Chinese Social Pragmatic Inference - Comment Only: leaderboard
Metric: Leave-writer-out Accuracy (%). Source: arxiv.org. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 9B | 33.58 |
| 2 | Llama 4 Maverick | 32.09 |
| 3 | DeepSeek V4 Pro | 31.2 |
| 4 | Mistral Medium 3 | 30.49 |
| 5 | MiniMax-M2.5 | 29.69 |
| 6 | GPT-5.5 | 28.4 |
| 7 | Claude Opus 4.7 | 26.79 |
| 8 | Kimi K2.6 | 24.91 |
Interactive version: theaggregate.ai/benchmark?slug=chinese-social-pragmatic-inference-comment-only · How It Works · Data refreshed daily, snapshot 2026-09-20.