Comment by Jatin Nainani (jatin-nainani) on SAEs are highly dataset dependent: a case study on the refusal direction · 2024-11-10T21:16:18.033Z · LW · GW

Makes sense! Thanks! In that case, we can potentially reduce the width, which might (along with a smaller dataset) help scale saes to understanding mechanisms in big models? 

Comment by Jatin Nainani (jatin-nainani) on SAEs are highly dataset dependent: a case study on the refusal direction · 2024-11-07T23:38:32.449Z · LW · GW

Great work! Is there something like too narrow of a dataset? For refusal, what do you think happens if we specifically train on a bunch of examples that show signs refusal?