RSS

Alan Cooney

Karma: 141

What Drives the Com­pli­ance Gap? A Three-Driver De­com­po­si­tion of Align­ment Faking

28 May 2026 10:50 UTC
22 points
0 comments8 min readLW link
(arxiv.org)

Misal­ign­ment clas­sifiers: Why they’re hard to eval­u­ate ad­ver­sar­i­ally, and why we’re study­ing them anyway

15 Aug 2025 11:48 UTC
68 points
3 comments17 min readLW link

White Box Con­trol at UK AISI—Up­date on Sand­bag­ging Investigations

10 Jul 2025 13:37 UTC
81 points
10 comments18 min readLW link