Official PyTorch implementation of SIGIR 2024 paper TRAD: Enhancing LLM Agents with Step-Wise Thought Retrieval and Aligned Decision
Tested on python 3.8 environment:
- openai == 0.27.8
- alfworld == 0.2.2
- lxml == 4.7.1
- langchain == 0.0.285
- pytorch == 2.0.1
- transformers == 4.33.2
- sentence-transformers == 2.2.2
See ALFWorld/ and Mind2Web/ folders for details about our experiments on these two environments.
Please cite our paper and star this repo if you use TRAD and find it interesting/helpful for your work, we'd appreciate it! Feel free to contact skyriver@sjtu.edu.cn or open an issue if you have any questions.
@inproceedings{zhou2024trad,
author={Ruiwen Zhou and Yingxuan Yang and Muning Wen and Ying Wen and Wenhao Wang and Chunling Xi and Guoqiang Xu and Yong Yu and Weinan Zhang},
title={{TRAD}: Enhancing LLM Agents with Step-Wise Thought Retrieval and Aligned Decision},
booktitle={Proceedings of the 47th International {ACM} {SIGIR} Conference on Research and Development in Information Retrieval ({SIGIR})},
year={2024},
}