I asked my agent to extract data from a PDF, but it returned the wrong value. I provided the correct figure and page number of the data, yet when I asked again, it still gave an incorrect answer or said the data wasn’t in the source. Any way to fix this?
At the moment IA don't have any memories like human. Some IA store few prompt like in GPT or copilot.
Then asking for a page number is not something reliable. IA at the moment are far from "star trek" they look for "small extract" of document and instruction like - go to "page 4" - it complicated : they don't have a "page 4" most of the time in the chunk they are analyzing(some time it work by accident because they have in the same chunk the page number and the data)
When about data extraction from document you have to provide a lot information on how to found the content or pattern (could you show the prompt you use ?) : data extraction from document/pdf/excel work very well in most case.
Do you use just the tchat or try with "prompt tool" from the tool tab ? wich is very better at document processing for data extraction ?
When asking for a "special" value in a document you have to think that if the IA give you something else it's because the prompt you made is not targeting enough with precision what you want. IA are not thinking like in movie, it's just math with "stats" and "probability".
Was this reply helpful?YesNo
Under review
Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.
4
Share
Report
All responses (
Answers (
