{
  "document_id": "lb_qasper_d3dac3676da5685f20bae39814f760368752f5bd8db93500",
  "family": "public_qasper",
  "split": "test",
  "document_hash": "8f673b48b0058e265bda13f9c98850a581b541a1e33ab67f6dc42bdddf9c135d",
  "questions": [
    {
      "question_id": "q00",
      "qtype": "qasper",
      "question": "What were their results on the new dataset?",
      "answer": "average recipe-level coherence scores of 1.78-1.82, human evaluators preferred personalized model outputs to baseline 63% of the time",
      "aliases": [],
      "answer_type": "qa_f1",
      "tolerance": 0.0,
      "supporting_fact_ids": [],
      "answer_position_frac": -1.0
    }
  ],
  "facts": []
}