StyleBench (Speaking Style) - Surprised (Turn 3): leaderboard
Metric: Valid sample percentage (%): share of the 160 surprised-emotion test dialogues in which human raters judge the turn-3 response to carry the distinct, intended emotional intensity, in three-turn spoken dialogues of StyleBench's 10% evaluation split, a neutral first turn followed by two conversational instructions that strengthen or weaken the style (a model completely unresponsive to emotional adjustment scores 0); higher is better. Source: arxiv.org. Saturation forecast: Around March 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen2.5-Omni-7B | 23.13 | #601 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=stylebench-speaking-style-surprised-turn-3 · How It Works · Data refreshed daily, snapshot 2026-10-11.