Upload 2 files

app.py

@@ -21,6 +21,7 @@ RESULTS_REPO = "RUC-NLPIR/GISA-leaderboard"
 META_FILE = "encrypted_question.jsonl"
 ANSWER_DIR = "answer"
 CACHE_DIR = "cache/answers"
+SEED_FILE = os.path.join(os.path.dirname(__file__), "seed.json")
 
 ASSETS_DIR = os.path.join(os.path.dirname(__file__), "assets")
 INDEX_HTML = os.path.join(ASSETS_DIR, "index.html")
@@ -105,10 +106,17 @@ def load_results_dataset():
 
 def build_leaderboard_rows() -> List[dict]:
     ds = load_results_dataset()
-    if ds is None:
-        return []
+    if ds is None or len(ds) == 0:
+        seed_rows = load_seed_rows()
+        if not seed_rows:
+            return []
+        return _rows_from_source(seed_rows)
+    return _rows_from_source(ds)
+
+
+def _rows_from_source(source) -> List[dict]:
     rows: List[dict] = []
-    for row in ds:
+    for row in source:
         rows.append(
             {
                 "model": row.get("model", "-"),
@@ -130,6 +138,17 @@ def build_leaderboard_rows() -> List[dict]:
     return rows
 
 
+def load_seed_rows() -> List[dict]:
+    if not os.path.exists(SEED_FILE):
+        return []
+    try:
+        with open(SEED_FILE, "r", encoding="utf-8") as f:
+            data = json.load(f)
+        return data if isinstance(data, list) else []
+    except Exception:
+        return []
+
+
 def render_page() -> str:
     html = _load_text(INDEX_HTML)
     html = html.replace("__LEADERBOARD_DATA__", "")
@@ -155,6 +174,47 @@ def ensure_results_repo():
     )
 
 
+def seed_results_if_needed():
+    seed_rows = load_seed_rows()
+    if not seed_rows:
+        return
+    ds = load_results_dataset()
+    if ds is not None and len(ds) > 0:
+        return
+    if not TOKEN:
+        return
+
+    entries = []
+    for row in seed_rows:
+        entries.append(
+            {
+                "model": row.get("model", "-"),
+                "org": row.get("org", "-"),
+                "framework": row.get("framework", "N/A"),
+                "date": row.get("date", "-"),
+                "overall_em": _safe_float(row.get("overall")),
+                "item_em": _safe_float(row.get("item_em")),
+                "set_em": _safe_float(row.get("set_em")),
+                "set_f1": _safe_float(row.get("set_f1")),
+                "list_em": _safe_float(row.get("list_em")),
+                "list_f1": _safe_float(row.get("list_f1")),
+                "list_order": _safe_float(row.get("list_order")),
+                "table_em": _safe_float(row.get("table_em")),
+                "table_row_f1": _safe_float(row.get("table_row_f1")),
+                "table_item_f1": _safe_float(row.get("table_item_f1")),
+                "url": row.get("url", ""),
+                "email": row.get("email", ""),
+                "username": row.get("username", "seed"),
+            }
+        )
+
+    try:
+        ensure_results_repo()
+        Dataset.from_list(entries).push_to_hub(RESULTS_REPO, token=TOKEN)
+    except Exception:
+        pass
+
+
 def _get_metric(summary: dict, qtype: str, key: str, fallback: float = 0.0) -> float:
     return float(summary.get(qtype, {}).get(key, fallback) or 0.0)
 
@@ -299,6 +359,7 @@ def add_new_eval(
     return format_log("Submission received! Please refresh the leaderboard to see your score.")
 
 
+seed_results_if_needed()
 leaderboard_data = build_leaderboard_rows()
 css = _load_text(STYLES_CSS)
 html = render_page()

seed.json

@@ -0,0 +1,258 @@
+[
+  {
+    "model": "Qwen3-235B-A22B (thinking)",
+    "org": "Alibaba Cloud",
+    "framework": "ReAct",
+    "date": "2025.7",
+    "overall": 9.65,
+    "item_em": 40.91,
+    "set_em": 18.0,
+    "set_f1": 52.37,
+    "list_em": 14.58,
+    "list_f1": 36.48,
+    "list_order": 35.96,
+    "table_em": 4.35,
+    "table_row_f1": 28.32,
+    "table_item_f1": 43.93
+  },
+  {
+    "model": "Claude 4.5 Sonnet (non-thinking)",
+    "org": "Anthropic AI",
+    "framework": "ReAct",
+    "date": "2025.9",
+    "overall": 16.36,
+    "item_em": 59.09,
+    "set_em": 26.0,
+    "set_f1": 60.87,
+    "list_em": 22.92,
+    "list_f1": 58.76,
+    "list_order": 57.78,
+    "table_em": 9.49,
+    "table_row_f1": 47.85,
+    "table_item_f1": 63.71
+  },
+  {
+    "model": "Claude 4.5 Sonnet (thinking)",
+    "org": "Anthropic AI",
+    "framework": "ReAct",
+    "date": "2025.9",
+    "overall": 19.3,
+    "item_em": 63.64,
+    "set_em": 28.0,
+    "set_f1": 64.86,
+    "list_em": 22.92,
+    "list_f1": 59.24,
+    "list_order": 56.42,
+    "table_em": 13.04,
+    "table_row_f1": 49.92,
+    "table_item_f1": 65.17
+  },
+  {
+    "model": "Gemini 3 Pro (low)",
+    "org": "Google",
+    "framework": "ReAct",
+    "date": "2025.11",
+    "overall": 14.74,
+    "item_em": 45.45,
+    "set_em": 28.0,
+    "set_f1": 63.82,
+    "list_em": 27.08,
+    "list_f1": 57.55,
+    "list_order": 56.37,
+    "table_em": 7.11,
+    "table_row_f1": 45.93,
+    "table_item_f1": 64.93
+  },
+  {
+    "model": "Gemini 3 Pro (high)",
+    "org": "Google",
+    "framework": "ReAct",
+    "date": "2025.11",
+    "overall": 15.28,
+    "item_em": 50.0,
+    "set_em": 22.0,
+    "set_f1": 62.66,
+    "list_em": 27.08,
+    "list_f1": 60.87,
+    "list_order": 60.12,
+    "table_em": 8.7,
+    "table_row_f1": 47.01,
+    "table_item_f1": 66.02
+  },
+  {
+    "model": "GPT-5.2 (thinking)",
+    "org": "OpenAI",
+    "framework": "ReAct",
+    "date": "2025.11",
+    "overall": 15.82,
+    "item_em": 63.64,
+    "set_em": 26.0,
+    "set_f1": 62.7,
+    "list_em": 16.67,
+    "list_f1": 54.11,
+    "list_order": 53.17,
+    "table_em": 9.49,
+    "table_row_f1": 43.04,
+    "table_item_f1": 60.2
+  },
+  {
+    "model": "DeepSeek-V3.2 (non-thinking)",
+    "org": "DeepSeek AI",
+    "framework": "ReAct",
+    "date": "2025.12",
+    "overall": 11.53,
+    "item_em": 22.73,
+    "set_em": 20.0,
+    "set_f1": 52.0,
+    "list_em": 22.92,
+    "list_f1": 56.02,
+    "list_order": 55.45,
+    "table_em": 6.72,
+    "table_row_f1": 44.14,
+    "table_item_f1": 62.24
+  },
+  {
+    "model": "DeepSeek-V3.2 (thinking)",
+    "org": "DeepSeek AI",
+    "framework": "ReAct",
+    "date": "2025.12",
+    "overall": 14.47,
+    "item_em": 63.64,
+    "set_em": 28.0,
+    "set_f1": 60.79,
+    "list_em": 20.83,
+    "list_f1": 62.25,
+    "list_order": 60.41,
+    "table_em": 6.32,
+    "table_row_f1": 43.44,
+    "table_item_f1": 62.42
+  },
+  {
+    "model": "GLM-4.7 (thinking)",
+    "org": "Z.AI",
+    "framework": "ReAct",
+    "date": "2025.12",
+    "overall": 14.21,
+    "item_em": 50.0,
+    "set_em": 22.0,
+    "set_f1": 59.44,
+    "list_em": 20.83,
+    "list_f1": 51.99,
+    "list_order": 50.97,
+    "table_em": 8.3,
+    "table_row_f1": 43.97,
+    "table_item_f1": 61.28
+  },
+  {
+    "model": "Seed-1.8 (thinking)",
+    "org": "ByteDance Seed",
+    "framework": "ReAct",
+    "date": "2025.12",
+    "overall": 13.4,
+    "item_em": 45.45,
+    "set_em": 32.0,
+    "set_f1": 56.77,
+    "list_em": 16.67,
+    "list_f1": 56.11,
+    "list_order": 53.54,
+    "table_em": 6.32,
+    "table_row_f1": 38.49,
+    "table_item_f1": 57.13
+  },
+  {
+    "model": "Qwen3-Max (thinking)",
+    "org": "Alibaba Cloud",
+    "framework": "ReAct",
+    "date": "2026.1",
+    "overall": 17.96,
+    "item_em": 59.09,
+    "set_em": 30.0,
+    "set_f1": 63.45,
+    "list_em": 25.0,
+    "list_f1": 66.51,
+    "list_order": 64.08,
+    "table_em": 10.67,
+    "table_row_f1": 48.48,
+    "table_item_f1": 66.86
+  },
+  {
+    "model": "Kimi K2.5 (thinking)",
+    "org": "Moonshot AI",
+    "framework": "ReAct",
+    "date": "2026.1",
+    "overall": 15.55,
+    "item_em": 68.18,
+    "set_em": 28.0,
+    "set_f1": 61.71,
+    "list_em": 18.75,
+    "list_f1": 50.52,
+    "list_order": 48.81,
+    "table_em": 7.91,
+    "table_row_f1": 45.19,
+    "table_item_f1": 61.23
+  },
+  {
+    "model": "GPT-4o Search Preview",
+    "org": "OpenAI",
+    "framework": "N/A",
+    "date": "2025.3",
+    "overall": 5.63,
+    "item_em": 13.64,
+    "set_em": 4.0,
+    "set_f1": 38.7,
+    "list_em": 8.33,
+    "list_f1": 36.65,
+    "list_order": 36.0,
+    "table_em": 4.74,
+    "table_row_f1": 29.59,
+    "table_item_f1": 45.61
+  },
+  {
+    "model": "OpenAI o4 Mini Deep Research",
+    "org": "OpenAI",
+    "framework": "N/A",
+    "date": "2025.6",
+    "overall": 7.78,
+    "item_em": 18.18,
+    "set_em": 14.0,
+    "set_f1": 63.03,
+    "list_em": 18.75,
+    "list_f1": 53.72,
+    "list_order": 52.59,
+    "table_em": 3.56,
+    "table_row_f1": 36.78,
+    "table_item_f1": 56.47
+  },
+  {
+    "model": "Perplexity Sonar Pro Search",
+    "org": "Perplexity AI",
+    "framework": "N/A",
+    "date": "-",
+    "overall": 7.51,
+    "item_em": 22.73,
+    "set_em": 20.0,
+    "set_f1": 47.04,
+    "list_em": 6.25,
+    "list_f1": 34.74,
+    "list_order": 33.16,
+    "table_em": 3.95,
+    "table_row_f1": 34.76,
+    "table_item_f1": 49.05
+  },
+  {
+    "model": "Google Search AI Mode",
+    "org": "Google",
+    "framework": "N/A",
+    "date": "-",
+    "overall": 9.38,
+    "item_em": 31.82,
+    "set_em": 20.0,
+    "set_f1": 46.34,
+    "list_em": 8.33,
+    "list_f1": 40.64,
+    "list_order": 39.36,
+    "table_em": 5.53,
+    "table_row_f1": 31.15,
+    "table_item_f1": 50.79
+  }
+]