Posts

Comments

Comment by Hao Huang (hao-huang) on How to use and interpret activation patching · 2024-12-25T05:42:09.957Z · LW · GW

find that denoising (1) “Nobel → L0H0” and (2) “Peace → L1N42” paths is sufficient.

Does it mean denoising the two paths simultaneously or separately?

Comment by Hao Huang (hao-huang) on How to use and interpret activation patching · 2024-12-25T03:46:08.793Z · LW · GW