Calibration
SimpleVQA → MedXpertQA (MM)
MedXpertQA (MM) is estimated from SimpleVQA with a offset logistic curve fitted on 6 models measured on both: y = 0.6501 + (0.7777 − 0.6501) / (1 + exp(−151.54·(x − 0.5920))), R² = 0.64, cross-validated error 7.1 pp. It is used for 1 estimate.
| Estimated model | SimpleVQA | MedXpertQA (MM) | Source |
|---|---|---|---|
| LFM2.5-VL-3B | 35.4% | 65.0% | estimated ± 7.1 pp, low confidence |
Fitted on
| Model | SimpleVQA | MedXpertQA (MM) |
|---|---|---|
| Gemini 3.1 Pro | 72.4% | 81.3% |
| GPT-5.4 | 61.1% | 77.1% |
| Grok 4.20 | 57.4% | 65.8% |
| Muse Spark | 71.3% | 78.4% |
| Qwen3.7 Plus | 81.7% | 71.0% |
| Qwen3.8 Max | 75.0% | 80.4% |