haystack
haystack copied to clipboard
Select 4 or 5 datasets
- Select datasets by looking at Notion page
- The datasets need to have the following properties:
- [X] at least one should be financial or legal and raw data needs to be in structured pdfs
- [x] least one should be about support/help centre
- [x] there should be one that has been used in other benchmarks (maybe based on wikipedia)
- [X] they should all have a set of labels so that we can get performance metrics from them