What exactly is the pipeline to use this data?

#2
by Ouz-G - opened

I am having hard time understanding this dataset. There is the entity column and info are related sources to that. What exactly is evaluated and how? Typically there is a question and expected answer. Is entity belonging to a question or answer or what is it?

The pipeline to use this dataset is to apply factscore to this. Our paper has more details: https://arxiv.org/abs/2407.17468

  1. Have LLMs to generate information about entities
  2. Use FactScore to fact check the information generated by the LLMs against the knowledge source we included in this dataset.

Sign up or log in to comment