Sepsis treatment on MIMIC-IV (test)
0.045WISBC
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| BCPreferences=[1.0, 0.0]2025.12 | 0.045 | — | — | — | — | — | |
| BCPreferences=[0.9, 0.1]2025.12 | 0.131 | — | — | — | — | — | |
| BCPreferences=[0.8, 0.2]2025.12 | 0.217 | — | — | — | — | — | |
| BCPreferences=[0.7, 0.3]2025.12 | 0.303 | — | — | — | — | — | |
| BCPreferences=[0.6, 0.4]2025.12 | 0.389 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.7, 0.3]2025.12 | 0.468 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.9, 0.1]2025.12 | 0.471 | — | — | — | — | — | |
| BCPreferences=[0.5, 0.5]2025.12 | 0.475 | — | — | — | — | — | |
| PEDA DTPreferences=[0.4, 0.6]2025.12 | 0.56 | — | — | — | — | — | |
| BCPreferences=[0.4, 0.6]2025.12 | 0.561 | — | — | — | — | — | |
| BCPreferences=[0.3, 0.7]2025.12 | 0.647 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.8, 0.2]2025.12 | 0.652 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.6, 0.4]2025.12 | 0.661 | — | — | — | — | — | |
| DDQNPreferences=[0.0, 1.0]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.1, 0.9]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.2, 0.8]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.3, 0.7]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.4, 0.6]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.5, 0.5]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.6, 0.4]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.7, 0.3]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.8, 0.2]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[0.9, 0.1]2025.12 | 0.664 | — | — | — | — | — | |
| DDQNPreferences=[1.0, 0.0]2025.12 | 0.664 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[1.0, 0.0]2025.12 | 0.707 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.5, 0.5]2025.12 | 0.714 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.6, 0.4]2025.12 | 0.725 | — | — | — | — | — | |
| BCPreferences=[0.2, 0.8]2025.12 | 0.733 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.8, 0.2]2025.12 | 0.738 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.4, 0.6]2025.12 | 0.74 | — | — | — | — | — | |
| PEDA DTPreferences=[0.2, 0.8]2025.12 | 0.746 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.9, 0.1]2025.12 | 0.76 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[1.0, 0.0]2025.12 | 0.76 | — | — | — | — | — | |
| PEDA DTPreferences=[0.3, 0.7]2025.12 | 0.767 | — | — | — | — | — | |
| BCPreferences=[0.1, 0.9]2025.12 | 0.819 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.2, 0.8]2025.12 | 0.82 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.3, 0.7]2025.12 | 0.831 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.2, 0.8]2025.12 | 0.857 | — | — | — | — | — | |
| PEDA DTPreferences=[0.1, 0.9]2025.12 | 0.864 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.5, 0.5]2025.12 | 0.874 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.7, 0.3]2025.12 | 0.88 | — | — | — | — | — | |
| PEDA DTPreferences=[0.5, 0.5]2025.12 | 0.881 | — | — | — | — | — | |
| CQLPreferences=[0.0, 1.0]2025.12 | 0.883 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.1, 0.9]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.1, 0.9]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.2, 0.8]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.3, 0.7]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.4, 0.6]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.5, 0.5]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.6, 0.4]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.7, 0.3]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.8, 0.2]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[0.9, 0.1]2025.12 | 0.883 | — | — | — | — | — | |
| CQLPreferences=[1.0, 0.0]2025.12 | 0.883 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.4, 0.6]2025.12 | 0.887 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.3, 0.7]2025.12 | 0.889 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.1, 0.9]2025.12 | 0.891 | — | — | — | — | — | |
| PEDA DTPreferences=[0.9, 0.1]2025.12 | 0.899 | — | — | — | — | — | |
| BCPreferences=[0.0, 1.0]2025.12 | 0.904 | — | — | — | — | — | |
| PEDA DTPreferences=[0.0, 1.0]2025.12 | 0.909 | — | — | — | — | — | |
| Conditioned CPQLPreferences=[0.0, 1.0]2025.12 | 0.911 | — | — | — | — | — | |
| Adaptive CPQLPreferences=[0.0, 1.0]2025.12 | 0.912 | — | — | — | — | — | |
| PEDA DTPreferences=[0.6, 0.4]2025.12 | 0.912 | — | — | — | — | — | |
| PEDA DTPreferences=[0.7, 0.3]2025.12 | 0.912 | — | — | — | — | — | |
| PEDA DTPreferences=[0.8, 0.2]2025.12 | 0.912 | — | — | — | — | — | |
| PEDA DTPreferences=[1.0, 0.0]2025.12 | 0.913 | — | — | — | — | — | |
| GPT4.1-mini + WMMethod Group=World-Model-Augmented LLMs, World Model (WM) Augmentation=true2026.05 | 5.21 | 7.32 | 17.09 | 94 | 0.58 | 1.6 | |
| Clinicians (Test Set)Method Group=Human Reference2026.05 | 5.27 | 5.06 | 10.82 | 94.76 | 0.35 | 0.19 | |
| GPT-OSS-120BMethod Group=Vanilla LLMs2026.05 | 6 | 8.25 | 21.17 | 79.42 | 1.06 | 1.11 | |
| DDQNMethod Group=Traditional RL2026.05 | 6.19 | 8.69 | 15.11 | 82.79 | 0.67 | 1.01 | |
| GPT4.1-miniMethod Group=Vanilla LLMs2026.05 | 6.59 | 6.13 | 10.82 | 80.59 | 0.66 | 2.18 | |
| Qwen3-4B-InstructMethod Group=SepsisAgent and Its Backbone, Backbone Selection=true2026.05 | 7.34 | 7.79 | 18.76 | 78 | 0.62 | 2.13 | |
| Gemini-3-Flash + WMMethod Group=World-Model-Augmented LLMs, World Model (WM) Augmentation=true2026.05 | 7.78 | 4.49 | 12.24 | 95.16 | 0.78 | 1.16 | |
| Gemini-3-ProMethod Group=Vanilla LLMs2026.05 | 8.59 | 5.84 | 19.68 | 96.74 | 0.09 | 1.62 | |
| Gemini-3-FlashMethod Group=Vanilla LLMs2026.05 | 9.09 | 8.17 | 13.98 | 96.43 | 1.19 | 2.58 | |
| o3Method Group=Vanilla LLMs2026.05 | 9.17 | 8.32 | 20.38 | 90.55 | 0.72 | 1.57 | |
| AI ClinicianMethod Group=Traditional RL2026.05 | 9.94 | 5.98 | 9.88 | 87.24 | 0.53 | 2.34 | |
| Deepseek-V3.2Method Group=Vanilla LLMs2026.05 | 10.25 | 8.63 | 15.07 | 81.8 | 0.19 | 0.41 | |
| o3 + WMMethod Group=World-Model-Augmented LLMs, World Model (WM) Augmentation=true2026.05 | 10.27 | 9.46 | 22.95 | 96.91 | 0.09 | 0.24 | |
| SepsisAgentMethod Group=SepsisAgent and Its Backbone, Backbone=Qwen3-4B-Instruct2026.05 | 11.14 | 10.01 | 23.4 | 97.95 | 0.08 | 0.14 | |
| WD3QNEMethod Group=Traditional RL2026.05 | 12.07 | 8.72 | 23.2 | 87.6 | 1.11 | 1.49 |