Reproduce “MolHop: Scaffold-Hopping Molecular Design with Agentic Reinforcement Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yanyi Su, Chenyan Wang, Yulei Gong, Yeyun Chen, Xiaohui Yang, Jun Cheng
MolHop is an agentic framework that navigates commercial-molecule scaffold hierarchies, edits candidates, and uses trajectory memory to match properties after structural changes. On the 200-molecule ChEMBL test set, frozen DeepSeek-Flash-V4.1’s joint@H1 success rises from 15.5% to 52.5% with the harness. The comparison combines scaffold search, tools, and feedback, so it does not isolate their effects or demonstrate biological activity or synthesis.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.