Efficient Membership Inference for Diffusion Models via Noise Injection
Discover an efficient method for membership inference in diffusion models using noise aggregation and low-intensity noise injection to enhance privacy prot...
DB-FGA-Net: Accurate Multi-Class Brain Tumor Classification
DB-FGA-Net uses dual backbone frequency gated attention and Grad-CAM for precise multi-class brain tumor classification with clinical interpretability.
STELA: Linguistics-Based Watermarking for LLMs
Discover STELA, a linguistics-aware watermarking method for LLMs that ensures robust, publicly verifiable AI content detection without model access.
Adaptive Entropy Regularization Boosts LLM Reinforcement Learning
Discover how Adaptive Entropy Regularization enhances exploration and reasoning in Large Language Models through dynamic entropy management.
Red-Teaming Web-Augmented Large Language Models Safely
Explore CREST-Search, a red-teaming framework to identify and mitigate risks in web-augmented large language models for safer AI search results.