Knowledge Regularized Negative Feature Tuning of Vision-Language Models for Out-of-Distribution Detection

What is it about?

To enhanced the OOD detection capabilities of vision-language models on unseen classes and styles, KR-NFT integrates an innovative adaptation architecture termed Negative Feature Tuning (NFT) and a corresponding knowledge-regularization (KR) optimization strategy.

Featured Image

Why is it important?

This work inspires future research not only on improving OOD detection performance on training data but also on enhancing generalization performance to unseen classes and styles.