Reproduce “Teaching Language Models Mechanistic Explainability Through MechSMILES” 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
Théo A. Neukomm, Zlatko Jončev, Philippe Schwaller
ChRIMP reconstructs electron-pushing mechanisms from supplied species and a product, letting reviewers inspect synthesis steps. Its MechSMILES representation and environment apply electron moves while enforcing atom and charge conservation. On FlowER Task 4, a T5 model recovered complete pathways for 93.16% of test reactions with greedy decoding and 97.58% with beam width 3; scores match annotations, not experimental feasibility.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.