Comment by pegasus
10 days ago
Look into RLVR (Reinforcement Learning with Verifiable Rewards). It happens during model post-training.
10 days ago
Look into RLVR (Reinforcement Learning with Verifiable Rewards). It happens during model post-training.
That’s not happening in run time, when you send the prompts for the next token to be generated. The comment seems to imply a run time phenomenon.