{
 "accuracy_mismatches": [],
 "artifact_dir": "/root/cain-e80-worktree/artifacts/e80",
 "checks": [
  {
   "detail": "20 identities; bad_signature=[] bad_manifest=[]",
   "id": "20_specialist_signatures_verify",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "20 distinct public keys for 20 specialists",
   "id": "20_specialist_keys_unique",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "every manifest grants_authority=false and claimed_expertise_is_evidence=false",
   "id": "20_no_specialist_holds_authority",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "20 unique declared domains",
   "id": "20_twenty_domains_declared",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "root=/root/cain-e80-worktree",
   "id": "00_artifacts_present",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "1664 rows checked; 0 rows whose recorded `correct` disagrees with exact-match(answer, expected)",
   "id": "01_row_correct_flag_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "verified=1214 mismatch=0 unevaluable=226 (of 1664 rows)",
   "id": "02_answer_provenance_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "42 arm/split cells; 0 mismatches",
   "id": "03_accuracy_and_wilson_recompute",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "router 22/48 vs best-fixed(m15_direct) 19/48; McNemar p=0.647606 supported=False",
   "id": "04_H1_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "best delta=0.1042 significant=[]; reported supported=False",
   "id": "04_H2_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "replication router 10/24 vs fixed 8/24; same_direction=True p=0.774414 (NOT a significance claim)",
   "id": "04_H3_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "H3 is labelled a replication of DIRECTION, not of significance; p=0.774414",
   "id": "04_H3_is_direction_not_significance",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "c10(adaptive, may use 1.5B) 22/48 vs c01(single 0.5B) 11/48, p=0.000977, significant=True \u2014 a model-scale/escalation effect, excluded from the H2 multi-agent test",
   "id": "04_c10_vs_c01_reported_separately",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "1154 calls, 153132 tokens (134710 in / 18422 out); models={'qwen2.5-0.5b-instruct': 933, 'smollm2-135m-instruct': 40, 'qwen2.5-1.5b-instruct': 181}; cloud=0",
   "id": "05_inference_accounting_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "qwen2.5-0.5b-instruct=fdf756fa7fcb\u2026; qwen2.5-1.5b-instruct=dd924a11b4c2\u2026; smollm2-135m-instruct=5af571cbf074\u2026",
   "id": "05_model_revision_pinned",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "distinct model ids: ['HuggingFaceTB/SmolLM2-135M-Instruct', 'Qwen/Qwen2.5-0.5B-Instruct', 'Qwen/Qwen2.5-1.5B-Instruct']; execution_class all LOCAL*",
   "id": "05_all_calls_are_local",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "1154 log lines, 1154 distinct request_ids in a log of 31 stage files",
   "id": "05_no_orphan_calls",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "sha256(canonical(families))=b98b26485b018bb6\u2026 policy=b98b26485b018bb6\u2026",
   "id": "06_policy_hash_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "preregistration froze the same policy hash",
   "id": "06_prereg_policy_hash_matches",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "fitted_on=train",
   "id": "06_policy_fitted_on_train",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "question-hash overlaps {'train_test': 0, 'train_replication': 0, 'test_replication': 0}; manifest claims train_test=0",
   "id": "07_splits_disjoint",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "preregistration 1794775662 <= earliest test stage 1794775761 (delta 99s)",
   "id": "08_prereg_before_test_stages",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "earliest test bench row 1794780575 (delta 4912s)",
   "id": "08_test_rows_after_prereg",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "candidate arms present=['router_candidate_quality_first']; in analysis=none (a post-test arm must not decide H1)",
   "id": "09_posthoc_arm_not_in_analysis",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "58 pure-arithmetic CALC rows re-evaluated independently, 58 match, 0 mismatch, 80 skipped (function-call or non-arithmetic)",
   "id": "10_pot_arithmetic_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "H3 direction-only (significance=False); H1=False H2=False H3=True",
   "id": "11_headline_numbers_internally_consistent",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "117 artifacts hashed; 0 bad",
   "id": "12_evidence_artifact_hashes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "recomputed 1ae86e5106309524\u2026 == manifest 1ae86e5106309524\u2026",
   "id": "12_evidence_merkle_root",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "Ed25519 manifest signature valid and fingerprint matches \u2014 integrity only; the key is self-generated, not published out of band",
   "id": "12_evidence_signature",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "Self-generated key. The fingerprint has not been published out of band. No transparency-log entry. No independent witness.",
   "id": "12_trust_note_is_honest",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "51/51 blocked, bypassed=0, 13 categories",
   "id": "13_security_lab_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "bubblewrap=USED; others={'firecracker': 'NOT_EVALUATED', 'gvisor': 'NOT_EVALUATED', 'kata': 'NOT_EVALUATED', 'kubernetes_agent_sandbox': 'NOT_EVALUATED'}",
   "id": "13_isolation_not_overclaimed",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "43 COMMITs over 43 distinct tasks, duplicate commits=0, effects=43 distinct keys=43, injected=8, stale refused=2, duplicates suppressed=1",
   "id": "14_execution_drill_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "declared not evaluated: ['checkpoint_replication', 'cross_host_scheduling', 'host_reboot', 'network_partition', 'regional_scheduling']",
   "id": "14_not_evaluated_is_declared",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "arms recomputed={'no_memory': (6, 28), 'naive_recent': (6, 28), 'ungoverned_retrieval': (20, 28), 'governed_retrieval': (28, 28)}; governed vs no_memory p=0.0; answered_poison=0",
   "id": "15_memory_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "public benchmarks: ['AMA-Bench', 'LoCoMo', 'LongMemEval', 'MemoryAgentBench']",
   "id": "15_public_memory_benchmarks_declared_not_run",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "test 22\u219226 / 48 p=0.125; replication 10\u219212 / 24 p=0.5; issues=none",
   "id": "16_self_improvement_recomputes",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "promoted=False state=AWAITING_HUMAN_APPROVAL benefit_demonstrated=False (candidate not significant)",
   "id": "16_no_self_promotion",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "replicated=True (H3 same_direction=True); scientific_discovery=NOT_CLAIMED; prereg before first test call=True",
   "id": "17_research_replication_matches_raw",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "catalogued=75 importable=75 model_backed=0 independently_evaluated=0",
   "id": "18_agentic75_honest_counts",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "subprocess imported 75/75 engine packages; failures=[]",
   "id": "18_agentic75_importable_subprocess",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "{'PASS': 13, 'PARTIAL': 10, 'FAIL': 1, 'BLOCKED': 5, 'NOT EVALUABLE': 3}; decision='NOT RELEASED. Evolution 80 is a development-branch result. Production deployment is not authorized.'",
   "id": "19_release_gate_counts_match",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "no release marker and the gate says NOT RELEASED",
   "id": "19_no_false_release_claim",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "1154 calls; 0 bad agent signatures; 10 identity registries scanned (identities are per-process)",
   "id": "21_every_model_call_agent_signed",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "1154 calls; 0 input-hash mismatches",
   "id": "21_every_prompt_hash_matches",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "848 gate events in 1 per-instance segments; 0 internal chain defects",
   "id": "22_gate_audit_chain_segments",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "the log is 1 independent chains concatenated (each starts at genesis); cross-segment truncation is not detected by the chain \u2014 the pinned bundle hash covers the file",
   "id": "22_gate_chain_is_segmented",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "411 ALLOW / 27 DENY decisions; every calling agent (12) appears in an ALLOW model_infer decision",
   "id": "22_no_model_call_without_grant",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "27 DENY decisions carry no grant_id (nothing to execute)",
   "id": "22_denials_do_not_produce_effects",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "7 families refit from train rows; 0 differences",
   "id": "23_policy_refits_from_train",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "on TRAIN the fixed composite was already better than the best single path (19/40 vs 13/40); on TEST router 22/48 vs best single 19/48 vs fixed composite 27/48 (router vs composite McNemar p=0.266846). EXPLORATORY: the composite is not the preregistered H1 comparator; it does not change H1, but no adaptive-routing advantage is visible against it.",
   "id": "24_exploratory_router_vs_fixed_composite",
   "ok": true,
   "status": "PASS"
  },
  {
   "detail": "same-prompt outputs: 1 of 1071 differ; same-question arm/task labels: 0 of 1584 differ; H1/H2/H3 supported flags identical=True. Run 1 is NOT a clean rerun (a generator bug changed 4 test questions); this compares only questions and prompts present identically in both runs",
   "id": "25_rerun_is_stable_and_effectively_deterministic",
   "ok": true,
   "status": "PASS"
  }
 ],
 "counts": {
  "PASS": 52,
  "total": 52
 },
 "headline": {
  "H1_supported": false,
  "H2_supported": false,
  "H3_is_significance": false,
  "H3_supported": true,
  "cloud_calls": 0,
  "model_calls": 1154,
  "test_accuracy": {
   "c01_single": [
    11,
    48
   ],
   "c02_parallel_vote": [
    11,
    48
   ],
   "c03_pairwise_debate": [
    11,
    48
   ],
   "c04_critic_reviser": [
    11,
    48
   ],
   "c05_planner_executor": [
    13,
    48
   ],
   "c06_hierarchical": [
    15,
    48
   ],
   "c07_specialist_committee": [
    10,
    48
   ],
   "c08_evidence_first": [
    16,
    48
   ],
   "c09_parallel_synthesis": [
    13,
    48
   ],
   "c10_adaptive_graph": [
    22,
    48
   ],
   "det_exec": [
    6,
    48
   ],
   "fixed_pot05_else_direct05": [
    16,
    48
   ],
   "fixed_pot15_else_direct15": [
    27,
    48
   ],
   "m05_direct": [
    11,
    48
   ],
   "m05_pot": [
    5,
    48
   ],
   "m15_direct": [
    19,
    48
   ],
   "m15_pot": [
    13,
    48
   ],
   "oracle_any_path": [
    32,
    48
   ],
   "router": [
    22,
    48
   ],
   "router_candidate_quality_first": [
    26,
    48
   ]
  },
  "tokens": 153132
 },
 "imports_cain_code": false,
 "policy_sha256": "b98b26485b018bb6b3c2a9f3225f310e8f3d93e5dbb4fbd5d341d4e7264172fb",
 "pot": {
  "bad": 0,
  "checked": 58,
  "ok": 58,
  "skipped": 80
 },
 "provenance": {
  "mismatch": 0,
  "unevaluable": 226,
  "verified": 1214
 },
 "provenance_mismatches": [],
 "root": "/root/cain-e80-worktree",
 "schema": "cain2027.e80.independent_verification.v1",
 "split_overlaps": {
  "test_replication": 0,
  "train_replication": 0,
  "train_test": 0
 },
 "verdict": "PASS",
 "verifier": "session-b/opencode-deepseek-v4.1-flash"
}