| Paper | Blog | Discord | X | Roblox Game
This repo features the backend code for the Roblox game we built, AI Space Escape, offering an unique experience to reason with AI. We design evaluation techniques to rank state-of-the-art large language models (LLMs). Our mission is to enable engaging gameplay while evaluating a variety of large-scale AI models and systems.
- Clone this repository:
git clone https://github.com/lmgame-org/game-arena-engine.git
cd game-arena-engine
- Install dependency:
python -m venv ./venv
.\venv\Scripts\activate.bat
pip install -r requirements.txt
- Navigate to
game-arena-engine/src/config, fill inYOUR_API_KEY.
- Launch your backend with
python ./src/serve.py
- Use ngrok to perform port forwarding to make this accessible on WAN.
If you find this repository helpful, Please kindly cite:
article{hu2024gamearena,
title={GameArena: Evaluating LLM Reasoning through Live Computer Games},
author={Hu, Lanxiang and Li, Qiyu and Xie, Anze and Jiang, Nan and Stoica, Ion and Jin, Haojian and Zhang, Hao},
journal={arXiv preprint arXiv:2412.06394},
year={2024}
}