Contributing to LLM Context Window Benchmark

Thanks for taking the time to contribute! This project aims to track and compare LLM coding capabilities through interactive demonstrations.

📋 Guidelines

Model Outputs:
- Provide only the generated index.html file in the model subdirectory.
- Ensure the output is relatively self-contained.
- If the model generated a single HTML file with CSS and JS included, that's preferred.
Structure:
- Follow the directory structure: benchmark/model/index.html.
- Include the matching prompt.txt for any new benchmark.
Model Identifiers: Use lowercase names for folders (e.g., gpt-4o, gemini-1.5-pro).
No Malicious Code: Submissions must not contain any scripts tracking users or conducting malicious activity.

Fork the repository and create your feature branch.
Add your model result or benchmark.
Run bash create_config.sh to update config.json.
Commit your changes (including the updated config.json).
Test Locally: Open index.html in your browser and verify your new entry appears correctly.
Submit a PR: Provide a brief description of the model or benchmark you added.

We use a simple bash script to generate the site index. Before submitting, ensure:

bash create_config.sh

runs without errors and the resulting config.json is valid.

If you have questions or want to discuss a specific benchmark, feel free to open an Issue!