Multimodal LLMs See Sentiment

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: da Silva, Neemias B., Harrison, John, Minetto, Rodrigo, Delgado, Myriam R., Nassu, Bogdan T., Silva, Thiago H.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914610321293312
author da Silva, Neemias B.
Harrison, John
Minetto, Rodrigo
Delgado, Myriam R.
Nassu, Bogdan T.
Silva, Thiago H.
author_facet da Silva, Neemias B.
Harrison, John
Minetto, Rodrigo
Delgado, Myriam R.
Nassu, Bogdan T.
Silva, Thiago H.
contents Understanding how visual content conveys sentiment is increasingly important in a digital landscape dominated by imagery. However, sentiment perception depends on complex scene-level semantics, making this a challenging task for computational models. This paper examines how Multimodal Large Language Models (MLLMs) perform sentiment analysis in images through a systematic, evaluation-driven study encompassing three perspectives: (i) direct sentiment classification from images using MLLMs; (ii) sentiment analysis on MLLM-generated descriptions using pre-trained LLMs; and (iii) fine-tuning these LLMs on sentiment-labeled descriptions to assess performance and generalization. Experiments on a recent benchmark show that a two-stage MLLM description-mediated pipeline can substantially improve prediction accuracy under several evaluation settings, particularly when the LLM component is fine-tuned. Across different agreement thresholds and sentiment granularities, the strongest configurations of this pipeline outperform lexicon-, CNN-, and Transformer-based baselines in our benchmark by up to 30.9%, 64.8%, and 42.4%, respectively. In cross-dataset evaluation, the proposed pipeline - without training or fine-tuning on the target dataset - still surpasses the best in-domain baseline by over 8%. Overall, the study provides a comprehensive assessment of MLLM description-mediated sentiment analysis, clarifying the conditions under which it is effective, the scenarios in which it fails, and its comparison with traditional vision-based approaches, while also providing a reproducible benchmark resource for future research.
format Preprint
id arxiv_https___arxiv_org_abs_2508_16873
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Multimodal LLMs See Sentiment
da Silva, Neemias B.
Harrison, John
Minetto, Rodrigo
Delgado, Myriam R.
Nassu, Bogdan T.
Silva, Thiago H.
Computer Vision and Pattern Recognition
Social and Information Networks
Understanding how visual content conveys sentiment is increasingly important in a digital landscape dominated by imagery. However, sentiment perception depends on complex scene-level semantics, making this a challenging task for computational models. This paper examines how Multimodal Large Language Models (MLLMs) perform sentiment analysis in images through a systematic, evaluation-driven study encompassing three perspectives: (i) direct sentiment classification from images using MLLMs; (ii) sentiment analysis on MLLM-generated descriptions using pre-trained LLMs; and (iii) fine-tuning these LLMs on sentiment-labeled descriptions to assess performance and generalization. Experiments on a recent benchmark show that a two-stage MLLM description-mediated pipeline can substantially improve prediction accuracy under several evaluation settings, particularly when the LLM component is fine-tuned. Across different agreement thresholds and sentiment granularities, the strongest configurations of this pipeline outperform lexicon-, CNN-, and Transformer-based baselines in our benchmark by up to 30.9%, 64.8%, and 42.4%, respectively. In cross-dataset evaluation, the proposed pipeline - without training or fine-tuning on the target dataset - still surpasses the best in-domain baseline by over 8%. Overall, the study provides a comprehensive assessment of MLLM description-mediated sentiment analysis, clarifying the conditions under which it is effective, the scenarios in which it fails, and its comparison with traditional vision-based approaches, while also providing a reproducible benchmark resource for future research.
title Multimodal LLMs See Sentiment
topic Computer Vision and Pattern Recognition
Social and Information Networks
url https://arxiv.org/abs/2508.16873