Reproduce “GI-RTL-Fuzz: Grammar-Informed RTL Generation with LLMs for End-to-End Differential Simulator Fuzzing” 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

Ziyue Zheng, Zhi Qu, Yangdi Lyu

GI-RTL-Fuzz generates SystemVerilog tests by steering a large language model toward undercovered grammar contexts, then compares their diagnostics and behavior across simulators. In 1,000-case GPT-4 runs, three-node guidance reached 51.92% grammar coverage and 41 behavioral mismatches, versus 35.12% and 23 without guidance. The paper reports 16 discrepancy entries, 15 marked new; mismatches are not necessarily bugs.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.