VeriMoA: A Mixture-of-Agents Framework for Spec-to-HDL Generation

2025-10-31

AI Summary

To address the challenges of domain-knowledge scarcity, limited reasoning space, and severe noise propagation in multi-agent collaboration for RTL-level hardware generation using large language models (LLMs), this paper proposes a training-free multi-agent framework. Our method introduces: (1) a quality-guided caching mechanism to accumulate and reuse cross-stage knowledge; (2) a multi-path generation strategy to enhance diversity in solution-space exploration; and (3) a collaborative translation and quality-ranked selection mechanism leveraging C++/Python intermediate representations. Evaluated on VerilogEval 2.0 and RTLLM 2.0 benchmarks, our approach achieves a 15–30% improvement in Pass@1, substantially narrowing the performance gap between small LLMs and larger or fine-tuned models. This work establishes a novel paradigm for trustworthy, resource-efficient RTL generation in low-resource settings.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Multiagent Learning
• Natural Language Processing: Generation

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Automation of Register Transfer Level (RTL) design can help developers meet increasing computational demands. Large Language Models (LLMs) show promise for Hardware Description Language (HDL) generation, but face challenges due to limited parametric knowledge and domain-specific constraints. While prompt engineering and fine-tuning have limitations in knowledge coverage and training costs, multi-agent architectures offer a training-free paradigm to enhance reasoning through collaborative generation. However, current multi-agent approaches suffer from two critical deficiencies: susceptibility to noise propagation and constrained reasoning space exploration. We propose VeriMoA, a training-free mixture-of-agents (MoA) framework with two synergistic innovations. First, a quality-guided caching mechanism to maintain all intermediate HDL outputs and enables quality-based ranking and selection across the entire generation process, encouraging knowledge accumulation over layers of reasoning. Second, a multi-path generation strategy that leverages C++ and Python as intermediate representations, decomposing specification-to-HDL translation into two-stage processes that exploit LLM fluency in high-resource languages while promoting solution diversity. Comprehensive experiments on VerilogEval 2.0 and RTLLM 2.0 benchmarks demonstrate that VeriMoA achieves 15--30% improvements in Pass@1 across diverse LLM backbones, especially enabling smaller models to match larger models and fine-tuned alternatives without requiring costly training.

Problem

• Automating Register Transfer Level design using Large Language Models
• Overcoming noise propagation in multi-agent hardware generation systems
• Enhancing reasoning space exploration for Hardware Description Language generation

Innovation

• Quality-guided caching mechanism for HDL ranking
• Multi-path generation using C++ and Python intermediates
• Training-free mixture-of-agents framework for HDL generation