JARVIS: A Multi-Agent Code Assistant for High-Quality EDA Script Generation

--This paper presents JARVIS, a novel multi-agent framework that leverages Large Language Models (LLMs) and domain expertise to generate high-quality scripts for specialized Electronic Design Automation (EDA) tasks. By combining a domain-specific LLM trained with synthetically generated data, a custom compiler for structural verification, rule enforcement, code fixing capabilities, and advanced retrieval mechanisms, our approach achieves significant improvements over state-of-the-art domain-specific models. Our framework addresses the challenges of data scarcity and hallucination errors in LLMs, demonstrating the potential of LLMs in specialized engineering domains. We evaluate our framework on multiple benchmarks and show that it outperforms existing models in terms of accuracy and reliability. Our work sets a new precedent for the application of LLMs in EDA and paves the way for future innovations in this field. Large Language Models (LLMs) have revolutionized software development by automating various coding tasks, streamlining repetitive processes, and enhancing developer productivity. Tools like Microsoft's Copilot and Meta's CodeLlama have demonstrated the potential of LLMs in generating boilerplate code, automating common patterns, and embedding best practices within generated outputs [3], [17], [21]. However, when applied to specialized fields like V ery-Large-Scale Integration (VLSI) design within Electronic Design Automation (EDA), LLM performance is hindered by the scarcity of relevant training data, leading to unreliable and inaccurate outputs. These models often misinterpret and hallucinate due to a lack of contextual depth, highlighting the need for domain-specific fine-tuning. Recent efforts have focused on enhancing the reasoning capabilities of LLM models using CoT [23] and agent-based frameworks [24] for general tasks.