DAR (Dynamic Affective Reasoning) - Emotion Accuracy: leaderboard
Metric: Emotion accuracy (%; a segment counts only if its predicted emotion matches the ground truth and its temporal IoU with the ground-truth segment is at least 0.5; 1,441 test videos of the DAR viewer-centric emotion dataset (27 emotion categories, emotions changing across consecutive events); the model segments the video into affective segments, labels each segment's emotion and writes a causal rationale; off-the-shelf models without fine-tuning on DAR). Source: arxiv.org. Saturation forecast: Around November 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 VL 7B Instruct | 14.9 |
| 2 | InternVL3.5-8B | 11.5 |
Interactive version: theaggregate.ai/benchmark?slug=dar-dynamic-affective-reasoning-emotion-accuracy · How It Works · Data refreshed daily, snapshot 2026-09-29.