Datasets:

dzigen
/

PAI-2_PreprocessedBenchmarks

license: mit
task_categories:
  - question-answering
language:
  - en
tags:
  - HotpotQA
  - TriviaQA
  - NaturalQuestions
  - MuSiQue
  - 2WikiMultihopQA
  - PAI-2
size_categories:
  - 1K<n<10K

This is a repository with a preprocessed well known benchmarks that were used for PAI-2 (GraphRAG-method from PAI framework) evaluation on QA tasks. More details in our paper: arxiv.org/abs/2605.13481.