MM-JudgeBias - Response Dominance: 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-Deviation from 0 to 1, the normalized score drop when both the query and the image are removed; higher is better.

Metric: Bias-Deviation (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 drop of the judge's score, normalized by the largest possible drop, when both the query and the image are replaced with null inputs; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 26 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 ProHigh0.988#145
2Gemini 2.5 Pro0.978#276
3Claude Opus 4.5Thinking0.917#160
4Qwen 3 VL 8B Instruct0.913#1029
5Gemini 2.5 Flash LiteThinking0.89#964
6Claude Sonnet 4.5Thinking0.831#261
7Qwen 3 VL 30B A3B (Thinking)0.806#719
8Gemini 2.5 Flash0.743#463
9Claude Haiku 4.5Thinking0.735#583
10Qwen 3 VL 30B A3B Instruct0.642#895
11O3High0.596#237
12Gemini 2.0 Flash Lite0.57#1050
13O4 MiniHigh0.518#368
14LLaVA-Critic-7B0.42
15Qwen 3 VL 8B (Thinking)0.412#910

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-response-dominance · How It Works · Data refreshed daily, snapshot 2026-11-21.