CALRK-Bench - Information Sufficiency Recognition: leaderboard
Metric: Accuracy (times 100) on the 279 Type-ISR questions of CALRK-Bench (49 sufficient, 123 insufficient and 107 partial conditions, weighted by count): given a legal consultation and some statutes, the model picks the missing statute or answers that no additional legal reference is needed, against co-cited hard-negative statutes; four-option multiple choice on Korean legal precedents and consultations (chance 25 percent); API models averaged over three seeded runs, open models one greedy run; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 (Low) | 75.4 | #91 (GPT-5) |
| 2 | Gemini 3 Flash (Preview) (Low) | 74.8 | #78 (Gemini 3 Flash (Preview)) |
| 3 | Gemini 3 Flash (Preview) (Minimal) | 73.7 | #78 (Gemini 3 Flash (Preview)) |
| 4 | Qwen 3 30B A3B 2507 (Thinking) | 72 | #366 (Qwen 3 30B A3B 2507) |
| 5 | Llama 3.3 70B Instruct | 68.5 | #520 |
| 6 | Qwen 3 30B A3B 2507 Instruct | 67.7 | #464 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=calrk-bench-information-sufficiency-recognition · How It Works · Data refreshed daily, snapshot 2026-10-11.