CLIR-Bench - Intervention Decision: leaderboard
The immediate intervention decision task of CLIR-Bench (2026): accuracy in percent at choosing the intervention category most consistent with the current clinical state and recent ICU history, against retrospective decision labels, 4-option multiple choice. Chance is 25; higher is better.
Metric: Accuracy (%) on immediate intervention decision, 4-option multiple-choice questions over serialized irregular ICU time series (Full-TS input), chance 25; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.6 | 91.67 | #178 |
| 2 | GPT-5.4 Mini | 76.67 | #389 |
| 3 | Gemini 2.5 Flash | 60 | #453 |
| 4 | DeepSeek V4 Flash | 41.67 | #258 |
| 5 | Gemma 3 27B (IT) | 39.67 | #1289 |
| 6 | Gemma 4 12B (IT) | 39.67 | #693 |
| 7 | Qwen 3.5 4B | 22 | #1175 |
| 8 | Qwen 3.5 9B | 20 | #803 |
| 9 | Qwen 3.6 27B | 14.33 | #341 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=clir-bench-intervention-decision · How It Works · Data refreshed daily, snapshot 2026-11-13.