RLPT

Stanford CS25: Transformers United V6 I From Next-Token Prediction to Next-Generation Intelligence

Stanford Online

A related paper and technique for Reinforcement Learning on Pretraining Data, compared against RLP.