E3mo-Bench (Expressed) - Video-Pair Perception (What-How): leaderboard
Metric: Accuracy (%; what or how multiple-choice questions comparing two videos, among 2,375 video-pair questions; 16 MLLMs on videos from eight emotion datasets, each assigned one affective perspective; subset: the emotion expressed by the people on screen). Source: arxiv.org. Saturation forecast: Around 2033. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-8B | 46.5 |
| 2 | GPT-5.4 | 44.9 |
| 3 | Qwen2.5-Omni-7B | 41.87 |
| 4 | Claude Sonnet 4.6 | 40.35 |
Interactive version: theaggregate.ai/benchmark?slug=e3mo-bench-expressed-video-pair-perception-what-how · How It Works · Data refreshed daily, snapshot 2026-09-29.