Jai Bhagat

• Jai Bhagat 31 Jul 2025 13:30 UTC

  1 point

  0

  on: Against Al­most Every The­ory of Im­pact of Interpretability

  Do any of these recent papers within the last year change your view on interp impact for these theories? :
  
  1. Understanding misalignment (at least some initial insights): https://​​arxiv.org/​​html/​​2502.17424v2

  2. Better prediction of future systems (interp for scaling):
  https://​​arxiv.org/​​abs/​​2303.13506

  3. Auditing to reveal hidden objectives:
  https://​​www.anthropic.com/​​research/​​auditing-hidden-objectives

• Jai Bhagat 27 Jul 2025 22:07 UTC

  4 points

  0

  on: Ac­ti­va­tion space in­ter­pretabil­ity may be doomed

  Nice post! Random thought—problem 1 seems a problem in systems neuroscience as well.

• Jai Bhagat 30 Jun 2025 7:30 UTC

  1 point

  0

  in reply to: Dan Braun’s comment on: Com­pressed Com­pu­ta­tion is (prob­a­bly) not Com­pu­ta­tion in Superposition

  Yes! But only if the mess is the residual stream, i.e. includes $x$! This is the heart of the necessary “feature mixing” we discuss

Com­pressed Com­pu­ta­tion is (prob­a­bly) not Com­pu­ta­tion in Superposition