Reasoning capabilities of GPT-4 based on what it already knows is impressive.
But when I ask GPT to provide insights using additional text, it provides inferior results, far from impressive. It can say superficial things but does not go deep and cannot get the underlying concepts, etc. I used langchain to manage the token limit but it didn’t help.
Does anyone experience similar issues and how did you solve that?