Understanding Retrieval-Augmented Task Adaptation for Vision-Language Models
This work addresses the adaptation problem for vision-language models in low-data regimes, providing systematic understanding for researchers in multimodal AI.
The paper investigates how retrieval-augmented adaptation improves vision-language models on fine-grained datasets with underrepresented categories, revealing insights into uni-modal and cross-modal retrieval and the importance of logit ensemble, supported by theoretical analysis.
Pre-trained contrastive vision-language models have demonstrated remarkable performance across a wide range of tasks. However, they often struggle on fine-trained datasets with categories not adequately represented during pre-training, which makes adaptation necessary. Recent works have shown promising results by utilizing samples from web-scale databases for retrieval-augmented adaptation, especially in low-data regimes. Despite the empirical success, understanding how retrieval impacts the adaptation of vision-language models remains an open research question. In this work, we adopt a reflective perspective by presenting a systematic study to understand the roles of key components in retrieval-augmented adaptation. We unveil new insights on uni-modal and cross-modal retrieval and highlight the critical role of logit ensemble for effective adaptation. We further present theoretical underpinnings that directly support our empirical observations.