Longitudinal Clinical Prediction on EHR Patient-Level Splits (test)
87.8AUROCOutcome-Aligned Representation Learning
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Outcome-Aligned Representation LearningParameter count=35M, Evaluation protocol=fine-tuned2026.03 | 87.8 | 45.1 | 0.131 | 0.028 | |
| Supervised TransformerParameter count=35M, Evaluation protocol=purely supervised2026.03 | 85.9 | 41.7 | 0.139 | 0.036 | |
| Autoregressive EHR ModelParameter count=35M, Evaluation protocol=fine-tuned2026.03 | 85.1 | 40.4 | 0.142 | 0.038 | |
| Contrastive PretrainingParameter count=35M, Evaluation protocol=fine-tuned2026.03 | 84.8 | 39.8 | 0.144 | 0.039 | |
| Masked EHR TransformerParameter count=35M, Evaluation protocol=fine-tuned2026.03 | 84.2 | 39.1 | 0.146 | 0.041 |