concept
active
concept:refusal-vectorRefusal Vector
Single linear direction mediating refusal behavior in LLMs, shown by Arditi et al.; related to but distinct from the Assistant Axis
Neighborhood — ranked by edge-count
Papers (1)
paper
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Extracted steering vector capturing semantic dimension of strategic deception in moral dilemmas in Experiment 1
- Arditi et al. 2024 finding that refusal behavior is mediated by one direction in LLM activations; exemplar of single-direction causal results
- The percentage of harmful requests that a model refuses to answer, a common safety metric.
- Used to classify responses with GPT-5.4-mini as LLM judge.
- Linear representations of emotion concepts extracted from activations.
- Models trained with amplification forward steering show reduced efficacy of refusal steering vectorsfinding0.737Fig. 10c.
- Vectors acquired during pretraining in Backpack LMs that have a multiplication effect on model generation
- Consistent with refusal cone work.