Using autogen assistants to talk over a dataset. But when im querying the dataset, assistantproxy agent is not able to filter the dataset according to the question and giving random records which are not part of my dataset [Bug]:

[PrashanthReddy1823]

Describe the bug

Hi team, Im using autogen multiple agents. Im passing a csv file while initiating the chat. But when im passing a message related to dataset i have passed its not giving correct answers and also display records those are not part of my dataset at all. How can we validate the output given by assistant agent?

Steps to reproduce

No response

Model Used

gpt-3.5-turbo

Expected Behavior

I need to work agent accurately according the user questions

Screenshots and logs

No response

Additional Information

using pyautogen==0.2.27.
I can provide the code snippets of required.

[ekzhu]

Your agents are hallucinating. Can you post your code?