Mask of Civility (Combined): leaderboard
Metric: Accuracy (%) of the combined (the definitions and the few-shot examples) prompt on the three-way classification (true politeness, impoliteness, mock politeness) of 100 Chinese dialogues with context (three fifths authentic TV-drama and interview excerpts, two fifths LLM-simulated), answer as one-line JSON; the Mask of Civility dataset; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2 0905 | 88 | #282 |
| 2 | DeepSeek V3.2 Exp | 87 | #227 |
| 3 | Seed-1.6 | 86 | #257 |
| 4 | GPT-5.1 | 84 | #131 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=mask-of-civility-combined · How It Works · Data refreshed daily, snapshot 2026-10-11.