Joschka Braun(@BraunJoschka):RL assumes that LLMs explore well during training. What if they choose not to? In our new ICML paper with @GoogleDeepMind, we train LLMs that strategically resist RL capability elicitation by under-exploring. We study this threat model, called exploration hacking.

Joschka Braun

@BraunJoschka

AI safety researcher @ApolloResearch | Science of Scheming | prev. @MATSprogram @kasl_ai @health_nlp @uni_tue

가입 April 2020

641 팔로잉 중 590 팬

Joschka Braun@BraunJoschka

2026.05.04 15:33

RL assumes that LLMs explore well during training. What if they choose not to? In our new ICML paper with @GoogleDeepMind, we train LLMs that strategically resist RL capability elicitation by under-exploring. We study this threat model, called exploration hacking.