Evidence-Guided Repository-Level RTL Repair

2026-10-02

AI Summary

This study addresses the challenges of fault localization and the absence of patch consistency verification in cross-file RTL repair by proposing an evidence-guided automated repair framework. Moving beyond the limitations of conventional source-code-only reasoning, this work introduces an execution-evidence-driven mechanism that achieves precise fault localization through failed run reproduction and waveform anchor guidance. Furthermore, it incorporates a consistency-aware module to enable closed-loop verification from local seeds to globally coherent patches. Evaluated on the HWE-Bench benchmark, the proposed approach significantly outperforms existing baselines, establishing a new paradigm for the automated debugging of complex hardware designs.

Abstract

Repository-level RTL repair must localize a failure that spans files, modules, and clock cycles, then propagate the fix consistently. Existing methods reason over source code, which reveals possible behaviors but not the failed execution, and cannot tell whether a local fix was propagated consistently. We therefore present an evidence-guided framework with three modules. Failure grounding converts a problem statement into a reproduced failing run and a failure anchor. Waveform-guided localization uses a localization toolbox to narrow the observed violation into a candidate mechanism and an evidence trail. Consistency-aware repair and validation then expand that seed into a coordinated patch and replay the same scenario to check that the violation disappears. We conducted experiments on HWE-Bench and achieved better performance than the baseline.

Problem

• RTL repair
• repository-level debugging
• failure localization
• hardware verification

Innovation

• Repository-Level RTL Repair
• Evidence-Guided Framework
• Waveform-Guided Localization
• Consistency-Aware Repair
• Failure Grounding