Skip to content

Latest commit

 

History

History
94 lines (68 loc) · 3.34 KB

README.md

File metadata and controls

94 lines (68 loc) · 3.34 KB

LLM Leaderboard Explorer

An interactive dashboard for exploring and visualizing merged data from LLM leaderboards, built with Gradio. Check it out our deployed HuggingFace Space: Link

📊 Overview

This application provides an interactive interface to view, filter, and compare Large Language Models (LLMs) based on aggregated data from prominent leaderboard sources:

  • LiveBench: Features performance metrics like Global Average, Reasoning, Coding, Mathematics, Data Analysis, Language, and Instruction Following scores.
  • LMSYS Chatbot Arena: Includes community-based Elo ratings (Arena Score), rankings, and voting data.

The dashboard allows users to easily navigate and compare models across various metrics and categories.

✨ Features

  • Interactive Data Tables: View LLM data organized into tabs:
    • Performance Metrics: Core benchmark scores from LiveBench.
    • Model Details: Information like Organization, License, Knowledge Cutoff, and links.
    • Community Stats: Data from the Chatbot Arena Leaderboard (Ranks, Score, Votes).
    • Model Mapping: Shows the unified model name alongside original names from LiveBench and Arena.
  • Filtering: Dynamically filter the displayed models by:
    • Search term (searches Model Name and Organization).
    • Organization.
    • Minimum Global Average score.
  • Detailed Model Card: Click on any row in the data tables to view a comprehensive card summarizing all metrics for that specific model.
  • Visualizations Tab:
    • Bar Chart: Compare the top 15 models based on a user-selected metric (e.g., Global Average, Arena Score, Coding Average).
    • Radar Chart: Select multiple models (up to 5) to compare their performance profile across key metrics (Reasoning, Coding, Math, Data Analysis, Language, IF Average, and scaled Arena Score).

💾 Data

The application uses a pre-merged CSV file (data/merged_leaderboards.csv) containing data aggregated from the sources mentioned above.

🚀 Getting Started

Prerequisites

  • Python 3.9+
  • pip (Python package installer)

Installation

  1. Clone the repository (Optional):

    # If you have the code in a git repository
git clone <your-repo-url>
cd <your-repo-directory>

    If you just have the files, navigate to the project directory in your terminal.

  2. Install Dependencies: Create a requirements.txt file with the following content:

    gradio==4.9.0
pandas
plotly
numpy

    Then, install the requirements:

    pip install -r requirements.txt

Running the Application

To run the application locally:

python app.py

The application will typically be available at http://127.0.0.1:7860 in your web browser.

📁 Project Structure

GTLLMZoo2
├─ app.py                  # Main Gradio application entry point
├─ requirements.txt        # Python dependencies
├─ data
│  └─ merged_leaderboards.csv # Merged leaderboard data
└─ src
   ├─ data_processing.py  # Data loading and filtering logic
   └─ ui.py               # Gradio UI definition and logic

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request if you have improvements or bug fixes.

📄 License

MIT License