CLIR-Bench - Missingness Awareness: leaderboard
The missingness awareness task of CLIR-Bench (2026): accuracy in percent at judging whether the sparse observations in a serialized ICU time series support a reliable answer, 4-option multiple choice. Chance is 25; higher is better.
Metric: Accuracy (%) on missingness awareness (whether the observations support a reliable answer), 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 | 73.33 | #178 |
| 2 | GPT-5.4 Mini | 45 | #389 |
| 3 | DeepSeek V4 Flash | 45 | #258 |
| 4 | Gemma 4 12B (IT) | 34.33 | #693 |
| 5 | Gemma 3 27B (IT) | 33.17 | #1289 |
| 6 | Qwen 3.5 4B | 24.67 | #1175 |
| 7 | Qwen 3.5 9B | 23.67 | #803 |
| 8 | Gemini 2.5 Flash | 23.33 | #453 |
| 9 | Qwen 3.6 27B | 14 | #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-missingness-awareness · How It Works · Data refreshed daily, snapshot 2026-11-13.