EvoDRC Automates Design Rule Check Violation Repair with Self-Evolving Agents.

Bing-Yue Wu, Chia-Tung Ho, Haoyu Yang, Brucek Khailany, Vidya A. Chhabria· July 23, 2026 View original

Summary

EvoDRC is a self-evolving agentic framework designed to automate the repair of Design Rule Check (DRC) violations in advanced-node physical design. It uses LLM repair agents with continuously evolving skills, achieving a 73.5% overall reduction in DRVs on benchmark designs.

Design Rule Check (DRC) closure remains a significant bottleneck in the physical design of advanced semiconductor nodes, often requiring extensive manual intervention despite rule-aware routers. Automating this process is complex due to the need to manage intricate geometric interactions, preserve circuit connectivity, and avoid introducing new violations. This research introduces EvoDRC, a novel skill-evolution framework for agentic block-level DRC repair. EvoDRC initializes layer-specific repair skills by distilling knowledge from reference designs and continuously refines these skills using traceable repair experiences gathered from the target design. The framework decomposes the layout into bounded repair regions, assigning a Large Language Model (LLM) repair agent to each. These agents receive feedback from local DRC analysis, connectivity-checking, and impact-preview tools. All repair operations and their resulting DRV changes are stored in a knowledge database, which then informs the evolution of the repair skills. Experiments on seven block-level designs from the DAC26 DRC Benchmark demonstrated that EvoDRC achieved an impressive 73.5% overall reduction in DRVs compared to reported baselines.

Why it matters

This breakthrough offers a significant acceleration in semiconductor design cycles by automating a highly complex and time-consuming manual task, directly impacting time-to-market and development costs.

How to implement this in your domain

  1. 1Evaluate current DRC repair workflows to identify bottlenecks and areas for automation.
  2. 2Explore integrating LLM-based agents into existing Electronic Design Automation (EDA) toolchains.
  3. 3Develop a knowledge database to capture and learn from historical DRC repair experiences.
  4. 4Pilot EvoDRC-like methodologies on non-critical design blocks to assess performance and integration challenges.
  5. 5Collaborate with EDA vendors or AI research teams to customize and deploy self-evolving repair agents.

Who benefits

SemiconductorElectronics ManufacturingHigh-Performance Computing

Key takeaways

  • EvoDRC automates complex Design Rule Check (DRC) violation repair in chip design.
  • It uses LLM repair agents with skills that evolve from traceable repair experience.
  • The framework decomposes layouts into regions for localized agent-based repair.
  • EvoDRC achieved a 73.5% reduction in DRVs on benchmark designs.

Original post by Bing-Yue Wu, Chia-Tung Ho, Haoyu Yang, Brucek Khailany, Vidya A. Chhabria

"arXiv:2607.20019v1 Announce Type: new Abstract: Design rule check (DRC) closure remains a major bottleneck in advanced-node physical design. Although detailed routers are rule-aware, residual design rule violations (DRVs) often require manual engineering change order iterations.…"

View on X

Originally posted by Bing-Yue Wu, Chia-Tung Ho, Haoyu Yang, Brucek Khailany, Vidya A. Chhabria on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses