Answer-first summary for fast verification
Answer: Add the section header as a prefix to chunks, Increase the document chunk size
The question focuses on improving RAG system performance when the retriever fails to capture all relevant context, specifically for structured HR policy documents. Option A (adding section headers as prefixes to chunks) provides semantic signals that help the retriever better match queries to the correct policy sections, which is crucial for structured documents. Option D (increasing document chunk size) ensures that complete policy details remain within single chunks rather than being split across multiple chunks, preventing fragmented retrieval. The community discussion with 100% consensus and 3 upvotes strongly supports AD, noting that B (sentence splitting) and E (response generation tuning) don't address retrieval gaps, while C (larger embedding models) was already tried without success.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
A Generative AI Engineer is building a RAG system for internal Q&A on structured HR policy documents. The system is returning incomplete and unstructured answers, and the issue appears to be that the retriever is failing to retrieve all relevant context. The engineer has already tried different embedding models and response-generation LLMs without success.
Which TWO of the following approaches could improve the quality of the responses? (Choose two.)
A
Add the section header as a prefix to chunks
B
Split the document by sentence
C
Use a larger embedding model
D
Increase the document chunk size
E
Fine tune the response generation model
No comments yet.