Hi, I’m trying to get relevant information snippets from several pdfs, although the assistant always just returns information from one pdf. If i then ask it to consider and retrieve information from all the pdf’s, it then returns parts from another one but it’s not efficient. How can i get it to return small quotes or sections from multiple documents and is there a limit on the number of documents it can perform retrieval on?
Hi! Are you using the file_search tool or the retrieval tool here?
Hi @nikunj – I also seem to be running into this issue. I am using file_search with GPT-3.5 Turbo.
I first upload my files (i.e. 3 PDFs), then create a VectorStore. I then pass the VectorStore’s ID to a call to create an Assistant with file_search.
Over around a dozen trials, they all seem to only be citing (pulling data?) from one of the PDFs. Any reason why this would be?
I believe expected behavior would be to gather multiple citations across the 3 documents, no?
We definitely get chunks for all documents (not just one). If you could email me more details at nikunj@openai.com, I can take a closer look.