MM-JudgeBias - Text Insertion: 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 query keywords are overlaid on the image; 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 query keywords or the query text are overlaid on the image; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 26 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1LLaVA-Critic-72B0.942
2Gemini 3 ProHigh0.933#145
3Gemini 2.5 Pro0.93#276
4GPT-5 MiniHigh0.927#364
5Gemini 2.5 Flash0.918#463
6O4 MiniHigh0.909#368
7GPT-5.1High0.908#240
8GPT-4.1 Mini0.904#789
9Claude Opus 4.5Thinking0.903#160
10Qwen 3 VL 8B (Thinking)0.898#910
11Claude Sonnet 4.5Thinking0.893#261
12O3High0.884#237
13Qwen 3 VL 30B A3B Instruct0.883#895
14Qwen 3 VL 8B Instruct0.88#1029
15Gemini 2.0 Flash Lite0.878#1050

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-text-insertion · How It Works · Data refreshed daily, snapshot 2026-11-21.