Reproduce “MicroEvo: Knowledge-Guided LLM Sampling for Efficient Microarchitecture Design Space Exploration” 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
Jia Xiong, Runkai Li, Chenxu Niu, Guangyuan Gao, Changwen Xing, Yifan Zhang, Xinlai Wan, Jieran Cui and 6 more
MicroEvo uses an LLM within a feedback-driven Monte Carlo Tree Search to propose and revise processor configurations, selecting candidates with Pareto-UCT and evaluating their performance, power, and area through simulation and modeling. In an out-of-order RISC-V evaluation, Gemini-3-pro achieved mean HV 0.7936 ± 0.0059 at 20 evaluations, versus 0.5826 ± 0.0439 for NSGA-II; results reflect simulated/modelled PPA, not fabricated processors.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.