PolyBench - Structural Understanding - Exact Match: leaderboard
Counting and other discrete-answer questions from PolyBench's structural understanding category (Mohanty et al., Stony Brook University, 2026), such as counting structural components of a polymer from its representation. The score is the percent of answers that exactly match the gold label; higher is better. Models suffixed PolyBench are the authors' fine-tunes on the benchmark's training split.
Metric: Exact Match (%; counting and discrete-output items). Source: arxiv.org. Saturation forecast: Around 2029. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Phi-4 | 67 | #1920 |
| 2 | GPT-4o | 60 | #741 |
| 3 | GPT-5.4 | 60 | #121 |
| 4 | Claude Sonnet 4.6 | 47 | #158 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=polybench-structural-understanding-exact-match · How It Works · Data refreshed daily, snapshot 2026-11-12.