Multiple document answering?

I was wondering if the Answers endpoint is single or multiple document?

I was thinking about this concept of document hybridization whereby you simply concatenate documents that are related and short enough in order to provide better answers.

I’m going to try some experiments with my cognitive architecture. I’m going to start recording very granular memories so they are smaller and easier to recombine for the sake of question answering. For instance, it will be easier to recall all memories about Bugs Bunny without much additional noise. As such, they will be more likely to fit within the 2048 token window.

1 Like

I just saw this article as well: https://www.marktechpost.com/2021/07/13/stanfords-ai-researchers-introduce-qa-gnn-that-jointly-reasons-with-language-models-and-knowledge-graphs/

Such a feature would be groundbreaking and disruptive. I am wondering if the OpenAI folks are working on this?

Here’s the ARXIV paper for the QA-GNN: https://arxiv.org/pdf/2104.06378.pdf

1 Like