Towards Understanding Modality Interaction in Multimodal Language Models via Partial Information Decomposition
2026-05-31 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors introduce a method called Partial Information Decomposition (PID) to better understand how multimodal large language models use different types of input, like images and text, when making decisions. They find that some tasks rely on the combined input from both modalities (synergy), while others depend mostly on language alone. The method also works for models with three inputs, such as video, audio, and language, highlighting that visual info often dominates. Lastly, the authors show that adjusting model weights based on PID insights can help improve performance on certain tasks.
Multimodal Large Language ModelsPartial Information DecompositionModality InteractionSynergyRedundancyVision-Language TasksMultimodal ReasoningSensory PIDAudio-Visual FusionModel Reweighting
Authors
Wanlong Fang, Tianle Zhang, Wen Tao, Alvin Chan
Abstract
Understanding modality interaction in multimodal large language models (MLLMs) is central to reliable deployment. We introduce Partial Information Decomposition (PID) as a decision-level framework that separates unique, redundant, and synergistic contributions of sensory and linguistic inputs, beyond representation alignment and outcome-based evaluation. Across vision--language benchmarks, PID reveals recurring modality-use profiles: reasoning and grounding-oriented tasks tend to exhibit high synergy, whereas expert and knowledge-oriented tasks show stronger language-unique reliance. These profiles generalize across model families and predict sensitivity to modality-level interventions. We further extend PID to tri-modal systems with Sensory PID, treating language as a control variable to decompose video--audio information gain. Applied to omni-modal models, Sensory PID reveals a sensory synergy bottleneck dominated by visual information even on audio--visual fusion tasks. Finally, PID-guided reweighting provides initial evidence for improving multimodal reasoning and grounding performance.