openreview8.0 / 10
Resisting RL Elicitation of Biosecurity Capabilities: Reasoning Models Exploration Hacking on WMDP
Abstract
Research paper by MATS scholars. Authors: Joschka Braun, Damon Falck, Yeonwoo Jang
Research area
post-trainingRLRL/RLHF
Published
—
Source
openreview
Org
MATS
Sign in to read and join the discussion.