Most turns in the QReCC dataset do not have "truth passages"

[lujiarui-iie]

Hello, for the rounds where the "truth passage" field is not annotated, is it due to missing annotations, or is there another reason? For example, in the first round of dialogue 1 in the test set:

{
    "Answer_URL": "https://explorehealthcareers.org/career/medicine/physician-assistant/",
    "Context": [],
    "Conversation_no": 1,
    "Conversation_source": "trec",
    "Question": "What is a physician's assistant?",
    "Transformer_rewrite": "What is a physician's assistant",
    "Truth_answer": "physician assistants are medical providers who are licensed to diagnose and treat illness and disease and to prescribe medication for patients",
    "Truth_passages": [],
    "Truth_rewrite": "What is a physician's assistant?",
    "Turn_no": 1
}

its truth answer corresponds to the sentence in the paragraph at http://web.archive.org/web/20200106012242id_/https://explorehealthcareers.org/career/medicine/physician-assistant/_p0. However, the test set does not have a truth passage labeled for it.

And, statis_info function shows that there are only 29,596 rounds with corresponding gold standard documents, in all 63,501 turns.