Peter Henderson(@PeterHndrsn):Now: Reward hacking by finding bugs in lean. Later: Reward hacking by introducing bugs in lean. Another lesson here is probably not to trust every lean proof that a model spits out as 100% solid.

Peter Henderson

@PeterHndrsn

Assistant Professor @ Princeton (RL+strategic decision-making+Law). Prev: Stanford (JD+PhD); Mila; FAIR; Amazon; Cal Supreme Court. Views are all only my own.

参加 February 2012

962 フォロー中 6.2K ファン

Peter Henderson@PeterHndrsn

2026.09.25 04:51

Now: Reward hacking by finding bugs in lean. Later: Reward hacking by introducing bugs in lean. Another lesson here is probably not to trust every lean proof that a model spits out as 100% solid.