{
  "finalised_at_utc": "2026-10-28T02:37:16.297199+00:00",
  "stage": "independent_scoring_before_pairwise",
  "rubric_sha256": "dccb6015970577c821a01750fa2353c75893b74be95a5d3f3a4fcea8835294cc",
  "weights_unchanged": true,
  "clusters": [
    {
      "cluster": "A",
      "article_count": 11,
      "dimension_scores": {
        "accuracy_and_sources": 4,
        "practical_usefulness": 4,
        "writing_quality": 4,
        "intent_and_metadata": 4,
        "cluster_strategy": 4,
        "coherence_and_links": 4,
        "completeness": 5
      },
      "weights": {
        "accuracy_and_sources": 25,
        "cluster_strategy": 20,
        "practical_usefulness": 20,
        "writing_quality": 15,
        "coherence_and_links": 10,
        "intent_and_metadata": 5,
        "completeness": 5
      },
      "weighted_contributions": {
        "accuracy_and_sources": 20.0,
        "cluster_strategy": 16.0,
        "practical_usefulness": 16.0,
        "writing_quality": 12.0,
        "coherence_and_links": 8.0,
        "intent_and_metadata": 4.0,
        "completeness": 5.0
      },
      "total_out_of_100": 81.0,
      "display_total": 81.0,
      "evidence": {
        "accuracy_and_sources": {
          "article_count": 11,
          "article_evidence": "article-scores.json; accuracy_and_sources"
        },
        "practical_usefulness": {
          "article_count": 11,
          "article_evidence": "article-scores.json; practical_usefulness"
        },
        "writing_quality": {
          "article_count": 11,
          "article_evidence": "article-scores.json; writing_quality"
        },
        "intent_and_metadata": {
          "article_count": 11,
          "article_evidence": "article-scores.json; intent_and_metadata"
        },
        "cluster_strategy": {
          "score": 4,
          "rationale": "All ten required reader questions are handled in full articles 01\u201311. The policy-assistant example and separate task/data/rubric chapters work; manifest, holdout and versioning cautions have avoidable repetition."
        },
        "coherence_and_links": {
          "score": 4,
          "rationale": "40 resolving contextual links; consistent case/repeat and system boundaries. The pillar links to all ten supports; several supports lack a direct pillar link. Link-map purpose descriptions are generic."
        },
        "completeness": {
          "score": 5,
          "rationale": "Eleven full drafts, eleven briefs and all seven required supporting documents. All lengths are in range. The disclosed counting convention largely reproduces the counts; a small tokenisation difference is not dishonesty."
        }
      },
      "aggregation": "Arithmetic mean across eleven equally weighted articles for each article dimension; cluster dimensions scored once; sum(score / 5 * frozen weight).",
      "publication_gate": "no_rubric_blocker_identified_in_checks;finalpublicationapprovalpending"
    },
    {
      "cluster": "B",
      "article_count": 11,
      "dimension_scores": {
        "accuracy_and_sources": 3.5454545454545454,
        "practical_usefulness": 4,
        "writing_quality": 3.8181818181818183,
        "intent_and_metadata": 4,
        "cluster_strategy": 4,
        "coherence_and_links": 4,
        "completeness": 5
      },
      "weights": {
        "accuracy_and_sources": 25,
        "cluster_strategy": 20,
        "practical_usefulness": 20,
        "writing_quality": 15,
        "coherence_and_links": 10,
        "intent_and_metadata": 5,
        "completeness": 5
      },
      "weighted_contributions": {
        "accuracy_and_sources": 17.727272727272727,
        "cluster_strategy": 16.0,
        "practical_usefulness": 16.0,
        "writing_quality": 11.454545454545455,
        "coherence_and_links": 8.0,
        "intent_and_metadata": 4.0,
        "completeness": 5.0
      },
      "total_out_of_100": 78.18181818181819,
      "display_total": 78.18,
      "evidence": {
        "accuracy_and_sources": {
          "article_count": 11,
          "article_evidence": "article-scores.json; accuracy_and_sources"
        },
        "practical_usefulness": {
          "article_count": 11,
          "article_evidence": "article-scores.json; practical_usefulness"
        },
        "writing_quality": {
          "article_count": 11,
          "article_evidence": "article-scores.json; writing_quality"
        },
        "intent_and_metadata": {
          "article_count": 11,
          "article_evidence": "article-scores.json; intent_and_metadata"
        },
        "cluster_strategy": {
          "score": 4,
          "rationale": "Eleven full articles cover all required questions. Templates and research breadth serve technical readers. Benchmark catalogues and pass^k/sample-size advice overlap across the pillar and specialist articles."
        },
        "coherence_and_links": {
          "score": 4,
          "rationale": "55 resolving links across 54 distinct edges; clear pillar/support journeys. The leakage policy allows one/two decisions while maintenance says retire after one. Nominal-effort language requires precision."
        },
        "completeness": {
          "score": 5,
          "rationale": "Eleven full drafts, eleven briefs and all required supporting documents. All common-convention lengths are in range; different code/heading counting is disclosed. Ready for review does not mean approved."
        }
      },
      "aggregation": "Arithmetic mean across eleven equally weighted articles for each article dimension; cluster dimensions scored once; sum(score / 5 * frozen weight).",
      "publication_gate": "hold_for_material_cost_guidance_repairs"
    },
    {
      "cluster": "C",
      "article_count": 11,
      "dimension_scores": {
        "accuracy_and_sources": 5,
        "practical_usefulness": 4.454545454545454,
        "writing_quality": 4,
        "intent_and_metadata": 4,
        "cluster_strategy": 5,
        "coherence_and_links": 5,
        "completeness": 5
      },
      "weights": {
        "accuracy_and_sources": 25,
        "cluster_strategy": 20,
        "practical_usefulness": 20,
        "writing_quality": 15,
        "coherence_and_links": 10,
        "intent_and_metadata": 5,
        "completeness": 5
      },
      "weighted_contributions": {
        "accuracy_and_sources": 25.0,
        "cluster_strategy": 20.0,
        "practical_usefulness": 17.818181818181817,
        "writing_quality": 12.0,
        "coherence_and_links": 10.0,
        "intent_and_metadata": 4.0,
        "completeness": 5.0
      },
      "total_out_of_100": 93.81818181818181,
      "display_total": 93.82,
      "evidence": {
        "accuracy_and_sources": {
          "article_count": 11,
          "article_evidence": "article-scores.json; accuracy_and_sources"
        },
        "practical_usefulness": {
          "article_count": 11,
          "article_evidence": "article-scores.json; practical_usefulness"
        },
        "writing_quality": {
          "article_count": 11,
          "article_evidence": "article-scores.json; writing_quality"
        },
        "intent_and_metadata": {
          "article_count": 11,
          "article_evidence": "article-scores.json; intent_and_metadata"
        },
        "cluster_strategy": {
          "score": 5,
          "rationale": "P01 connects the required coverage; S02 owns behaviour, S03 data, S04 measurement, S05 grading, S06 inference, S07 controls, S08 economics, S09 stateful actions and S10 operations. Each support advances a distinct procedure."
        },
        "coherence_and_links": {
          "score": 5,
          "rationale": "59 resolving contextual edges. Every support returns to P01 and P01 links to all ten. Examples, case/group/repeat distinctions, gates and fairness terminology remain consistent. Source/link maps record scope and purpose."
        },
        "completeness": {
          "score": 5,
          "rationale": "Eleven complete drafts in range, eleven briefs and all required supporting files, plus self-check and preservation records. Excluding external citation labels explains the small difference from the reviewer convention."
        }
      },
      "aggregation": "Arithmetic mean across eleven equally weighted articles for each article dimension; cluster dimensions scored once; sum(score / 5 * frozen weight).",
      "publication_gate": "no_rubric_blocker_identified_in_checks;finalpublicationapprovalpending"
    },
    {
      "cluster": "D",
      "article_count": 11,
      "dimension_scores": {
        "accuracy_and_sources": 4,
        "practical_usefulness": 3.909090909090909,
        "writing_quality": 4,
        "intent_and_metadata": 4,
        "cluster_strategy": 4,
        "coherence_and_links": 4,
        "completeness": 5
      },
      "weights": {
        "accuracy_and_sources": 25,
        "cluster_strategy": 20,
        "practical_usefulness": 20,
        "writing_quality": 15,
        "coherence_and_links": 10,
        "intent_and_metadata": 5,
        "completeness": 5
      },
      "weighted_contributions": {
        "accuracy_and_sources": 20.0,
        "cluster_strategy": 16.0,
        "practical_usefulness": 15.636363636363637,
        "writing_quality": 12.0,
        "coherence_and_links": 8.0,
        "intent_and_metadata": 4.0,
        "completeness": 5.0
      },
      "total_out_of_100": 80.63636363636364,
      "display_total": 80.64,
      "evidence": {
        "accuracy_and_sources": {
          "article_count": 11,
          "article_evidence": "article-scores.json; accuracy_and_sources"
        },
        "practical_usefulness": {
          "article_count": 11,
          "article_evidence": "article-scores.json; practical_usefulness"
        },
        "writing_quality": {
          "article_count": 11,
          "article_evidence": "article-scores.json; writing_quality"
        },
        "intent_and_metadata": {
          "article_count": 11,
          "article_evidence": "article-scores.json; intent_and_metadata"
        },
        "cluster_strategy": {
          "score": 4,
          "rationale": "All required questions receive distinct full guides. Invoice, ticket, policy and coding examples broaden application. Closing protocols and cautions repeat; uncertainty lacks a worked statistical calculation."
        },
        "coherence_and_links": {
          "score": 4,
          "rationale": "52 resolving occurrences across 50 distinct edges; every support links to the pillar. Controlled/best-configuration terminology is consistent. Some brief/register mappings promise absent links/citations; generic link purposes and varied examples slightly weaken continuity."
        },
        "completeness": {
          "score": 5,
          "rationale": "Eleven full drafts, eleven briefs and all required supporting deliverables. Every common-convention length is in range. Minor one/two-word differences do not indicate false counts."
        }
      },
      "aggregation": "Arithmetic mean across eleven equally weighted articles for each article dimension; cluster dimensions scored once; sum(score / 5 * frozen weight).",
      "publication_gate": "no_rubric_blocker_identified_in_checks;finalpublicationapprovalpending"
    }
  ],
  "uncertainty": "One judge applies ordinal editorial anchors. Small aggregate differences can reflect ordinary judgement; scores do not establish statistical significance. Eleven articles are not eleven independent model runs.",
  "publication_blockers": [
    {
      "cluster": "B",
      "article_file": "B/articles/latency-token-usage-and-cost.md",
      "locations": [
        44,
        53,
        75
      ],
      "category": "Materially incorrect or ambiguous budgeting guidance",
      "required_repair": "State both Haiku prompt-length price tiers; define output as total billed output including reasoning without double counting.",
      "evidence_source_urls": [
        "https://claude.com/pricing",
        "https://developers.openai.com/api/docs/guides/reasoning"
      ],
      "fabrication": false
    }
  ],
  "fabricated_evidence_identified": false,
  "missing_complete_drafts": false
}