RSS

RobertKirk

Karma: 321

Research Scientist on the Safeguards team at UK AI Security Institute

A Sober Look at Steer­ing Vec­tors for LLMs

Joschka Braun, Dmitrii Krasheninnikov, Usman Anwar, RobertKirk, Daniel Tan and David Scott Krueger (formerly: capybaralet)
23 Nov 2024 17:30 UTC
40 points
0 comments5 min readLW link

Spec­u­la­tive in­fer­ences about path de­pen­dence in LLM su­per­vised fine-tun­ing from re­sults on lin­ear mode con­nec­tivity and model souping

RobertKirk20 Jul 2023 9:56 UTC
39 points
2 comments5 min readLW link

Causal con­fu­sion as an ar­gu­ment against the scal­ing hypothesis

RobertKirk and David Scott Krueger (formerly: capybaralet)
20 Jun 2022 10:54 UTC
86 points
30 comments15 min readLW link

Spar­sity and in­ter­pretabil­ity?

Ada Böhm, RobertKirk and Tomáš Gavenčiak
1 Jun 2020 13:25 UTC
41 points
3 comments7 min readLW link

How can In­ter­pretabil­ity help Align­ment?

RobertKirk and Tomáš Gavenčiak
23 May 2020 16:16 UTC
37 points
3 comments9 min readLW link

What is In­ter­pretabil­ity?

RobertKirk, Tomáš Gavenčiak and Ada Böhm
17 Mar 2020 20:23 UTC
39 points
1 comment11 min readLW link