RollingQ: Reviving the Cooperation Dynamics in Multimodal Transformer

TL;DR

This paper identifies a bias in multimodal Transformers where the attention mechanism favors one modality, and introduces RollingQ, a simple method to restore dynamic cooperation among modalities, improving multimodal fusion performance.

Contribution

The paper reveals the diminishing dynamic adaptability in existing self-attention models and proposes RollingQ to effectively balance attention across modalities, enhancing multimodal Transformer capabilities.

Findings

RollingQ successfully restores the cooperation dynamics in multimodal Transformers.

Experiments show improved multimodal fusion performance with RollingQ.

The method is validated across various multimodal scenarios.

Abstract

Multimodal learning faces challenges in effectively fusing information from diverse modalities, especially when modality quality varies across samples. Dynamic fusion strategies, such as attention mechanism in Transformers, aim to address such challenge by adaptively emphasizing modalities based on the characteristics of input data. However, through amounts of carefully designed experiments, we surprisingly observed that the dynamic adaptability of widely-used self-attention models diminishes. Model tends to prefer one modality regardless of data characteristics. This bias triggers a self-reinforcing cycle that progressively overemphasizes the favored modality, widening the distribution gap in attention keys across modalities and deactivating attention mechanism's dynamic properties. To revive adaptability, we propose a simple yet effective method Rolling Query (RollingQ), which balances attention allocation by rotating the query to break the self-reinforcing cycle and mitigate the key distribution gap. Extensive experiments on various multimodal scenarios validate the effectiveness of RollingQ and the restoration of cooperation dynamics is pivotal for enhancing the broader capabilities of widely deployed multimodal Transformers. The source code is available at https://github.com/GeWu-Lab/RollingQ_ICML2025.