Upload 7 files

.gitignore

@@ -0,0 +1 @@
+__pycache__

Dockerfile

@@ -0,0 +1,36 @@
+# --- Base image ---
+FROM python:3.11-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PIP_NO_CACHE_DIR=on
+
+# OS deps (git useful for HF repos; curl for HEALTHCHECK)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    git curl && \
+    rm -rf /var/lib/apt/lists/*
+
+# Non-root user
+RUN useradd -m -u 1000 user
+USER user
+ENV PATH="/home/user/.local/bin:$PATH"
+
+WORKDIR /app
+
+# Python deps
+COPY --chown=user ./requirements.txt requirements.txt
+RUN pip install --upgrade pip && pip install -r requirements.txt
+
+# App source
+COPY --chown=user . /app
+
+EXPOSE 7860
+
+# Healthcheck pings your FastAPI health endpoint
+HEALTHCHECK --interval=30s --timeout=3s --start-period=20s --retries=3 \
+  CMD curl -fsS http://localhost:7860/api/health || exit 1
+
+# Production server (Gunicorn + Uvicorn worker)
+#CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
+
+CMD ["gunicorn", "-k", "uvicorn.workers.UvicornWorker", "-w", "2", "-b", "0.0.0.0:7860", "server:app"]

README.md

@@ -1,13 +1,9 @@
----
-title: NL2SQL
-emoji: ⚡
-colorFrom: yellow
-colorTo: red
-sdk: gradio
-sdk_version: 6.2.0
-app_file: app.py
-pinned: false
-license: apache-2.0
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+---
+title: Table_discovery
+emoji: 📈
+colorFrom: red
+colorTo: blue
+sdk: docker
+pinned: false
+---
+

data.json

@@ -0,0 +1,88 @@
+{
+  "datasets": {
+    "birddev": {
+      "name": "BIRD Dev",
+      "metrics": [
+        "execution_ability",
+        "correct_rate",
+        "execution_efficiency_on_executable_sql",
+        "table_precision_on_executable_sql",
+        "table_recall_on_executable_sql",
+        "table_f1_on_executable_sql",
+        "column_precision_on_executable_sql",
+        "column_recall_on_executable_sql",
+        "column_f1_on_executable_sql"
+      ],
+      "algorithms": [
+        {
+          "name": "Ferry Qwen3-4B",
+          "execution_ability": 0.3123,
+          "correct_rate": 0.146,
+          "execution_efficiency_on_executable_sql": 1.0454,
+          "table_precision_on_executable_sql": 0.9426,
+          "table_recall_on_executable_sql": 0.8432,
+          "table_f1_on_executable_sql": 0.8959,
+          "column_precision_on_executable_sql": 0.8524,
+          "column_recall_on_executable_sql": 0.7012,
+          "column_f1_on_executable_sql": 0.7878
+        },
+        {
+          "name": "Ferry Qwen3-8B",
+          "execution_ability": 0.0632,
+          "correct_rate": 0.0332,
+          "execution_efficiency_on_executable_sql": 0.9659,
+          "table_precision_on_executable_sql": 1,
+          "table_recall_on_executable_sql": 0.9485,
+          "table_f1_on_executable_sql": 0.9656,
+          "column_precision_on_executable_sql": 0.9273,
+          "column_recall_on_executable_sql": 0.7618,
+          "column_f1_on_executable_sql": 0.8387
+        },
+        {
+          "name": "Ferry Qwen3-30B",
+          "execution_ability": 0.8598,
+          "correct_rate": 0.4993,
+          "execution_efficiency_on_executable_sql": 0.9939,
+          "table_precision_on_executable_sql": 0.9644,
+          "table_recall_on_executable_sql": 0.9325,
+          "table_f1_on_executable_sql": 0.9476,
+          "column_precision_on_executable_sql": 0.8959,
+          "column_recall_on_executable_sql": 0.8386,
+          "column_f1_on_executable_sql": 0.8734
+        }
+      ]
+    },
+    "birdtrain": {
+      "name": "BIRD Train",
+      "metrics": [
+        "execution_ability",
+        "correct_rate",
+        "execution_efficiency_on_executable_sql",
+        "table_precision_on_executable_sql",
+        "table_recall_on_executable_sql",
+        "table_f1_on_executable_sql",
+        "column_precision_on_executable_sql",
+        "column_recall_on_executable_sql",
+        "column_f1_on_executable_sql"
+      ],
+      "algorithms": []
+    },
+    "spider": {
+      "name": "Spider",
+      "metrics": [
+        "execution_ability",
+        "correct_rate",
+        "execution_efficiency_on_executable_sql",
+        "table_precision_on_executable_sql",
+        "table_recall_on_executable_sql",
+        "table_f1_on_executable_sql",
+        "column_precision_on_executable_sql",
+        "column_recall_on_executable_sql",
+        "column_f1_on_executable_sql"
+      ],
+      "algorithms": [
+      ]
+    }
+  },
+  "last_updated": "2025-12-23T00:00:00"
+}

index.html

@@ -0,0 +1,899 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>NL2SQL Leaderboard</title>
+    <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.4.0/css/all.min.css">
+    <style>
+        * {
+            margin: 0;
+            padding: 0;
+            box-sizing: border-box;
+            font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, sans-serif;
+        }
+
+        body {
+            background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+            min-height: 100vh;
+            padding: 20px;
+        }
+
+        .container {
+            max-width: 1400px;
+            margin: 0 auto;
+            background: #f8fafc;
+            border-radius: 20px;
+            box-shadow: 0 20px 60px rgba(0, 0, 0, 0.3);
+            overflow: hidden;
+        }
+
+        header {
+            background: linear-gradient(to right, #1e40af, #3b82f6);
+            color: white;
+            padding: 30px 40px;
+            text-align: center;
+        }
+
+        h1 {
+            font-size: 2.8rem;
+            margin-bottom: 10px;
+            font-weight: 800;
+        }
+
+        .subtitle {
+            font-size: 1.2rem;
+            opacity: 0.9;
+            margin-bottom: 20px;
+        }
+
+        .tabs {
+            display: flex;
+            background: rgba(255, 255, 255, 0.1);
+            border-radius: 12px;
+            padding: 5px;
+            margin: 20px auto 0;
+            max-width: 800px;
+        }
+
+        .tab {
+            flex: 1;
+            padding: 15px;
+            text-align: center;
+            font-weight: 600;
+            font-size: 1.1rem;
+            cursor: pointer;
+            border-radius: 8px;
+            transition: all 0.3s ease;
+        }
+
+        .tab:hover {
+            background: rgba(255, 255, 255, 0.2);
+        }
+
+        .tab.active {
+            background: white;
+            color: #3b82f6;
+            box-shadow: 0 4px 12px rgba(0, 0, 0, 0.1);
+        }
+
+        main {
+            padding: 30px;
+        }
+
+        .controls {
+            background: white;
+            padding: 30px;
+            border-radius: 12px;
+            box-shadow: 0 4px 6px rgba(0, 0, 0, 0.05);
+            margin-bottom: 30px;
+        }
+
+        .dataset-selector {
+            margin-bottom: 30px;
+            padding-bottom: 30px;
+            border-bottom: 1px solid #e2e8f0;
+        }
+
+        .metric-selector {
+            margin-bottom: 20px;
+        }
+
+        .dataset-label, .metric-label {
+            font-weight: 600;
+            color: #475569;
+            font-size: 1.1rem;
+            margin-bottom: 15px;
+            display: flex;
+            align-items: center;
+            gap: 10px;
+        }
+
+        .dataset-label i, .metric-label i {
+            color: #3b82f6;
+        }
+
+        .dataset-buttons {
+            display: flex;
+            flex-wrap: wrap;
+            gap: 10px;
+        }
+
+        .dataset-btn {
+            padding: 12px 24px;
+            background: white;
+            border: 2px solid #cbd5e1;
+            border-radius: 8px;
+            font-weight: 600;
+            color: #475569;
+            cursor: pointer;
+            transition: all 0.2s ease;
+        }
+
+        .dataset-btn:hover {
+            background: #f1f5f9;
+            transform: translateY(-2px);
+        }
+
+        .dataset-btn.active {
+            background: #3b82f6;
+            color: white;
+            border-color: #3b82f6;
+            box-shadow: 0 4px 12px rgba(59, 130, 246, 0.3);
+        }
+
+        .metric-checkboxes {
+            display: grid;
+            grid-template-columns: repeat(auto-fill, minmax(250px, 1fr));
+            gap: 15px;
+            margin-top: 10px;
+        }
+
+        .metric-checkbox {
+            display: flex;
+            align-items: center;
+            gap: 12px;
+            padding: 10px 15px;
+            background: #f8fafc;
+            border-radius: 8px;
+            transition: all 0.2s ease;
+            border: 1px solid #e2e8f0;
+        }
+
+        .metric-checkbox:hover {
+            background: #f1f5f9;
+            transform: translateY(-2px);
+            box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);
+        }
+
+        .metric-checkbox input {
+            width: 18px;
+            height: 18px;
+            cursor: pointer;
+            accent-color: #3b82f6;
+        }
+
+        .metric-checkbox label {
+            font-weight: 500;
+            color: #475569;
+            cursor: pointer;
+            font-size: 0.95rem;
+            flex: 1;
+        }
+
+        .metric-hint {
+            font-size: 0.85rem;
+            color: #64748b;
+            margin-top: 5px;
+            font-style: italic;
+        }
+
+        .table-container {
+            overflow-x: auto;
+            border-radius: 12px;
+            box-shadow: 0 10px 25px rgba(0, 0, 0, 0.1);
+            background: white;
+            margin-bottom: 30px;
+        }
+
+        table {
+            width: 100%;
+            border-collapse: collapse;
+            min-width: 1200px;
+        }
+
+        th {
+            background: linear-gradient(to right, #1e40af, #3b82f6);
+            color: white;
+            text-align: left;
+            padding: 18px 20px;
+            font-weight: 600;
+            font-size: 1.1rem;
+            position: sticky;
+            top: 0;
+        }
+
+        th.algorithm-header {
+            min-width: 200px;
+        }
+
+        td {
+            padding: 16px 20px;
+            border-bottom: 1px solid #e2e8f0;
+            font-size: 1rem;
+        }
+
+        tr:hover {
+            background: #f8fafc;
+        }
+
+        tr:nth-child(even) {
+            background: #f8fafc;
+        }
+
+        tr:nth-child(even):hover {
+            background: #f1f5f9;
+        }
+
+        .algorithm-name {
+            font-weight: 600;
+            color: #1e293b;
+            font-size: 1.1rem;
+        }
+
+        .metric-value {
+            font-weight: 600;
+            color: #0f172a;
+            text-align: center;
+        }
+
+        .highlight {
+            background: linear-gradient(120deg, #f0f9ff 0%, #e0f2fe 100%);
+            border-left: 4px solid #0ea5e9;
+        }
+
+        .best-value {
+            background: linear-gradient(120deg, #dcfce7 0%, #bbf7d0 100%);
+            font-weight: 700;
+            color: #166534;
+            position: relative;
+        }
+
+        .best-value::after {
+            content: "🏆";
+            position: absolute;
+            right: 5px;
+            top: 50%;
+            transform: translateY(-50%);
+            font-size: 0.9rem;
+        }
+
+        .no-data {
+            text-align: center;
+            padding: 60px 20px;
+            color: #64748b;
+            font-size: 1.2rem;
+        }
+
+        .no-data i {
+            font-size: 3rem;
+            margin-bottom: 20px;
+            opacity: 0.5;
+        }
+
+        /* Submit Data Styles */
+        .submit-container {
+            padding: 40px;
+            text-align: center;
+            background: white;
+            border-radius: 12px;
+            box-shadow: 0 10px 25px rgba(0, 0, 0, 0.1);
+            margin-bottom: 30px;
+        }
+
+        .submit-icon {
+            font-size: 5rem;
+            color: #cbd5e1;
+            margin-bottom: 20px;
+        }
+
+        .submit-title {
+            font-size: 2.5rem;
+            color: #64748b;
+            margin-bottom: 15px;
+        }
+
+        .submit-description {
+            font-size: 1.2rem;
+            color: #94a3b8;
+            max-width: 600px;
+            margin: 0 auto 30px;
+            line-height: 1.6;
+        }
+
+        .coming-soon {
+            display: inline-block;
+            background: #f1f5f9;
+            color: #64748b;
+            padding: 10px 25px;
+            border-radius: 30px;
+            font-weight: 600;
+            font-size: 1.1rem;
+        }
+
+        .submit-form {
+            max-width: 600px;
+            margin: 30px auto;
+            background: #f8fafc;
+            padding: 25px;
+            border-radius: 12px;
+            text-align: left;
+        }
+
+        .form-group {
+            margin-bottom: 20px;
+        }
+
+        .form-group label {
+            display: block;
+            font-weight: 600;
+            color: #475569;
+            margin-bottom: 8px;
+        }
+
+        .form-group input,
+        .form-group select,
+        .form-group textarea {
+            width: 100%;
+            padding: 12px;
+            border: 2px solid #cbd5e1;
+            border-radius: 8px;
+            font-size: 1rem;
+            transition: border-color 0.2s ease;
+        }
+
+        .form-group input:focus,
+        .form-group select:focus,
+        .form-group textarea:focus {
+            outline: none;
+            border-color: #3b82f6;
+        }
+
+        .submit-btn {
+            background: #3b82f6;
+            color: white;
+            border: none;
+            padding: 12px 30px;
+            border-radius: 8px;
+            font-weight: 600;
+            font-size: 1.1rem;
+            cursor: pointer;
+            transition: all 0.2s ease;
+        }
+
+        .submit-btn:hover {
+            background: #2563eb;
+            transform: translateY(-2px);
+            box-shadow: 0 4px 12px rgba(37, 99, 235, 0.3);
+        }
+
+        /* Footer */
+        footer {
+            text-align: center;
+            padding: 20px;
+            color: #94a3b8;
+            font-size: 0.9rem;
+            border-top: 1px solid #e2e8f0;
+            margin-top: 30px;
+        }
+
+        .controls-header {
+            display: flex;
+            justify-content: space-between;
+            align-items: center;
+            margin-bottom: 25px;
+            padding-bottom: 15px;
+            border-bottom: 1px solid #e2e8f0;
+        }
+
+        .controls-title {
+            font-size: 1.4rem;
+            color: #1e293b;
+            font-weight: 700;
+        }
+
+        .reset-btn {
+            background: #f1f5f9;
+            color: #475569;
+            border: 1px solid #cbd5e1;
+            padding: 8px 16px;
+            border-radius: 6px;
+            font-weight: 500;
+            cursor: pointer;
+            transition: all 0.2s ease;
+        }
+
+        .reset-btn:hover {
+            background: #e2e8f0;
+        }
+
+        /* Responsive */
+        @media (max-width: 768px) {
+            .container {
+                border-radius: 10px;
+            }
+            
+            h1 {
+                font-size: 2rem;
+            }
+            
+            .tab {
+                padding: 12px 15px;
+                font-size: 1rem;
+            }
+            
+            .controls {
+                padding: 20px;
+            }
+            
+            .metric-checkboxes {
+                grid-template-columns: 1fr;
+            }
+            
+            th, td {
+                padding: 12px 15px;
+            }
+            
+            .submit-container {
+                padding: 20px;
+            }
+            
+            .submit-title {
+                font-size: 1.8rem;
+            }
+            
+            .controls-header {
+                flex-direction: column;
+                align-items: flex-start;
+                gap: 10px;
+            }
+            
+            .reset-btn {
+                align-self: flex-end;
+            }
+        }
+    </style>
+</head>
+<body>
+    <div class="container">
+        <header>
+            <h1>NL2SQL Leaderboard</h1>
+            <div class="subtitle">Benchmarking Natural Language to SQL Conversion Models and Agents</div>
+            <div class="tabs">
+                <div class="tab active" onclick="switchTab('leaderboard')">Leaderboard</div>
+                <div class="tab" onclick="switchTab('submit')">Submit Results</div>
+                <div class="tab" onclick="switchTab('about')">About</div>
+            </div>
+        </header>
+
+        <main id="leaderboard-content">
+            <div class="controls">
+                <div class="controls-header">
+                    <div class="controls-title">
+                        <i class="fas fa-sliders-h"></i> Filter Controls
+                    </div>
+                    <button class="reset-btn" onclick="resetFilters()">
+                        <i class="fas fa-redo"></i> Reset Filters
+                    </button>
+                </div>
+                
+                <div class="dataset-selector">
+                    <div class="dataset-label">
+                        <i class="fas fa-database"></i>
+                        Select Dataset
+                    </div>
+                    <div class="dataset-buttons" id="dataset-buttons">
+                        <!-- 动态生成数据集按钮 -->
+                    </div>
+                </div>
+                
+                <div class="metric-selector">
+                    <div class="metric-label">
+                        <i class="fas fa-chart-bar"></i>
+                        Select Metrics to Display
+                    </div>
+                    <div class="metric-checkboxes" id="metric-checkboxes">
+                        <!-- 动态生成指标复选框 -->
+                    </div>
+                    <div class="metric-hint">
+                        Tip: Click on metrics to show/hide them in the table. The best value for each metric is highlighted with a 🏆. Metrics Precision/Recall/F1 are calculated only among the sqls that are executable.
+                    </div>
+                </div>
+            </div>
+
+            <div class="table-container">
+                <table id="leaderboard-table">
+                    <thead>
+                        <tr id="table-header">
+                            <!-- 动态生成表头 -->
+                        </tr>
+                    </thead>
+                    <tbody id="table-body">
+                        <!-- 动态生成表格内容 -->
+                    </tbody>
+                </table>
+            </div>
+
+            <div class="no-data" id="no-data-message" style="display: none;">
+                <i class="fas fa-database"></i>
+                <h3>No Data Available</h3>
+                <p>This dataset currently has no results. Be the first to submit!</p>
+                <button class="submit-btn" onclick="switchTab('submit')" style="margin-top: 20px;">
+                    <i class="fas fa-cloud-upload-alt"></i> Submit Results
+                </button>
+            </div>
+        </main>
+
+        <div id="submit-content" style="display: none;">
+            <div class="submit-container">
+                <div class="submit-icon">
+                    <i class="fas fa-cloud-upload-alt"></i>
+                </div>
+                <h2 class="submit-title">Submit Your Results</h2>
+                <p class="submit-description">
+                    Submit your NL2SQL model results to be included in the leaderboard. 
+                    Please ensure your results follow the required format.
+                </p>
+                <div class="coming-soon">
+                    <i class="fas fa-tools"></i> Submission Portal Under Development
+                </div>
+
+                <div class="submit-form">
+                    <div class="form-group">
+                        <label for="dataset-select"><i class="fas fa-database"></i> Select Dataset</label>
+                        <select id="dataset-select" class="dataset-select">
+                            <option value="birddev">BIRD Dev</option>
+                            <option value="birdtrain">BIRD Train</option>
+                            <option value="spider">Spider</option>
+                        </select>
+                    </div>
+
+                    <div class="form-group">
+                        <label for="algorithm-name"><i class="fas fa-robot"></i> Algorithm/Model Name</label>
+                        <input type="text" id="algorithm-name" placeholder="Enter your algorithm name">
+                    </div>
+
+                    <div class="form-group">
+                        <label for="results-json"><i class="fas fa-code"></i> Results (JSON format)</label>
+                        <textarea id="results-json" rows="6" placeholder='{
+  "execution_ability": 0.85,
+  "correct_rate": 0.78,
+  "execution_efficiency_on_executable_sql": 0.92,
+  "table_precision_on_executable_sql": 0.87,
+  "table_recall_on_executable_sql": 0.85,
+  "table_f1_on_executable_sql": 0.86,
+  "column_precision_on_executable_sql": 0.83,
+  "column_recall_on_executable_sql": 0.80,
+  "column_f1_on_executable_sql": 0.815
+}'></textarea>
+                    </div>
+
+                    <div style="text-align: center;">
+                        <button class="submit-btn" onclick="submitResults()">
+                            <i class="fas fa-paper-plane"></i> Submit Results (Demo)
+                        </button>
+                    </div>
+                </div>
+            </div>
+        </div>
+
+        <div id="about-content" style="display: none;">
+            <div class="submit-container">
+                <h2 class="submit-title">About NL2SQL Leaderboard</h2>
+                <p class="submit-description">
+                    This leaderboard tracks the performance of various Natural Language to SQL (NL2SQL) models 
+                    across multiple benchmarks and metrics.
+                </p>
+                
+                <div style="text-align: left; max-width: 800px; margin: 40px auto;">
+                    <h3 style="color: #475569; margin-bottom: 15px;">Metrics Explained:</h3>
+                    <ul style="list-style-type: none; padding: 0;">
+                        <li style="margin-bottom: 10px; padding: 10px; background: #f1f5f9; border-radius: 6px;">
+                            <strong>Execution Ability:</strong> Ability to generate executable SQL queries
+                        </li>
+                        <li style="margin-bottom: 10px; padding: 10px; background: #f1f5f9; border-radius: 6px;">
+                            <strong>Correct Rate:</strong> Percentage of correctly generated SQL queries
+                        </li>
+                        <li style="margin-bottom: 10px; padding: 10px; background: #f1f5f9; border-radius: 6px;">
+                            <strong>Execution Efficiency:</strong> Performance on executable SQL queries
+                        </li>
+                        <li style="margin-bottom: 10px; padding: 10px; background: #f1f5f9; border-radius: 6px;">
+                            <strong>Table/Column Precision/Recall/F1:</strong> Accuracy in selecting correct tables and columns when the sql is executable
+                        </li>
+                    </ul>
+                    
+                    <h3 style="color: #475569; margin-top: 30px; margin-bottom: 15px;">Supported Datasets:</h3>
+                    <ul style="list-style-type: none; padding: 0;">
+                        <li style="margin-bottom: 10px; padding: 10px; background: #f1f5f9; border-radius: 6px;">
+                            <strong>BIRD Dev:</strong> Benchmark for large-scale database tasks in the real world
+                        </li>
+                        <li style="margin-bottom: 10px; padding: 10px; background: #f1f5f9; border-radius: 6px;">
+                            <strong>BIRD Train:</strong> Training set for BIRD benchmark
+                        </li>
+                        <li style="margin-bottom: 10px; padding: 10px; background: #f1f5f9; border-radius: 6px;">
+                            <strong>Spider:</strong> Large-scale complex and cross-domain semantic parsing and text-to-SQL dataset
+                        </li>
+                    </ul>
+                </div>
+            </div>
+        </div>
+
+        <footer>
+            <p>NL2SQL Leaderboard • Last Updated: <span id="last-updated">Loading...</span></p>
+            <p style="margin-top: 5px; font-size: 0.8rem;">
+                <i class="fas fa-code"></i> Built with FastAPI &bull; 
+                <i class="fas fa-server"></i> Docker Ready &bull; 
+                <i class="fas fa-chart-line"></i> Real-time Updates
+            </p>
+        </footer>
+    </div>
+
+    <script>
+        let currentDataset = 'birddev';
+        let datasets = {};
+        let selectedMetrics = new Set();
+        let datasetMetrics = {};
+
+        // 页面加载时获取数据
+        document.addEventListener('DOMContentLoaded', async function() {
+            await loadLeaderboardData();
+            updateLastUpdated();
+        });
+
+        // 加载排行榜数据
+        async function loadLeaderboardData() {
+            try {
+                const response = await fetch('/api/data');
+                const data = await response.json();
+                datasets = data.datasets;
+                
+                // 为每个数据集存储指标
+                for (const [key, dataset] of Object.entries(datasets)) {
+                    datasetMetrics[key] = dataset.metrics;
+                }
+                
+                // 初始化数据集选择按钮
+                initDatasetButtons();
+                
+                // 显示当前数据集
+                showDataset(currentDataset);
+                
+            } catch (error) {
+                console.error('Error loading data:', error);
+                alert('Failed to load leaderboard data');
+            }
+        }
+
+        // 更新最后更新时间
+        async function updateLastUpdated() {
+            try {
+                const response = await fetch('/api/data');
+                const data = await response.json();
+                const date = new Date(data.last_updated);
+                document.getElementById('last-updated').textContent = 
+                    date.toLocaleString();
+            } catch (error) {
+                console.error('Error updating timestamp:', error);
+            }
+        }
+
+        // 初始化数据集按钮
+        function initDatasetButtons() {
+            const container = document.getElementById('dataset-buttons');
+            container.innerHTML = '';
+            
+            for (const [key, dataset] of Object.entries(datasets)) {
+                const button = document.createElement('button');
+                button.className = `dataset-btn ${key === currentDataset ? 'active' : ''}`;
+                button.textContent = dataset.name;  // 移除了图标
+                button.onclick = () => {
+                    currentDataset = key;
+                    showDataset(key);
+                    updateDatasetButtons();
+                };
+                container.appendChild(button);
+            }
+        }
+
+        // 更新数据集按钮状态
+        function updateDatasetButtons() {
+            const buttons = document.querySelectorAll('.dataset-btn');
+            buttons.forEach(button => {
+                // 移除所有active类
+                button.classList.remove('active');
+                
+                // 找到对应的数据集键
+                const buttonText = button.textContent.toLowerCase().replace(/\s+/g, '');
+                let foundKey = null;
+                for (const [key, dataset] of Object.entries(datasets)) {
+                    if (dataset.name.toLowerCase().replace(/\s+/g, '').includes(buttonText.replace(/icon/g, ''))) {
+                        foundKey = key;
+                        break;
+                    }
+                }
+                if (foundKey === currentDataset) {
+                    button.classList.add('active');
+                }
+            });
+        }
+
+        // 显示指定数据集
+        function showDataset(datasetKey) {
+            const dataset = datasets[datasetKey];
+            if (!dataset) return;
+            
+            // 初始化指标选择（如果这是第一次显示这个数据集）
+            if (!selectedMetrics.size || !Array.from(selectedMetrics).some(m => dataset.metrics.includes(m))) {
+                // 重置为选中所有指标
+                selectedMetrics = new Set(dataset.metrics || []);
+            }
+            
+            // 更新指标复选框
+            updateMetricCheckboxes(datasetKey);
+            
+            // 更新表格
+            updateTable(dataset);
+            
+            // 显示/隐藏无数据消息
+            const noDataMsg = document.getElementById('no-data-message');
+            if (dataset.algorithms.length === 0) {
+                noDataMsg.style.display = 'block';
+                document.querySelector('.table-container').style.display = 'none';
+            } else {
+                noDataMsg.style.display = 'none';
+                document.querySelector('.table-container').style.display = 'block';
+            }
+        }
+
+        // 更新指标复选框
+        function updateMetricCheckboxes(datasetKey) {
+            const container = document.getElementById('metric-checkboxes');
+            container.innerHTML = '';
+            
+            const metrics = datasetMetrics[datasetKey] || [];
+            
+            metrics.forEach(metric => {
+                const checkbox = document.createElement('div');
+                checkbox.className = 'metric-checkbox';
+                
+                const inputId = `metric-${metric}`;
+                const isChecked = selectedMetrics.has(metric);
+                
+                checkbox.innerHTML = `
+                    <input type="checkbox" id="${inputId}" ${isChecked ? 'checked' : ''} 
+                        onchange="toggleMetric('${metric}', this.checked)">
+                    <label for="${inputId}">${formatMetricName(metric)}</label>  <!-- 移除了图标 -->
+                `;
+                container.appendChild(checkbox);
+            });
+        }
+
+        // 切换指标显示
+        function toggleMetric(metric, isChecked) {
+            if (isChecked) {
+                selectedMetrics.add(metric);
+            } else {
+                selectedMetrics.delete(metric);
+            }
+            
+            // 重新显示当前数据集（会刷新表格）
+            showDataset(currentDataset);
+        }
+
+        // 格式化指标名称
+        function formatMetricName(metric) {
+            // 先替换_on_executable_sql
+            let name = metric.replace(/_on_executable_sql/g, '');
+            // 将下划线替换为空格
+            name = name.replace(/_/g, ' ');
+            // 首字母大写
+            name = name.replace(/\b\w/g, l => l.toUpperCase());
+            return name;
+        }
+
+        // 重置过滤器
+        function resetFilters() {
+            // 重置为默认数据集
+            currentDataset = 'birddev';
+            
+            // 重置指标选择
+            const dataset = datasets[currentDataset];
+            selectedMetrics = new Set(dataset.metrics || []);
+            
+            // 更新显示
+            showDataset(currentDataset);
+            updateDatasetButtons();
+        }
+
+        // 切换标签页（如果需要，也可以移除控制面板标题中的图标）
+        function switchTab(tabName) {
+            // 更新标签页状态
+            document.querySelectorAll('.tab').forEach(tab => {
+                tab.classList.toggle('active', tab.textContent.toLowerCase().includes(tabName));
+            });
+            
+            // 显示对应的内容
+            document.getElementById('leaderboard-content').style.display = 
+                tabName === 'leaderboard' ? 'block' : 'none';
+            document.getElementById('submit-content').style.display = 
+                tabName === 'submit' ? 'block' : 'none';
+            document.getElementById('about-content').style.display = 
+                tabName === 'about' ? 'block' : 'none';
+        }
+
+        // 更新表格
+        function updateTable(dataset) {
+            const algorithms = dataset.algorithms;
+            const metrics = dataset.metrics || [];
+            
+            // 更新表头
+            updateTableHeader(metrics);
+            
+            // 更新表格内容
+            updateTableBody(algorithms, metrics);
+        }
+
+        // 更新表头
+        function updateTableHeader(metrics) {
+            const header = document.getElementById('table-header');
+            header.innerHTML = '<th class="algorithm-header">Algorithm</th>';  // 移除了图标
+            
+            // 只添加选中的指标
+            metrics.forEach(metric => {
+                if (selectedMetrics.has(metric)) {
+                    header.innerHTML += `<th>${formatMetricName(metric)}</th>`;  // 移除了图标
+                }
+            });
+        }
+
+        // 更新表格内容
+        function updateTableBody(algorithms, metrics) {
+            const tbody = document.getElementById('table-body');
+            tbody.innerHTML = '';
+            
+            if (algorithms.length === 0) return;
+            
+            // 计算每个指标的最佳值
+            const bestValues = {};
+            metrics.forEach(metric => {
+                if (selectedMetrics.has(metric)) {
+                    const values = algorithms.map(a => a[metric] || 0);
+                    bestValues[metric] = Math.max(...values);
+                }
+            });
+            
+            // 填充数据行
+            algorithms.forEach(algorithm => {
+                const row = document.createElement('tr');
+                let rowHtml = `<td class="algorithm-name">${algorithm.name}</td>`;  // 移除了图标
+                
+                metrics.forEach(metric => {
+                    if (selectedMetrics.has(metric)) {
+                        const value = algorithm[metric];
+                        const isBest = value === bestValues[metric] && value !== undefined;
+                        const cellClass = isBest ? 'metric-value best-value' : 'metric-value';
+                        const displayValue = value !== undefined ? value.toFixed(3) : 'N/A';
+                        rowHtml += `<td class="${cellClass}">${displayValue}</td>`;
+                    }
+                });
+                
+                row.innerHTML = rowHtml;
+                tbody.appendChild(row);
+            });
+        }
+
+        // 每30秒刷新数据
+        setInterval(updateLastUpdated, 30000);
+    </script>
+</body>
+</html>

requirements.txt

@@ -0,0 +1,8 @@
+fastapi==0.104.1
+uvicorn[standard]==0.24.0
+gunicorn==21.2.0
+python-multipart==0.0.6
+pydantic==2.5.0
+pydantic-settings==2.1.0
+aiofiles==23.2.1
+Jinja2==3.1.3

server.py

@@ -0,0 +1,225 @@
+from fastapi import FastAPI, Request, HTTPException, UploadFile, File
+from fastapi.responses import HTMLResponse, JSONResponse
+from fastapi.staticfiles import StaticFiles
+from fastapi.templating import Jinja2Templates
+import json
+import os
+from pathlib import Path
+from typing import Dict, List, Optional
+import logging
+from datetime import datetime
+
+# 设置日志
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+app = FastAPI(title="NL2SQL Leaderboard", version="1.0.0")
+
+# 创建必要的目录
+Path("static").mkdir(exist_ok=True)
+
+# 挂载静态文件
+app.mount("/static", StaticFiles(directory="static"), name="static")
+
+# 模板
+templates = Jinja2Templates(directory=".")
+
+# 数据文件路径
+DATA_FILE = "data.json"
+
+def load_data() -> Dict:
+    """加载排行榜数据"""
+    try:
+        with open(DATA_FILE, "r", encoding="utf-8") as f:
+            return json.load(f)
+    except FileNotFoundError:
+        # 返回默认数据结构
+        return {
+            "datasets": {
+                "birddev": {
+                    "name": "BIRD Dev",
+                    "metrics": [
+                        "execution_ability",
+                        "correct_rate", 
+                        "execution_efficiency_on_executable_sql",
+                        "table_precision_on_executable_sql",
+                        "table_recall_on_executable_sql",
+                        "table_f1_on_executable_sql",
+                        "column_precision_on_executable_sql",
+                        "column_recall_on_executable_sql", 
+                        "column_f1_on_executable_sql"
+                    ],
+                    "algorithms": [
+                        {
+                            "name": "Algorithm A",
+                            "execution_ability": 0.85,
+                            "correct_rate": 0.78,
+                            "execution_efficiency_on_executable_sql": 0.92,
+                            "table_precision_on_executable_sql": 0.87,
+                            "table_recall_on_executable_sql": 0.85,
+                            "table_f1_on_executable_sql": 0.86,
+                            "column_precision_on_executable_sql": 0.83,
+                            "column_recall_on_executable_sql": 0.80,
+                            "column_f1_on_executable_sql": 0.815
+                        },
+                        {
+                            "name": "Algorithm B", 
+                            "execution_ability": 0.82,
+                            "correct_rate": 0.75,
+                            "execution_efficiency_on_executable_sql": 0.90,
+                            "table_precision_on_executable_sql": 0.85,
+                            "table_recall_on_executable_sql": 0.82,
+                            "table_f1_on_executable_sql": 0.835,
+                            "column_precision_on_executable_sql": 0.80,
+                            "column_recall_on_executable_sql": 0.78,
+                            "column_f1_on_executable_sql": 0.79
+                        },
+                        {
+                            "name": "Algorithm C",
+                            "execution_ability": 0.88,
+                            "correct_rate": 0.81,
+                            "execution_efficiency_on_executable_sql": 0.94,
+                            "table_precision_on_executable_sql": 0.89,
+                            "table_recall_on_executable_sql": 0.87,
+                            "table_f1_on_executable_sql": 0.88,
+                            "column_precision_on_executable_sql": 0.85,
+                            "column_recall_on_executable_sql": 0.83,
+                            "column_f1_on_executable_sql": 0.84
+                        }
+                    ]
+                },
+                "birdtrain": {
+                    "name": "BIRD Train",
+                    "metrics": [
+                        "execution_ability",
+                        "correct_rate", 
+                        "execution_efficiency_on_executable_sql",
+                        "table_precision_on_executable_sql",
+                        "table_recall_on_executable_sql",
+                        "table_f1_on_executable_sql",
+                        "column_precision_on_executable_sql",
+                        "column_recall_on_executable_sql", 
+                        "column_f1_on_executable_sql"
+                    ],
+                    "algorithms": []  # 空数据，只有架子
+                },
+                "spider": {
+                    "name": "Spider",
+                    "metrics": [
+                        "execution_ability",
+                        "correct_rate"
+                    ],
+                    "algorithms": [
+                        {
+                            "name": "Model X",
+                            "execution_ability": 0.91,
+                            "correct_rate": 0.85
+                        },
+                        {
+                            "name": "Model Y",
+                            "execution_ability": 0.89,
+                            "correct_rate": 0.83
+                        }
+                    ]
+                }
+            },
+            "last_updated": datetime.now().isoformat()
+        }
+
+def save_data(data: Dict):
+    """保存数据到文件"""
+    data["last_updated"] = datetime.now().isoformat()
+    with open(DATA_FILE, "w", encoding="utf-8") as f:
+        json.dump(data, f, indent=2, ensure_ascii=False)
+
+@app.get("/", response_class=HTMLResponse)
+async def read_root(request: Request):
+    """渲染主页面"""
+    data = load_data()
+    return templates.TemplateResponse(
+        "index.html", 
+        {"request": request, "datasets": data["datasets"]}
+    )
+
+@app.get("/api/data")
+async def get_data():
+    """获取排行榜数据"""
+    return load_data()
+
+@app.get("/api/health")
+async def health_check():
+    """健康检查端点"""
+    return {"status": "healthy", "timestamp": datetime.now().isoformat()}
+
+@app.post("/api/submit")
+async def submit_results(
+    dataset: str,
+    algorithm_name: str,
+    results: str  # JSON字符串
+):
+    """
+    提交新结果的端点（架子功能）
+    实际使用时应该验证和解析results
+    """
+    try:
+        # 这里只是架子，实际应该验证和存储数据
+        results_data = json.loads(results)
+        logger.info(f"Received submission for {dataset} - {algorithm_name}")
+        
+        # 在实际实现中，这里应该：
+        # 1. 验证数据格式
+        # 2. 存储到数据库或文件
+        # 3. 更新排行榜
+        
+        return {
+            "status": "success",
+            "message": "Submission received (demo mode - not actually saved)",
+            "dataset": dataset,
+            "algorithm": algorithm_name,
+            "received_results": results_data
+        }
+    except json.JSONDecodeError:
+        raise HTTPException(status_code=400, detail="Invalid JSON format")
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+@app.post("/api/submit/file")
+async def submit_results_file(
+    dataset: str,
+    algorithm_name: str,
+    file: UploadFile = File(...)
+):
+    """
+    通过文件提交结果的端点
+    """
+    try:
+        # 读取文件内容
+        content = await file.read()
+        results = json.loads(content)
+        
+        logger.info(f"Received file submission for {dataset} - {algorithm_name}")
+        
+        return {
+            "status": "success",
+            "message": "File submission received (demo mode)",
+            "dataset": dataset,
+            "algorithm": algorithm_name,
+            "filename": file.filename,
+            "file_size": len(content)
+        }
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+@app.get("/api/datasets")
+async def get_datasets():
+    """获取所有数据集信息"""
+    data = load_data()
+    return {
+        "datasets": list(data["datasets"].keys()),
+        "details": {k: {"name": v["name"], "count": len(v["algorithms"])} 
+                   for k, v in data["datasets"].items()}
+    }
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run("server:app", host="0.0.0.0", port=7860, reload=True)