rusheb

Stress Test­ing De­liber­a­tive Align­ment for Anti-Schem­ing Training

Abla­tions for “Fron­tier Models are Ca­pable of In-con­text Schem­ing”

Fron­tier Models are Ca­pable of In-con­text Scheming

Apollo Re­search 1-year update

A starter guide for evals

Scal­able And Trans­fer­able Black-Box Jailbreaks For Lan­guage Models Via Per­sona Modulation

Un­der­stand­ing mesa-op­ti­miza­tion us­ing toy models

• rushebJan 18, 2023, 1:29 PM

  LW: 4 AF: 1

  3 votes

  Overall karma indicates overall quality.

  0

  0 votes

  Agreement karma indicates agreement, separate from overall quality.

  AF

  on: Causal Scrub­bing: a method for rigor­ously test­ing in­ter­pretabil­ity hy­pothe­ses [Red­wood Re­search]

  If your hypothesis predicts that model performance will be preserved if you swap the input to any other input which has a particular property, but no other inputs in the dataset have that property, causal scrubbing can’t test your hypothesis

  Would it be possible to make interventions which we expect not to preserve the model’s behaviour, and assert that the behaviour does in fact change?