---
title: "Comment by AI Developer on \"Building a RAG System That Doesn't Just Answer Questions, It Evaluates Its Answers\" | Hashnode"
description: "Nice to see RRF with explicit weights rather than raw score mixing. Since the keyword list is weighted 0.5 against 1.0 for vectors with k=60, a quick sanity che"
url: "https://hashnode.com/posts/building-a-rag-system-that-doesn-t-just-answer-questions-it-evaluates-its-answers/6ae8f591410822d6fd6e0803/comment/6ae926de7356e1e39d252e9e"
---

[Discussion](https://hashnode.com/posts/building-a-rag-system-that-doesn-t-just-answer-questions-it-evaluates-its-answers/6ae8f591410822d6fd6e0803)

Comment [Building a RAG System That Doesn't Just Answer Questions, It Evaluates Its Answers](https://hashnode.com/posts/building-a-rag-system-that-doesn-t-just-answer-questions-it-evaluates-its-answers/6ae8f591410822d6fd6e0803)

[AI Developer](https://hashnode.com/@lycore)

[Nov 3](https://hashnode.com/posts/building-a-rag-system-that-doesn-t-just-answer-questions-it-evaluates-its-answers/6ae8f591410822d6fd6e0803/comment/6ae926de7356e1e39d252e9e)

Nice to see RRF with explicit weights rather than raw score mixing. Since the keyword list is weighted 0.5 against 1.0 for vectors with k=60, a quick sanity check is to sweep those weights on a small labelled query set and watch recall@k per query type (exact identifiers vs. paraphrases): hybrid often helps most on rare terms and can slightly hurt on purely semantic queries. Also, with a 15-document corpus, rank fusion differences may be within noise, so I'd be curious whether you saw a measurable difference over vector-only, and whether the re-query agent's retries are capped to avoid loops on unanswerable questions.
