Misalign, Contrast then Distill: Rethinking Misalignments in Language-Image Pretraining
This addresses data-efficiency issues in multimodal AI for researchers and practitioners, though it is incremental as it builds on existing contrastive pretraining frameworks.
The paper tackles the problem of image-text misalignments caused by random augmentations in contrastive language-image pretraining, proposing a method that uses these misalignments as training signals to achieve state-of-the-art transferability in classification and retrieval datasets.
Contrastive Language-Image Pretraining has emerged as a prominent approach for training vision and text encoders with uncurated image-text pairs from the web. To enhance data-efficiency, recent efforts have introduced additional supervision terms that involve random-augmented views of the image. However, since the image augmentation process is unaware of its text counterpart, this procedure could cause various degrees of image-text misalignments during training. Prior methods either disregarded this discrepancy or introduced external models to mitigate the impact of misalignments during training. In contrast, we propose a novel metric learning approach that capitalizes on these misalignments as an additional training source, which we term "Misalign, Contrast then Distill (MCD)". Unlike previous methods that treat augmented images and their text counterparts as simple positive pairs, MCD predicts the continuous scales of misalignment caused by the augmentation. Our extensive experimental results show that our proposed MCD achieves state-of-the-art transferability in multiple classification and retrieval downstream datasets.