method
active
method:contrastive-pair-activation-subtractionContrastive pair activation subtraction
Technique for obtaining concept vectors by presenting model with two scenarios differing in one respect and subtracting activations to isolate conceptual difference.
Neighborhood — ranked by edge-count
Concepts (1)
concept
- Concept InjectionimplementsTechnique of injecting activation patterns associated with specific concepts into a model's internal states to test whether self-reports reflect ground truth.
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.
- Method formalized by Rimsky et al. (2024) that this paper adopts for extracting persona vectors
- Pairs of prompts at different reflection levels used to compute steering vectors.
- Applying reverse steering vector to suppress reflective behavior at inference time.
- Pairs of statements with opposite truth values used as input to CCS; e.g., cities and neg_cities paired statements
- Performance gains over CAA in steering tasks.
- An existing activation steering method used as comparative baseline.
- Experimental design where injection strengths are swapped between sentences in two parts of each trial to cancel positional preferences
- Core technique: takes mean difference of model activations on contrastive prompts and adds the resulting vector to the residual stream at inference time.