MMOOC - Misleading Premise (Yes/No): leaderboard
MMOOC also tests answering under shifted context, where the question is still answerable from the image. This board covers yes or no questions where the question carries a false or irrelevant premise while the visual query stays answerable, and averages accuracy with a judge rating of correctness and evidence consistency in the answer.
Metric: Shifted in-context score (%; mean of accuracy on the answerable Misleading Premise Yes/No questions and answer rationality, the mean 0-1 judge score of the answer's correctness and evidence consistency 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 February 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 | 88.25 | #102 |
| 2 | Qwen 3.5 27B | 88.25 | #359 |
| 3 | GPT-4o | 82.5 | #741 |
| 4 | Gemma 4 26B | 81.75 | #612 |
| 5 | Qwen 3.5 122B A10B | 80.25 | #331 |
| 6 | O1 | 79.25 | #393 |
| 7 | O3 | 78.25 | #228 |
| 8 | Llama 4 Maverick | 76 | #1099 |
| 9 | Gemma 4 31B | 75.75 | #394 |
| 10 | Ministral 3 14B | 75.25 | #1845 |
| 11 | Ministral 3 8B | 73.25 | #2023 |
| 12 | InternVL3-8B | 71.75 | #1688 |
| 13 | Gemini 3.1 Pro (Preview) | 70.75 | #94 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mmooc-misleading-premise-yes-no · How It Works · Data refreshed daily, snapshot 2026-11-13.