CAREBench - Appraisal Rating: leaderboard
Cognitive appraisal of emotion (CAREBench, Sun et al., 2026): the model reads a first-person narrative of a recent emotional event and answers as the narrator would, zero-shot, against the annotations the narrator gave for 1,000 narratives. The model rates 22 appraisal statements (relevance, certainty, congruence, control and accountability) on a 5-point scale; normalized root mean square error pooled over the ratings, from 0 to 1, lower is better.
Metric: Normalized RMSE (0-1; pooled over the 22 appraisal ratings, predicted from the first-person narrative against the narrator's own rating; 1,000 narratives, zero-shot, mean of five runs for open models). Source: arxiv.org. Saturation forecast: Around 2029. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 (Non-reasoning) | 0.36 | #158 (Claude Sonnet 4.6) |
| 2 | GPT-5.2 (Non-reasoning) | 0.37 | #172 (GPT-5.2) |
| 3 | Llama 3.1 8B Instruct | 0.4 | #4277 |
| 4 | Qwen 3.5 9B (Non-reasoning) | 0.43 | #803 (Qwen 3.5 9B) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=carebench-appraisal-rating · How It Works · Data refreshed daily, snapshot 2026-11-12.