From Error Prediction to Causal Influence: Auditing SAE Features in Multiple-Choice QA

arXiv:2610.38067v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) can predict internal features associated with model errors, but association alone does not establish causal influence. We study this distinction in multiple-choice question answering, where answer preference can be measured b

arXiv cs.CLFreitag 05:00 · 30.10.1 Min. Lesezeit