3 papers
cs.LG2024
Obfuscated Activations Bypass LLM Latent-Space Defenses
Luke Bailey, Alex Serrano, Abhay Sheshadri +7
Recent latent-space monitoring techniques have shown promise as defenses against LLM attacks. These defenses act as scanners that seek to detect harmful activations before they lea…
cs.LG2024
Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning
Dan Braun, Jordan Taylor, Nicholas Goldowsky-Dill +1
Identifying the features learned by neural networks is a core challenge in mechanistic interpretability. Sparse autoencoders (SAEs), which learn a sparse, overcomplete dictionary t…
cs.LG2024
An introduction to graphical tensor notation for mechanistic interpretability
Jordan K. Taylor
Graphical tensor notation is a simple way of denoting linear operations on tensors, originating from physics. Modern deep learning consists almost entirely of operations on or betw…