Can Agents Design Better Chips with a Higher Level Abstraction?
Source
Zijian Ding, Yang Zou, Yizhou Sun, Jason Cong
Author
Zijian Ding, Yang Zou, Yizhou Sun, Jason Cong
Date
Key takeaways · AI-distilled
The study compares four workflows - Direct RTL Design, AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition →-based HLS Design, Post-Compiler HLS Refinement, and Post-HLS RTL Refinement - then combines agent-based HLS with RTL refinement into what it calls AHRR.
AHRR's 2.6x geometric-mean speedup over direct RTL design was measured across an 11-task benchmarkA standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.Full definition → suite, using FPGAs as the evalA repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.Full definition → platform.
Case studies attribute the gain to two complementary effects: HLS distills design knowledge into abstractions agents can actually use, while the later RTL-refinement pass recovers lower-level optimizations HLS alone misses.
The authors say the design-flow tradeoffs they studied are largely independent of the target technology, even though FPGAs, chosen for ease of end-to-end deployment, were the only platform tested.
Terms in this piece · Glossary
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
Why it matters
Demonstrates a transferable agent-design pattern: letting agents work at a higher abstraction level and then refine the lower-level output beats generating low-level code directly, a lesson relevant beyond chip design.
articleDo AI Agents Understand Computer Architecture?Ambika Sharan, Grigory Chirkov, Soheil Abbasloo
articleCallability Is Not Operability: Controlled Interface Interventions for LLM AgentsZihao Wang
articleLabAgent: Customize Any Research Hubs for Scientific Discoveries Using AI AgentsLei Liu, Yikun Zhang, Jialin Chen, Wanjia Zhao, Rex Ying, Wengong Jin, Hua Xu, James Zou, Tianyu Liu, Hongyu Zhao