Reproduce “A Survey on LLM-Integrated Hardware Design Verification” 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
Hao Zheng, Jaime Rafael Imperial, Bardia Nadimi, Xiangfei Kong
This survey organizes research on integrating large language models into hardware functional verification, covering assertion and testbench generation, debugging and RTL repair, and formal verification. It finds that LLMs are most useful for interpreting requirements, proposing candidates, and coordinating workflows when their outputs are checked by simulation, coverage, or formal tools; it also highlights persistent gaps in specification fidelity and evaluation comparability.
Key insights
- Executable feedback can improve generated verification artifacts, but successful compilation, simulation, or proof does not by itself show that an artifact captures the intended specification.
- Assertions, testbenches, and repairs face distinct semantic challenges: properties may be misaligned with requirements, coverage may not reveal missed bugs, and patches may pass limited tests while causing other errors.
- Formal tools provide checkable evidence for specified models and assumptions; LLM-generated explanations or proof strategies are not themselves proofs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.