Enregistré dans:
Détails bibliographiques
Auteurs principaux: Oh, Changdae, Li, Jiatong, Im, Shawn, Li, Sharon
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2505.13946
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866917025422508032
author Oh, Changdae
Li, Jiatong
Im, Shawn
Li, Sharon
author_facet Oh, Changdae
Li, Jiatong
Im, Shawn
Li, Sharon
contents Despite widespread adoption, multimodal large language models (MLLMs) suffer performance degradation when encountering unfamiliar queries under distribution shifts. Existing methods to improve MLLM generalization typically require either more instruction data or larger advanced model architectures, both of which incur non-trivial human labor or computational costs. In this work, we take an alternative approach to enhance the generalization and robustness of MLLMs under distribution shifts, from a representation learning perspective. Inspired by information bottleneck (IB) principle, we derive a variational lower bound of the IB for MLLMs and devise a practical implementation, Visual Instruction Bottleneck Tuning (Vittle). We then provide a theoretical justification of Vittle by revealing its connection to an information-theoretic robustness metric of MLLM. Empirical validation of multiple MLLMs on open-ended and closed-form question answering and object hallucination detection tasks over 45 datasets, including 30 shift scenarios, demonstrates that Vittle consistently improves the MLLM's robustness under shifts by pursuing the learning of a minimal sufficient representation.
format Preprint
id arxiv_https___arxiv_org_abs_2505_13946
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Visual Instruction Bottleneck Tuning
Oh, Changdae
Li, Jiatong
Im, Shawn
Li, Sharon
Artificial Intelligence
Despite widespread adoption, multimodal large language models (MLLMs) suffer performance degradation when encountering unfamiliar queries under distribution shifts. Existing methods to improve MLLM generalization typically require either more instruction data or larger advanced model architectures, both of which incur non-trivial human labor or computational costs. In this work, we take an alternative approach to enhance the generalization and robustness of MLLMs under distribution shifts, from a representation learning perspective. Inspired by information bottleneck (IB) principle, we derive a variational lower bound of the IB for MLLMs and devise a practical implementation, Visual Instruction Bottleneck Tuning (Vittle). We then provide a theoretical justification of Vittle by revealing its connection to an information-theoretic robustness metric of MLLM. Empirical validation of multiple MLLMs on open-ended and closed-form question answering and object hallucination detection tasks over 45 datasets, including 30 shift scenarios, demonstrates that Vittle consistently improves the MLLM's robustness under shifts by pursuing the learning of a minimal sufficient representation.
title Visual Instruction Bottleneck Tuning
topic Artificial Intelligence
url https://arxiv.org/abs/2505.13946