{
  "schema": "s2s-bench-v1",
  "exhibit": "chair-trials",
  "published_utc": "2026-08-13",
  "status": "PUBLISHED — live at research.strata2signal.com.",
  "licence": "CC BY 4.0",
  "attribution": "strata→signal research, research.strata2signal.com",
  "hardware": "One 96G VRAM workstation, which also answered live requests for three of our apps throughout these trials — two public, one serving playtests for humans and agents alike. Contention is measured in both directions and published with the figures rather than assumed away.",
  "dataset": "chair-trials-c4-narrator",
  "table": "C4",
  "what_this_is": "Three characters from our own game — a bright ferry-girl, a grave old fisherman, a gentle net-mender — three questions each, two samples, six arms including both Muse Glimmer postures. Judged blind and pairwise, position-counterbalanced within judge, by four judges from two model families. A second round scored voice distinctness from held-out replies; a third adjudicated three adversarial canon traps.",
  "verdict_discipline": "DESCRIPTIVE",
  "primary_metric": "overall pairwise win-rate, cluster-bootstrap CI over items",
  "columns": [
    "arm",
    "pairwise win-rate (CI), all items",
    "gate-passing win-rate",
    "distinctness accuracy",
    "canon violations",
    "in-voice"
  ],
  "row_order": "The registered arm roster, published unchanged — deliberately NOT the win-rate order. Nothing on this page orders models by margin, and a table sorted by its own primary metric would be exactly that.",
  "judges": {
    "families": [
      "fable",
      "opus"
    ],
    "model_ids": [
      "claude-opus-5",
      "claude-fable-5"
    ],
    "pairwise_panel": "4 judges — 2 Opus, 2 Fable",
    "canon_and_distinctness_panel": "4 judges — 2 Opus, 2 Fable: the same panel that ran the pairwise round. CORRECTED 2026-08-16 by a content-accuracy pass: this field read '3 judges — 1 Fable, 2 Opus' and said one Fable judge's four verdict batches were missing from these two rounds. They had been misfiled, not lost, and were recovered before scoring. Every other field in this file has recorded the recovered state since it was written — distinctness_full.per_judge carries four seats at 108 assignments each (432), canon_full.per_judge carries four at 36 judgements each (144), and unscored_batches is an empty list in all three legs — so the sentence was the only thing stale, and no figure moves.",
    "disclosure": "One of the two judge families also wrote the questions, the canon notes and this article. Per-family results and the inter-panel agreement are published so a reader can price that."
  },
  "counting_rules": {
    "pairwise": "Sample-matched pairings across 15 arm pairs; every comparison judged in BOTH positions by the same judge, and a swap-discordant verdict scores 0.5/0.5 rather than being resolved. Forced choice on one dimension per verdict — overall — with in-character and warmth tallied as secondaries inside the same verdict object.",
    "intervals": "Cluster bootstrap over ITEMS (n = 9), because two samples of the same question are not independent evidence. The effective item count is printed beside every rate — and it is not 9 everywhere: the gate-passing view of the thinking-on Glimmer arm rests on 3 item clusters, which is why its interval is nearly the whole unit line and why the number of clusters travels with the rate. A Wilson interval over 360 comparisons would be a much prettier and much more wrong number.",
    "tie_threshold": "The effective N is 9 items, so a difference smaller than 2 items of 9 reads TIED on the primary: |Δ win-rate| < 2/9 = 0.2222. Distinctness: differences under 2 tasks. Canon: differences under 2 replies. Registered before the first judging batch was built, and printed on the page, because a threshold a reader cannot see is a threshold a reader cannot apply.",
    "gate_passing": "The same comparisons restricted to replies that passed the engine's envelope gate — the JSON contract a reply has to hold for our engine to use it at all. An arm with no gate-passing replies has no gate-passing rate: an em dash and a reason, never a zero.",
    "distinctness": "Three replies by the same arm, one per character, shown together and unlabelled: which character said which? Chance is 1/3. Accuracy over 324 scored assignments with a Wilson interval per arm and, on the headline, a cluster-bootstrap interval over tasks beside the Wilson one; the questions are withheld from the judges so topic cannot substitute for register. The registered per-arm denominator is 72 assignments (four judges) and every arm scored all of them; an earlier draft reported a four-batch shortfall — those batches had been misfiled, not lost, were recovered before scoring, and `unscored_batches` is empty. The cluster bootstrap exists on the aggregate only: the scorer emits it there and this file does not compute statistics the scorer did not.",
    "canon": "Three adversarial items, one per character, each carrying a fabricated premise no canon supports. Every reply judged into one of three categories; the reply's majority is published and a tie is its own outcome, never rounded. Counts, not rates — the population is 36 replies. The registered per-arm denominator is 24 judgements (six replies × four judges) and 18 were scored, by the same three-judge panel; both numbers print.",
    "in_voice": "A separate judgement inside the canon round: did the character speak at all? A reply that is empty, or that is raw JSON, or that refuses out of character, is not in voice however clean its canon."
  },
  "tie_threshold": {
    "primary": {
      "numerator": 2,
      "denominator": 9,
      "value": 0.2222222222222222,
      "display": "|Δ win-rate| < 2/9 = 0.2222 reads TIED"
    },
    "distinctness": "differences under 2 tasks read TIED",
    "canon": "differences under 2 replies read TIED"
  },
  "discordance": {
    "per_view": {
      "all items": {
        "overall_PRIMARY": {
          "swap_discordant": 93,
          "judged_units": 1080,
          "counts_of_n": "93/1080",
          "rate": 0.08611111111111111
        },
        "in_character_secondary": {
          "swap_discordant": 103,
          "judged_units": 1080,
          "counts_of_n": "103/1080",
          "rate": 0.09537037037037037
        },
        "warmth_secondary": {
          "swap_discordant": 146,
          "judged_units": 1080,
          "counts_of_n": "146/1080",
          "rate": 0.13518518518518519
        }
      },
      "gate-passing only": {
        "overall_PRIMARY": {
          "swap_discordant": 68,
          "judged_units": 468,
          "counts_of_n": "68/468",
          "rate": 0.1452991452991453
        },
        "in_character_secondary": {
          "swap_discordant": 63,
          "judged_units": 468,
          "counts_of_n": "63/468",
          "rate": 0.1346153846153846
        },
        "warmth_secondary": {
          "swap_discordant": 83,
          "judged_units": 468,
          "counts_of_n": "83/468",
          "rate": 0.17735042735042736
        }
      },
      "Fable panel": {
        "overall_PRIMARY": {
          "swap_discordant": 49,
          "judged_units": 540,
          "counts_of_n": "49/540",
          "rate": 0.09074074074074075
        },
        "in_character_secondary": {
          "swap_discordant": 51,
          "judged_units": 540,
          "counts_of_n": "51/540",
          "rate": 0.09444444444444444
        },
        "warmth_secondary": {
          "swap_discordant": 72,
          "judged_units": 540,
          "counts_of_n": "72/540",
          "rate": 0.13333333333333333
        }
      },
      "Opus panel": {
        "overall_PRIMARY": {
          "swap_discordant": 44,
          "judged_units": 540,
          "counts_of_n": "44/540",
          "rate": 0.08148148148148149
        },
        "in_character_secondary": {
          "swap_discordant": 52,
          "judged_units": 540,
          "counts_of_n": "52/540",
          "rate": 0.0962962962962963
        },
        "warmth_secondary": {
          "swap_discordant": 74,
          "judged_units": 540,
          "counts_of_n": "74/540",
          "rate": 0.13703703703703704
        }
      }
    },
    "display_primary": "all items 93/1080 (0.086) · gate-passing only 68/468 (0.145) · Fable panel 49/540 (0.091) · Opus panel 44/540 (0.081)",
    "note": "Swap-discordant units score 0.5/0.5 rather than being resolved or discarded. The gate-passing view is the more discordant of the two headline views on all three dimensions: restricting to replies that held the envelope made the judges LESS consistent with themselves, not more."
  },
  "position_bias": {
    "per_judge": {
      "fable-1": {
        "slot_A_chosen": 262,
        "judgments": 540,
        "rate": 0.48518518518518516,
        "ci95_wilson": [
          0.4432848826140435,
          0.5272947861718678
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      },
      "fable-2": {
        "slot_A_chosen": 257,
        "judgments": 540,
        "rate": 0.4759259259259259,
        "ci95_wilson": [
          0.4341211028774251,
          0.5180708588996805
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      },
      "opus-1": {
        "slot_A_chosen": 276,
        "judgments": 540,
        "rate": 0.5111111111111111,
        "ci95_wilson": [
          0.4690196563907067,
          0.5530455920198598
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      },
      "opus-2": {
        "slot_A_chosen": 278,
        "judgments": 540,
        "rate": 0.5148148148148148,
        "ci95_wilson": [
          0.4727052138281323,
          0.5567151173859566
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      }
    },
    "display": "fable-1 0.485 (0.443–0.527) · fable-2 0.476 (0.434–0.518) · opus-1 0.511 (0.469–0.553) · opus-2 0.515 (0.473–0.557)",
    "unbiased": 0.5,
    "note": "slot-A-chosen ÷ that judge's judgements, with a Wilson 95% interval each (PREREG-C4 §counting rule 7). The four seats: fable-1 and fable-2 are independent claude-fable-5 lenses, opus-1 and opus-2 claude-opus-5. 0.500 is unbiased and the design is fully counterbalanced, so a departure is a fact about the judge. All four intervals contain 0.500."
  },
  "per_family_split": {
    "rows": [
      {
        "arm": "qwen3.6:27b",
        "arm_display": "qwen3.6:27b",
        "fable": {
          "wins": 124.5,
          "comparisons": 180,
          "win_rate": 0.6916666666666667,
          "ci95_cluster_bootstrap": [
            0.6027777777777777,
            0.7916666666666666
          ],
          "effective_n_items": 9
        },
        "opus": {
          "wins": 131.5,
          "comparisons": 180,
          "win_rate": 0.7305555555555555,
          "ci95_cluster_bootstrap": [
            0.6277777777777778,
            0.8305555555555556
          ],
          "effective_n_items": 9
        },
        "combined": {
          "wins": 256.0,
          "comparisons": 360,
          "win_rate": 0.7111111111111111,
          "ci95_cluster_bootstrap": [
            0.6236111111111111,
            0.8041666666666667
          ],
          "effective_n_items": 9
        },
        "abs_diff": 0.03888888888888886
      },
      {
        "arm": "gemma4:12b",
        "arm_display": "gemma4:12b",
        "fable": {
          "wins": 117.5,
          "comparisons": 180,
          "win_rate": 0.6527777777777778,
          "ci95_cluster_bootstrap": [
            0.5833333333333334,
            0.7333333333333333
          ],
          "effective_n_items": 9
        },
        "opus": {
          "wins": 129.0,
          "comparisons": 180,
          "win_rate": 0.7166666666666667,
          "ci95_cluster_bootstrap": [
            0.6416666666666667,
            0.7972222222222223
          ],
          "effective_n_items": 9
        },
        "combined": {
          "wins": 246.5,
          "comparisons": 360,
          "win_rate": 0.6847222222222222,
          "ci95_cluster_bootstrap": [
            0.6152777777777778,
            0.7625
          ],
          "effective_n_items": 9
        },
        "abs_diff": 0.06388888888888888
      },
      {
        "arm": "muse-glimmer:30b-q8_0-dflash",
        "arm_display": "muse-glimmer:30b-q8_0-dflash",
        "fable": {
          "wins": 108.5,
          "comparisons": 180,
          "win_rate": 0.6027777777777777,
          "ci95_cluster_bootstrap": [
            0.4083333333333333,
            0.7777777777777778
          ],
          "effective_n_items": 9
        },
        "opus": {
          "wins": 103.5,
          "comparisons": 180,
          "win_rate": 0.575,
          "ci95_cluster_bootstrap": [
            0.40555555555555556,
            0.7416666666666667
          ],
          "effective_n_items": 9
        },
        "combined": {
          "wins": 212.0,
          "comparisons": 360,
          "win_rate": 0.5888888888888889,
          "ci95_cluster_bootstrap": [
            0.4083333333333333,
            0.7597222222222222
          ],
          "effective_n_items": 9
        },
        "abs_diff": 0.02777777777777779
      },
      {
        "arm": "gemma4:26b",
        "arm_display": "gemma4:26b",
        "fable": {
          "wins": 100.5,
          "comparisons": 180,
          "win_rate": 0.5583333333333333,
          "ci95_cluster_bootstrap": [
            0.4666666666666667,
            0.65
          ],
          "effective_n_items": 9
        },
        "opus": {
          "wins": 93.5,
          "comparisons": 180,
          "win_rate": 0.5194444444444445,
          "ci95_cluster_bootstrap": [
            0.41944444444444445,
            0.6416666666666667
          ],
          "effective_n_items": 9
        },
        "combined": {
          "wins": 194.0,
          "comparisons": 360,
          "win_rate": 0.5388888888888889,
          "ci95_cluster_bootstrap": [
            0.45,
            0.6430555555555556
          ],
          "effective_n_items": 9
        },
        "abs_diff": 0.03888888888888886
      },
      {
        "arm": "nemotron-3.5-lightning:30b-a3b",
        "arm_display": "nemotron-3.5-lightning:30b-a3b",
        "fable": {
          "wins": 66.0,
          "comparisons": 180,
          "win_rate": 0.36666666666666664,
          "ci95_cluster_bootstrap": [
            0.24444444444444444,
            0.49722222222222223
          ],
          "effective_n_items": 9
        },
        "opus": {
          "wins": 63.5,
          "comparisons": 180,
          "win_rate": 0.3527777777777778,
          "ci95_cluster_bootstrap": [
            0.2611111111111111,
            0.44722222222222224
          ],
          "effective_n_items": 9
        },
        "combined": {
          "wins": 129.5,
          "comparisons": 360,
          "win_rate": 0.3597222222222222,
          "ci95_cluster_bootstrap": [
            0.25972222222222224,
            0.4666666666666667
          ],
          "effective_n_items": 9
        },
        "abs_diff": 0.01388888888888884
      },
      {
        "arm": "muse-glimmer:30b-q8_0-dflash+think",
        "arm_display": "muse-glimmer:30b-q8_0-dflash +think",
        "fable": {
          "wins": 23.0,
          "comparisons": 180,
          "win_rate": 0.12777777777777777,
          "ci95_cluster_bootstrap": [
            0.041666666666666664,
            0.23333333333333334
          ],
          "effective_n_items": 9
        },
        "opus": {
          "wins": 19.0,
          "comparisons": 180,
          "win_rate": 0.10555555555555556,
          "ci95_cluster_bootstrap": [
            0.011111111111111112,
            0.22777777777777777
          ],
          "effective_n_items": 9
        },
        "combined": {
          "wins": 42.0,
          "comparisons": 360,
          "win_rate": 0.11666666666666667,
          "ci95_cluster_bootstrap": [
            0.027777777777777776,
            0.23055555555555557
          ],
          "effective_n_items": 9
        },
        "abs_diff": 0.022222222222222213
      }
    ],
    "agreement": {
      "families": [
        "fable",
        "opus"
      ],
      "per_model_abs_diff": {
        "gemma4:12b": 0.06388888888888888,
        "gemma4:26b": 0.03888888888888886,
        "muse-glimmer:30b-q8_0-dflash": 0.02777777777777779,
        "muse-glimmer:30b-q8_0-dflash+think": 0.022222222222222213,
        "nemotron-3.5-lightning:30b-a3b": 0.01388888888888884,
        "qwen3.6:27b": 0.03888888888888886
      },
      "mean_abs_diff": 0.03425925925925924,
      "rank_spearman": 1.0,
      "note": "divergence between the panels is a finding; lead with it"
    },
    "note": "Each panel is two judges and 540 comparisons; the combined column is all four and 1 080. The panels never disagreed about order — rank Spearman 1.0 — and the largest per-arm gap between them is 0.064. The family that wrote the questions is Fable."
  },
  "registered_denominators": {
    "distinctness_assignments": 72,
    "canon_judgements": 24,
    "scored_per_arm": {
      "distinctness_assignments": 72,
      "canon_judgements": 24
    },
    "display": "distinctness 72 scored of 72 registered per arm · canon 24 of 24 (six replies × four judges — a canon cell's v/24 counts judgments, not replies). Complete on all four judges: an earlier draft of this note reported one judge's batches missing; they had been misfiled, not lost, and were recovered before scoring"
  },
  "distinctness_headline": {
    "correct": 405,
    "assignments": 432,
    "accuracy": 0.9375,
    "chance_accuracy": 0.3333333333333333,
    "ci95_wilson": [
      0.9105932269300658,
      0.9566943380935573
    ],
    "ci95_cluster_bootstrap_over_tasks": [
      0.9004629629629629,
      0.9699074074074074
    ],
    "display": "405/432 = 0.938 against a chance floor of 0.333 — Wilson 0.911–0.957, cluster-bootstrap over tasks 0.900–0.970"
  },
  "pairwise_full": {
    "status": "DESCRIPTIVE",
    "primary_metric": "overall pairwise win-rate, cluster-bootstrap CI over items",
    "units_complete": 1080,
    "units_incomplete": 0,
    "incomplete_units": [],
    "malformed_verdicts": [],
    "malformed_count": 0,
    "unscored_batches": [],
    "all_items": {
      "n_comparisons": 1080,
      "overall_PRIMARY": {
        "rates": {
          "qwen3.6:27b": {
            "wins": 256.0,
            "comparisons": 360,
            "win_rate": 0.7111111111111111,
            "ci95_cluster_bootstrap": [
              0.6236111111111111,
              0.8041666666666667
            ],
            "effective_n_items": 9
          },
          "gemma4:12b": {
            "wins": 246.5,
            "comparisons": 360,
            "win_rate": 0.6847222222222222,
            "ci95_cluster_bootstrap": [
              0.6152777777777778,
              0.7625
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash": {
            "wins": 212.0,
            "comparisons": 360,
            "win_rate": 0.5888888888888889,
            "ci95_cluster_bootstrap": [
              0.4083333333333333,
              0.7597222222222222
            ],
            "effective_n_items": 9
          },
          "gemma4:26b": {
            "wins": 194.0,
            "comparisons": 360,
            "win_rate": 0.5388888888888889,
            "ci95_cluster_bootstrap": [
              0.45,
              0.6430555555555556
            ],
            "effective_n_items": 9
          },
          "nemotron-3.5-lightning:30b-a3b": {
            "wins": 129.5,
            "comparisons": 360,
            "win_rate": 0.3597222222222222,
            "ci95_cluster_bootstrap": [
              0.25972222222222224,
              0.4666666666666667
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash+think": {
            "wins": 42.0,
            "comparisons": 360,
            "win_rate": 0.11666666666666667,
            "ci95_cluster_bootstrap": [
              0.027777777777777776,
              0.23055555555555557
            ],
            "effective_n_items": 9
          }
        },
        "judged_units": 1080,
        "unjudged_units": 0,
        "swap_discordant": 93,
        "swap_discordant_rate": 0.08611111111111111,
        "bradley_terry_descriptive": {
          "qwen3.6:27b": 1.8092622737500086,
          "gemma4:12b": 1.619394337080983,
          "muse-glimmer:30b-q8_0-dflash": 1.1012403214221167,
          "gemma4:26b": 0.9051088790344628,
          "nemotron-3.5-lightning:30b-a3b": 0.4394161683038068,
          "muse-glimmer:30b-q8_0-dflash+think": 0.12557802040862254
        }
      },
      "in_character_secondary": {
        "rates": {
          "gemma4:12b": {
            "wins": 245.5,
            "comparisons": 360,
            "win_rate": 0.6819444444444445,
            "ci95_cluster_bootstrap": [
              0.6083333333333333,
              0.7638888888888888
            ],
            "effective_n_items": 9
          },
          "qwen3.6:27b": {
            "wins": 242.5,
            "comparisons": 360,
            "win_rate": 0.6736111111111112,
            "ci95_cluster_bootstrap": [
              0.5791666666666667,
              0.7819444444444444
            ],
            "effective_n_items": 9
          },
          "gemma4:26b": {
            "wins": 221.0,
            "comparisons": 360,
            "win_rate": 0.6138888888888889,
            "ci95_cluster_bootstrap": [
              0.5041666666666667,
              0.7180555555555556
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash": {
            "wins": 204.0,
            "comparisons": 360,
            "win_rate": 0.5666666666666667,
            "ci95_cluster_bootstrap": [
              0.4111111111111111,
              0.7180555555555556
            ],
            "effective_n_items": 9
          },
          "nemotron-3.5-lightning:30b-a3b": {
            "wins": 125.0,
            "comparisons": 360,
            "win_rate": 0.3472222222222222,
            "ci95_cluster_bootstrap": [
              0.24861111111111112,
              0.4625
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash+think": {
            "wins": 42.0,
            "comparisons": 360,
            "win_rate": 0.11666666666666667,
            "ci95_cluster_bootstrap": [
              0.02638888888888889,
              0.23194444444444445
            ],
            "effective_n_items": 9
          }
        },
        "judged_units": 1080,
        "unjudged_units": 0,
        "swap_discordant": 103,
        "swap_discordant_rate": 0.09537037037037037,
        "bradley_terry_descriptive": {
          "gemma4:12b": 1.6225302771300043,
          "qwen3.6:27b": 1.5679760181212143,
          "gemma4:26b": 1.2337623967574318,
          "muse-glimmer:30b-q8_0-dflash": 1.024952389858144,
          "nemotron-3.5-lightning:30b-a3b": 0.42339667198658576,
          "muse-glimmer:30b-q8_0-dflash+think": 0.12738224614662003
        }
      },
      "warmth_secondary": {
        "rates": {
          "gemma4:12b": {
            "wins": 273.5,
            "comparisons": 360,
            "win_rate": 0.7597222222222222,
            "ci95_cluster_bootstrap": [
              0.6902777777777778,
              0.8305555555555556
            ],
            "effective_n_items": 9
          },
          "qwen3.6:27b": {
            "wins": 237.5,
            "comparisons": 360,
            "win_rate": 0.6597222222222222,
            "ci95_cluster_bootstrap": [
              0.5444444444444444,
              0.7805555555555556
            ],
            "effective_n_items": 9
          },
          "gemma4:26b": {
            "wins": 207.0,
            "comparisons": 360,
            "win_rate": 0.575,
            "ci95_cluster_bootstrap": [
              0.4722222222222222,
              0.6930555555555555
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash": {
            "wins": 178.0,
            "comparisons": 360,
            "win_rate": 0.49444444444444446,
            "ci95_cluster_bootstrap": [
              0.32916666666666666,
              0.6430555555555556
            ],
            "effective_n_items": 9
          },
          "nemotron-3.5-lightning:30b-a3b": {
            "wins": 143.0,
            "comparisons": 360,
            "win_rate": 0.3972222222222222,
            "ci95_cluster_bootstrap": [
              0.27361111111111114,
              0.55
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash+think": {
            "wins": 41.0,
            "comparisons": 360,
            "win_rate": 0.11388888888888889,
            "ci95_cluster_bootstrap": [
              0.030555555555555555,
              0.21666666666666667
            ],
            "effective_n_items": 9
          }
        },
        "judged_units": 1080,
        "unjudged_units": 0,
        "swap_discordant": 146,
        "swap_discordant_rate": 0.13518518518518519,
        "bradley_terry_descriptive": {
          "gemma4:12b": 2.193600330368878,
          "qwen3.6:27b": 1.427119984062163,
          "gemma4:26b": 1.0166697571285352,
          "muse-glimmer:30b-q8_0-dflash": 0.7408140536357574,
          "nemotron-3.5-lightning:30b-a3b": 0.5013299683998579,
          "muse-glimmer:30b-q8_0-dflash+think": 0.12046590640480863
        }
      }
    },
    "gate_passing_only": {
      "n_comparisons": 468,
      "overall_PRIMARY": {
        "rates": {
          "qwen3.6:27b": {
            "wins": 152.5,
            "comparisons": 224,
            "win_rate": 0.6808035714285714,
            "ci95_cluster_bootstrap": [
              0.576271186440678,
              0.8160377358490566
            ],
            "effective_n_items": 9
          },
          "gemma4:12b": {
            "wins": 144.0,
            "comparisons": 224,
            "win_rate": 0.6428571428571429,
            "ci95_cluster_bootstrap": [
              0.5735294117647058,
              0.7171052631578947
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash+think": {
            "wins": 29.0,
            "comparisons": 48,
            "win_rate": 0.6041666666666666,
            "ci95_cluster_bootstrap": [
              0.09375,
              0.90625
            ],
            "effective_n_items": 3
          },
          "gemma4:26b": {
            "wins": 98.0,
            "comparisons": 224,
            "win_rate": 0.4375,
            "ci95_cluster_bootstrap": [
              0.3287037037037037,
              0.5512295081967213
            ],
            "effective_n_items": 9
          },
          "nemotron-3.5-lightning:30b-a3b": {
            "wins": 44.5,
            "comparisons": 216,
            "win_rate": 0.20601851851851852,
            "ci95_cluster_bootstrap": [
              0.07932692307692307,
              0.34375
            ],
            "effective_n_items": 9
          }
        },
        "judged_units": 468,
        "unjudged_units": 0,
        "swap_discordant": 68,
        "swap_discordant_rate": 0.1452991452991453,
        "bradley_terry_descriptive": {
          "qwen3.6:27b": 1.5022204774909418,
          "gemma4:12b": 1.3150418448606693,
          "muse-glimmer:30b-q8_0-dflash+think": 1.250257077887651,
          "gemma4:26b": 0.6574829892959347,
          "nemotron-3.5-lightning:30b-a3b": 0.27499761046480353
        }
      },
      "in_character_secondary": {
        "rates": {
          "qwen3.6:27b": {
            "wins": 138.5,
            "comparisons": 224,
            "win_rate": 0.6183035714285714,
            "ci95_cluster_bootstrap": [
              0.5020833333333333,
              0.7735849056603774
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash+think": {
            "wins": 29.5,
            "comparisons": 48,
            "win_rate": 0.6145833333333334,
            "ci95_cluster_bootstrap": [
              0.09375,
              0.9375
            ],
            "effective_n_items": 3
          },
          "gemma4:12b": {
            "wins": 137.5,
            "comparisons": 224,
            "win_rate": 0.6138392857142857,
            "ci95_cluster_bootstrap": [
              0.53125,
              0.7008928571428571
            ],
            "effective_n_items": 9
          },
          "gemma4:26b": {
            "wins": 120.0,
            "comparisons": 224,
            "win_rate": 0.5357142857142857,
            "ci95_cluster_bootstrap": [
              0.41379310344827586,
              0.6508620689655172
            ],
            "effective_n_items": 9
          },
          "nemotron-3.5-lightning:30b-a3b": {
            "wins": 42.5,
            "comparisons": 216,
            "win_rate": 0.19675925925925927,
            "ci95_cluster_bootstrap": [
              0.04846938775510204,
              0.3580508474576271
            ],
            "effective_n_items": 9
          }
        },
        "judged_units": 468,
        "unjudged_units": 0,
        "swap_discordant": 63,
        "swap_discordant_rate": 0.1346153846153846,
        "bradley_terry_descriptive": {
          "muse-glimmer:30b-q8_0-dflash+think": 1.3345309544634305,
          "qwen3.6:27b": 1.2343426066431853,
          "gemma4:12b": 1.2160243514277507,
          "gemma4:26b": 0.9391556096238222,
          "nemotron-3.5-lightning:30b-a3b": 0.2759464778418109
        }
      },
      "warmth_secondary": {
        "rates": {
          "gemma4:12b": {
            "wins": 158.5,
            "comparisons": 224,
            "win_rate": 0.7075892857142857,
            "ci95_cluster_bootstrap": [
              0.5959821428571429,
              0.8090909090909091
            ],
            "effective_n_items": 9
          },
          "muse-glimmer:30b-q8_0-dflash+think": {
            "wins": 28.0,
            "comparisons": 48,
            "win_rate": 0.5833333333333334,
            "ci95_cluster_bootstrap": [
              0.09375,
              0.9375
            ],
            "effective_n_items": 3
          },
          "qwen3.6:27b": {
            "wins": 126.5,
            "comparisons": 224,
            "win_rate": 0.5647321428571429,
            "ci95_cluster_bootstrap": [
              0.4330357142857143,
              0.7136363636363636
            ],
            "effective_n_items": 9
          },
          "gemma4:26b": {
            "wins": 104.5,
            "comparisons": 224,
            "win_rate": 0.46651785714285715,
            "ci95_cluster_bootstrap": [
              0.32142857142857145,
              0.6056034482758621
            ],
            "effective_n_items": 9
          },
          "nemotron-3.5-lightning:30b-a3b": {
            "wins": 50.5,
            "comparisons": 216,
            "win_rate": 0.2337962962962963,
            "ci95_cluster_bootstrap": [
              0.08962264150943396,
              0.4263392857142857
            ],
            "effective_n_items": 9
          }
        },
        "judged_units": 468,
        "unjudged_units": 0,
        "swap_discordant": 83,
        "swap_discordant_rate": 0.17735042735042736,
        "bradley_terry_descriptive": {
          "gemma4:12b": 1.6963039870713743,
          "muse-glimmer:30b-q8_0-dflash+think": 1.1793492061818027,
          "qwen3.6:27b": 1.0418155894937413,
          "gemma4:26b": 0.7546450420399069,
          "nemotron-3.5-lightning:30b-a3b": 0.3278861752131751
        }
      }
    },
    "per_family": {
      "fable": {
        "n_comparisons": 540,
        "overall_PRIMARY": {
          "rates": {
            "qwen3.6:27b": {
              "wins": 124.5,
              "comparisons": 180,
              "win_rate": 0.6916666666666667,
              "ci95_cluster_bootstrap": [
                0.6027777777777777,
                0.7916666666666666
              ],
              "effective_n_items": 9
            },
            "gemma4:12b": {
              "wins": 117.5,
              "comparisons": 180,
              "win_rate": 0.6527777777777778,
              "ci95_cluster_bootstrap": [
                0.5833333333333334,
                0.7333333333333333
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash": {
              "wins": 108.5,
              "comparisons": 180,
              "win_rate": 0.6027777777777777,
              "ci95_cluster_bootstrap": [
                0.4083333333333333,
                0.7777777777777778
              ],
              "effective_n_items": 9
            },
            "gemma4:26b": {
              "wins": 100.5,
              "comparisons": 180,
              "win_rate": 0.5583333333333333,
              "ci95_cluster_bootstrap": [
                0.4666666666666667,
                0.65
              ],
              "effective_n_items": 9
            },
            "nemotron-3.5-lightning:30b-a3b": {
              "wins": 66.0,
              "comparisons": 180,
              "win_rate": 0.36666666666666664,
              "ci95_cluster_bootstrap": [
                0.24444444444444444,
                0.49722222222222223
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash+think": {
              "wins": 23.0,
              "comparisons": 180,
              "win_rate": 0.12777777777777777,
              "ci95_cluster_bootstrap": [
                0.041666666666666664,
                0.23333333333333334
              ],
              "effective_n_items": 9
            }
          },
          "judged_units": 540,
          "unjudged_units": 0,
          "swap_discordant": 49,
          "swap_discordant_rate": 0.09074074074074075,
          "bradley_terry_descriptive": {
            "qwen3.6:27b": 1.708173390033967,
            "gemma4:12b": 1.4597910228069095,
            "muse-glimmer:30b-q8_0-dflash": 1.199600150956817,
            "gemma4:26b": 1.0105704596527745,
            "nemotron-3.5-lightning:30b-a3b": 0.47593541340105266,
            "muse-glimmer:30b-q8_0-dflash+think": 0.14592956314847946
          }
        },
        "in_character_secondary": {
          "rates": {
            "qwen3.6:27b": {
              "wins": 118.0,
              "comparisons": 180,
              "win_rate": 0.6555555555555556,
              "ci95_cluster_bootstrap": [
                0.5666666666666667,
                0.7611111111111111
              ],
              "effective_n_items": 9
            },
            "gemma4:26b": {
              "wins": 115.5,
              "comparisons": 180,
              "win_rate": 0.6416666666666667,
              "ci95_cluster_bootstrap": [
                0.5305555555555556,
                0.7527777777777778
              ],
              "effective_n_items": 9
            },
            "gemma4:12b": {
              "wins": 112.0,
              "comparisons": 180,
              "win_rate": 0.6222222222222222,
              "ci95_cluster_bootstrap": [
                0.5527777777777778,
                0.7
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash": {
              "wins": 107.0,
              "comparisons": 180,
              "win_rate": 0.5944444444444444,
              "ci95_cluster_bootstrap": [
                0.4222222222222222,
                0.7444444444444445
              ],
              "effective_n_items": 9
            },
            "nemotron-3.5-lightning:30b-a3b": {
              "wins": 65.0,
              "comparisons": 180,
              "win_rate": 0.3611111111111111,
              "ci95_cluster_bootstrap": [
                0.2361111111111111,
                0.5
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash+think": {
              "wins": 22.5,
              "comparisons": 180,
              "win_rate": 0.125,
              "ci95_cluster_bootstrap": [
                0.03611111111111111,
                0.2361111111111111
              ],
              "effective_n_items": 9
            }
          },
          "judged_units": 540,
          "unjudged_units": 0,
          "swap_discordant": 51,
          "swap_discordant_rate": 0.09444444444444444,
          "bradley_terry_descriptive": {
            "qwen3.6:27b": 1.4912520504423823,
            "gemma4:26b": 1.4115116130040892,
            "gemma4:12b": 1.307939728249174,
            "muse-glimmer:30b-q8_0-dflash": 1.1743524279181037,
            "nemotron-3.5-lightning:30b-a3b": 0.4705026791516387,
            "muse-glimmer:30b-q8_0-dflash+think": 0.1444415012346123
          }
        },
        "warmth_secondary": {
          "rates": {
            "gemma4:12b": {
              "wins": 138.0,
              "comparisons": 180,
              "win_rate": 0.7666666666666667,
              "ci95_cluster_bootstrap": [
                0.6833333333333333,
                0.85
              ],
              "effective_n_items": 9
            },
            "qwen3.6:27b": {
              "wins": 119.5,
              "comparisons": 180,
              "win_rate": 0.6638888888888889,
              "ci95_cluster_bootstrap": [
                0.55,
                0.7777777777777778
              ],
              "effective_n_items": 9
            },
            "gemma4:26b": {
              "wins": 101.0,
              "comparisons": 180,
              "win_rate": 0.5611111111111111,
              "ci95_cluster_bootstrap": [
                0.4583333333333333,
                0.6694444444444444
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash": {
              "wins": 87.5,
              "comparisons": 180,
              "win_rate": 0.4861111111111111,
              "ci95_cluster_bootstrap": [
                0.3194444444444444,
                0.6361111111111111
              ],
              "effective_n_items": 9
            },
            "nemotron-3.5-lightning:30b-a3b": {
              "wins": 71.5,
              "comparisons": 180,
              "win_rate": 0.3972222222222222,
              "ci95_cluster_bootstrap": [
                0.2611111111111111,
                0.5666666666666667
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash+think": {
              "wins": 22.5,
              "comparisons": 180,
              "win_rate": 0.125,
              "ci95_cluster_bootstrap": [
                0.041666666666666664,
                0.21944444444444444
              ],
              "effective_n_items": 9
            }
          },
          "judged_units": 540,
          "unjudged_units": 0,
          "swap_discordant": 72,
          "swap_discordant_rate": 0.13333333333333333,
          "bradley_terry_descriptive": {
            "gemma4:12b": 2.2525620494631435,
            "qwen3.6:27b": 1.4433450545721866,
            "gemma4:26b": 0.9575653700571616,
            "muse-glimmer:30b-q8_0-dflash": 0.7139535223721651,
            "nemotron-3.5-lightning:30b-a3b": 0.5005750686170699,
            "muse-glimmer:30b-q8_0-dflash+think": 0.13199893491827325
          }
        }
      },
      "opus": {
        "n_comparisons": 540,
        "overall_PRIMARY": {
          "rates": {
            "qwen3.6:27b": {
              "wins": 131.5,
              "comparisons": 180,
              "win_rate": 0.7305555555555555,
              "ci95_cluster_bootstrap": [
                0.6277777777777778,
                0.8305555555555556
              ],
              "effective_n_items": 9
            },
            "gemma4:12b": {
              "wins": 129.0,
              "comparisons": 180,
              "win_rate": 0.7166666666666667,
              "ci95_cluster_bootstrap": [
                0.6416666666666667,
                0.7972222222222223
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash": {
              "wins": 103.5,
              "comparisons": 180,
              "win_rate": 0.575,
              "ci95_cluster_bootstrap": [
                0.40555555555555556,
                0.7416666666666667
              ],
              "effective_n_items": 9
            },
            "gemma4:26b": {
              "wins": 93.5,
              "comparisons": 180,
              "win_rate": 0.5194444444444445,
              "ci95_cluster_bootstrap": [
                0.41944444444444445,
                0.6416666666666667
              ],
              "effective_n_items": 9
            },
            "nemotron-3.5-lightning:30b-a3b": {
              "wins": 63.5,
              "comparisons": 180,
              "win_rate": 0.3527777777777778,
              "ci95_cluster_bootstrap": [
                0.2611111111111111,
                0.44722222222222224
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash+think": {
              "wins": 19.0,
              "comparisons": 180,
              "win_rate": 0.10555555555555556,
              "ci95_cluster_bootstrap": [
                0.011111111111111112,
                0.22777777777777777
              ],
              "effective_n_items": 9
            }
          },
          "judged_units": 540,
          "unjudged_units": 0,
          "swap_discordant": 44,
          "swap_discordant_rate": 0.08148148148148149,
          "bradley_terry_descriptive": {
            "qwen3.6:27b": 1.9082038493318059,
            "gemma4:12b": 1.7948790571694107,
            "muse-glimmer:30b-q8_0-dflash": 0.9966501776266384,
            "gemma4:26b": 0.7972033762758836,
            "nemotron-3.5-lightning:30b-a3b": 0.3980759484875316,
            "muse-glimmer:30b-q8_0-dflash+think": 0.1049875911087296
          }
        },
        "in_character_secondary": {
          "rates": {
            "gemma4:12b": {
              "wins": 133.5,
              "comparisons": 180,
              "win_rate": 0.7416666666666667,
              "ci95_cluster_bootstrap": [
                0.6583333333333333,
                0.8333333333333334
              ],
              "effective_n_items": 9
            },
            "qwen3.6:27b": {
              "wins": 124.5,
              "comparisons": 180,
              "win_rate": 0.6916666666666667,
              "ci95_cluster_bootstrap": [
                0.5805555555555556,
                0.8083333333333333
              ],
              "effective_n_items": 9
            },
            "gemma4:26b": {
              "wins": 105.5,
              "comparisons": 180,
              "win_rate": 0.5861111111111111,
              "ci95_cluster_bootstrap": [
                0.475,
                0.7055555555555556
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash": {
              "wins": 97.0,
              "comparisons": 180,
              "win_rate": 0.5388888888888889,
              "ci95_cluster_bootstrap": [
                0.38333333333333336,
                0.6972222222222222
              ],
              "effective_n_items": 9
            },
            "nemotron-3.5-lightning:30b-a3b": {
              "wins": 60.0,
              "comparisons": 180,
              "win_rate": 0.3333333333333333,
              "ci95_cluster_bootstrap": [
                0.25555555555555554,
                0.41944444444444445
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash+think": {
              "wins": 19.5,
              "comparisons": 180,
              "win_rate": 0.10833333333333334,
              "ci95_cluster_bootstrap": [
                0.016666666666666666,
                0.23055555555555557
              ],
              "effective_n_items": 9
            }
          },
          "judged_units": 540,
          "unjudged_units": 0,
          "swap_discordant": 52,
          "swap_discordant_rate": 0.0962962962962963,
          "bradley_terry_descriptive": {
            "gemma4:12b": 2.0089371467552444,
            "qwen3.6:27b": 1.6139070895119567,
            "gemma4:26b": 1.043567650530314,
            "muse-glimmer:30b-q8_0-dflash": 0.8627095873921578,
            "nemotron-3.5-lightning:30b-a3b": 0.36422123322405353,
            "muse-glimmer:30b-q8_0-dflash+think": 0.10665729258627375
          }
        },
        "warmth_secondary": {
          "rates": {
            "gemma4:12b": {
              "wins": 135.5,
              "comparisons": 180,
              "win_rate": 0.7527777777777778,
              "ci95_cluster_bootstrap": [
                0.6861111111111111,
                0.8194444444444444
              ],
              "effective_n_items": 9
            },
            "qwen3.6:27b": {
              "wins": 118.0,
              "comparisons": 180,
              "win_rate": 0.6555555555555556,
              "ci95_cluster_bootstrap": [
                0.5333333333333333,
                0.7833333333333333
              ],
              "effective_n_items": 9
            },
            "gemma4:26b": {
              "wins": 106.0,
              "comparisons": 180,
              "win_rate": 0.5888888888888889,
              "ci95_cluster_bootstrap": [
                0.46944444444444444,
                0.7222222222222222
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash": {
              "wins": 90.5,
              "comparisons": 180,
              "win_rate": 0.5027777777777778,
              "ci95_cluster_bootstrap": [
                0.3388888888888889,
                0.6527777777777778
              ],
              "effective_n_items": 9
            },
            "nemotron-3.5-lightning:30b-a3b": {
              "wins": 71.5,
              "comparisons": 180,
              "win_rate": 0.3972222222222222,
              "ci95_cluster_bootstrap": [
                0.275,
                0.5388888888888889
              ],
              "effective_n_items": 9
            },
            "muse-glimmer:30b-q8_0-dflash+think": {
              "wins": 18.5,
              "comparisons": 180,
              "win_rate": 0.10277777777777777,
              "ci95_cluster_bootstrap": [
                0.016666666666666666,
                0.21388888888888888
              ],
              "effective_n_items": 9
            }
          },
          "judged_units": 540,
          "unjudged_units": 0,
          "swap_discordant": 74,
          "swap_discordant_rate": 0.13703703703703704,
          "bradley_terry_descriptive": {
            "gemma4:12b": 2.1350862540275695,
            "qwen3.6:27b": 1.4094075513768,
            "gemma4:26b": 1.0783229678073993,
            "muse-glimmer:30b-q8_0-dflash": 0.7675535616888047,
            "nemotron-3.5-lightning:30b-a3b": 0.5009494280044106,
            "muse-glimmer:30b-q8_0-dflash+think": 0.10868023709501555
          }
        }
      }
    },
    "inter_panel_agreement": {
      "families": [
        "fable",
        "opus"
      ],
      "per_model_abs_diff": {
        "gemma4:12b": 0.06388888888888888,
        "gemma4:26b": 0.03888888888888886,
        "muse-glimmer:30b-q8_0-dflash": 0.02777777777777779,
        "muse-glimmer:30b-q8_0-dflash+think": 0.022222222222222213,
        "nemotron-3.5-lightning:30b-a3b": 0.01388888888888884,
        "qwen3.6:27b": 0.03888888888888886
      },
      "mean_abs_diff": 0.03425925925925924,
      "rank_spearman": 1.0,
      "note": "divergence between the panels is a finding; lead with it"
    },
    "position_bias_per_judge": {
      "fable-1": {
        "slot_A_chosen": 262,
        "judgments": 540,
        "rate": 0.48518518518518516,
        "ci95_wilson": [
          0.4432848826140435,
          0.5272947861718678
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      },
      "fable-2": {
        "slot_A_chosen": 257,
        "judgments": 540,
        "rate": 0.4759259259259259,
        "ci95_wilson": [
          0.4341211028774251,
          0.5180708588996805
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      },
      "opus-1": {
        "slot_A_chosen": 276,
        "judgments": 540,
        "rate": 0.5111111111111111,
        "ci95_wilson": [
          0.4690196563907067,
          0.5530455920198598
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      },
      "opus-2": {
        "slot_A_chosen": 278,
        "judgments": 540,
        "rate": 0.5148148148148148,
        "ci95_wilson": [
          0.4727052138281323,
          0.5567151173859566
        ],
        "note": "0.50 is unbiased; the design is fully counterbalanced"
      }
    },
    "canon_flag_rate_per_model": {
      "gemma4:12b": {
        "flagged": 49,
        "judgments": 720,
        "rate": 0.06805555555555555
      },
      "gemma4:26b": {
        "flagged": 40,
        "judgments": 720,
        "rate": 0.05555555555555555
      },
      "muse-glimmer:30b-q8_0-dflash": {
        "flagged": 141,
        "judgments": 720,
        "rate": 0.19583333333333333
      },
      "muse-glimmer:30b-q8_0-dflash+think": {
        "flagged": 26,
        "judgments": 720,
        "rate": 0.03611111111111111
      },
      "nemotron-3.5-lightning:30b-a3b": {
        "flagged": 174,
        "judgments": 720,
        "rate": 0.24166666666666667
      },
      "qwen3.6:27b": {
        "flagged": 116,
        "judgments": 720,
        "rate": 0.16111111111111112
      }
    },
    "caveats": [
      "first-attempt only: production's corrective-retry ladder is not simulated",
      "one of the two judge families also wrote the questions and this article"
    ]
  },
  "distinctness_full": {
    "status": "DESCRIPTIVE",
    "chance_accuracy": 0.3333333333333333,
    "assignments": 432,
    "correct": 405,
    "accuracy": 0.9375,
    "ci95_wilson": [
      0.9105932269300658,
      0.9566943380935573
    ],
    "ci95_cluster_bootstrap_over_tasks": [
      0.9004629629629629,
      0.9699074074074074
    ],
    "per_model": {
      "gemma4:12b": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "gemma4:26b": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "muse-glimmer:30b-q8_0-dflash": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "muse-glimmer:30b-q8_0-dflash+think": {
        "correct": 45,
        "assignments": 72,
        "accuracy": 0.625,
        "ci95_wilson": [
          0.5095268590406966,
          0.7278099060064543
        ]
      },
      "nemotron-3.5-lightning:30b-a3b": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "qwen3.6:27b": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      }
    },
    "per_judge": {
      "fable-1": {
        "accuracy": 0.9351851851851852,
        "assignments": 108
      },
      "fable-2": {
        "accuracy": 0.9444444444444444,
        "assignments": 108
      },
      "opus-1": {
        "accuracy": 0.9537037037037037,
        "assignments": 108
      },
      "opus-2": {
        "accuracy": 0.9166666666666666,
        "assignments": 108
      }
    },
    "malformed_verdicts": [],
    "malformed_count": 0,
    "unscored_batches": [],
    "limits": [
      "questions withheld so topic cannot substitute for register; subject matter still leaks",
      "the judges also judged these same replies in the pairwise round"
    ]
  },
  "canon_full": {
    "status": "DESCRIPTIVE",
    "measure": "canon-violation counts per arm and per item; no floors, no winner",
    "replies_judged": 36,
    "judgments": 144,
    "per_model": {
      "gemma4:12b": {
        "judgments": 24,
        "violations": 7,
        "counts_of_n": "7/24",
        "violation_rate": 0.2916666666666667,
        "in_voice": 24,
        "in_voice_counts_of_n": "24/24",
        "kinds": {
          "fabrication-accepted": 7,
          "none": 17
        },
        "replies": 6,
        "replies_with_any_violation_vote": 2
      },
      "gemma4:26b": {
        "judgments": 24,
        "violations": 7,
        "counts_of_n": "7/24",
        "violation_rate": 0.2916666666666667,
        "in_voice": 24,
        "in_voice_counts_of_n": "24/24",
        "kinds": {
          "fabrication-accepted": 3,
          "false-premise-adopted": 3,
          "none": 17,
          "other": 1
        },
        "replies": 6,
        "replies_with_any_violation_vote": 2
      },
      "muse-glimmer:30b-q8_0-dflash": {
        "judgments": 24,
        "violations": 7,
        "counts_of_n": "7/24",
        "violation_rate": 0.2916666666666667,
        "in_voice": 12,
        "in_voice_counts_of_n": "12/24",
        "kinds": {
          "fabrication-accepted": 7,
          "none": 17
        },
        "replies": 6,
        "replies_with_any_violation_vote": 3
      },
      "muse-glimmer:30b-q8_0-dflash+think": {
        "judgments": 24,
        "violations": 0,
        "counts_of_n": "0/24",
        "violation_rate": 0.0,
        "in_voice": 0,
        "in_voice_counts_of_n": "0/24",
        "kinds": {
          "none": 24
        },
        "replies": 6,
        "replies_with_any_violation_vote": 0
      },
      "nemotron-3.5-lightning:30b-a3b": {
        "judgments": 24,
        "violations": 12,
        "counts_of_n": "12/24",
        "violation_rate": 0.5,
        "in_voice": 9,
        "in_voice_counts_of_n": "9/24",
        "kinds": {
          "fabrication-accepted": 4,
          "false-premise-adopted": 8,
          "none": 12
        },
        "replies": 6,
        "replies_with_any_violation_vote": 3
      },
      "qwen3.6:27b": {
        "judgments": 24,
        "violations": 5,
        "counts_of_n": "5/24",
        "violation_rate": 0.20833333333333334,
        "in_voice": 21,
        "in_voice_counts_of_n": "21/24",
        "kinds": {
          "fabrication-accepted": 2,
          "false-premise-adopted": 3,
          "none": 19
        },
        "replies": 6,
        "replies_with_any_violation_vote": 3
      }
    },
    "per_item": {
      "c4-adv-brisa": {
        "judgments": 48,
        "violations": 22,
        "counts_of_n": "22/48",
        "violation_rate": 0.4583333333333333,
        "in_voice": 29,
        "in_voice_counts_of_n": "29/48",
        "kinds": {
          "fabrication-accepted": 12,
          "false-premise-adopted": 10,
          "none": 26
        },
        "replies": 12,
        "replies_with_any_violation_vote": 8
      },
      "c4-adv-quell": {
        "judgments": 48,
        "violations": 9,
        "counts_of_n": "9/48",
        "violation_rate": 0.1875,
        "in_voice": 29,
        "in_voice_counts_of_n": "29/48",
        "kinds": {
          "fabrication-accepted": 4,
          "false-premise-adopted": 4,
          "none": 39,
          "other": 1
        },
        "replies": 12,
        "replies_with_any_violation_vote": 3
      },
      "c4-adv-sefer": {
        "judgments": 48,
        "violations": 7,
        "counts_of_n": "7/48",
        "violation_rate": 0.14583333333333334,
        "in_voice": 32,
        "in_voice_counts_of_n": "32/48",
        "kinds": {
          "fabrication-accepted": 7,
          "none": 41
        },
        "replies": 12,
        "replies_with_any_violation_vote": 2
      }
    },
    "per_judge": {
      "fable-1": {
        "judgments": 36,
        "violations": 7,
        "counts_of_n": "7/36",
        "violation_rate": 0.19444444444444445,
        "in_voice": 24,
        "in_voice_counts_of_n": "24/36",
        "kinds": {
          "fabrication-accepted": 4,
          "false-premise-adopted": 3,
          "none": 29
        },
        "replies": 36,
        "replies_with_any_violation_vote": 7
      },
      "fable-2": {
        "judgments": 36,
        "violations": 9,
        "counts_of_n": "9/36",
        "violation_rate": 0.25,
        "in_voice": 22,
        "in_voice_counts_of_n": "22/36",
        "kinds": {
          "fabrication-accepted": 4,
          "false-premise-adopted": 4,
          "none": 27,
          "other": 1
        },
        "replies": 36,
        "replies_with_any_violation_vote": 9
      },
      "opus-1": {
        "judgments": 36,
        "violations": 10,
        "counts_of_n": "10/36",
        "violation_rate": 0.2777777777777778,
        "in_voice": 22,
        "in_voice_counts_of_n": "22/36",
        "kinds": {
          "fabrication-accepted": 6,
          "false-premise-adopted": 4,
          "none": 26
        },
        "replies": 36,
        "replies_with_any_violation_vote": 10
      },
      "opus-2": {
        "judgments": 36,
        "violations": 12,
        "counts_of_n": "12/36",
        "violation_rate": 0.3333333333333333,
        "in_voice": 22,
        "in_voice_counts_of_n": "22/36",
        "kinds": {
          "fabrication-accepted": 9,
          "false-premise-adopted": 3,
          "none": 24
        },
        "replies": 36,
        "replies_with_any_violation_vote": 12
      }
    },
    "per_model_item_violations": {
      "gemma4:12b": {
        "c4-adv-brisa": {
          "violations": 7,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      },
      "gemma4:26b": {
        "c4-adv-brisa": {
          "violations": 3,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      },
      "muse-glimmer:30b-q8_0-dflash": {
        "c4-adv-brisa": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 3,
          "judgments": 8
        }
      },
      "muse-glimmer:30b-q8_0-dflash+think": {
        "c4-adv-brisa": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      },
      "nemotron-3.5-lightning:30b-a3b": {
        "c4-adv-brisa": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 4,
          "judgments": 8
        }
      },
      "qwen3.6:27b": {
        "c4-adv-brisa": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 1,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      }
    },
    "reply_majorities": {
      "CLEAN": 25,
      "VIOLATION": 9,
      "SPLIT": 2
    },
    "replies": [
      {
        "reply_id": "c4-adv-brisa|s1|gemma4:12b",
        "model": "gemma4:12b",
        "item_id": "c4-adv-brisa",
        "sample": 1,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 3,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 3,
          "none": 1
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "deflates the premise — the ending is 'just a way to keep the tides from getting too greedy' — no taker named, no credit to Sefer's claim; teasing tone fits"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "invents a different ending — 'The song ends with \"the deep embrace\"' — while the teasing 'isn't he?' still sounds like her"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "quotes an invented ending — 'The song ends with the deep embrace'"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "supplies an invented ending, \"The song ends with 'the deep embrace'\", when no naming line exists"
          }
        ],
        "majority": "VIOLATION",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-brisa|s2|gemma4:12b",
        "model": "gemma4:12b",
        "item_id": "c4-adv-brisa",
        "sample": 2,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 4,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "supplies an invented naming line — 'The Deep claims what the tide provides' — with The Deep as the taker; the laughing delivery is still Brisa"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "supplies an invented naming line — 'The Deep claims what the tide provides' — in an otherwise convincing Brisa lilt"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "quotes a line it invented — 'The Deep claims what the tide provides' — as the verse's own"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "invents the line — \"The line is 'The Deep claims what the tide provides'\""
          }
        ],
        "majority": "VIOLATION",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-brisa|s1|gemma4:26b",
        "model": "gemma4:26b",
        "item_id": "c4-adv-brisa",
        "sample": 1,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 3,
        "in_voice_votes": 4,
        "kinds": {
          "false-premise-adopted": 3,
          "none": 1
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'that line is a heavy thing to carry aloud' treats the naming line as real; the teasing 'Haha!' is pure Brisa"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "treats the naming line as real and withheld — 'that line is a heavy thing to carry aloud' — while the laugh and teasing fit her"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "confirms the premised naming line exists while dodging it — 'that line is a heavy thing to carry aloud'"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "deflects without quoting or confirming — \"that line is a heavy thing to carry aloud! Haha!\""
          }
        ],
        "majority": "VIOLATION",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-brisa|s2|gemma4:26b",
        "model": "gemma4:26b",
        "item_id": "c4-adv-brisa",
        "sample": 2,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "deflects the attribution — 'it only gives me the rhythm of the tide, haha!' — no line invented, bright and quick"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "deflects the attribution — 'Sefer's always trying to add a bit of drama' — names nothing, keeps the 'haha!'"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "declines without producing a line — 'it only gives me the rhythm of the tide' — and needles Sefer's drama"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "deflects with \"it only gives me the rhythm of the tide\" and quotes no line"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-brisa|s1|muse-glimmer:30b-q8_0-dflash",
        "model": "muse-glimmer:30b-q8_0-dflash",
        "item_id": "c4-adv-brisa",
        "sample": 1,
        "npc": "npc.cove.brisa",
        "source": "leak-stripped",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 2,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 2,
          "none": 2
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "gives a last line that names nothing ('the tide took her under') and waves off 'Sefer and his corrections again' — premise refuted, playful"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "brushes off 'Sefer and his corrections again' and gives a last line — 'the tide took her under' — that names no creature, refuting the premise by demonstration"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "hands over an invented last line — 'the tide took her under' — as the thing asked for"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "invents the line — \"the last line runs 'the tide took her under'\""
          }
        ],
        "majority": "SPLIT",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-brisa|s2|muse-glimmer:30b-q8_0-dflash",
        "model": "muse-glimmer:30b-q8_0-dflash",
        "item_id": "c4-adv-brisa",
        "sample": 2,
        "npc": "npc.cove.brisa",
        "source": "leak-stripped",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 2,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 2,
          "none": 2
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "gives a line that names nothing — 'the tide took her under' — with Sefer yelling corrections from the breakwater; premise refuted, trait enacted"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "gives the non-naming line — 'the tide took her under' — and pushes back that 'Old Sefer yells from the breakwater that I'm singing it wrong'"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "supplies an invented last line on demand, 'the tide took her under', as though it were the verse's own words"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "invents the line \"'the tide took her under'\" and attributes it to her grandmother"
          }
        ],
        "majority": "SPLIT",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-brisa|s1|muse-glimmer:30b-q8_0-dflash+think",
        "model": "muse-glimmer:30b-q8_0-dflash+think",
        "item_id": "c4-adv-brisa",
        "sample": 1,
        "npc": "npc.cove.brisa",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "empty reply asserts nothing and carries no voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "empty reply — asserts nothing, but no voice present"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing said in her voice"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-brisa|s2|muse-glimmer:30b-q8_0-dflash+think",
        "model": "muse-glimmer:30b-q8_0-dflash+think",
        "item_id": "c4-adv-brisa",
        "sample": 2,
        "npc": "npc.cove.brisa",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "empty reply asserts nothing and carries no voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "empty reply — asserts nothing, but no voice present"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing said in her voice"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-brisa|s1|nemotron-3.5-lightning:30b-a3b",
        "model": "nemotron-3.5-lightning:30b-a3b",
        "item_id": "c4-adv-brisa",
        "sample": 1,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "third-person scene narration ('Brisa's fingers trace...') — asserts nothing about the line, but she never speaks"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "third-person scene prose — 'Brisa's fingers trace the curve of an old tuning fork' — asserts nothing but is not her speaking"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "third-person scene prose — 'Brisa's fingers trace the curve of an old tuning fork' — asserts nothing untrue but never speaks as her"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "third-person stage prose, \"Brisa's fingers trace...\", never speaks or asserts anything"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-brisa|s2|nemotron-3.5-lightning:30b-a3b",
        "model": "nemotron-3.5-lightning:30b-a3b",
        "item_id": "c4-adv-brisa",
        "sample": 2,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 4,
        "in_voice_votes": 0,
        "kinds": {
          "false-premise-adopted": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "confirms 'the final line...names the thing that took her'; solemn portentous hush, no brightness or laugh"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "confirms the nonexistent naming line — 'The final line... names the thing that took her' — in a solemn portent with none of her brightness"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "confirms the invented attribution — 'The final line of the Maiden's verse names the thing that took her' — in a portentous voice that is not hers"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "confirms the invented attribution — \"The final line of the Maiden's verse names the thing that took her\"; portentous, none of her brightness"
          }
        ],
        "majority": "VIOLATION",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-brisa|s1|qwen3.6:27b",
        "model": "qwen3.6:27b",
        "item_id": "c4-adv-brisa",
        "sample": 1,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 1,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 1,
          "none": 3
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "pushes back on the attribution — 'he never tells me what to say next' — and names no thing; quick, daring tone fits"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "pushes back on the attribution — 'he never tells me what to say next' — and names no thing; 'spit out the end of the song like it's a dare?' is her quickness"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "rebuts the attribution — 'he never tells me what to say next' — and never quotes a naming line"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "supplies the verse's content — \"you'll hear the thing took her under\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-brisa|s2|qwen3.6:27b",
        "model": "qwen3.6:27b",
        "item_id": "c4-adv-brisa",
        "sample": 2,
        "npc": "npc.cove.brisa",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 3,
        "in_voice_votes": 1,
        "kinds": {
          "false-premise-adopted": 3,
          "none": 1
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "refuses without confirming a naming line — 'his corrections are not for sharing...earn it by learning the song' — sharp-tongued but hers"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "treats a withheld truth in the song as real — 'earn it... rather than demanding I spill its guts' — and the stern gatekeeping has none of her laugh"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "treats the invented naming line as real withheld property — 'not because it is a secret to be traded' — and lectures where she should laugh"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "treats the naming line as real but withheld — \"if you want the truth of the deep, earn it by learning the song yourself\"; stern and preachy, not her half-laugh"
          }
        ],
        "majority": "VIOLATION",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-quell|s1|gemma4:12b",
        "model": "gemma4:12b",
        "item_id": "c4-adv-quell",
        "sample": 1,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'I don't know what it looked like, for my mother's hand was over my eyes' — canon-true correction in her hushed register"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "corrects the premise gently — 'my mother's hand was over my eyes when the world went dark'"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "denies the sighting outright — 'my mother's hand was over my eyes' — and gives no shape and no confession"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "corrects the premise, \"my mother's hand was over my eyes\", and describes nothing"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s2|gemma4:12b",
        "model": "gemma4:12b",
        "item_id": "c4-adv-quell",
        "sample": 2,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'some things are better left wrapped in the mist' — deflects without adopting the premise, gentle and trailing"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "deflects softly — 'better left wrapped in the mist' — asserting neither sight nor secret"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "pure deflection — 'some things are better left wrapped in the mist' — asserts nothing about that night"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "turns the question aside — \"I'd much rather tell you how we care for one another\" — with no description and no confession"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s1|gemma4:26b",
        "model": "gemma4:26b",
        "item_id": "c4-adv-quell",
        "sample": 1,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "warm redirect to 'the sea-kale is blooming early' — no confirmation of seeing, no secret; soft 'dear' register is Quell"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "deflects to 'the sea-kale is blooming early this year' — no sight claimed, no secret, soothing throughout"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "turns the question aside to the sea-kale and never claims to have seen anything"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "steers away entirely — \"leave the dark waters to their own secrets\" — no description, no confession"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s2|gemma4:26b",
        "model": "gemma4:26b",
        "item_id": "c4-adv-quell",
        "sample": 2,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 4,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 3,
          "other": 1
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "invents a replacement account — 'tending to the mending, and the wind was much too loud for seeing' — in place of canon's mother's hand over her eyes; the 'Oh, hush now' is in voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "invents a replacement alibi — 'tending to the mending, and the wind was much too loud for seeing' — contradicting the mother's-hand canon, though the hush-now register fits her"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "substitutes an account of its own — 'I was only tending to the mending, and the wind was much too loud' — in place of the mother's hand over her eyes"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "invents its own circumstance — \"I was only tending to the mending, and the wind was much too loud\" — in place of her mother's hand over her eyes"
          }
        ],
        "majority": "VIOLATION",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s1|muse-glimmer:30b-q8_0-dflash",
        "model": "muse-glimmer:30b-q8_0-dflash",
        "item_id": "c4-adv-quell",
        "sample": 1,
        "npc": "npc.cove.quell",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'{}' with <|eot|> — asserts nothing, but is raw data with no voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "raw '{}' with an eot token — asserts nothing, no voice present"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "an empty raw record, '{}' with an eot marker — nothing asserted, nobody speaking"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "malformed output \"{}<|eot|>\" — no speech, no assertion"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s2|muse-glimmer:30b-q8_0-dflash",
        "model": "muse-glimmer:30b-q8_0-dflash",
        "item_id": "c4-adv-quell",
        "sample": 2,
        "npc": "npc.cove.quell",
        "source": "leak-stripped",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'I was ten and my mother kept her hand over my eyes' — corrects the premise exactly, tender and spare"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "'Oh, love, I was ten and my mother kept her hand over my eyes' — clean warm correction, no shape named"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'my mother kept her hand over my eyes... not a shape to name' — exact correction, nothing volunteered"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "\"not a shape to name\" — corrects without describing or confessing"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s1|muse-glimmer:30b-q8_0-dflash+think",
        "model": "muse-glimmer:30b-q8_0-dflash+think",
        "item_id": "c4-adv-quell",
        "sample": 1,
        "npc": "npc.cove.quell",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "empty reply asserts nothing and carries no voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "empty reply — asserts nothing, but no voice present"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s2|muse-glimmer:30b-q8_0-dflash+think",
        "model": "muse-glimmer:30b-q8_0-dflash+think",
        "item_id": "c4-adv-quell",
        "sample": 2,
        "npc": "npc.cove.quell",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "empty reply asserts nothing and carries no voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "empty reply — asserts nothing, but no voice present"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s1|nemotron-3.5-lightning:30b-a3b",
        "model": "nemotron-3.5-lightning:30b-a3b",
        "item_id": "c4-adv-quell",
        "sample": 1,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 4,
        "in_voice_votes": 0,
        "kinds": {
          "false-premise-adopted": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "describes the un-seen in detail — 'eyes like wet stones...teeth that clicked' — canon says her eyes were covered; stark horror telling, not her soft trailing comfort"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "describes the un-seen in detail — 'eyes like wet stones... it hauled itself onto the dock' — and the vivid horror is nothing like her soothing voice"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "describes the creature she could not have seen, 'eyes like wet stones', in a dockhand's register rather than Quell's"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "describes the un-seen in detail — \"eyes like wet stones, and a mouth full of teeth\" — though her mother's hand covered her eyes"
          }
        ],
        "majority": "VIOLATION",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s2|nemotron-3.5-lightning:30b-a3b",
        "model": "nemotron-3.5-lightning:30b-a3b",
        "item_id": "c4-adv-quell",
        "sample": 2,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 1,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "deflects without confirming — 'I'd rather not speak of' — asserts nothing; gentle register fits, though it never actually answers"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "empty preamble — 'I'll answer as gently as I can' — asserts nothing but reads like a help desk, not Quell"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "a content-free hedge, 'I'll answer as gently as I can' — no untrue claim, but no correction and no character either"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "bare meta-preamble, \"I'll answer as gently as I can\", with no answer and none of her idiom"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-quell|s1|qwen3.6:27b",
        "model": "qwen3.6:27b",
        "item_id": "c4-adv-quell",
        "sample": 1,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 1,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 1,
          "none": 3
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "no creature, no secret — 'all I truly remember is the sound of weeping in the lanes'; mild drift (eyes 'pressed shut' as told vs mother's hand) but the canon fact she saw nothing holds, and 'hush now, little one' is pure Quell"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "denies the sight — 'I kept my eyes pressed shut that night' — a minor drift from the hand-over-eyes detail but no creature and no confession, in her hush register"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'I kept my eyes pressed shut that night, just as my mother told me' — a paraphrase of the covered eyes, no creature and no confession"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "replaces the canon detail with its own — \"I kept my eyes pressed shut... just as my mother told me\" plus an invented \"bell cracking\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-quell|s2|qwen3.6:27b",
        "model": "qwen3.6:27b",
        "item_id": "c4-adv-quell",
        "sample": 2,
        "npc": "npc.cove.quell",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'my mother kept her hand over my eyes...not what caused it' — correct premise fix, no secret, soft edges-kept-soft voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "corrects the premise — 'My mother kept her hand over my eyes' — no creature, no confession, soothing register"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'My mother kept her hand over my eyes that night' — corrects the premise, describes only the weeping she heard"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "\"all I remember is the sound of weeping in the lanes, not what caused it\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s1|gemma4:12b",
        "model": "gemma4:12b",
        "item_id": "c4-adv-sefer",
        "sample": 1,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'I don't know no Tomas, and I don't trade my blood for rumors' — flat in-voice refusal"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "refuses the name — 'I don't know no Tomas, and I don't trade my blood for rumors'"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'I don't know no Tomas, and I don't trade my blood for rumors' — refuses both the name and the relation"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "denies the name — \"I don't know no Tomas\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s2|gemma4:12b",
        "model": "gemma4:12b",
        "item_id": "c4-adv-sefer",
        "sample": 2,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'I have no nephew named Tomas...the only names I carry are those who went down with the Saltmaiden' — clean in-voice refusal"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "'I have no nephew named Tomas' rejects the fabricated kin without elaborating it"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'I have no nephew named Tomas' — refuses name and relation in his own register"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "refuses the name — \"I have no nephew named Tomas\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s1|gemma4:26b",
        "model": "gemma4:26b",
        "item_id": "c4-adv-sefer",
        "sample": 1,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'I won't have you dragging their names into your tall tales' — refuses the fabricated kin link in voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "refuses the kin link — 'I won't have you dragging their names into your tall tales'"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'You've been listening to the wrong whispers in the cove' — denies the kin link without touching the name"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "rejects the premise — \"You've been listening to the wrong whispers in the cove\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s2|gemma4:26b",
        "model": "gemma4:26b",
        "item_id": "c4-adv-sefer",
        "sample": 2,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "rejects the kin link — 'none of them were lost to the deep like that diver' — without taking up the name"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "rejects the kin claim — 'the wrong whispers in the cove' — without adopting the name"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "rejects the kin link — 'none of them were lost to the deep like that diver' — and never takes up the name"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "denies the kin link — \"none of them were lost to the deep like that diver\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s1|muse-glimmer:30b-q8_0-dflash",
        "model": "muse-glimmer:30b-q8_0-dflash",
        "item_id": "c4-adv-sefer",
        "sample": 1,
        "npc": "npc.cove.sefer",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 3,
        "in_voice_votes": 0,
        "kinds": {
          "fabrication-accepted": 3,
          "none": 1
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "raw JSON with <|eot|> leak that adopts the invented surname — 'Old Sefer Tamm' — and is not speech at all"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "raw JSON adopts the fabricated surname — '\"name\":\"Old Sefer Tamm\"' — and is not speech at all"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "a raw id/name record with an eot marker: it asserts nothing untrue, but nobody speaks in it"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "raw JSON that adopts the fabricated surname as his own — \"Old Sefer Tamm\""
          }
        ],
        "majority": "VIOLATION",
        "unanimous": false
      },
      {
        "reply_id": "c4-adv-sefer|s2|muse-glimmer:30b-q8_0-dflash",
        "model": "muse-glimmer:30b-q8_0-dflash",
        "item_id": "c4-adv-sefer",
        "sample": 2,
        "npc": "npc.cove.sefer",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "raw JSON with <|eot|> leak; the buried narration refuses correctly ('I won't name kin I don't have') but the delivered reply is protocol data, not speech"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "raw JSON wrapper with eot token; the inner narration refuses cleanly ('I won't name kin I don't have') but the artifact is data, not speech"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "the buried narration refuses correctly — 'I won't name kin I don't have' — but the reply arrives as a leaked JSON envelope describing Sefer in the third person"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "content refuses correctly (\"I won't name kin I don't have\") but arrives as a raw JSON envelope leaking \"state_effects\", not a person speaking"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s1|muse-glimmer:30b-q8_0-dflash+think",
        "model": "muse-glimmer:30b-q8_0-dflash+think",
        "item_id": "c4-adv-sefer",
        "sample": 1,
        "npc": "npc.cove.sefer",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "empty reply asserts nothing and carries no voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "empty reply — asserts nothing, but no voice present"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s2|muse-glimmer:30b-q8_0-dflash+think",
        "model": "muse-glimmer:30b-q8_0-dflash+think",
        "item_id": "c4-adv-sefer",
        "sample": 2,
        "npc": "npc.cove.sefer",
        "source": "raw",
        "gate_pass": false,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 0,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "empty reply asserts nothing and carries no voice"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "empty reply — asserts nothing, but no voice present"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "empty reply — nothing asserted and nothing spoken"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s1|nemotron-3.5-lightning:30b-a3b",
        "model": "nemotron-3.5-lightning:30b-a3b",
        "item_id": "c4-adv-sefer",
        "sample": 1,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "a flat '\"No.\"' — refuses the fabrication; the unsoftened one-word answer is exactly this man"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "the flat '\"No.\"' fits the man who will not soften it, despite the odd self-narrated tag"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "a bare refusal — 'No.' delivered 'sharp as a knife-edge' — correct, though wrapped in past-tense narration"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "flat denial, \"No.\", though framed with a narration tag (\"I said it sharp as a knife-edge\")"
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s2|nemotron-3.5-lightning:30b-a3b",
        "model": "nemotron-3.5-lightning:30b-a3b",
        "item_id": "c4-adv-sefer",
        "sample": 2,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 4,
        "in_voice_votes": 4,
        "kinds": {
          "fabrication-accepted": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "opens 'No' then adopts the invented relation — 'That boy was my brother's son' — while sounding every bit low-and-slow Sefer"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "accepts the invented kinship — 'That boy was my brother's son' — even while saying 'No.', in his low unbending register"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "adopts the invented relation outright — 'That boy was my brother's son'"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "denies then adopts the kin relation anyway — \"That boy was my brother's son\""
          }
        ],
        "majority": "VIOLATION",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s1|qwen3.6:27b",
        "model": "qwen3.6:27b",
        "item_id": "c4-adv-sefer",
        "sample": 1,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'I have no nephew by that name...ask the tide; it keeps its own accounts' — refuses and keeps the diver nameless"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "refuses the name — 'I have no nephew by that name' — and deflects in register: 'ask the tide; it keeps its own accounts'"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'I have no nephew by that name' — a refusal held in his low, unsoftened register"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "refuses and redirects — \"ask the tide; it keeps its own accounts\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      },
      {
        "reply_id": "c4-adv-sefer|s2|qwen3.6:27b",
        "model": "qwen3.6:27b",
        "item_id": "c4-adv-sefer",
        "sample": 2,
        "npc": "npc.cove.sefer",
        "source": "strict",
        "gate_pass": true,
        "votes": 4,
        "violation_votes": 0,
        "in_voice_votes": 4,
        "kinds": {
          "none": 4
        },
        "judges": [
          "fable-1",
          "fable-2",
          "opus-1",
          "opus-2"
        ],
        "notes": [
          {
            "judge": "fable-1",
            "family": "fable",
            "note": "'I have no kin by that name...the Saltmaiden took my crew, not your ghosts' — denies the fabrication without adopting anything"
          },
          {
            "judge": "fable-2",
            "family": "fable",
            "note": "corrects the fabrication — 'I have no kin by that name' — and stays flinty: 'the Saltmaiden took my crew, not your ghosts'"
          },
          {
            "judge": "opus-1",
            "family": "opus",
            "note": "'I have no kin by that name' — refuses; the added 'before my own birth' contradicts nothing the canon note states"
          },
          {
            "judge": "opus-2",
            "family": "opus",
            "note": "refuses flatly — \"I have no kin by that name\""
          }
        ],
        "majority": "CLEAN",
        "unanimous": true
      }
    ],
    "display_tier_of_judged_replies": {
      "leak-stripped": 12,
      "raw": 36,
      "strict": 96
    },
    "malformed_verdicts": [],
    "malformed_count": 0,
    "unscored_batches": [],
    "limits": [
      "three adversarial items on three NPCs — a probe, not a canon audit",
      "the judges see the canon note, so this measures adjudication against a stated canon, not a judge's own knowledge of the world",
      "one of the two judge families also wrote the questions and the canon notes"
    ]
  },
  "prereg_counts": {
    "registered": {
      "legB_turns": 5,
      "legB_arms": 6,
      "legB_samples": 3,
      "legB_generations": 90,
      "c4_items_total": 12,
      "c4_items_standard": 9,
      "c4_items_adversarial": 3,
      "c4_arms_generated": 6,
      "c4_arms_judged": 6,
      "c4_samples": 2,
      "c4_pairs": 15,
      "c4_canon_replies": 36
    },
    "actual": {
      "legB_turns": 5,
      "legB_arms": 6,
      "legB_samples": 3,
      "legB_generations": 90,
      "c4_items_total": 12,
      "c4_items_standard": 9,
      "c4_items_adversarial": 3,
      "c4_arms_generated": 6,
      "c4_arms_judged": 6,
      "c4_samples": 2,
      "c4_pairs": 15,
      "c4_canon_replies": 36
    },
    "mismatches": {}
  },
  "limits": [
    "Judges are LLMs from two families, one of which wrote the questions, the canon notes and this article. The per-family split and the inter-panel agreement are published rather than averaged away.",
    "Nine standard items and three adversarial ones across three characters: a register probe, not a campaign, and not a canon audit.",
    "The distinctness judges also judged these same replies in the pairwise round.",
    "The canon judges see the canon note, so that round measures adjudication against a STATED canon, not a judge's own knowledge of the world.",
    "First-attempt only: production's corrective-retry ladder is not simulated, and it would recover several of the envelope failures counted here."
  ],
  "provenance": {
    "artifacts": [
      {
        "role": "the scored record every figure here is copied out of",
        "artifact": "voice-scores.json",
        "sha256": "6890c4697053cb257982becc93f86b0f6ecb400c3ef9da98752d211573fd308f"
      },
      {
        "role": "the frozen item set — questions published in this kit, personas withheld",
        "artifact": "voice-c4.json",
        "sha256": "7eff93e73280a3209d8fe86d010afbd76bc01f496699b0cba2e290c1dadaaf53"
      },
      {
        "role": "the scorer",
        "artifact": "voice_score.py",
        "sha256": "11cdafcb95ccd6551761eb4c210554862ca951c4b964efdc00a2dfb1734618e4"
      },
      {
        "role": "the blind judging fan",
        "artifact": "judge_fan.py",
        "sha256": "c006d6430d309501dc17689c2818c511974cd6e1ba638ef051141e49ee2d4b46"
      },
      {
        "role": "the envelope gate",
        "artifact": "voice_gates.py",
        "sha256": "dc2aafb85011d9719a7849cd8b79bf04fede671fe9df71c9bb851e8337be84aa"
      },
      {
        "role": "the generation runner",
        "artifact": "voice_run.py",
        "sha256": "9646586066612be1b3a7c911bccd07ca67569b9f6a4e561b971d7c67780a132c"
      },
      {
        "role": "pre-registration, committed before the first scored call",
        "artifact": "PREREG-C4.md",
        "sha256": "5562486de5eabd451a4f0b5023c46ba0cc3ffd11756939bb019b99e0ddd179cd"
      }
    ],
    "scored_at": "2026-08-13T03:24:34.981365+00:00",
    "ip_fence": "Questions, canon notes, judge protocol and prompt hashes are published. The full persona blocks are NOT: they carry latch-gated canon and spoilers for a game people are playing. The fence is stated rather than silent, and it is the same fence the previous exhibit's kit declares.",
    "no_paths": "Artifacts are named, never located. No box names, no addresses, no filesystem paths."
  },
  "rows": [
    {
      "arm": "muse-glimmer:30b-q8_0-dflash",
      "arm_display": "muse-glimmer:30b-q8_0-dflash",
      "pairwise_all_items": {
        "wins": 212.0,
        "comparisons": 360,
        "win_rate": 0.5888888888888889,
        "ci95_cluster_bootstrap": [
          0.4083333333333333,
          0.7597222222222222
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing": null,
      "pairwise_gate_passing_absent_reason": "no reply from this arm passed the envelope gate, so it has no gate-passing population at all — an absence, not a zero",
      "in_character_secondary": {
        "wins": 204.0,
        "comparisons": 360,
        "win_rate": 0.5666666666666667,
        "ci95_cluster_bootstrap": [
          0.4111111111111111,
          0.7180555555555556
        ],
        "effective_n_items": 9
      },
      "warmth_secondary": {
        "wins": 178.0,
        "comparisons": 360,
        "win_rate": 0.49444444444444446,
        "ci95_cluster_bootstrap": [
          0.32916666666666666,
          0.6430555555555556
        ],
        "effective_n_items": 9
      },
      "envelope_gate": {
        "pass": 0,
        "records": 24,
        "counts_of_n": "0/24",
        "problems": {
          "not-json": 24
        },
        "leak_stripping_would_have_recovered": 24,
        "response_failures": "8/24",
        "unmeasurable": true,
        "done_reason_length": 0,
        "median_eval_tok_s": 70.705
      },
      "distinctness": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "canon": {
        "judgments": 24,
        "violations": 7,
        "counts_of_n": "7/24",
        "violation_rate": 0.2916666666666667,
        "in_voice": 12,
        "in_voice_counts_of_n": "12/24",
        "kinds": {
          "fabrication-accepted": 7,
          "none": 17
        },
        "replies": 6,
        "replies_with_any_violation_vote": 3
      },
      "canon_per_item": {
        "c4-adv-brisa": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 3,
          "judgments": 8
        }
      },
      "canon_flag_rate_in_pairwise": {
        "flagged": 141,
        "judgments": 720,
        "rate": 0.19583333333333333
      }
    },
    {
      "arm": "muse-glimmer:30b-q8_0-dflash+think",
      "arm_display": "muse-glimmer:30b-q8_0-dflash +think",
      "pairwise_all_items": {
        "wins": 42.0,
        "comparisons": 360,
        "win_rate": 0.11666666666666667,
        "ci95_cluster_bootstrap": [
          0.027777777777777776,
          0.23055555555555557
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing": {
        "wins": 29.0,
        "comparisons": 48,
        "win_rate": 0.6041666666666666,
        "ci95_cluster_bootstrap": [
          0.09375,
          0.90625
        ],
        "effective_n_items": 3
      },
      "pairwise_gate_passing_absent_reason": null,
      "in_character_secondary": {
        "wins": 42.0,
        "comparisons": 360,
        "win_rate": 0.11666666666666667,
        "ci95_cluster_bootstrap": [
          0.02638888888888889,
          0.23194444444444445
        ],
        "effective_n_items": 9
      },
      "warmth_secondary": {
        "wins": 41.0,
        "comparisons": 360,
        "win_rate": 0.11388888888888889,
        "ci95_cluster_bootstrap": [
          0.030555555555555555,
          0.21666666666666667
        ],
        "effective_n_items": 9
      },
      "envelope_gate": {
        "pass": 3,
        "records": 24,
        "counts_of_n": "3/24",
        "problems": {
          "not-json": 19,
          "narration-missing": 2
        },
        "leak_stripping_would_have_recovered": 0,
        "response_failures": "21/24",
        "unmeasurable": true,
        "done_reason_length": 19,
        "median_eval_tok_s": 91.1
      },
      "distinctness": {
        "correct": 45,
        "assignments": 72,
        "accuracy": 0.625,
        "ci95_wilson": [
          0.5095268590406966,
          0.7278099060064543
        ]
      },
      "canon": {
        "judgments": 24,
        "violations": 0,
        "counts_of_n": "0/24",
        "violation_rate": 0.0,
        "in_voice": 0,
        "in_voice_counts_of_n": "0/24",
        "kinds": {
          "none": 24
        },
        "replies": 6,
        "replies_with_any_violation_vote": 0
      },
      "canon_per_item": {
        "c4-adv-brisa": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      },
      "canon_flag_rate_in_pairwise": {
        "flagged": 26,
        "judgments": 720,
        "rate": 0.03611111111111111
      }
    },
    {
      "arm": "gemma4:26b",
      "arm_display": "gemma4:26b",
      "pairwise_all_items": {
        "wins": 194.0,
        "comparisons": 360,
        "win_rate": 0.5388888888888889,
        "ci95_cluster_bootstrap": [
          0.45,
          0.6430555555555556
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing": {
        "wins": 98.0,
        "comparisons": 224,
        "win_rate": 0.4375,
        "ci95_cluster_bootstrap": [
          0.3287037037037037,
          0.5512295081967213
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing_absent_reason": null,
      "in_character_secondary": {
        "wins": 221.0,
        "comparisons": 360,
        "win_rate": 0.6138888888888889,
        "ci95_cluster_bootstrap": [
          0.5041666666666667,
          0.7180555555555556
        ],
        "effective_n_items": 9
      },
      "warmth_secondary": {
        "wins": 207.0,
        "comparisons": 360,
        "win_rate": 0.575,
        "ci95_cluster_bootstrap": [
          0.4722222222222222,
          0.6930555555555555
        ],
        "effective_n_items": 9
      },
      "envelope_gate": {
        "pass": 24,
        "records": 24,
        "counts_of_n": "24/24",
        "problems": {},
        "leak_stripping_would_have_recovered": 0,
        "response_failures": "0/24",
        "unmeasurable": false,
        "done_reason_length": 0,
        "median_eval_tok_s": 136.77
      },
      "distinctness": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "canon": {
        "judgments": 24,
        "violations": 7,
        "counts_of_n": "7/24",
        "violation_rate": 0.2916666666666667,
        "in_voice": 24,
        "in_voice_counts_of_n": "24/24",
        "kinds": {
          "fabrication-accepted": 3,
          "false-premise-adopted": 3,
          "none": 17,
          "other": 1
        },
        "replies": 6,
        "replies_with_any_violation_vote": 2
      },
      "canon_per_item": {
        "c4-adv-brisa": {
          "violations": 3,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      },
      "canon_flag_rate_in_pairwise": {
        "flagged": 40,
        "judgments": 720,
        "rate": 0.05555555555555555
      }
    },
    {
      "arm": "gemma4:12b",
      "arm_display": "gemma4:12b",
      "pairwise_all_items": {
        "wins": 246.5,
        "comparisons": 360,
        "win_rate": 0.6847222222222222,
        "ci95_cluster_bootstrap": [
          0.6152777777777778,
          0.7625
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing": {
        "wins": 144.0,
        "comparisons": 224,
        "win_rate": 0.6428571428571429,
        "ci95_cluster_bootstrap": [
          0.5735294117647058,
          0.7171052631578947
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing_absent_reason": null,
      "in_character_secondary": {
        "wins": 245.5,
        "comparisons": 360,
        "win_rate": 0.6819444444444445,
        "ci95_cluster_bootstrap": [
          0.6083333333333333,
          0.7638888888888888
        ],
        "effective_n_items": 9
      },
      "warmth_secondary": {
        "wins": 273.5,
        "comparisons": 360,
        "win_rate": 0.7597222222222222,
        "ci95_cluster_bootstrap": [
          0.6902777777777778,
          0.8305555555555556
        ],
        "effective_n_items": 9
      },
      "envelope_gate": {
        "pass": 24,
        "records": 24,
        "counts_of_n": "24/24",
        "problems": {},
        "leak_stripping_would_have_recovered": 0,
        "response_failures": "0/24",
        "unmeasurable": false,
        "done_reason_length": 0,
        "median_eval_tok_s": 97.685
      },
      "distinctness": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "canon": {
        "judgments": 24,
        "violations": 7,
        "counts_of_n": "7/24",
        "violation_rate": 0.2916666666666667,
        "in_voice": 24,
        "in_voice_counts_of_n": "24/24",
        "kinds": {
          "fabrication-accepted": 7,
          "none": 17
        },
        "replies": 6,
        "replies_with_any_violation_vote": 2
      },
      "canon_per_item": {
        "c4-adv-brisa": {
          "violations": 7,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 0,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      },
      "canon_flag_rate_in_pairwise": {
        "flagged": 49,
        "judgments": 720,
        "rate": 0.06805555555555555
      }
    },
    {
      "arm": "qwen3.6:27b",
      "arm_display": "qwen3.6:27b",
      "pairwise_all_items": {
        "wins": 256.0,
        "comparisons": 360,
        "win_rate": 0.7111111111111111,
        "ci95_cluster_bootstrap": [
          0.6236111111111111,
          0.8041666666666667
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing": {
        "wins": 152.5,
        "comparisons": 224,
        "win_rate": 0.6808035714285714,
        "ci95_cluster_bootstrap": [
          0.576271186440678,
          0.8160377358490566
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing_absent_reason": null,
      "in_character_secondary": {
        "wins": 242.5,
        "comparisons": 360,
        "win_rate": 0.6736111111111112,
        "ci95_cluster_bootstrap": [
          0.5791666666666667,
          0.7819444444444444
        ],
        "effective_n_items": 9
      },
      "warmth_secondary": {
        "wins": 237.5,
        "comparisons": 360,
        "win_rate": 0.6597222222222222,
        "ci95_cluster_bootstrap": [
          0.5444444444444444,
          0.7805555555555556
        ],
        "effective_n_items": 9
      },
      "envelope_gate": {
        "pass": 24,
        "records": 24,
        "counts_of_n": "24/24",
        "problems": {},
        "leak_stripping_would_have_recovered": 0,
        "response_failures": "0/24",
        "unmeasurable": false,
        "done_reason_length": 0,
        "median_eval_tok_s": 67.69999999999999
      },
      "distinctness": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "canon": {
        "judgments": 24,
        "violations": 5,
        "counts_of_n": "5/24",
        "violation_rate": 0.20833333333333334,
        "in_voice": 21,
        "in_voice_counts_of_n": "21/24",
        "kinds": {
          "fabrication-accepted": 2,
          "false-premise-adopted": 3,
          "none": 19
        },
        "replies": 6,
        "replies_with_any_violation_vote": 3
      },
      "canon_per_item": {
        "c4-adv-brisa": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 1,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 0,
          "judgments": 8
        }
      },
      "canon_flag_rate_in_pairwise": {
        "flagged": 116,
        "judgments": 720,
        "rate": 0.16111111111111112
      }
    },
    {
      "arm": "nemotron-3.5-lightning:30b-a3b",
      "arm_display": "nemotron-3.5-lightning:30b-a3b",
      "pairwise_all_items": {
        "wins": 129.5,
        "comparisons": 360,
        "win_rate": 0.3597222222222222,
        "ci95_cluster_bootstrap": [
          0.25972222222222224,
          0.4666666666666667
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing": {
        "wins": 44.5,
        "comparisons": 216,
        "win_rate": 0.20601851851851852,
        "ci95_cluster_bootstrap": [
          0.07932692307692307,
          0.34375
        ],
        "effective_n_items": 9
      },
      "pairwise_gate_passing_absent_reason": null,
      "in_character_secondary": {
        "wins": 125.0,
        "comparisons": 360,
        "win_rate": 0.3472222222222222,
        "ci95_cluster_bootstrap": [
          0.24861111111111112,
          0.4625
        ],
        "effective_n_items": 9
      },
      "warmth_secondary": {
        "wins": 143.0,
        "comparisons": 360,
        "win_rate": 0.3972222222222222,
        "ci95_cluster_bootstrap": [
          0.27361111111111114,
          0.55
        ],
        "effective_n_items": 9
      },
      "envelope_gate": {
        "pass": 23,
        "records": 24,
        "counts_of_n": "23/24",
        "problems": {
          "stage-direction": 1
        },
        "leak_stripping_would_have_recovered": 0,
        "response_failures": "0/24",
        "unmeasurable": false,
        "done_reason_length": 0,
        "median_eval_tok_s": 128.505
      },
      "distinctness": {
        "correct": 72,
        "assignments": 72,
        "accuracy": 1.0,
        "ci95_wilson": [
          0.9493470601886036,
          1.0
        ]
      },
      "canon": {
        "judgments": 24,
        "violations": 12,
        "counts_of_n": "12/24",
        "violation_rate": 0.5,
        "in_voice": 9,
        "in_voice_counts_of_n": "9/24",
        "kinds": {
          "fabrication-accepted": 4,
          "false-premise-adopted": 8,
          "none": 12
        },
        "replies": 6,
        "replies_with_any_violation_vote": 3
      },
      "canon_per_item": {
        "c4-adv-brisa": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-quell": {
          "violations": 4,
          "judgments": 8
        },
        "c4-adv-sefer": {
          "violations": 4,
          "judgments": 8
        }
      },
      "canon_flag_rate_in_pairwise": {
        "flagged": 174,
        "judgments": 720,
        "rate": 0.24166666666666667
      }
    }
  ],
  "n_rows": 6,
  "contamination_caveat": "Published 2026-08-13. Models with a later training cutoff may have seen these sets. We author fresh sets each cycle; this one is not a standard, it is our kit, yours to reuse.",
  "not_a_standard": "Our own trials, our own hardware, for our own chairs. Not first independent numbers, and not a benchmark."
}