intermediateSpring AI — Core, ChatClient & RAG
What are chunking strategies in RAG? How does chunk size affect quality?
Fixed-size (split by token count), sentence-based (better coherence), recursive (paragraphs → sentences → words), and semantic (split on topic shifts, more expensive). Too small loses context; too large adds irrelevant noise — overlap between chunks preserves context across boundaries.
This is a Pro question
Sign in, then upgrade to Pro or Power to unlock this question and the full Interview Prep bank.
What are chunking strategies in RAG? How does chunk size affect quality?