RTL-BenchMT: Agentic Framework for Dynamic RTL Benchmark Maintenance
RTL-BenchMT is an agentic framework designed to dynamically maintain RTL generation benchmarks used in EDA research. It addresses two key challenges: flawed benchmark cases and overfitting to benchmarks, which are difficult to resolve manually. The framework automatically identifies and revises flawed cases and detects and updates overfitting cases. The refined benchmark suite will be open-sourced.
Key facts
- RTL-BenchMT is an agentic framework for dynamic maintenance of RTL generation benchmarks.
- It addresses flawed cases and overfitting in current RTL benchmarks.
- The framework automates identification and revision of flawed cases.
- It also detects and updates overfitting cases.
- The refined benchmark suite will be open-sourced.
- Large Language Models (LLMs) assist automated RTL generation in EDA research.
- Manual engineering effort alone is insufficient to resolve benchmark challenges.
- RTL-BenchMT reduces human maintenance costs.
Entities
Institutions
- arXiv