The Role of Linguistic Priors in Measuring Compositional Generalization of Vision-Language Models
This addresses the problem of accurately assessing compositional generalization in multi-modal models for researchers, but it is incremental as it focuses on metric development rather than model improvement.
The paper identifies that current methods for improving compositional generalization in vision-language models rely on linguistic priors rather than image information, and proposes a new metric to measure compositionality without these priors.
Compositionality is a common property in many modalities including natural languages and images, but the compositional generalization of multi-modal models is not well-understood. In this paper, we identify two sources of visual-linguistic compositionality: linguistic priors and the interplay between images and texts. We show that current attempts to improve compositional generalization rely on linguistic priors rather than on information in the image. We also propose a new metric for compositionality without such linguistic priors.