AI Engineer Problems › Vector Databases & RAG

Document Chunking Trade-offs in Graph Extraction

Medium · AI Engineering · Vector Databases & RAG · Asked at Microsoft, OpenAI

You are chunking your source PDFs before feeding them to the LLM for entity extraction.

Why does choosing a small chunk size (e.g., 200 tokens) lead to a fragmented graph compared to a larger chunk size (e.g., 1200 tokens)?

Solve it in the browser on GenAIDojo