Hacker News new | past | comments | ask | show | jobs | submit login

So is GRPO that much better because it ascribes feedback to a whole tight band of ‘quality’ ranges of on-policy answers while the band tends towards improvement in the aggregate, or is it just faster algorithm = more updates for a given training duration?



Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: