Mortality prediction on MIMIC-III v1.4 (1000 patients)
13.5AUPRC+ threshold
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| + thresholdPKG=Filtered, Zero-shot=true, LLM=gpt-4.1-mini, Confidence threshold=0.82025.12 | 13.5 | 57.5 | |
| + confidence scorePKG=Confidence-aware, Zero-shot=true, LLM=gpt-4.1-mini2025.12 | 11.3 | 52.2 | |
| BaselinePKG=Historical, Zero-shot=true, LLM=gpt-4.1-mini2025.12 | 10.6 | 48.7 |