Towards Reliable Zero Shot Classification in Self-Supervised Models with Conformal Prediction
This addresses the challenge of ensuring reliable zero-shot classification for users in domains like medicine where caption compatibility is uncertain, representing an incremental improvement.
The paper tackles the problem of unreliable zero-shot classification in self-supervised models like CLIP by framing it as an outlier detection task and using conformal prediction to assess caption compatibility, showing improved reliability in a real-world medical example.
Self-supervised models trained with a contrastive loss such as CLIP have shown to be very powerful in zero-shot classification settings. However, to be used as a zero-shot classifier these models require the user to provide new captions over a fixed set of labels at test time. In many settings, it is hard or impossible to know if a new query caption is compatible with the source captions used to train the model. We address these limitations by framing the zero-shot classification task as an outlier detection problem and develop a conformal prediction procedure to assess when a given test caption may be reliably used. On a real-world medical example, we show that our proposed conformal procedure improves the reliability of CLIP-style models in the zero-shot classification setting, and we provide an empirical analysis of the factors that may affect its performance.