Discover a novel weak supervision method to distill hallucination signals into transformer models for efficient, internal hallucination detection without e...
Enhance zero-shot generalization in visual unsupervised RL with saliency-guided representation and consistency policy learning for better task performance.