Phun-Bench - Rhyming Sentence Generation (2-Rhyme): leaderboard
Phun-Bench (2026) tests whether text-only language models understand the sound of Chinese words, using tasks that cannot be solved by memorized pinyin alone. This harder setting requires the last two characters of the new sentence to rhyme with those of a reference lyric line; exact-match rate in percent, higher is better.
Metric: Exact-match rate (%) of the rhyme groups of the generated sentence's last two characters with the reference lyric line (1,000 lines; copying the reference ending disqualifies); thinking models run on 300 sampled items per setting, DeepSeek-V3 and GPT-4o on 500, the others on the full set; recommended sampling temperatures; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 0528 | 84.8 | #428 |
| 2 | DeepSeek V3 (0324) | 8.3 | #709 |
| 3 | GPT-4o (2024-11-20) | 7.2 | #866 |
| 4 | Qwen 3 8B (Non-reasoning) | 5.1 | #1830 (Qwen 3 8B) |
| 5 | Llama 3.2 3B Instruct | 1.4 | #7147 |
| 6 | GLM-4 9B Chat | 0.9 | #2759 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=phun-bench-rhyming-sentence-generation-2-rhyme · How It Works · Data refreshed daily, snapshot 2026-11-13.