MATRIX (Materials Science) - Hypothesis Generation: leaderboard
Metric: Judge score (% of the maximum) on MATRIX hypothesis generation: text-only hypothesis generation from the early sections of research papers; text only, no image; GPT-5.1 rubric judge scoring each long-form answer 0, 0.25, 0.5, 0.75 or 1 (scientific correctness, reasoning quality, consistency with the context), mean shown times 100; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 8B Instruct | 76.5 | #401 |
| 2 | Ministral-3-8B-Instruct-2512 | 68.5 | #644 |
| 3 | InternVL2-8B | 51 | #826 |
| 4 | Qwen 2 VL 7B Instruct | 49.5 | #816 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=matrix-materials-science-hypothesis-generation · How It Works · Data refreshed daily, snapshot 2026-10-11.