Search papers, labs, and topics across Lattice.
This paper introduces RF-Agent, a framework that leverages knowledge distillation from RF textbooks to create a novel RF-domain reasoning dataset and benchmark, addressing the lack of domain-specific resources in RF circuit design. By implementing a multi-agent Question-Thinking-Solution-Answer (QTSA) pipeline, the authors generated over 11,000 samples, enabling a comparative study of adaptation strategies including supervised fine-tuning (SFT) and retrieval-augmented generation (RAG). The results reveal that domain-specific SFT significantly enhances RF reasoning capabilities, particularly in smaller models, while semantic retrieval outperforms other RAG configurations, underscoring the importance of context alignment in RF design tasks.
Domain-specific fine-tuning boosts RF reasoning in LLMs, especially for smaller models, while semantic retrieval proves superior for context alignment.
Large language models (LLMs) have driven rapid progress in electronic design automation (EDA), yet their application to radio-frequency (RF) circuit design remains limited by the scarcity of domain-specific datasets and standardized benchmarks. We present RF-Agent, which addresses this gap through textbook-driven knowledge distillation. A multi-agent Question-Thinking-Solution-Answer (QTSA) pipeline converts a subsection-level corpus from seven canonical RF textbooks into the first-of-its-kind RF-domain reasoning dataset (over 11,000 samples) with a dedicated multiple-choice benchmark. On this benchmark we study two adaptation strategies: supervised fine-tuning (SFT) and three retrieval-augmented generation (RAG) configurations (semantic, keyword, hybrid). Across multiple LLM families, domain-specific SFT significantly improves RF reasoning, especially for small and medium-sized models; among RAG configurations, semantic retrieval performs best, indicating embedding-based context alignment suits RF reasoning better than naive fusion. The dataset and benchmark provide a reusable foundation for future work on LLM-aided RF circuit design.
