Hey everyone, looking for some help please! I am trying to parse a generic set of files to return a specific response. Currently, I am using the assistants API with some files and knowledge retrieval and it works appropriately, but for other file formats that aren’t so straightforward, it gets confused and doesn’t return the correct response. I am thinking about fine-tuning, but since I don’t know the exact context it is receiving (since it’s using knowledge retrieval), I don’t know how to create my dataset. Does anyone have any advice for me, I am new to AI! Thank you!
Fine-tune models cannot be used with Assistants.
The enfeeblement of gpt-3.5 models that can use tools to the point where they barely perform retrieval would make this undisclosed tool easily broken by any model fine tuning.
The context it is receiving as a result of running the myfiles_browser tool is top results with line numbered text for then calling to traverse and explore the documents and mark retention snippets for citation. In an ideal world of AI understanding. Normally just gives poor answers from a happenstance document injection or only the search alone.