Calibration
Video-MME (with subtitle) → DynaMath
DynaMath is estimated from Video-MME (with subtitle) with a linear curve fitted on 7 models measured on both: y = 0.3849·x + 0.5178, R² = 0.82, cross-validated error 2.1 pp. It is used for 1 estimate.
| Estimated model | Video-MME (with subtitle) | DynaMath | Source |
|---|---|---|---|
| MiniMax M3 | 85.4% | 84.7% | estimated ± 2.1 pp, medium confidence |
Fitted on
| Model | Video-MME (with subtitle) | DynaMath |
|---|---|---|
| Claude Opus 4.5 | 77.6% | 79.7% |
| Gemini 3 Pro | 88.4% | 85.1% |
| GPT-5.2 | 86.0% | 86.8% |
| Kimi K2.5 | 87.4% | 84.4% |
| LFM3-VL-2B | 69.0% | 79.2% |
| Qwen3.5 397B | 87.5% | 86.3% |
| Qwen3.6-27B | 87.7% | 85.6% |