CorrectRTL: Boost RTL Generation Quality Through Automated Self-Correction with Instruction Decomposition
Large language models (LLMs) have shown strong promise in specification-to-RTL generation.
ProofPaper ↗
Key points
- Among the explorations, the self-correction paradigm has significantly advanced LLM-based RTL generation.
- To address the challenge in methodology, we propose CorrectRTL, an agentic framework that realizes a new decompose-then-correct paradigm, in which the LLM agent explicitly decouples scaffold construction from detailed implementation.
- CorrectRTL contributes three core techniques: (1) Structure-oriented decomposition, (2) Snippet-based RAG, and (3) SVA-based self-correction.
- On ScaleBench, CorrectRTL improves Pass@1 from 20.5% to 49.0% (+28.5%).
Sources (1)
- [1]CorrectRTL: Boost RTL Generation Quality Through Automated Self-Correction with Instruction DecompositionarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 03:15 AM
Large language models (LLMs) have shown strong promise in specification-to-RTL generation.
Among the explorations, the self-correction paradigm has significantly advanced LLM-based RTL generation.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 17, 2026A Cross-Domain Evaluation of Prompting Techniques for Large Language Models
- Oct 16, 2026Characterising and Analysing Retrieval-Augmented Generation Techniques
- Oct 16, 2026Language Models Learn to Improve Language Models
- Oct 15, 2026Granite Switch 4.1: Chat, Decisions, and Speech from One Endpoint
- Oct 15, 2026Private Retrieval Augmented Generation via Random Projection
- Oct 15, 2026NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal Reasoning