A paratransit-booking policy binder will not fit in one prompt, so the shop breaks it into pieces of about 256–512 tokens. Why is that chunking done?
Select an answer to reveal the explanation.
Short Explanation
A fat binder will not fit through a narrow mail slot — so you mail chapters. Chunking splits large documents into pieces (often about 256–512 tokens) for focused retrieval and context-window fit. It is preparation for RAG, not a token-counting lab exercise.
Full Explanation
Chunking divides large documents into smaller units so retrieval can focus on relevant pieces and stay within the model’s context window. Syllabus examples often cite roughly 256–512 token chunks. The goal is searchable, window-compatible segments — not fine-tuning jobs or treating the model as the SUT.