MM-JudgeBias - Visual Transformation: leaderboard
MM-JudgeBias (2026) tests multimodal LLM judges for compositional bias: the judge scores image, query and response triplets from 1 to 10 before and after a controlled perturbation. Bias-Conformity from 0 to 1, the stability of the score when the image gets meaning-preserving augmentations; higher is better.
Metric: Bias-Conformity (0 to 1) on MM-JudgeBias (about 200 image-query-response triplets per bias type, 1,804 in all, each judged on a 1-10 scale before and after a controlled perturbation; mean of three runs): mean score invariance (1 minus the absolute score change over its largest possible value) when the image receives meaning-preserving augmentations; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 26 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | LLaVA-Critic-72B | 0.931 | ||
| 2 | Gemini 3 Pro | High | 0.889 | #145 |
| 3 | GPT-4.1 Mini | 0.887 | #789 | |
| 4 | GPT-5 Mini | High | 0.877 | #364 |
| 5 | Qwen 3 VL 8B Instruct | 0.87 | #1029 | |
| 6 | LLaVA-Critic-7B | 0.869 | ||
| 7 | GPT-5.1 | High | 0.867 | #240 |
| 8 | Qwen 3 VL 8B (Thinking) | 0.865 | #910 | |
| 9 | Gemini 2.5 Flash | 0.864 | #463 | |
| 10 | Gemini 2.0 Flash Lite | 0.862 | #1050 | |
| 11 | Gemini 2.5 Pro | 0.861 | #276 | |
| 12 | Qwen 2.5 VL 72B Instruct | 0.855 | #841 | |
| 13 | Qwen 3 VL 30B A3B Instruct | 0.854 | #895 | |
| 14 | O4 Mini | High | 0.854 | #368 |
| 15 | Qwen 3 VL 30B A3B (Thinking) | 0.85 | #719 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=mm-judgebias-visual-transformation · How It Works · Data refreshed daily, snapshot 2026-11-21.