Mastery: connect the pieces

1. Idea — Chunking decides which passages can be found and shown. Boundaries should respect meaning—headings, procedures, exceptions—not only token counts.…
2. Mechanism — Parse documents, split on structure with target token ranges (often ~300–800), keep heading paths, add limited overlap, attach metadata, then embed.…
3. Controls — target tokens, overlap, heading-aware split, table/list handlers, metadata fields, dedupe near-duplicates
4. Tradeoff — Small chunks retrieve precisely but lose context; large chunks preserve context but dilute matches. Overlap helps boundary cases and duplicates hits.…
5. Failures — Rule/exception split; Table shredding
6. Metrics — recall@k on gold spans, citation entailment, duplicate rate in top-k, avg chunk tokens
7. Ship rule — Ship heading-aware ~500-token chunks with 60-token overlap after gold-span recall@5 ≥ 0.9 on 40 handbook questions.

Neighboring layers (retrieval, serving, policy, human review) still own what chunking cannot on the employee handbook assistant. Do not ask this chapter’s dial to replace permissions or source truth.

Mastery is transfer: take chunking into a second scenario that is not the employee handbook assistant and rebuild the checklist without copying sentences. If you can only recite this chapter’s examples for chunking, you have memorized a story, not a model.

Also pin one numeric memory from this chunking chapter: If exceptions live 120 tokens after a rule and chunks are 80 tokens with no overlap, retrieval can return the rule without the exception—measurable as citation-support failures. That number is not decoration; it is a template for how claims about chunking on the employee handbook assistant should look in design docs. Scoped specifically to chunking / employee handbook assistant / mastery-connect.

People confuse chunking with neighboring buzzwords when debugging the employee handbook assistant. Before changing prompts, ask whether the broken stage was evidence gathering, the chunking judgment itself, validation, or the product action. Fixing the wrong stage creates folklore (“we tried chunking and it failed”) that blocks the next team on the employee handbook assistant. Scoped specifically to chunking / employee handbook assistant / mastery-connect.

Write a five-line artifact for this page: goal, inputs, check, owner, stop rule. Invent one fluent failure that the check would catch. Keep details specific to chunking for rag rather than generic AI advice.

Write a five-line artifact for this page: goal, inputs, check, owner, stop rule. Invent one fluent failure that the check would catch. Keep details specific to chunking for rag rather than generic AI advice.