Skip to content

Good summaries were ignored #9

@nasirus

Description

@nasirus

Hi

I'm using RetrievalQA to do question answering on vector store, below are the config I used

llm=ChatOpenAI chatgpt 3.5 model
embeddings = OpenAIEmbeddings
retriever=Pinecone
This is one of the messages

Given the following extracted parts of a long document and a question, create a final answer.\nIf you don't know the answer, just say that you don't know. Don't try to make up an answer.\n______________________\n
I set the results from Pinecone to k=4, it did summarize for documents and put into the final prompt messages. The strange thing is that the first summary is good and the rest of the summaries all said 'here is no information provided in the given portion of the document something like this. But the expected answer should use the first summary.

And then the final answer I got is like there is no information about xxxx. This happened since last Friday after 4~5pm, and I can get the correct answer before.

Do you know if something changed on openai chat completion API?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions