COMEX - Explanation METEOR: leaderboard
COMEX (ZTE, 2026) is a composition-grounded benchmark for explainable aesthetic image cropping; this board covers the zero-shot general-purpose vision-language models, not the classic or fine-tuned cropping systems the paper also reports. METEOR measures the similarity of the generated composition-grounded crop explanation to the reference on a native 0 to 1 scale, higher is better.
Metric: METEOR (0-1, native; similarity of the generated composition-grounded crop explanation to the reference explanation; zero-shot). Source: arxiv.org. Saturation forecast: Around 2035. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 9B | 0.27 | #803 |
| 2 | Qwen 3.5 4B | 0.24 | #1180 |
| 3 | MiniMax-M2.5 | 0.23 | #590 |
| 4 | Seed 2.0 Pro | 0.22 | #209 |
| 5 | Seed-1.6 | 0.22 | #447 |
| 6 | Seed 1.8 | 0.22 | #239 |
| 7 | Qwen 3.5 2B | 0.19 | #4849 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=comex-explanation-meteor · How It Works · Data refreshed daily, snapshot 2026-11-12.