Visual Instruction Bottleneck Tuning
This addresses the challenge of enhancing MLLM generalization without costly data or architecture changes, though it is an incremental method based on information bottleneck principles.
The paper tackles the problem of multimodal large language models (MLLMs) suffering performance degradation under distribution shifts by proposing Visual Instruction Bottleneck Tuning (Vittle), which improves robustness across 45 datasets and 30 shift scenarios.
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.