APUN-Bench (Real-World) - Homographic Pun Inference: leaderboard
Metric: Homographic pun-inference accuracy (%), both predicted meanings above a similarity threshold to the gold pair, on the real-world set (893 human-recorded clips); temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.0 Flash | 47.36 | #331 |
| 2 | Qwen2.5-Omni-7B | 35.78 | #601 |
| 3 | MERaLiON-2-10B | 32.28 | #567 |
| 4 | SALMONN-13B | 23.15 | #1237 |
| 5 | Qwen2-Audio-7B-Instruct | 21.59 | #859 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=apun-bench-real-world-homographic-pun-inference · How It Works · Data refreshed daily, snapshot 2026-10-11.