The model is working well now, but it probably could be refined by tuning the prompt we've given to the LLM agent.
Approach
As part of end-to-end testing, we should try various ways of answering questions, including typos (go wild in testing), and determine if we want to give any more specific instructions to the LLM agent.
Overview
The model is working well now, but it probably could be refined by tuning the prompt we've given to the LLM agent.
Approach
As part of end-to-end testing, we should try various ways of answering questions, including typos (go wild in testing), and determine if we want to give any more specific instructions to the LLM agent.