:00",
+  "reevaluated_date": "2026-01-22T22:19:59.654603+00:00",
   "source_predictions_file": "OpenAI_-_KAIST/GPT-5.2_(2024-08-06)_with_HEAVEN_Retrieval_predictions_20260109_175639.jsonl",
   "result_file_path": "OpenAI_-_KAIST/GPT-5.2_(2024-08-06)_with_HEAVEN_Retrieval_results_20260109_175639.json"
 }

eval/reevaluated_results/Z.AI/GLM-4.6V_Flash_with_BM25_Search_Tool_results_20260110_132227.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-10T13:22:27.811792+00:00",
   "results": {
     "overall": {
-      "semantic": 44.57895382601939,
+      "semantic": 46.020408163265316,
       "semantic_ci": [
-        40.138241872067184,
-        49.01966577997159
+        41.56988099714986,
+        50.470935329380765
       ],
-      "anls": 30.17090068718362,
-      "page_f1": 28.991793110029583,
-      "doc_f1": 51.58650634602539,
-      "kuiper": 26.255020080321316
+      "anls": 30.43927742127836,
+      "page_f1": 28.93380952380953,
+      "doc_f1": 51.483333333333334,
+      "kuiper": 27.468937875751397
     },
     "single_evidence": {
-      "semantic": 50.88024220677282,
-      "anls": 35.972008760814205,
-      "n": 364
+      "semantic": 52.278445624825274,
+      "anls": 36.186768374440895,
+      "n": 365
     },
     "multi_evidence_same_doc": {
-      "semantic": 16.399416909621,
-      "anls": 12.405045351473925,
+      "semantic": 18.221574344023328,
+      "anls": 13.000283446712018,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 46.018407362945176,
+      "semantic": 47.0188075230092,
       "anls": 18.02832244008715,
       "n": 51
     },
@@ -47,32 +47,32 @@
         "n": 15
       },
       "Education": {
-        "semantic": 57.97773654916514,
+        "semantic": 60.29684601113172,
         "anls": 34.34782608695652,
         "n": 22
       },
       "Events": {
-        "semantic": 72.27891156462587,
-        "anls": 52.92922722985768,
+        "semantic": 74.4047619047619,
+        "anls": 55.012560563191016,
         "n": 24
       },
       "Financial": {
-        "semantic": 34.38331854480922,
-        "anls": 23.538822057620244,
+        "semantic": 36.601597160603376,
+        "anls": 24.625778579359373,
         "n": 92
       },
       "Financial/Tax": {
-        "semantic": 19.132653061224488,
+        "semantic": 22.321428571428573,
         "anls": 21.39516129032258,
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 37.99392097264438,
-        "anls": 29.5464725643897,
+        "semantic": 40.16500217108119,
+        "anls": 29.574468085106382,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 53.50920856147338,
+        "semantic": 49.77600796416127,
         "anls": 37.17815890071988,
         "n": 41
       },
@@ -87,17 +87,22 @@
         "n": 25
       },
       "Misc": {
-        "semantic": 68.02721088435374,
+        "semantic": 70.1530612244898,
         "anls": 48.707026404394824,
         "n": 24
       },
+      "Other": {
+        "semantic": 100.0,
+        "anls": 0.0,
+        "n": 1
+      },
       "Reference": {
-        "semantic": 40.22762951334379,
-        "anls": 23.25877926421405,
+        "semantic": 43.171114599686035,
+        "anls": 23.509615384615383,
         "n": 52
       },
       "Reports": {
-        "semantic": 42.85714285714285,
+        "semantic": 44.897959183673464,
         "anls": 25.79399206429042,
         "n": 75
       },
@@ -107,10 +112,10 @@
         "n": 23
       }
     },
-    "n_evaluated": 499,
-    "n_unmatched": 1767
+    "n_evaluated": 500,
+    "n_unmatched": 1766
   },
-  "reevaluated_date": "2026-01-16T16:19:24.691340+00:00",
+  "reevaluated_date": "2026-01-22T22:21:10.223032+00:00",
   "source_predictions_file": "Z.AI/GLM-4.6V_Flash_with_BM25_Search_Tool_predictions_20260110_132227.jsonl",
   "result_file_path": "Z.AI/GLM-4.6V_Flash_with_BM25_Search_Tool_results_20260110_132227.json"
 }

eval/reevaluated_results/Z.AI/GLM-4.6V_with_BM25_Search_Tool_results_20260110_131826.json

@@ -15,28 +15,28 @@
   "submission_date": "2026-01-10T13:18:26.686587+00:00",
   "results": {
     "overall": {
-      "semantic": 65.64148705574414,
+      "semantic": 66.12244897959185,
       "semantic_ci": [
-        61.35270759605711,
-        69.93026651543116
+        61.850797145071965,
+        70.39410081411174
       ],
-      "anls": 59.661893537203156,
-      "page_f1": 66.02347552247352,
-      "doc_f1": 86.7908978129419,
-      "kuiper": 50.23185483870943
+      "anls": 60.25345765733381,
+      "page_f1": 65.89142857142856,
+      "doc_f1": 86.61731601731603,
+      "kuiper": 51.43661971830999
     },
     "single_evidence": {
-      "semantic": 71.34447185467593,
-      "anls": 64.16155076739506,
-      "n": 364
+      "semantic": 72.2672630696114,
+      "anls": 64.8225984463954,
+      "n": 365
     },
     "multi_evidence_same_doc": {
       "semantic": 37.0505344995141,
-      "anls": 36.01688306453112,
+      "anls": 36.61212115976921,
       "n": 84
     },
     "multi_evidence_multi_doc": {
-      "semantic": 72.02881152460986,
+      "semantic": 70.0280112044818,
       "anls": 66.49141604533267,
       "n": 51
     },
@@ -52,32 +52,32 @@
         "n": 22
       },
       "Events": {
-        "semantic": 78.65646258503402,
-        "anls": 67.87290397408577,
+        "semantic": 80.78231292517005,
+        "anls": 69.9562373074191,
         "n": 24
       },
       "Financial": {
         "semantic": 53.79325643300798,
-        "anls": 51.19993983845437,
+        "anls": 52.97859596493658,
         "n": 92
       },
       "Financial/Tax": {
-        "semantic": 63.775510204081634,
+        "semantic": 70.1530612244898,
         "anls": 62.5648667601683,
         "n": 16
       },
       "Government/Regulatory": {
-        "semantic": 75.98784194528876,
-        "anls": 70.93589720557641,
+        "semantic": 77.07338254450717,
+        "anls": 71.3587742867764,
         "n": 47
       },
       "HR/Employment": {
-        "semantic": 67.19761075161773,
-        "anls": 59.735891761304075,
+        "semantic": 64.70881035340967,
+        "anls": 60.27789718135828,
         "n": 41
       },
       "Legal": {
-        "semantic": 70.0047460844803,
+        "semantic": 67.63170384432841,
         "anls": 55.536175710594314,
         "n": 43
       },
@@ -87,30 +87,35 @@
         "n": 25
       },
       "Misc": {
-        "semantic": 74.4047619047619,
+        "semantic": 72.27891156462587,
         "anls": 75.10160446706249,
         "n": 24
       },
+      "Other": {
+        "semantic": 100.0,
+        "anls": 86.66666666666667,
+        "n": 1
+      },
       "Reference": {
-        "semantic": 59.85086342229201,
-        "anls": 60.632124141167864,
+        "semantic": 63.775510204081634,
+        "anls": 60.88296026156921,
         "n": 52
       },
       "Reports": {
-        "semantic": 63.94557823129252,
+        "semantic": 65.30612244897961,
         "anls": 56.89167319856098,
         "n": 75
       },
       "Technical": {
-        "semantic": 68.76663708961844,
+        "semantic": 66.54835847382431,
         "anls": 51.450020851943364,
         "n": 23
       }
     },
-    "n_evaluated": 499,
-    "n_unmatched": 1767
+    "n_evaluated": 500,
+    "n_unmatched": 1766
   },
-  "reevaluated_date": "2026-01-16T16:20:28.331848+00:00",
+  "reevaluated_date": "2026-01-22T22:22:05.057646+00:00",
   "source_predictions_file": "Z.AI/GLM-4.6V_with_BM25_Search_Tool_predictions_20260110_131826.jsonl",
   "result_file_path": "Z.AI/GLM-4.6V_with_BM25_Search_Tool_results_20260110_131826.json"
 }