Ashish Kharbanda·4d agoMechanistic Interpretability of LLM Refusal BehaviourWhere does a language model keep its “refusals”?A response icon1
Shayan Bemanian·Oct 27Reading LLMs Minds — Natural Language Autoencoders ExplainedHow can you train a neural network to translate activations into English when nobody knows what the activations mean in the first place?
Alborz Nazari·Oct 19Your Firewall Can’t Read English: What Linux and Windows Hardening Get Right, and Where It Breaks…1. The claim and its premises
Pancham_B·Sep 15Building model_probe: An Open-Source Framework for Adversarial Safety Testing of LLMsI recently released model_probe v0.1.0 (https://github.com/panchambanerjee/model_probe), an open-source Python framework for adversarial…
InLunit Team BlogbyCedric Caruzzo·Jul 20Deceptive Grounding: Entity Attribution Failure in Clinical Retrieval-Augmented GenerationTL;DR. We found a failure mode in clinical retrieval-augmented generation (RAG) that is invisible to every standard check. The model…
