LESSWRONGTags
LW

Activation Engineering

•

Applied to Avoiding jailbreaks by discouraging their representation in activation space by Guido Bergman 18d ago

•

Applied to Validating / finding alignment-relevant concepts using neural data by Bogdan Ionut Cirstea 24d ago

•

Applied to [Paper] Programming Refusal with Conditional Activation Steering by Bruce W. Lee 1mo ago

•

Applied to Activation Engineering Theories of Impact by kubanetics 3mo ago

•

Applied to I found >800 orthogonal "write code" steering vectors by Jacob G-W 3mo ago

•

Applied to An Introduction to Representation Engineering - an activation-based paradigm for controlling LLMs by Jan Wehner 3mo ago

•

Applied to Control Vectors as Dispositional Traits by Jan Wehner 3mo ago

•

Applied to Representation Tuning by Jan Wehner 3mo ago

•

Applied to LLMs Universally Learn a Feature Representing Token Frequency / Rarity by Sean Osier 4mo ago

•

Applied to Jailbreak steering generalization by Nina Panickssery 4mo ago

•

Applied to Self-Control of LLM Behaviors by Compressing Suffix Gradient into Prefix Controller by Henry Cai 4mo ago

•

Applied to Introducing SARA: a new activation steering technique by Alejandro Tlaie 4mo ago

•

Applied to Mechanistically Eliciting Latent Behaviors in Language Models by TurnTrout 6mo ago

•

Applied to How well do truth probes generalise? by mishajw 8mo ago

•

Applied to Causal Graphs of GPT-2-Small's Residual Stream by David Udell 8mo ago