MicroEvo couples LLMs with Monte Carlo Tree Search to optimize microarchitecture design spaces, improving Pareto-front quality by 36.2% over NSGA-II. The framework uses evolutionary operators and active knowledge accumulation to manage expensive PPA evaluations.
HOW THIS AFFECTS YOU
●
builderYou can use this method to automate and accelerate hardware design space exploration.
●
researcherYou can apply MCTS-coupled LLM sampling to complex, multi-objective optimization problems.