Saved in:
Bibliographic Details
Main Authors: Bennett, Liam, Clark, Mason, Anderson, Lucas, Satou, Hana, Martinez, Olivia
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.12733
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909649832247296
author Bennett, Liam
Clark, Mason
Anderson, Lucas
Satou, Hana
Martinez, Olivia
author_facet Bennett, Liam
Clark, Mason
Anderson, Lucas
Satou, Hana
Martinez, Olivia
contents Multimodal foundation models have achieved impressive progress across a wide range of vision-language tasks. However, existing approaches often adopt fixed or task-specific fusion strategies, neglecting the intrinsic variability of modality reliability and sample complexity. In this paper, we propose Modality-Aware Adaptive Fusion Scheduling (MA-AFS), a general framework that learns to dynamically modulate the contribution of each modality on a per-instance basis. MA-AFS introduces a lightweight neural scheduler that predicts modality fusion weights by integrating visual and textual entropy signals along with cross-modal agreement cues. This enables the model to adaptively emphasize more reliable modalities, especially under noisy, missing, or misaligned inputs. We formulate the fusion process as a differentiable scheduling mechanism, analyze its theoretical consistency and regularization effect, and demonstrate that it improves robustness without increasing model capacity significantly. Extensive experiments on image-text retrieval, captioning, and visual question answering show that MA-AFS achieves consistent performance gains over strong baselines such as CLIP, ALBEF, and BLIP. Moreover, MA-AFS exhibits improved robustness under modality corruption and enhanced generalization under domain shifts. Our work highlights the importance of adaptive fusion and opens a promising direction toward reliable and uncertainty-aware multimodal learning.
format Preprint
id arxiv_https___arxiv_org_abs_2506_12733
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning to Fuse: Modality-Aware Adaptive Scheduling for Robust Multimodal Foundation Models
Bennett, Liam
Clark, Mason
Anderson, Lucas
Satou, Hana
Martinez, Olivia
Computer Vision and Pattern Recognition
Multimodal foundation models have achieved impressive progress across a wide range of vision-language tasks. However, existing approaches often adopt fixed or task-specific fusion strategies, neglecting the intrinsic variability of modality reliability and sample complexity. In this paper, we propose Modality-Aware Adaptive Fusion Scheduling (MA-AFS), a general framework that learns to dynamically modulate the contribution of each modality on a per-instance basis. MA-AFS introduces a lightweight neural scheduler that predicts modality fusion weights by integrating visual and textual entropy signals along with cross-modal agreement cues. This enables the model to adaptively emphasize more reliable modalities, especially under noisy, missing, or misaligned inputs. We formulate the fusion process as a differentiable scheduling mechanism, analyze its theoretical consistency and regularization effect, and demonstrate that it improves robustness without increasing model capacity significantly. Extensive experiments on image-text retrieval, captioning, and visual question answering show that MA-AFS achieves consistent performance gains over strong baselines such as CLIP, ALBEF, and BLIP. Moreover, MA-AFS exhibits improved robustness under modality corruption and enhanced generalization under domain shifts. Our work highlights the importance of adaptive fusion and opens a promising direction toward reliable and uncertainty-aware multimodal learning.
title Learning to Fuse: Modality-Aware Adaptive Scheduling for Robust Multimodal Foundation Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.12733