jordine

• jordineApr 16, 2025, 11:21 AM

  1 point

  0

  in reply to: Cleo Nardo’s comment on: Can SAE steer­ing re­veal sand­bag­ging?

  Refusals were mostly 1-2%, so ignoring them doesn’t change results significantly. Ignoring gibberish does change results, but since we are measuring correct answers this shouldn’t matter

• jordineJan 2, 2025, 5:02 AM

  1 point

  0

  in reply to: GriebelGrutjes’s comment on: Shal­low re­view of tech­ni­cal AI safety, 2024

  fixed! edited hyperlink.

• jordineDec 30, 2024, 4:30 AM

  1 point

  0

  in reply to: Satron’s comment on: Shal­low re­view of tech­ni­cal AI safety, 2024

  edited, thanks for catching this!