Hello! I’m using the OPENAI Assistant with file_search, basically what I’m trying to do is answer questions based on the files uploaded, the answer has to be like: “the number 1234 says: Exact transcription of what the number 1234 says in the file”.
I have a lot of good answers but in some cases(some numbers) it answers wrong, I tried to fix the instructions, the chunks and overlap of the vector_store but I don’t seem to find better responses.
Currently, I have this configuration for my assistant:
Model: gpt-4o
Temperature: 0.3
Top_P: 0.7
File_Search max num results: 10
Can someone explain me why this happens or how can I find a solution to this problem?