MCQ Diagnostic Accuracy on Stanford Multimodal (test)
73.7AccuracyMARCUS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MARCUSN=382026.03 | 73.7 | 60.5 | — | |
| Gemini 2.5 ProN=342026.03 | 29.4 | 14.7 | — | |
| GPT-5N=402026.03 | 22.5 | 10 | — | |
| MARCUS vs Gemini (McNemar p)N=322026.03 | — | — | 0.001 | |
| MARCUS vs GPT-5 (McNemar p)N=382026.03 | — | — | 0.001 |