English

On Robustness in Multimodal Learning

Machine Learning 2023-04-12 v2

Abstract

Multimodal learning is defined as learning over multiple heterogeneous input modalities such as video, audio, and text. In this work, we are concerned with understanding how models behave as the type of modalities differ between training and deployment, a situation that naturally arises in many applications of multimodal learning to hardware platforms. We present a multimodal robustness framework to provide a systematic analysis of common multimodal representation learning methods. Further, we identify robustness short-comings of these approaches and propose two intervention techniques leading to 1.5×1.5\times-4×4\times robustness improvements on three datasets, AudioSet, Kinetics-400 and ImageNet-Captions. Finally, we demonstrate that these interventions better utilize additional modalities, if present, to achieve competitive results of 44.244.2 mAP on AudioSet 20K.

Keywords

Cite

@article{arxiv.2304.04385,
  title  = {On Robustness in Multimodal Learning},
  author = {Brandon McKinzie and Joseph Cheng and Vaishaal Shankar and Yinfei Yang and Jonathon Shlens and Alexander Toshev},
  journal= {arXiv preprint arXiv:2304.04385},
  year   = {2023}
}
R2 v1 2026-06-28T09:56:43.672Z