Phun-Bench - Homophone Recall: leaderboard

Metric: Accuracy (%) of recovering a four-character chengyu from an automatically generated homophone of it (1,098 items); 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

#ModelScore
1DeepSeek R1 052897.7
2DeepSeek V3 (0324)66.4
3GPT-4o (2024-11-20)61.8
4Qwen 3 8B (Thinking)56.3
5Qwen 3 8B (Non-reasoning)15.2
6GLM-4 9B Chat12.4
7Llama 3.2 3B Instruct0.7

Interactive version: theaggregate.ai/benchmark?slug=phun-bench-homophone-recall · How It Works · Data refreshed daily, snapshot 2026-09-29.