Why Is Prompt Tuning for Vision-Language Models Robust to Noisy Labels?
This addresses the problem of label noise in adapting vision-language models for classification, offering insights for researchers in robust machine learning, though it is incremental in exploring existing model properties.
The paper investigates why few-shot prompt tuning for vision-language models like CLIP is robust to noisy labels, finding that fixed classname tokens and pre-trained embeddings reduce gradient noise and provide strong priors, and demonstrates that using noisy zero-shot predictions can enhance accuracy in unsupervised settings.
Vision-language models such as CLIP learn a generic text-image embedding from large-scale training data. A vision-language model can be adapted to a new classification task through few-shot prompt tuning. We find that such a prompt tuning process is highly robust to label noises. This intrigues us to study the key reasons contributing to the robustness of the prompt tuning paradigm. We conducted extensive experiments to explore this property and find the key factors are: 1) the fixed classname tokens provide a strong regularization to the optimization of the model, reducing gradients induced by the noisy samples; 2) the powerful pre-trained image-text embedding that is learned from diverse and generic web data provides strong prior knowledge for image classification. Further, we demonstrate that noisy zero-shot predictions from CLIP can be used to tune its own prompt, significantly enhancing prediction accuracy in the unsupervised setting. The code is available at https://github.com/CEWu/PTNL.