joanv

Karma: 155

www.joanvelja.com

joanv’s Shortform

joanv8 Apr 2026 9:21 UTC

2 points

10 comments1 min readLW link

[Paper] When can we trust untrusted monitoring? A safety case sketch across collusion strategies

Nelson Gardner-Challis, Morgan S, J Bostock, BarAltiva, joanv and Charlie Griffin

10 Mar 2026 17:28 UTC

46 points

2 comments6 min readLW link

[Paper] Hidden in Plain Text: Emergence and Mitigation of Steganographic Collusion in LLMs

Yohan Mathew, joanv, robert mccarthy, ollie, Nandi and Dylan Cope

25 Sep 2024 14:52 UTC

37 points

2 comments4 min readLW link

(arxiv.org)