Using Curate you can use the Large Language Model capability of Composer to ask questions of documents in a Curate document set. A license for Composer is required for this functionality to be enabled.
First, create and save a new document set. Instructions can be found in the following link:
Prompt the GPT
Once you create a document set, you can Prompt the GPT from the Actions menu.
In the modal that appears, type the question you would like answered.
Click the Show Advanced Settings link.
In the Prompt the GPT modal, under the DEEP LEARNING section, there are multiple options: BioMegaTron SQuAD v2, Large Language Model, Meta-Llama, and GPT-5.
BioMegaTron SQuAD v2 is an extractive model: it reads through documents to find the exact line of text that includes the answer to your question
The Large Language Model is abstractive: it reads through documents and provides either a summary of the text to answer your question or infers the answer from the text
Meta-Llama is another advanced language model that uses a combination of extractive and abstractive techniques to provide accurate answers. It has been trained on a vast amount of text data and can understand nuances in language.
GPT-5 is a state-of-the-art language model that uses transformer architecture to generate human-like responses. It can provide detailed answers by summarizing or inferring information from the text.
You will also find the custom models you've retrained and published here.
đĄ Pro Tip - Before attempting to apply a newly developed query on a larger set of documents, it is recommended to perform prompt engineering on a smaller subset of around 20 documents in Curate.