EmoTrans - Emotion Transition Reasoning: leaderboard

EmoTrans (2026) emotion transition reasoning: the model explains why a person's emotion changed or stayed the same in a video clip, and a GPT-5 judge checks agreement with the reference explanation; LLM-Score (%) over 724 questions, higher is better.

Metric: LLM-Score (%) on Emotion Transition Reasoning (724 questions: a GPT-5 judge checks whether the model's explanation of why the emotion changed or stayed matches the reference explanation), 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. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.485.7#135
2Gemini 2.5 FlashNon-reasoning82.7#463
3Gemini 3 Pro82.3#145
4Gemini 3 FlashNon-reasoning82.3#175
5Seed 2.0 Lite81.4#300
6GLM-4.5V79.5#754
7GLM-4.6V79#710
8Qwen 3 VL 32B Instruct78.9#606
9Qwen 3.5 27B77.9#375
10Qwen 3.5 122B A10B77.3#346
11Qwen 2.5 VL 72B Instruct76.4#841
12Qwen 3 Omni 30B A3B75.5#847
13Qwen 3 VL 8B75.4#917
14InternVL3.5-8B70.2#1604
15InternVL3.5-14B65.8#1504

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=emotrans-emotion-transition-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-21.