PostEDA-Bench (ReAct) - PPA-Multi: leaderboard
Multi-objective PPA set of PostEDA-Bench (Liu et al., University of Minnesota, 2026): a ReAct agent must move an OpenROAD design from inside the design space to a target point on a two- or three-objective Pareto front of power, performance and area without regressing the constrained metrics, with functional equivalence checked. Success rate in percent over 40 tasks, each the share of 5 temperature-0 runs that meet every target within 16 agent actions; higher is better.
Metric: Success rate (%; 40 Pareto-targeted multi-objective PPA tasks, mean of 5 temperature-0 runs per task, ReAct agent). Source: arxiv.org. Saturation forecast: Around May 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 18.5 | #141 |
| 2 | GPT-5 | 11.5 | #163 |
| 3 | Gemma 4 31B (IT) (Thinking) | 11 | #343 (Gemma 4 31B (IT)) |
| 4 | GPT-5 Mini | 10 | #349 |
| 5 | Qwen 3.5 122B A10B | 4 | #329 |
| 6 | Qwen 3.5 9B | 1 | #803 |
| 7 | DeepSeek V3.2 | 0 | #403 |
| 8 | Qwen 3.5 27B | 0 | #358 |
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-ppa-multi · How It Works · Data refreshed daily, snapshot 2026-11-12.