AGILE

AGILE: AGent that Interacts and Learns from Environments
Code for NeurIPS 2024 paper: AGILE: A Novel Reinforcement Learning Framework of LLM Agents

Getting Started

For a simple run of AGILE on ProductQA dataset

sh examples/test_productQA.sh

Prerequisites

We recommend to use the following environment:

GPU=A100
CUDA_version=12.1
python=3.9.2
torch=2.1.0+cu121
transformers=4.37.2
accelerate=0.26.1
sqlite3

Contributing

Peiyuan Feng, Yichen He, Guanhua Huang, Yuan Lin

Versioning

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Citation

If you find it useful, please cite our work.

@inproceedings{
fengpeiyuan2024agile,
title={{AGILE}: A Novel Reinforcement Learning Framework of {LLM} Agents},
author={FengPeiyuan and Yichen He and Guanhua Huang and Yuan Lin and Hanchong Zhang and Yuchen Zhang and Hang Li},
booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
year={2024},
url={https://openreview.net/forum?id=Ul3lDYo3XQ}
}

Acknowledgments