Spaces:

CamoAiLab
/

InferenceNet-Leaderboard

Running

App Files Files Community

zsss0316 commited on 9 days ago

Commit

4368125

verified ·

1 Parent(s): 549676f

Update index.html

Browse files

Files changed (1) hide show

index.html +114 -60

index.html CHANGED Viewed

@@ -1,75 +1,129 @@
-<html>
 <head>
-    <meta charset="UTF-8">
-    <title>Infernet Leaderboard</title>
-    <script type="module" src="https://cdn.jsdelivr.net/npm/@gradio/lite/dist/lite.js"></script>
-    <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/@gradio/lite/dist/lite.css" />
-    <style>
-        body { background-color: #f3f4f6; padding-top: 20px; }
-        .gradio-container { border-radius: 15px !important; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1 ) !important; }
-    </style>
 </head>
 <body>
-    <gradio-lite>
-        <gradio-file name="app.py" entrypoint>
 import gradio as gr
-# 1. 核心数据（原生列表格式，无需 pandas）
-# 格式：[模型名, 编译%, 复现%, 方向%, 综合分]
-RAW_DATA = [
-    ["Econometrics-Agent (Ours)", 92.5, 88.0, 85.0, 88.5],
-    ["GPT-4o", 98.0, 82.5, 75.1, 85.2],
-    ["Claude 3.5 Sonnet", 96.2, 80.4, 72.8, 83.1],
-    ["Llama-3-70B", 85.0, 65.0, 60.5, 70.2]
-]
-HEADERS = ["Model", "Compilation %", "Replication %", "Direction %", "Overall Score"]
-def get_leaderboard(query=""):
-    # 过滤与排序逻辑
-    filtered = [row for row in RAW_DATA if query.lower() in row[0].lower()]
-    # 按最后一列（综合分）降序排列
-    sorted_data = sorted(filtered, key=lambda x: x[4], reverse=True)
-    return sorted_data
-# 2. 界面构建
-with gr.Blocks(title="Infernet Leaderboard") as demo:
-    gr.HTML("<h1 style='text-align: center; color: #111827;'>🏆 Infernet Econometrics Leaderboard</h1>")
-    gr.HTML("<p style='text-align: center; color: #4b5563; margin-bottom: 20px;'>CamoAiLab | Browser-based Empirical AI Benchmark</p>")
     with gr.Row():
-        gr.Number(label="Total Models", value=len(RAW_DATA), interactive=False)
-        gr.Number(label="Best Score (%)", value=max(row[4] for row in RAW_DATA), interactive=False)
-    gr.Markdown("---")
     with gr.Tabs():
         with gr.TabItem("📊 Main Leaderboard"):
-            search = gr.Textbox(placeholder="🔍 Search for a model...", label=None, show_label=False)
-            # 使用原生列表渲染表格
             table = gr.Dataframe(
-                headers=HEADERS,
-                value=get_leaderboard(),
-                datatype=["str", "number", "number", "number", "number"],
-                interactive=False
             )
-            search.change(fn=get_leaderboard, inputs=search, outputs=table)
         with gr.TabItem("📖 Metric Definitions"):
             gr.Markdown("""
-            ### 🔍 计量经济学评测指标说明
-            1. **Compilation %**: Agent 生成的代码是否能成功运行通过。
-            2. **Replication %**: 回归系数等结果与原始论文的重合度。
-            3. **Direction %**: **核心指标**。回归系数的正负号预测是否正确。
-            """)
-        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 ) 社区。")
-demo.launch()
-        </gradio-file>
-    </gradio-lite>
 </body>
-</html>

+<!DOCTYPE html>
+<html lang="en">
 <head>
+<meta charset="utf-8">
+<title>Infernet Econometrics Benchmark Leaderboard</title>
+<script type="module" crossorigin src="https://cdn.jsdelivr.net/npm/@gradio/lite/dist/lite.js"></script>
+<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/@gradio/lite/dist/lite.css"/>
+<style>
+body { margin:0; background-color:#f7f8fa; }
+</style>
 </head>
 <body>
+<gradio-lite>
 import gradio as gr
+import pandas as pd
+CSS = """
+.gradio-container {
+    max-width: 1280px !important;
+    margin: 24px auto !important;
+    padding: 0 20px !important;
+    background-color: #ffffff;
+    border-radius: 12px;
+    box-shadow: 0 2px 12px rgba(0,0,0,0.06);
+}
+.stat-card {
+    background: linear-gradient(135deg,#f0f4f9,#ffffff);
+    padding:20px 12px;
+    border-radius:12px;
+    text-align:center;
+    border:1px solid #e2e8f0;
+    box-shadow:0 1px 4px rgba(0,0,0,0.04);
+}
+.stat-val {
+    font-size:28px;
+    font-weight:700;
+    color:#2b549c;
+    margin-bottom:4px;
+}
+.stat-label {
+    font-size:14px;
+    color:#475569;
+}
+h1 {
+    color:#1e293b;
+}
+"""
+def load_data():
+    df = pd.read_csv("results.csv")
+    df["Overall Score"] = df[["Compilation %","Replication %","Direction %"]].mean(axis=1).round(2)
+    return df.sort_values("Overall Score", ascending=False).reset_index(drop=True)
+with gr.Blocks(css=CSS, title="Infernet Leaderboard") as demo:
+    gr.HTML("""
+    <div style="text-align:center;padding-top:28px;padding-bottom:12px;">
+        <h1 style="margin:0 0 8px 0;color:#0f172a;">🏆 Infernet Econometrics Benchmark Leaderboard</h1>
+        <p style="color:#475569;font-size:15px;">
+            CamoAiLab, HKU&nbsp;&nbsp;|&nbsp;&nbsp;
+            Dataset: <a target="_blank" href="https://huggingface.co/datasets/CamoAiLab/Infernet">CamoAiLab/Infernet</a>
+            &nbsp;&nbsp;Paper: <a target="_blank" href="https://arxiv.org/abs/2506.00856">arXiv:2506.00856</a>
+        </p>
+    </div>
+    """)
+    df_init = load_data()
     with gr.Row():
+        with gr.Column():
+            gr.HTML(f'<div class="stat-card"><div class="stat-val">{len(df_init)}</div><div class="stat-label">Total Evaluated Models</div></div>')
+        with gr.Column():
+            gr.HTML(f'<div class="stat-card"><div class="stat-val">{df_init["Overall Score"].max()}%</div><div class="stat-label">Best Overall Score</div></div>')
+        with gr.Column():
+            gr.HTML(f'<div class="stat-card"><div class="stat-val">1000</div><div class="stat-label">Test Instances</div></div>')
+    gr.Markdown("<hr style='margin:24px 0'>")
     with gr.Tabs():
         with gr.TabItem("📊 Main Leaderboard"):
+            search_box = gr.Textbox(
+                placeholder="🔍 Search model name ...",
+                label=None,
+                show_label=False
+            )
             table = gr.Dataframe(
+                value=df_init,
+                interactive=False,
+                wrap=True
             )
+            def filter_data(query):
+                full = load_data()
+                if not query:
+                    return full
+                return full[full["Model"].str.contains(query, case=False)]
+            search_box.change(filter_data, inputs=search_box, outputs=table)
         with gr.TabItem("📖 Metric Definitions"):
             gr.Markdown("""
+### Evaluation Metrics Explanation
+1. **Compilation %**
+Code compilation success rate: proportion of generated econometric code that runs without runtime errors.
+2. **Replication % (Partial Replication)**
+Partial replication rate: proportion of tasks where model partially reproduces target estimation outputs.
+3. **Direction % (Correct Coefficient Direction)**
+Core metric: accuracy for predicting correct positive/negative sign of target regression coefficients.
+4. **Overall Score**
+Simple average of above three metrics for comprehensive comparison.
+""")
+        with gr.TabItem("✉️ Submit Results"):
+            gr.Markdown("""
+To submit your model results:
+1. Open a **Discussion** on dataset page: [CamoAiLab/Infernet](https://huggingface.co/datasets/CamoAiLab/Infernet)
+2. Provide model name, scores and reproduction evidence/logs.
+3. After verification, we will update the leaderboard.
+""")
+demo
+</gradio-lite>
+<gradio-file name="results.csv" url="./results.csv"/>
 </body>
+</html>