CLIR-Bench: leaderboard
Metric: Macro-average accuracy (%) over the 11 tasks, 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 November 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 Mini | 50.15 |
| 2 | Kimi K2.6 | 49.09 |
| 3 | Gemini 2.5 Flash | 40.3 |
| 4 | DeepSeek V4 Flash | 35.91 |
| 5 | Gemma 3 27B (IT) | 30.87 |
| 6 | Gemma 4 12B (IT) | 29.82 |
| 7 | Qwen 3.5 9B | 26.64 |
| 8 | Qwen 3.5 4B | 25.73 |
| 9 | Qwen 3.6 27B | 25.01 |
Interactive version: theaggregate.ai/benchmark?slug=clir-bench · How It Works · Data refreshed daily, snapshot 2026-09-29.