AICA-Bench - Emotion Reasoning: leaderboard

AICA-Bench emotion reasoning: explain why an image evokes its labeled emotion, rated 1 to 5 by a scoring model trained on human ratings for alignment, descriptiveness and causal soundness, rescaled to 20-100. Higher is better.

Metric: Emotion reasoning score on a percent scale: the model explains why an image evokes its labeled emotion, scored by the AICA-Bench scoring model (Qwen2.5-VL-7B fine-tuned on human 1-5 ratings) on emotion alignment and descriptiveness and causal soundness, ratings rescaled as s/5 x 100 (20-100), over AICA-Bench (8,086 affective images from nine public emotion datasets with GPT-4o-generated instructions; closed-source models through their APIs at standard settings, open models on A100 GPUs); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 23 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro79.08#280
2GPT-4o77.81#754
3Qwen-VL-Max (AICA-Bench checkpoint unspecified)77.75
4Gemini 2.5 Flash76.55#466
5GPT-4o Mini76.45#1680
6Qwen 2.5 VL 7B Instruct74.5#1919
7LLaVA-1.6-7B73.81#4990
8LLaVA-1.6-13B73.57#6010
9Qwen-VL-Plus (AICA-Bench checkpoint unspecified)72.39
10Gemini 2.0 Flash71.05#731
11MiniCPM-o-2.670.16#2442
12Qwen 2.5 VL 3B Instruct69.68#2720
13Ovis2-8B68.89#2059
14InternVL3-14B-Instruct68.27#1810
15Ovis2-16B68.24#1608

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=aica-bench-emotion-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-22.