Calibration
Video-MME (with subtitle) → MathVision
MathVision is estimated from Video-MME (with subtitle) with a linear curve fitted on 8 models measured on both: y = 1.8310·x + -0.7260, R² = 0.96, cross-validated error 4.5 pp. It is used for 0 estimates.
Fitted on
| Model | Video-MME (with subtitle) | MathVision |
|---|---|---|
| Claude Opus 4.5 | 77.6% | 74.3% |
| Gemini 3 Pro | 88.4% | 86.6% |
| GPT-5.2 | 86.0% | 83.0% |
| Kimi K2.5 | 87.4% | 84.2% |
| LFM3-VL-2B | 69.0% | 51.6% |
| Qwen3.5 397B | 87.5% | 88.6% |
| Qwen3.7 Plus | 88.0% | 90.3% |
| Qwen3.8 Max | 90.4% | 95.2% |