RealICU (RAG) - Patient Status: leaderboard
RealICU (Shen et al., 2026) replays MIMIC-IV intensive-care stays in 30-minute windows and scores an LLM against hindsight labels from senior ICU physicians on the 930 windows of RealICU-Gold. Patient Status: accuracy (0-1) of the patient-status assessment at each window, higher is better. RAG protocol: the model sees the current window and the five most relevant windows of the stay so far, retrieved with PubMedBERT embeddings.
Metric: Accuracy (0-1; patient-status class per 30-minute window against hindsight physician labels, 930 RealICU-Gold windows; RAG strategy, the top-5 prior windows retrieved with PubMedBERT). Source: arxiv.org. Saturation forecast: Around August 2028. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 0.4 | #91 |
| 2 | Qwen 3 235B A22B | 0.32 | #622 |
| 3 | GPT-5.4 | 0.29 | #121 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=realicu-rag-patient-status · How It Works · Data refreshed daily, snapshot 2026-11-12.