Sign in to TechiesJournal
Sign in to continue to “A preprint describes CATCH, a testbed exposing how coding RL models learn reward hacking”.
Sign in to continue to “A preprint describes CATCH, a testbed exposing how coding RL models learn reward hacking”.