CALRK-Bench - Temporal Context Recognition: leaderboard
Metric: Accuracy (times 100) on the 100 Type-TCR questions of CALRK-Bench: given a judgment summary and a later constitutional invalidation, statutory amendment or precedent change, the model picks the combination of act and trial dates under which the same conclusion holds; mean of the 50 pre-change and 50 post-change questions; 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. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) (Low) | 77 | #78 (Gemini 3 Flash (Preview)) |
| 2 | GPT-5 (High) | 63 | #91 (GPT-5) |
| 3 | GPT-5 (Medium) | 62.3 | #91 (GPT-5) |
| 4 | GPT-5 (Low) | 58.7 | #91 (GPT-5) |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 53 | #366 (Qwen 3 30B A3B 2507) |
| 6 | Gemini 3 Flash (Preview) (Minimal) | 46.3 | #78 (Gemini 3 Flash (Preview)) |
| 7 | GPT-5 (Minimal) | 32.7 | #91 (GPT-5) |
| 8 | Llama 3.3 70B Instruct | 31 | #520 |
| 9 | Qwen 3 30B A3B 2507 Instruct | 31 | #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-temporal-context-recognition · How It Works · Data refreshed daily, snapshot 2026-10-11.