arXiv：cs.AI· Pengju Liu, Nuo Xu, Jinwei Tang, Yu Cao, Caiwen Ding·· 3 小时前
PostEDA-Bench：面向 PPA 收敛与 DRC 修复的分层基准
Bridging the Last Mile of Circuit Design: PostEDA-Bench, a Hierarchical Benchmark for PPA Convergence and DRC Fixing
AI 导读
研究者提出 PostEDA-Bench，一个包含 145 个任务的分层基准，覆盖 DRC-Essential、DRC-Reasoning、PPA-Mono 和 PPA-Multi 四类，用于评估 LLM 智能体在 EDA 签核后 DRC 违规修复与 PPA 收敛中的表现。
正文
Abstract:LLM-based agents are increasingly applied to the "last mile" of Electronic Design Automation (EDA): repairing residual sign-off Design Rule Check (DRC) violations and converging Power-Performance-Area (PPA) targets after tool runs. Existing EDA-LLM benchmarks, however, omit DRC fixing entirely and rely on flat hierarchies tied to a single toolchain. We introduce PostEDA-Bench, a hierarchical benchmark with 145 tasks across DRC-Essential, DRC-Reasoning, PPA-Mono, and PPA-Multi, supported by EDA toolchains with machine-checkable evaluation. Across eight commercial and open-source LLMs under multiple agent scaffolds, we find that agents handle synthetic DRC-Essential and single-objective PPA-Mono reasonably well but degrade sharply on the more practical DRC-Reasoning, where the best success rate is 36.66%, and PPA-Multi, where the best success rate is 20.00%; vision augmentation consistently enhances DRC-Bench; and trade-off reasoning, rather than knob knowledge, is the dominant PPA-Multi bottleneck.
| Subjects: | Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2605.06936 [cs.AR] |
| (or arXiv:2605.06936v4 [cs.AR] for this version) | |
| https://doi.org/10.48550/arXiv.2605.06936 arXiv-issued DOI via DataCite |
Submission history
From: Pengju Liu [view email]
[v1]
Thu, 7 May 2026 20:54:07 UTC (515 KB)
[v2]
Thu, 21 May 2026 20:26:07 UTC (515 KB)
[v3]
Tue, 21 Jul 2026 15:06:04 UTC (512 KB)
[v4]
Wed, 7 Oct 2026 22:33:36 UTC (508 KB)
来源：arXiv：cs.AI · arxiv.org