Reproduce “Reaction precedents extend retrosynthetic prediction beyond training coverage” 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
Jiaxing Sun, Xuli Shen, Jiang Wu, Ruiyi Ding, Ruikang Zhang, Xiao An, Xibao Li, Yubin Wang and 12 more
The authors test whether inference-time reaction retrieval can extend retrosynthesis coverage without updating a trained model. The model receives five product-similar reaction precedents, and a fixed-checkpoint experiment changes only the retrieval corpus. Across three template partitions, withheld-template Top-1 exact-match accuracy rose from 13.0% to 58.0% historically and 4.9% to 23.1% in 2025, while retained-template accuracy dipped modestly.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.