Re:Dynamic Abliteration: Non-Destructive Refusal Suppression...
by phatak-dev on Thursday September 24, 2026 @11:56AM
(#49832527)
Makes sense.. Just updated my comment to reflect it.. Thanks for making it clear.
Residual stream steering is what the authors do, the orthogonalized weights are downstream of that. Those weights, of an obliterated model, are all computed w.r.t. the computed residual stream vector. The original paper focuses on steering, but the community loves the simplicity of not needing to make changes at test-time.