CCiV (Direct) - Structural Accuracy: leaderboard
Metric: Structural accuracy (%, 0-100): a poem counts when its line count and the character count of every line exactly match the target template (non-Chinese characters removed) of the standard form or of any documented historical variant of the requested Cipai; direct prompt: only the Cipai name and the theme; 300 prompts (10 historical themes for each of the 30 most frequent Cipai tune patterns), three samples each at temperature 0.7 and top-p 0.95, averaged; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 68 | #333 |
| 2 | DeepSeek R1 | 67.33 | #245 |
| 3 | GLM-4 Plus | 51 | #373 |
| 4 | Qwen Max | 50.67 | #367 |
| 5 | DeepSeek R1 Distill Qwen 32B | 38 | #640 |
| 6 | QwQ-32B | 14.67 | #410 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=cciv-direct-structural-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-11.