Researchers Unveil GoGoTB, an AI Agent Framework for Automated Chip Verification Achieving 98%+ Coverage

Researchers have developed GoGoTB, an agentic framework that uses large language models to automate functional verification of integrated circuit designs—a process that currently dominates engineering timelines and where a single missed bug can trigger costly silicon respins. The system combines agentic execution control, an evolvable knowledge base, and specification-grounded coverage tracking to achieve end-to-end verification closure, reaching 98.4% line coverage and 97.2% branch coverage across eight RTL designs without human intervention. GoGoTB's key innovation is maintaining shared context across verification components and anchoring coverage metrics to named specification behaviors, allowing it to pinpoint and remedy coverage gaps systematically.

Why it matters: This represents a potential breakthrough in automating one of semiconductor design's most labor-intensive and error-critical processes, with implications for reducing both development cycles and the risk of costly manufacturing errors.