finding
active
finding:models-trained-with-amplification-forward-steering-show-reduced-efficacy-of-refusal-steering-vectorsModels trained with amplification forward steering show reduced efficacy of refusal steering vectors
Fig. 10c.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Sidharth Baskaran · Max Loeffler +13
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Model may fight chosen-response steering.
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.
- Steering vectors used to reduce eval awareness can inadvertently introduce alternative user personasfinding0.792A side effect observed when applying activation steering: the model's response persona changed unexpectedly.
- Mechanistic interpretation of how activation steering induces deception through the model's reasoning process
- Steering toward deflection vectors makes speaker deny the emotion rather than increasing itfinding0.789Deflection not internal state.
- Central claim of the paper; supported by the model organism ground-truth approach.
- Steering vectors computed over multiple shots and norm-normalized have weaker effect than 0-shot vectorsfinding0.780Counter-intuitive result.
- Validates that steering vectors capture reflection semantics by finding tokens reported in related work.
- Practical application proposed based on mechanistic findings
- Key asymmetry finding: suppressing reflection is easier than inducing it.