Brief

Researchers have developed a new method called DRIFT for detecting AI-generated images, which adapts to unseen image generators. This approach formulates detection as learning an invariance manifold of real images using one-class supervision. DRIFT utilizes lightweight projection heads to separate image representation space into robust and fragile subspaces, enabling detection by testing for violations of learned invariances. AI

Researchers have introduced DRIFT, a new framework designed to improve the efficiency of training large language models for multi-turn interactions. DRIFT addresses the trade-off between costly online reinforcement learning and less effective offline supervised fine-tuning. By decoupling trajectory sampling from optimization and using importance weights, DRIFT achieves performance comparable to reinforcement learning while maintaining the simplicity and efficiency of supervised fine-tuning. AI