A web application to display intelligence evaluation scores for open-source AI models, similar to the structure on artificialanalysis.ai/models/open-source.

• 📊 Display model performance across multiple benchmarks (MMLU, AALCR, SciCode, τ²-Bench, Telecom, etc.)
• 🔍 Search and filter models
• 📈 Color-coded scores (high/medium/low performance)
• 🎨 Modern, responsive UI
• 🔄 Refresh data functionality

Test_AA/
├── index.html          # Main web interface
├── styles.css          # Styling
├── app.js             # Frontend JavaScript
├── data_fetcher.py    # Data fetching script (sample data)
├── api_client.py      # API client for artificialanalysis.ai
├── web_scraper.py     # Web scraper for the website
├── evaluations_data.json  # Data file (generated)
└── README.md          # This file

1. Install dependencies:

   pip3 install requests beautifulsoup4

2. Fetch data:

   # Try API first (recommended)
   python3 api_client.py
   
   # Or try web scraping (for JavaScript-rendered pages)
   python3 web_scraper.py
   
   # Or use Selenium scraper (requires ChromeDriver)
   pip install selenium
   python3 selenium_scraper.py
   
   # Or use sample data (fallback)
   python3 data_fetcher.py

3. Open the web interface:

   • Option 1: Use the included server (recommended):

     python3 server.py

     This will automatically open your browser

   • Option 2: Use Python's built-in server:

     python3 -m http.server 8000

     Then visit http://localhost:8000

   • Option 3: Simply open index.html directly in your browser

Your API key is configured in the scripts:

• api_client.py: aa_OXmwOTJvjVHpPnJQsOgimbFMwsPoVgOT
• data_fetcher.py: aa_OXmwOTJvjVHpPnJQsOgimbFMwsPoVgOT

• MMLU - Massive Multitask Language Understanding
• AALCR - AALCR Benchmark
• SciCode - Scientific Code Understanding
• τ²-Bench - Tau Squared Benchmark
• Telecom - Telecom Benchmark
• HellaSwag - HellaSwag Benchmark
• ARC - AI2 Reasoning Challenge
• TruthfulQA - TruthfulQA Benchmark
• GSM8K - Grade School Math 8K
• Winogrande - Winogrande Benchmark

The evaluations_data.json file follows this structure:

{
  "benchmarks": [
    {
      "id": "mmlu",
      "name": "MMLU",
      "full_name": "Massive Multitask Language Understanding"
    }
  ],
  "models": [
    {
      "id": "deepseek-v3.2",
      "name": "DeepSeek V3.2",
      "provider": "DeepSeek",
      "scores": {
        "mmlu": 0.85,
        "aalcr": 0.69,
        "scicode": 0.42
      }
    }
  ]
}

1. View Evaluations:

   • Open index.html in your browser
   • The table displays all models with their scores

2. Search Models:

   • Use the search box to filter models by name or provider

3. Refresh Data:

   • Click the "Refresh Data" button to reload from evaluations_data.json

To get actual scores from artificialanalysis.ai:

1. API Method:

   • Update api_client.py with correct API endpoints if they differ
   • Run python3 api_client.py

2. Scraping Method:

   • If the website uses JavaScript rendering, you may need Selenium/Playwright
   • Run python3 web_scraper.py

3. Manual Method:

   • Manually update evaluations_data.json with real scores
   • The web interface will automatically display the updated data

• Add more benchmarks: Edit the benchmarks list in data_fetcher.py or evaluations_data.json
• Add more models: Add entries to the models array in evaluations_data.json
• Styling: Modify styles.css to change colors, fonts, or layout
• Functionality: Extend app.js for additional features

• The API endpoints may require authentication or have different URLs
• If the website structure changes, update web_scraper.py accordingly
• Scores are displayed as percentages (0.85 = 85%)
• Color coding: Green (≥80%), Yellow (60-79%), Red (<60%)

This project is for educational and evaluation purposes.