AIBullisharXiv โ CS AI ยท 5h ago1
๐ง
NExT-Guard: Training-Free Streaming Safeguard without Token-Level Labels
Researchers introduce NExT-Guard, a training-free framework for real-time AI safety monitoring that uses Sparse Autoencoders to detect unsafe content in streaming language models. The system outperforms traditional supervised training methods while requiring no token-level annotations, making it more cost-effective and scalable for deployment.