Reproduce “DS2SC-Agent: A Multi-Agent Automated Pipeline for Rapid Chiplet Model Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yiwei Wu, Yifan Wu, Yunhao Xiong, Dengwei Zhao, Jiaxuan Shen, Jianfei Jiang, Guanghui He, Shikui Tu and 1 more
The DS2SC-Agent is a multi-agent automated pipeline designed to convert unstructured industrial datasheets directly into functionally correct SystemC chiplet models. It successfully generated high-fidelity models for digital, analog, and RF chiplets, achieving computational errors bounded within 10e-16 for digital and maximum absolute errors below 1dB for RF characteristics.
Key insights
- A multi-agent architecture with specialized roles and structured prompts effectively mitigates common LLM limitations when processing complex, lengthy, and multimodal industrial datasheets.
- The use of a "mixed-fill" Specification Intermediate Representation (Spec IR) template significantly improves the fidelity and functional correctness of generated hardware models by standardizing unstructured documentation and guiding the LLM.
- Closed-loop automated debugging, incorporating Chain-of-Thought reasoning and detailed error analysis, enables reliable self-repair of generated code, ensuring high functional accuracy critical for trustworthy automation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.