README.md

metadata

language:
  - en
library_name: transformers
tags:
  - chemistry
  - molecular-editing
  - agentic-rl
license: apache-2.0
base_model:
  - Qwen/Qwen2.5-7B-Instruct

MolEditAgent-7B

MolEditAgent is the Stage-1 model of the MolAct framework. It is fine-tuned to master basic molecular editing primitives within an agentic loop.

Key Features

• Actionable Operations: Learns to perform Add, Delete, and Substitute operations on SMILES.
• Tool-Augmented: Trained to interact with chemical tools for validity checking and similarity control.
• Agentic Reasoning: Unlike static generators, it can "Think" and "Call" tools to refine the editing process.

Training & Usage

This model serves as the foundation for property-driven optimization.

• Code: https://github.com/little1d/MolAct
• ArXiv https://arxiv.org/abs/2512.20135

If you use MolAct in your research, please cite:

@article{molact2025,
  title={MolAct: An Agentic RL Framework for Molecular Editing and Property Optimization},
  author={Zhuo Yang and Yeyun Chen and Jiaqing Xie and Ben Gao and Shuaike Shen and Wanhao Liu and Liujia Yang and Beilun Wang and Tianfan Fu and Yuqiang Li},
  year={2025},
  eprint={2512.20135},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2512.20135}
}