Scaling pretraining affects RL sample efficiencyrunrl.com1 point·ag8··0 commentsOpen articleSaveView on HN