{
  "dataset": "Prompt and response pairs",
  "note": "Illustrative sample. It shows the structure, fields and metadata of this dataset with invented values, not real recordings or records. A real sample pack is sent on request.",
  "records": [
    {
      "id": "tx_0001",
      "language": "English",
      "turns": "6",
      "domain": "LLM alignment",
      "speaker_roles": "writer, reviewer",
      "excerpt": "Customer reports a delayed refund and asks for a status update.",
      "label": "resolved",
      "reviewer_agreement": "agree",
      "qa_status": "accepted"
    },
    {
      "id": "tx_0002",
      "language": "English",
      "turns": "12",
      "domain": "LLM alignment",
      "speaker_roles": "writer, reviewer",
      "excerpt": "Agent confirms identity and explains what happens next.",
      "label": "escalated",
      "reviewer_agreement": "agree",
      "qa_status": "accepted"
    },
    {
      "id": "tx_0003",
      "language": "English",
      "turns": "4",
      "domain": "LLM alignment",
      "speaker_roles": "writer, reviewer",
      "excerpt": "User asks a follow-up that depends on the previous turn.",
      "label": "resolved",
      "reviewer_agreement": "split, adjudicated",
      "qa_status": "accepted"
    },
    {
      "id": "tx_0004",
      "language": "English",
      "turns": "9",
      "domain": "LLM alignment",
      "speaker_roles": "writer, reviewer",
      "excerpt": "Customer switches topic and the agent re-confirms the goal.",
      "label": "clarified",
      "reviewer_agreement": "agree",
      "qa_status": "accepted"
    },
    {
      "id": "tx_0005",
      "language": "English",
      "turns": "7",
      "domain": "LLM alignment",
      "speaker_roles": "writer, reviewer",
      "excerpt": "Conversation closes with a summary and a confirmation.",
      "label": "resolved",
      "reviewer_agreement": "agree",
      "qa_status": "accepted"
    }
  ]
}