The Geometry of Harmful Intent: Training-Free Anomaly Detection via Angular Deviation in LLM Residual Streams
arXiv:2603.27412v1 Announce Type: new
Abstract: We present LatentBiopsy, a training-free method for detecting harmful prompts by analysing the geometry of residual-stream activations in large language models. Given 200 safe normative prompts, LatentBi…