RL-MemoryAgent-14B

Model Description

The RL-MemAgent-14B is a part of the MemAgent framework, which enables Large Language Models (LLMs) to process arbitrarily long texts through end-to-end Reinforcement Learning without altering their core architecture.

Usage

This model is ideal for tasks requiring the understanding and processing of very long documents, such as comprehensive question answering, summarizing extensive reports, or analyzing large codebases.

For detailed instructions on how to use, evaluate, and train models within the MemAgent framework, please refer to the main MemAgent GitHub repository.

Links

• Paper: https://arxiv.org/abs/2507.02259
• Blog: https://memagent-sialab.github.io/
• GitHub: https://github.com/BytedTsinghua-SIA/MemAgent

Citation

If you find this work useful, please consider citing our paper:

@article{yu2025memagent,
  title={MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent},
  author={Yu, Hongli and Chen, Tinghong and Feng, Jiangtao and Chen, Jiangjie and Dai, Weinan and Yu, Qiying and Zhang, Ya-Qin and Ma, Wei-Ying and Liu, Jingjing and Wang, Mingxuan and others},
  journal={arXiv preprint arXiv:2507.02259},
  year={2025}
}