An Empirical Study on Configuring In-Context Learning Demonstrations for Unleashing MLLMs’ Sentimental Perception Capability

Daiqing Wu, Dongbao Yang, Sicheng Zhao, Can Ma, Yu Zhou
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:67986-67999, 2025.

Abstract

The advancements in Multimodal Large Language Models (MLLMs) have enabled various multimodal tasks to be addressed under a zero-shot paradigm. This paradigm sidesteps the cost of model fine-tuning, emerging as a dominant trend in practical application. Nevertheless, Multimodal Sentiment Analysis (MSA), a pivotal challenge in the quest for general artificial intelligence, fails to accommodate this convenience. The zero-shot paradigm exhibits undesirable performance on MSA, casting doubt on whether MLLMs can perceive sentiments as competent as supervised models. By extending the zero-shot paradigm to In-Context Learning (ICL) and conducting an in-depth study on configuring demonstrations, we validate that MLLMs indeed possess such capability. Specifically, three key factors that cover demonstrations’ retrieval, presentation, and distribution are comprehensively investigated and optimized. A sentimental predictive bias inherent in MLLMs is also discovered and later effectively counteracted. By complementing each other, the devised strategies for three factors result in average accuracy improvements of 15.9% on six MSA datasets against the zero-shot paradigm and 11.2% against the random ICL baseline.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-wu25ao, title = {An Empirical Study on Configuring In-Context Learning Demonstrations for Unleashing {MLLM}s’ Sentimental Perception Capability}, author = {Wu, Daiqing and Yang, Dongbao and Zhao, Sicheng and Ma, Can and Zhou, Yu}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {67986--67999}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/wu25ao/wu25ao.pdf}, url = {https://proceedings.mlr.press/v267/wu25ao.html}, abstract = {The advancements in Multimodal Large Language Models (MLLMs) have enabled various multimodal tasks to be addressed under a zero-shot paradigm. This paradigm sidesteps the cost of model fine-tuning, emerging as a dominant trend in practical application. Nevertheless, Multimodal Sentiment Analysis (MSA), a pivotal challenge in the quest for general artificial intelligence, fails to accommodate this convenience. The zero-shot paradigm exhibits undesirable performance on MSA, casting doubt on whether MLLMs can perceive sentiments as competent as supervised models. By extending the zero-shot paradigm to In-Context Learning (ICL) and conducting an in-depth study on configuring demonstrations, we validate that MLLMs indeed possess such capability. Specifically, three key factors that cover demonstrations’ retrieval, presentation, and distribution are comprehensively investigated and optimized. A sentimental predictive bias inherent in MLLMs is also discovered and later effectively counteracted. By complementing each other, the devised strategies for three factors result in average accuracy improvements of 15.9% on six MSA datasets against the zero-shot paradigm and 11.2% against the random ICL baseline.} }
Endnote
%0 Conference Paper %T An Empirical Study on Configuring In-Context Learning Demonstrations for Unleashing MLLMs’ Sentimental Perception Capability %A Daiqing Wu %A Dongbao Yang %A Sicheng Zhao %A Can Ma %A Yu Zhou %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-wu25ao %I PMLR %P 67986--67999 %U https://proceedings.mlr.press/v267/wu25ao.html %V 267 %X The advancements in Multimodal Large Language Models (MLLMs) have enabled various multimodal tasks to be addressed under a zero-shot paradigm. This paradigm sidesteps the cost of model fine-tuning, emerging as a dominant trend in practical application. Nevertheless, Multimodal Sentiment Analysis (MSA), a pivotal challenge in the quest for general artificial intelligence, fails to accommodate this convenience. The zero-shot paradigm exhibits undesirable performance on MSA, casting doubt on whether MLLMs can perceive sentiments as competent as supervised models. By extending the zero-shot paradigm to In-Context Learning (ICL) and conducting an in-depth study on configuring demonstrations, we validate that MLLMs indeed possess such capability. Specifically, three key factors that cover demonstrations’ retrieval, presentation, and distribution are comprehensively investigated and optimized. A sentimental predictive bias inherent in MLLMs is also discovered and later effectively counteracted. By complementing each other, the devised strategies for three factors result in average accuracy improvements of 15.9% on six MSA datasets against the zero-shot paradigm and 11.2% against the random ICL baseline.
APA
Wu, D., Yang, D., Zhao, S., Ma, C. & Zhou, Y.. (2025). An Empirical Study on Configuring In-Context Learning Demonstrations for Unleashing MLLMs’ Sentimental Perception Capability. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:67986-67999 Available from https://proceedings.mlr.press/v267/wu25ao.html.

Related Material