Towards Unified Vision-Language Models with Incomplete Multi-Modal Inputs
It addresses the practical problem of missing modalities in video-language models, which has been largely neglected, but the approach is incremental as it builds on existing models.
The paper proposes the first unified video-language model to handle incomplete multi-modal inputs (e.g., missing video or language) caused by real-world sensor failures, and shows it can be used as a plug-and-play module to improve performance across multiple tasks.
Video-Language Models (VLMs) have demonstrated impressive multi-modal reasoning capabilities across diverse computer vision applications. However, these VLMs are task-specific and assume that both video and language inputs are complete. However, real-world VLM applications might face challenges due to deactivated sensors (e.g., cameras are unavailable due to data privacy), yielding modality-incomplete data and leading to inconsistency between training and testing data. While straightforward incomplete input can boast training generalization-ability and lead to training failure, its potential risks to VLMs regarding safety and trustworthiness have been largely neglected. To this end, we make the first attempt to propose a unified incomplete video-language model to process the incomplete multi-modal inputs. Extensive experimental results show that our method can serve as a plug-and-play module for previous works to improve their performance in various multi-modal tasks.