A new technical paper, Agentic Hardware Design as Repository-Level Code Evolution, was published by researchers at Nvidia Research.
Abstract
“We present HORIZON, a self-evolving agent framework that treats hardware design as repository-level code evolution. A Markdown harness is compiled into a project pack containing domain knowledge, an executable evaluator, an acceptance predicate, and a git/runtime policy; a hands-free agent loop then evolves an isolated git worktree, using repository operations for state management, tracing, and replay. This extends prior works of repository-scale self-evolution from EDA software systems, to hardware-design artifacts themselves. We evaluate our approach on ChipBench, RTLLM, Verilog-Eval, and nine CVDP categories, achieving 100% benchmark completion across all suites with a fully hands-free agentic loop. However, we do not claim that agentic AI for hardware design is solved: these benchmarks are controlled proxies for a much broader engineering problem in chip design. Section 5 examines the limitations of the current study and highlights open research challenges.”
Find the technical paper here. June 2026 preprint.
Yu, Cunxi, Chenhui Deng, Nathaniel Pinckney, and Brucek Khailany. “Agentic Hardware Design as Repository-Level Code Evolution.” arXiv preprint arXiv:2606.28279 (2026).
Leave a Reply