PostEDA-Bench (ReAct) - DRC-Reasoning: leaderboard
Practical design-rule repair set of PostEDA-Bench (Liu et al., University of Minnesota, 2026): a ReAct agent fixes violations that remain after full OpenROAD or commercial place-and-route flows and built-in repair, which need geometric reasoning and one to several layout edits. Success rate in percent over 30 tasks, each the share of 5 temperature-0 runs that end DRC-clean within 8 agent actions; higher is better.
Metric: Success rate (%; 30 residual post-flow DRC repair tasks, mean of 5 temperature-0 runs per task, ReAct agent). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 36.66 | #141 |
| 2 | GPT-5 | 31.33 | #163 |
| 3 | Gemma 4 31B (IT) (Thinking) | 27.99 | #343 (Gemma 4 31B (IT)) |
| 4 | Qwen 3.5 27B | 27.32 | #358 |
| 5 | GPT-5 Mini | 20.66 | #349 |
| 6 | DeepSeek V3.2 | 16.66 | #403 |
| 7 | Qwen 3.5 122B A10B | 13.33 | #329 |
| 8 | Qwen 3.5 9B | 5.99 | #803 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=posteda-bench-react-drc-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-12.