Eyon Jang(@eyonjang):Can LLMs learn to resist RL training? We empirically study exploration hacking: models controlling behavior during RL to prevent unwanted capabilities from being reinforced. Joint work with @GoogleDeepMind and @MATSprogram. More in the thread below 👇

Eyon Jang

@eyonjang

AGI safety researcher (MATS 8)

加入 December 2012

1.2K 正在關注 131 粉絲

Eyon Jang@eyonjang

2026.05.04 18:55

Can LLMs learn to resist RL training? We empirically study exploration hacking: models controlling behavior during RL to prevent unwanted capabilities from being reinforced. Joint work with @GoogleDeepMind and @MATSprogram. More in the thread below 👇