Publication
Chem-R: Learning to Reason as a Chemist
In brief
A three-phase chemical reasoning model spanning molecular- and reaction-level tasks.
Cite this work
Citation
Weida Wang, Benteng Chen, Di Zhang, Wanhao Liu, Shuchen Pu, Ben Gao, Jin Zeng, Xiaoyong Wei, Tianshu Yu, Shuzhou Sun, Tianfan Fu, Wanli Ouyang, Lei Bai, Jiatong Li, Zifu Wang, Yuqiang Li, Shufei Zhang. (2026). Chem-R: Learning to Reason as a Chemist. KDD 2026.
About this work
Chem-R learns chemical reasoning through three training stages: chemical foundation training, reasoning-protocol distillation, and multi-task Group Relative Policy Optimization.
The model is evaluated on molecular and reaction tasks. The reported gains over leading language models, including Gemini-2.5-Pro and DeepSeek-R1, reach up to 32% on molecular tasks and 48% on reaction tasks.