Guidance Needed: Ensuring GPT-4 Utilizes All JSON Records and Tracks Unused Entries

API
1

Hi everyone,

I’m working with GPT-4 and passing it a JSON list containing document names along with other details, as well as a question. The goal is to generate a final response based on the JSON list.

  1. How can I ensure the LLM uses all the records in the JSON list while generating the response?
  2. If this isn’t feasible, how can I instruct the LLM to accurately return the document names it did not use from the list?

I’d greatly appreciate any guidance or suggestions. Thanks in advance!