MMOOC - Visual False Premises (Yes/No): leaderboard
Metric: Out-of-context score (%; mean of the refusal rate, the share of the Visual False Premises Yes/No questions, which the image and context cannot answer, that the model correctly declines, and refusal rationality, the mean 0-1 judge score of its abstention reasoning in steps of 0.25; judge scores averaged over GPT-5.6, Claude Opus 5 and DeepSeek-V4-Pro). Source: arxiv.org. Saturation forecast: Around 2029. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O1 | 71.75 |
| 2 | Claude Opus 4.6 | 67.75 |
| 3 | Gemma 4 26B | 66.25 |
| 4 | Qwen 3.5 122B A10B | 65.25 |
| 5 | Qwen 3.5 27B | 63.25 |
| 6 | Ministral 3 8B | 60.5 |
| 7 | GPT-4o | 59.25 |
| 8 | InternVL3-8B | 55.5 |
| 9 | Gemma 4 31B | 54.25 |
| 10 | Ministral 3 14B | 52.75 |
| 11 | Llama 4 Maverick | 51.25 |
| 12 | O3 | 35.5 |
| 13 | Gemini 3.1 Pro (Preview) | 9 |
Interactive version: theaggregate.ai/benchmark?slug=mmooc-visual-false-premises-yes-no · How It Works · Data refreshed daily, snapshot 2026-09-29.