Action Recommendation on RealICU-Scale (test)
57.5Hit@5ICU-Evo
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ICU-EvoBackbone=GPT-5.4 [22], System=ICU-Evo2026.05 | 57.5 | 36.8 | |
| ICU-EvoBackbone=Qwen3-235B [34], System=ICU-Evo2026.05 | 51.5 | 32.7 | |
| ICU-EvoBackbone=Gemini-3.1-pro [8], System=ICU-Evo2026.05 | 51.4 | 33 | |
| RAGBackbone=GPT-5.4 [22], System=RAG2026.05 | 50.9 | 43.5 | |
| RAGBackbone=Gemini-3.1-pro [8], System=RAG2026.05 | 46.6 | 33.1 | |
| Full-contextBackbone=Qwen3-235B [34], System=Full-context2026.05 | 45.5 | 29.9 | |
| Local-windowBackbone=GPT-5.4 [22], System=Local-window2026.05 | 45.1 | 30.8 | |
| Local-windowBackbone=Gemini-3.1-pro [8], System=Local-window2026.05 | 44.7 | 30.7 | |
| RAGBackbone=Qwen3-235B [34], System=RAG2026.05 | 44.6 | 34.2 | |
| Local-windowBackbone=Qwen3-235B [34], System=Local-window2026.05 | 44 | 29.5 |