Refusal Direction in LLMs

Prior finding that LLM refusal is mediated by a single latent direction, analogous to this paper's reflection direction.

Neighborhood — ranked by edge-count

concept

Refusal direction
related_to
Arditi et al. 2024 finding that refusal behavior is mediated by one direction in LLM activations; exemplar of single-direction causal results
Latent Direction of Reflection
analogous_to
The paper's central construct: a vector in LLM activation space encoding the transition between reflection levels.

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Truth direction in LLMsconcept0.832
Linear direction in LLM activations associated with truthfulness, identified by Burns et al. 2022 and Azaria & Mitchell 2023
Reflection in LLMsconcept0.775
The core phenomenon studied: the ability of LLMs to evaluate and revise their own reasoning.
Persona Assignment to LLMsconcept0.769
The practice of providing LLMs with a persona description to shape their generated responses
Agentic behavior in LLMsconcept0.765
Nine traits characterizing LLM-as-agent deployments; all nine are natural in both models
Model view of LLM individuationconcept0.761
The view that the individual is the abstract function defined by a given architecture and weight matrix
Behavioral organization of LLMsconcept0.761
The layered structure of which behaviors a model defaults to, can be steered toward, or resists—the central object of study
LLM judge evaluationmethod0.757
Using Claude Sonnet 4 as a grader to categorize model responses according to predefined criteria.
Truth Direction in LLM Latent Spaceconcept0.757
A specific direction in an LLM's residual stream that encodes the truth or falsehood of factual statements