Multiclass diagnostic prediction on MC-MED
88.2AUROCMedM2T
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MedM2TModality=Multimodal, Model Architecture Ablations=w/o Shared Encoder2025.10 | 88.2 | 39.5 | |
| MedM2TModality=Multimodal2025.10 | 88 | 39.2 | |
| MedM2TModality=Multimodal, Model Architecture Ablations=w/o BiModal Attention2025.10 | 87.8 | 39.2 | |
| MedM2TModality=Multimodal, Model Architecture Ablations=w/o Pre-trained Encoder2025.10 | 83.5 | 34.1 | |
| LabsModality=Unimodal, Input Data Source=Labs2025.10 | 80.2 | 31.9 | |
| StaticModality=Unimodal, Input Data Source=Static2025.10 | 74.4 | 27.7 | |
| RadiologyModality=Unimodal, Input Data Source=Radiology2025.10 | 73 | 25.3 | |
| Sparse VitalsModality=Unimodal, Input Data Source=Sparse Vitals2025.10 | 69.8 | 25 | |
| Dense VitalsModality=Unimodal, Input Data Source=Dense Vitals2025.10 | 64.5 | 22.6 | |
| ECGModality=Unimodal, Input Data Source=ECG2025.10 | 59.4 | 22 |