RSS

Connor Kissane

Karma: 350

SAEs are highly dataset de­pen­dent: a case study on the re­fusal direction

7 Nov 2024 5:22 UTC
62 points
4 comments14 min readLW link

Open Source Repli­ca­tion of An­thropic’s Cross­coder pa­per for model-diffing

27 Oct 2024 18:46 UTC
38 points
4 comments5 min readLW link

Base LLMs re­fuse too

29 Sep 2024 16:04 UTC
60 points
20 comments10 min readLW link

SAEs (usu­ally) Trans­fer Between Base and Chat Models

18 Jul 2024 10:29 UTC
65 points
0 comments10 min readLW link

At­ten­tion Out­put SAEs Im­prove Cir­cuit Analysis

21 Jun 2024 12:56 UTC
31 points
0 comments19 min readLW link

We In­spected Every Head In GPT-2 Small us­ing SAEs So You Don’t Have To

6 Mar 2024 5:03 UTC
58 points
0 comments12 min readLW link

At­ten­tion SAEs Scale to GPT-2 Small

3 Feb 2024 6:50 UTC
77 points
4 comments8 min readLW link

Sparse Au­toen­coders Work on At­ten­tion Layer Outputs

16 Jan 2024 0:26 UTC
83 points
9 comments18 min readLW link