Chunking splits documents into passages small enough to retrieve and fit in context. Chunk size and overlap trade recall against precision: too large and you dilute the context, too small and you lose meaning. Chunking quality often shows up as a retrieval problem long before it shows up as a model problem.
Why it matters
Retrieval can only return chunks that exist. Chunking choices decide whether the right information is reachable and whether it still makes sense when it arrives.
How it works
Documents are split into passages, often with overlap so meaning is not cut at boundaries. Smaller chunks improve precision; larger chunks preserve context. Chunking problems usually surface as retrieval problems, not model problems.