data/metadata/dataset_summary.json

{
  "project": "LegalScope",
  "snapshot": "paper-submission benchmark",
  "release_status": "public research repository; full workbook, paper draft, model outputs, and human review sheets withheld",
  "public_release_boundary": {
    "full_workbook_public": false,
    "full_prompt_matrix_public": false,
    "full_reference_answers_public": false,
    "full_model_outputs_public": false,
    "human_review_sheets_public": false,
    "non_deidentified_judgments_public": false,
    "paper_draft_public": false
  },
  "dataset_modules": [
    {
      "name": "Bar exam data",
      "type": "Dataset",
      "source": "Public legal exams",
      "rows": 868,
      "model_groups": 20,
      "evaluations": 17360
    },
    {
      "name": "CN judgment data",
      "type": "Dataset",
      "source": "Real cases",
      "rows": 76,
      "model_groups": 20,
      "evaluations": 1520
    },
    {
      "name": "Human bar exam",
      "type": "Human scored",
      "source": "Public legal exams",
      "rows": 80,
      "model_groups": 20,
      "evaluations": 1600
    },
    {
      "name": "Human CN judgments",
      "type": "Human scored",
      "source": "Real cases",
      "rows": 10,
      "model_groups": 20,
      "evaluations": 200
    }
  ],
  "counts": {
    "public_exam_items": 868,
    "real_case_issue_stance_prompts": 76,
    "dataset_items_total": 944,
    "model_groups": 20,
    "public_exam_model_responses": 17360,
    "real_case_model_responses": 1520,
    "dataset_model_responses_total": 18880,
    "human_public_exam_items": 80,
    "human_real_case_prompts": 10,
    "human_validation_items_total": 90,
    "human_public_exam_responses": 1600,
    "human_real_case_responses": 200,
    "human_validation_responses_total": 1800,
    "deidentified_chinese_civil_judgments": 15,
    "real_case_legal_issues": 38
  },
  "transfer_summary": {
    "model_judge_public_exam_to_real_case_pearson_r": 0.835,
    "model_judge_public_exam_to_real_case_spearman_rho": 0.661,
    "human_public_exam_to_real_case_pearson_r": 0.846,
    "human_public_exam_to_real_case_spearman_rho": 0.784
  }
}