Neural Cleanse
PulseAugur coverage of Neural Cleanse — every cluster mentioning Neural Cleanse across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New TARE method accurately measures AI backdoor defense costs
A new research paper introduces TARE, a method designed to more accurately assess the cost of backdoor defenses in AI models. Traditional methods measure the drop in clean accuracy, which can be misleading as it doesn't…
-
New backdoor attack exploits Vision MoE capacity overflow
Researchers have identified a new vulnerability in Mixture-of-Experts (MoE) architectures for Vision Transformers, termed 'Capacity Overflow'. This vulnerability stems from the batch-dependent token dispatch mechanism u…
-
New BadBone attack targets visual prompt learning models
Researchers have introduced BadBone, a novel backdoor attack targeting visual prompt learning models. This attack compromises the backbone model, causing downstream tasks that use prompt learning to inherit vulnerabilit…