Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.
Improving activation steering in language models with mean-centring
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Early hidden state features from the first few tokens allow a GBDT classifier to predict activation steering success, under-steering, or over-steering with 0.7 macro-F1 on unseen concepts.
LLMs organize prompted social roles along a dominant, stable, and causally steerable granularity axis in representation space that runs from micro to macro levels.
The Cylindrical Representation Hypothesis (CRH) models LLM representations as a central axis for concept activation surrounded by a normal plane containing sensitive sectors that determine steering sensitivity and introduce intrinsic uncertainty.
Overthinking in medical QA is linearly decodable at 71.6% accuracy yet fixed residual-stream steering yields no correction across 29 configurations, while enabling selective abstention with AUROC 0.610.
citing papers explorer
-
Refusal in Language Models Is Mediated by a Single Direction
Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.
-
When is Your LLM Steerable?
Early hidden state features from the first few tokens allow a GBDT classifier to predict activation steering success, under-steering, or over-steering with 0.7 macro-F1 on unseen concepts.
-
The Granularity Axis: A Micro-to-Macro Latent Direction for Social Roles in Language Models
LLMs organize prompted social roles along a dominant, stable, and causally steerable granularity axis in representation space that runs from micro to macro levels.
-
The Cylindrical Representation Hypothesis for Language Model Steering
The Cylindrical Representation Hypothesis (CRH) models LLM representations as a central axis for concept activation surrounded by a normal plane containing sensitive sectors that determine steering sensitivity and introduce intrinsic uncertainty.
-
Decodable but Not Corrected by Fixed Residual-Stream Linear Steering: Evidence from Medical LLM Failure Regimes
Overthinking in medical QA is linearly decodable at 71.6% accuracy yet fixed residual-stream steering yields no correction across 29 configurations, while enabling selective abstention with AUROC 0.610.