I was using QA summary graph as my query engine. One question that I want to ask my document is a value extraction question, which uses the vecstore index. The response i get is incorrect but when I went back and checked the log. I realized it got it right in the initial attempt but messed up with the refine template. I know this issue is kinda vague but are there ways for us to stop refining after getting it right in the first place.
There are some settings you can tune to hopefully avoid refine, but no actual toggle
You can also customize the refine template as well. If you are usin gpt-3.5, this can be especailly helpful (tbh, that model is... not great lately lol)