MM-JudgeBias - Text Insertion: leaderboard
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. 30 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (High) | 0.93 |
| 2 | Gemini 2.5 Pro | 0.93 |
| 3 | GPT-5 Mini (High) | 0.93 |
| 4 | Gemini 2.5 Flash | 0.92 |
| 5 | O4 Mini (High) | 0.91 |
| 6 | GPT-5.1 (High) | 0.91 |
| 7 | GPT-4.1 Mini | 0.9 |
| 8 | Claude Opus 4.5 (Thinking) | 0.9 |
| 9 | Qwen 3 VL 8B (Thinking) | 0.9 |
| 10 | Claude Sonnet 4.5 (Thinking) | 0.89 |
| 11 | Claude Opus 4.5 | 0.89 |
| 12 | O3 (High) | 0.88 |
| 13 | Qwen 3 VL 30B A3B Instruct | 0.88 |
| 14 | Qwen 3 VL 8B Instruct | 0.88 |
| 15 | Gemini 2.0 Flash Lite | 0.88 |
Interactive version: theaggregate.ai/benchmark?slug=mm-judgebias-text-insertion · How It Works · Data refreshed daily, snapshot 2026-10-07.