Not really, LLM Decoders(without RLHF) are usually trained by teacher forcing, that means they just get learning singals from just predicting one token.