CAREBench - Negative Emotion Intensity: leaderboard

Metric: Normalized RMSE (0-1; negative emotion intensity, 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 May 2027. 6 models tracked.

Top models

#ModelScore
1Claude Sonnet 4.6 (Non-reasoning)0.24
2Llama 3.1 8B Instruct0.25
3Qwen 3.5 9B (Non-reasoning)0.25
4GPT-5.2 (Non-reasoning)0.25

Interactive version: theaggregate.ai/benchmark?slug=carebench-negative-emotion-intensity · How It Works · Data refreshed daily, snapshot 2026-09-26.