LLM-FSM: Revolutionizing Hardware Design with Advanced AI

This research introduces LLM-FSM, a groundbreaking benchmark leveraging the power of 大規模言語モデル (LLMs) to automate and enhance hardware design. The automated pipeline generates finite-state machine (FSM) problems, allowing for more efficient evaluation of LLMs in translating natural-language specifications into RTL code. This opens exciting new possibilities for automating complex engineering tasks.

• •LLM-FSM automates FSM problem generation for LLM evaluation in hardware design.
• •The benchmark reveals the performance limitations of existing LLMs as FSM complexity grows.
• •Training-time scaling via 修正 (Fine-tuning) improves out-of-distribution (OOD) task performance.