Manifold steering along activation geometry induces behavioral trajectories matching the natural manifold of outputs, while linear steering produces off-manifold unnatural behaviors.
Steering off course: Reliability challenges in steering language models
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 5verdicts
UNVERDICTED 5roles
background 1polarities
unclear 1representative citing papers
Activation steering produces synthetic safety-violating data that improves downstream classifiers over prompting on most tested concepts when a harmonic mean of alignment, coherence, and diversity is optimized.
Order is distinct from control, where control is defined as a local receiver-gated response law demonstrated across biological circuits and LLM response panels with reported prediction accuracies of 72-84%.
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.
The survey organizes mechanistic interpretability techniques into a Locate-Steer-Improve framework to enable actionable improvements in LLM alignment, capability, and efficiency.
citing papers explorer
-
Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior
Manifold steering along activation geometry induces behavioral trajectories matching the natural manifold of outputs, while linear steering produces off-manifold unnatural behaviors.
-
Activation Steering for Synthetic Data Generation: The Role of Diversity in Downstream Safety Detection
Activation steering produces synthetic safety-violating data that improves downstream classifiers over prompting on most tested concepts when a harmonic mean of alignment, coherence, and diversity is optimized.
-
Order Is Not Control
Order is distinct from control, where control is defined as a local receiver-gated response law demonstrated across biological circuits and LLM response panels with reported prediction accuracies of 72-84%.
-
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.
-
Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models
The survey organizes mechanistic interpretability techniques into a Locate-Steer-Improve framework to enable actionable improvements in LLM alignment, capability, and efficiency.