RTL-BenchMT: Dynamic Maintenance of RTL Generation Benchmark Through Agent-Assisted Analysis and Revision

2026-05-14

AI Summary

This work addresses critical limitations in existing RTL generation benchmarks, which suffer from erroneous test cases, model overfitting, and reliance on inefficient manual curation. To overcome these issues, the authors propose the first agent-based automated framework that integrates large language models with formal verification tools to dynamically identify and correct faulty examples while detecting and updating samples that induce overfitting. This approach enables continuous, self-improving benchmark refinement with substantially reduced human intervention. The resulting system not only enhances the reliability and generalization capacity of RTL benchmarks in a principled manner but also introduces an open-sourced, high-quality benchmark suite that has undergone extensive optimization.

Abstract

This paper introduces RTL-BenchMT, an agentic framework for dynamically maintaining RTL generation benchmarks. Large Language Models (LLMs) assisted automated RTL generation is one of the most important directions in EDA research. However, current RTL benchmarks face two critical challenges: (1) flawed cases in the benchmarks and (2) overfitting to the benchmarks. Both challenges are difficult to resolve purely by manual engineering effort. To address these issues and systematically reduce human maintenance costs, we propose an automated agentic framework, RTL-BenchMT. RTL-BenchMT focuses on two key applications: (1) automatically identifying and revising flawed benchmark cases and (2) automatically detecting and updating overfitting cases. With the assistance of RTL-BenchMT, we conduct a thorough, in-depth analysis of flawed and overfitting cases and produce a refined benchmark suite that will be open-sourced to the community.

Problem

• RTL generation
• benchmark maintenance
• flawed cases
• overfitting
• EDA

Innovation

• RTL generation
• benchmark maintenance
• agent-assisted analysis
• overfitting detection
• automated revision