Find answers from the community

Updated last year

Alright, so I have tried to get my

Alright, so I have tried to get my program to answer queries regarding legal documentation but the result is not what I want. I have used fine-tuning, coupled with ToolRetrieverRouterQueryEngine and SubQueryEngine to try and answer queries about my documents. The response is seldom correct or just absolutely incorrect. Does anyone else have an idea on how to improve retrieval?

2 comments

TTeemu

Did these changes improve retrieval over the baseline of just making a simple vector index and querying it? https://gpt-index.readthedocs.io/en/stable/getting_started/starter_example.html

TTren Hard Eat Clen!

The baseline did so well but only on a very small set of similar data. I want to do hetrogenous data. The Multidoc agent doesnt work at all.

Add a reply