No, that's how base model pretraining works. Claude's behavior is more based on its constitution and RLVR feedback, because that's the most recent thing that happened to it.