Reproduce “SpecEvo: Speculative Evolution with Large Language Models for Cost-Efficient Scientific Discovery” 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
Ha Minh Hieu, Tongyao Zhu, Do Xuan Long, Min-Yen Kan
SpecEvo coordinates evolutionary program search with an archive of competitive programs and a history of rejected candidates and execution failures. Lightweight models use this evidence to guide proposals, while a frontier model periodically redirects search. Under 500 model calls, it performs best on five of six tasks, using the frontier model for 2.4% of calls; LEVI’s low share shows frequency alone is insufficient.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.