Calibration
Video-MME (with subtitle) → We-Math
We-Math is estimated from Video-MME (with subtitle) with a Hill curve fitted on 5 models measured on both: y = 0.3349 + (1.2000 − 0.3349)·x^6.00 / (0.82056^6.00 + x^6.00), R² = 0.89, cross-validated error 3.7 pp. It is used for 4 estimates.
| Estimated model | Video-MME (with subtitle) | We-Math | Source |
|---|---|---|---|
| MiMo-V2.5 | 87.7% | 85.3% | estimated ± 3.7 pp, medium confidence |
| Qwen3.6-27B | 87.7% | 85.3% | estimated ± 3.7 pp, medium confidence |
| Qwen3.6-35B-A3B | 86.6% | 83.7% | estimated ± 3.7 pp, medium confidence |
| MiniMax M3 | 85.4% | 81.9% | estimated ± 3.7 pp, medium confidence |