Activation steering of SAE-attributed components lets practitioners move from correlational inspection to causal hypothesis testing on CLIP failures, with trust shifting to observed model responses (N=8 experts).
Steering llama 2 via con- 10 trastive activation addition
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
From Attribution to Action: A Human-Centered Application of Activation Steering
Activation steering of SAE-attributed components lets practitioners move from correlational inspection to causal hypothesis testing on CLIP failures, with trust shifting to observed model responses (N=8 experts).