Understanding Transferable Representation Learning and Zero-shot Transfer in CLIP
This work addresses the problem of improving multi-modal learning for researchers and practitioners by providing theoretical insights and a more effective method, though it is incremental as it builds directly on CLIP.
The paper tackles the lack of theoretical understanding of CLIP's transferable representation learning and zero-shot transfer, proposing a new CLIP-type approach that achieves better performance than CLIP and other state-of-the-art methods on benchmark datasets.
Multi-modal learning has become increasingly popular due to its ability to leverage information from different data sources (e.g., text and images) to improve the model performance. Recently, CLIP has emerged as an effective approach that employs vision-language contrastive pretraining to learn joint image and text representations and exhibits remarkable performance in zero-shot learning and text-guided natural image generation. Despite the huge practical success of CLIP, its theoretical understanding remains elusive. In this paper, we formally study transferrable representation learning underlying CLIP and demonstrate how features from different modalities get aligned. We also analyze its zero-shot transfer performance on the downstream tasks. Inspired by our analysis, we propose a new CLIP-type approach, which achieves better performance than CLIP and other state-of-the-art methods on benchmark datasets.