CLIP Models are Few-shot Learners: Empirical Studies on VQA and Visual Entailment
This work addresses the problem of adapting pre-trained models for few-shot learning in vision-language tasks, offering a parameter-efficient fine-tuning strategy, though it is incremental as it builds on existing CLIP capabilities.
The paper demonstrates that CLIP, originally a zero-shot visual encoder, can function as a strong few-shot learner for vision-language tasks like visual question answering and visual entailment, achieving competitive results without additional pre-training.
CLIP has shown a remarkable zero-shot capability on a wide range of vision tasks. Previously, CLIP is only regarded as a powerful visual encoder. However, after being pre-trained by language supervision from a large amount of image-caption pairs, CLIP itself should also have acquired some few-shot abilities for vision-language tasks. In this work, we empirically show that CLIP can be a strong vision-language few-shot learner by leveraging the power of language. We first evaluate CLIP's zero-shot performance on a typical visual question answering task and demonstrate a zero-shot cross-modality transfer capability of CLIP on the visual entailment task. Then we propose a parameter-efficient fine-tuning strategy to boost the few-shot performance on the vqa task. We achieve competitive zero/few-shot results on the visual question answering and visual entailment tasks without introducing any additional pre-training procedure.