Reproduce “CircuitWeave: Topology-Behavior Alignment for Executable Multimodal RTL Generation” 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
Jiahao Feng, Haiyan Qin, Zhiwei Xie, Wang Kang
CircuitWeave generates RTL by extracting topology from a schematic and behavior from text, then reconciling both into a contract that conditions generation. In a two-state Moore FSM example, its image-conditioned output follows the current state while the text-only output lags one cycle. On VerilogEval-Human pass@1, adapted Qwen3.5-4B scores 46.60% with schematics versus 38.14% with text; this does not establish visual grounding.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.