AICA-Bench: Holistically Examining the Capabilities of VLMs in Affective Image Content Analysis

TL;DR

AICA-Bench introduces a comprehensive benchmark for evaluating vision-language models in holistic affective image content analysis, addressing perception, reasoning, and generation tasks.

Contribution

The paper presents AICA-Bench, a new benchmark with three core tasks and proposes GAT Prompting, a training-free framework to improve VLMs' affective understanding.

Findings

GAT reduces intensity calibration errors.

GAT enhances descriptive depth in open-ended tasks.

Evaluation of 23 VLMs reveals key limitations in affective content analysis.

Abstract

Vision-Language Models (VLMs) have demonstrated strong capabilities in perception, yet holistic Affective Image Content Analysis (AICA), which integrates perception, reasoning, and generation into a unified framework, remains underexplored. To address this gap, we introduce AICA-Bench, a comprehensive benchmark with three core tasks: Emotion Understanding (EU), Emotion Reasoning (ER), and Emotion-Guided Content Generation (EGCG). We evaluate 23 VLMs and identify two major limitations: weak intensity calibration and shallow open-ended descriptions. To address these issues, we propose Grounded Affective Tree (GAT) Prompting, a training-free framework that combines visual scaffolding with hierarchical reasoning. Experiments show that GAT reduces intensity errors and improves descriptive depth, providing a strong baseline for future research on affective multimodal understanding and generation.