BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//OneJoy//events//EN
BEGIN:VEVENT
UID:onejoy-event-86438@onejoy
DTSTAMP:20261027T203551Z
DTSTART:20261028T013000Z
DTEND:20261028T033000Z
SUMMARY:Cosmin goes over Score Centering Stabilizes Off-policy Reinforcement Learning
LOCATION:
DESCRIPTION:Cosmin goes over Score Centering Stabilizes Off-policy Reinforcement Learning\n\nThis Tuesday Cosmin will cover "Score Centering Stabilizes Off-policy Reinforcement Learning". This method is used in the recently thinking machines rl platform update.\n\nResources\nPaper https://arxiv.org/abs/2609.20807\nAuthor thread about the relation to the thinker StableReinforce objective https://x.com/mrtnm/status/2114145574987702525\n\nOther papers and blogs\nhttp://joschu.net/blog/kl-approx.html\nhttps://arxiv.org/a\nhttps://onejoy.io/event/86438
URL:https://onejoy.io/event/86438
END:VEVENT
END:VCALENDAR