A raw next-token model completes “The lower gate” with more memoir text; after an instruction stage it instead follows “Summarize this lock log in two bullets.” What did that stage teach?
Select an answer to reveal the explanation.
Short Explanation
A raw next-token model just keeps writing the memoir. After an instruction stage it follows “Summarize this lock log in two bullets.” That stage is instruction-style fine-tuning, not RAG, not a new architecture, and not a dock-cluster algorithm.
Full Explanation
Instruction tuning teaches a pretrained model to follow requests instead of only continuing text. It is not retrieval, not a new block design, and not a graph algorithm.