MATEO (Image+Text, ICL): leaderboard
Metric: Consistent temporal-execution-order accuracy (%) on MATEO's test recipes from step texts and photos, prompted with the rules plus three text-only worked examples per relation; a prediction counts only when it is consistent across the original and swapped step order (before then after for dependent pairs, independent both ways); greedy decoding, single run; GPT-5.1 was scored on a stratified 20% sample of the test set; printed as fractions and shown times 100; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 (2025-11-13) (Non-reasoning) | 70 | #97 (GPT-5.1 (2025-11-13)) |
| 2 | Qwen 2.5 VL 72B Instruct | 68 | #364 |
| 3 | Qwen 2.5 VL 7B Instruct | 3 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=mateo-image-plustext-icl · How It Works · Data refreshed daily, snapshot 2026-10-11.