Create app.py

app.py

@@ -0,0 +1,73 @@
+import gradio as gr
+import pandas as pd
+import os
+
+# --- 1. 视觉样式配置 ---
+CSS = """
+.gradio-container { max-width: 1200px !important; }
+.stat-card { 
+    background: #f8f9fa; 
+    padding: 15px; 
+    border-radius: 10px; 
+    text-align: center; 
+    border: 1px solid #e0e0e0;
+}
+.stat-val { font-size: 24px; font-weight: bold; color: #1f77b4; }
+"""
+
+# --- 2. 数据加载逻辑 ---
+def load_data():
+    # 优先读取本地 results.csv，如果没有则使用演示数据
+    if os.path.exists("results.csv"):
+        df = pd.read_csv("results.csv")
+    else:
+        df = pd.DataFrame([
+            {"Model": "Econometrics-Agent (Ours)", "Compilation %": 92.5, "Replication %": 88.0, "Direction %": 85.0},
+            {"Model": "GPT-4o", "Compilation %": 98.0, "Replication %": 82.5, "Direction %": 75.1},
+            {"Model": "Claude 3.5 Sonnet", "Compilation %": 96.2, "Replication %": 80.4, "Direction %": 72.8},
+        ])
+    
+    # 计算平均分作为排序依据
+    df["Overall Score"] = df[["Compilation %", "Replication %", "Direction %"]].mean(axis=1).round(2)
+    return df.sort_values("Overall Score", ascending=False)
+
+# --- 3. 界面构建 ---
+with gr.Blocks(css=CSS, title="Infernet Leaderboard") as demo:
+    gr.HTML("<h1 style='text-align: center;'>🏆 Infernet Econometrics Leaderboard</h1>")
+    gr.HTML("<p style='text-align: center; color: #666;'>CamoAiLab | Evaluating AI Agents in Empirical Social Science Research</p>")
+    
+    # 顶部统计卡片
+    with gr.Row():
+        df_init = load_data()
+        gr.HTML(f"<div class='stat-card'><div class='stat-val'>{len(df_init)}</div><div>Total Models</div></div>")
+        gr.HTML(f"<div class='stat-card'><div class='stat-val'>{df_init['Overall Score'].max()}%</div><div>Best Overall</div></div>")
+        gr.HTML(f"<div class='stat-card'><div class='stat-val'>1,000</div><div>Test Tasks</div></div>")
+
+    gr.Markdown("---")
+    
+    with gr.Tabs():
+        # 标签页 1：排行榜
+        with gr.TabItem("📊 Leaderboard"):
+            search = gr.Textbox(placeholder="🔍 Search for a model...", label=None, show_label=False)
+            table = gr.Dataframe(value=df_init, interactive=False)
+            
+            def filter_table(query):
+                full_df = load_data()
+                return full_df[full_df["Model"].str.contains(query, case=False)] if query else full_df
+            
+            search.change(fn=filter_table, inputs=search, outputs=table)
+
+        # 标签页 2：指标定义
+        with gr.TabItem("📖 Metric Definitions"):
+            gr.Markdown("""
+            ### 🔍 核心评测指标定义
+            1. **Compilation % (编译成功率)**: 生成的代码是否能直接运行通过。
+            2. **Replication % (部分复现率)**: 回归系数等统计量与原始论文的重合度。
+            3. **Direction % (系数方向正确率)**: **核心指标**。回归系数的正负号是否预测正确。
+            """)
+
+        # 标签页 3：参与方式
+        with gr.TabItem("✉️ Submission"):
+            gr.Markdown("请将您的预测文件上传至 [CamoAiLab/Infernet](https://huggingface.co/datasets/CamoAiLab/Infernet ) 的 Community 讨论区。")
+
+demo.launch()