Saved in:
Bibliographic Details
Main Authors: Zhang, Dengming, You, Weitao, Li, Jingxiong, Lin, Weishen, Shi, Wenda, Zhao, Xue, Zuo, Heda, Wu, Junxian, Sun, Lingyun
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.12077
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918223544320000
author Zhang, Dengming
You, Weitao
Li, Jingxiong
Lin, Weishen
Shi, Wenda
Zhao, Xue
Zuo, Heda
Wu, Junxian
Sun, Lingyun
author_facet Zhang, Dengming
You, Weitao
Li, Jingxiong
Lin, Weishen
Shi, Wenda
Zhao, Xue
Zuo, Heda
Wu, Junxian
Sun, Lingyun
contents Emotion understanding is critical for making Large Language Models (LLMs) more general, reliable, and aligned with humans. Art conveys emotion through the joint design of visual and auditory elements, yet most prior work is human-centered or single-modality, overlooking the emotion intentionally expressed by the artwork. Meanwhile, current Audio-Visual Language Models (AVLMs) typically require large-scale audio pretraining to endow Visual Language Models (VLMs) with hearing, which limits scalability. We present Vision Anchored Audio-Visual Emotion LLM (VAEmotionLLM), a two-stage framework that teaches a VLM to hear by seeing with limited audio pretraining and to understand emotion across modalities. In Stage 1, Vision-Guided Audio Alignment (VG-Align) distills the frozen visual pathway into a new audio pathway by aligning next-token distributions of the shared LLM on synchronized audio-video clips, enabling hearing without a large audio dataset. In Stage 2, a lightweight Cross-Modal Emotion Adapter (EmoAdapter), composed of the Emotion Enhancer and the Emotion Supervisor, injects emotion-sensitive residuals and applies emotion supervision to enhance cross-modal emotion understanding. We also construct ArtEmoBenchmark, an art-centric emotion benchmark that evaluates content and emotion understanding under audio-only, visual-only, and audio-visual inputs. VAEmotionLLM achieves state-of-the-art results on ArtEmoBenchmark, outperforming audio-only, visual-only, and audio-visual baselines. Ablations show that the proposed components are complementary.
format Preprint
id arxiv_https___arxiv_org_abs_2511_12077
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning to Hear by Seeing: It's Time for Vision Language Models to Understand Artistic Emotion from Sight and Sound
Zhang, Dengming
You, Weitao
Li, Jingxiong
Lin, Weishen
Shi, Wenda
Zhao, Xue
Zuo, Heda
Wu, Junxian
Sun, Lingyun
Computer Vision and Pattern Recognition
Emotion understanding is critical for making Large Language Models (LLMs) more general, reliable, and aligned with humans. Art conveys emotion through the joint design of visual and auditory elements, yet most prior work is human-centered or single-modality, overlooking the emotion intentionally expressed by the artwork. Meanwhile, current Audio-Visual Language Models (AVLMs) typically require large-scale audio pretraining to endow Visual Language Models (VLMs) with hearing, which limits scalability. We present Vision Anchored Audio-Visual Emotion LLM (VAEmotionLLM), a two-stage framework that teaches a VLM to hear by seeing with limited audio pretraining and to understand emotion across modalities. In Stage 1, Vision-Guided Audio Alignment (VG-Align) distills the frozen visual pathway into a new audio pathway by aligning next-token distributions of the shared LLM on synchronized audio-video clips, enabling hearing without a large audio dataset. In Stage 2, a lightweight Cross-Modal Emotion Adapter (EmoAdapter), composed of the Emotion Enhancer and the Emotion Supervisor, injects emotion-sensitive residuals and applies emotion supervision to enhance cross-modal emotion understanding. We also construct ArtEmoBenchmark, an art-centric emotion benchmark that evaluates content and emotion understanding under audio-only, visual-only, and audio-visual inputs. VAEmotionLLM achieves state-of-the-art results on ArtEmoBenchmark, outperforming audio-only, visual-only, and audio-visual baselines. Ablations show that the proposed components are complementary.
title Learning to Hear by Seeing: It's Time for Vision Language Models to Understand Artistic Emotion from Sight and Sound
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.12077