EmoTrans - Emotion Change Detection: leaderboard
Metric: Accuracy (%) on Emotion Change Detection (1,000 questions: whether each target person's emotion changes), on EmoTrans, built from 1,000 manually annotated real-world video clips (OpenHumanVid and CH-SIMS v2) over 12 scenarios with one or more target persons; zero-shot, video with its text (and audio where the model accepts it); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 75.4 |
| 2 | Gemini 3 Pro | 74.7 |
| 3 | Qwen 3.5 122B A10B | 74 |
| 4 | Qwen 2.5 VL 72B Instruct | 73.9 |
| 5 | Qwen 3.5 27B | 72.8 |
| 6 | Qwen 3 VL 32B Instruct | 71.5 |
| 7 | GLM-4.6V | 70 |
| 8 | Gemini 2.5 Flash (Non-reasoning) | 69.8 |
| 9 | GLM-4.5V | 66.2 |
| 10 | Seed 2.0 Lite | 64.9 |
Interactive version: theaggregate.ai/benchmark?slug=emotrans-emotion-change-detection · How It Works · Data refreshed daily, snapshot 2026-10-07.