{
  "schema_version": 1,
  "audit_date_utc": "2026-10-06",
  "evidence_kind": "post-hoc alternative endpoint audit of retained historical outputs; not new generation or causal improvement",
  "source_aggregate_sha256": "ccb127d738c6c21cd20b479c68fb3aeccf6096b23eda8a7495dc6c9b4790cbd7",
  "root_reconciliation_sha256": "c955af70673e675e1f83418fdedea73a6aba692b10908d89389d05ed6ed492d1",
  "runtime_source_sha256": {
    "PROTOCOL.md": "678ee0e03f6caccfde557b5cba5433f1f6abd7f4172a9466f1f51c1715cb02a8",
    "audit.py": "8a3377d5cee5c7c684a94d47223b350ecf6b76c64c43e15ddaab131d6773f9a9",
    "supervisor.py": "324dba8040255664bf3d2ced63c61833649db4c99ba161004db0d35820f73211",
    "test_guards.py": "af88ee8b39c4ff73cd0707c7e89299d7fe91d8d22e27ac69d228675221fa94fa",
    "test_integration.py": "73efa3de37b522b8670c619c57e6d8d932e8d44cd55892cb6ad77dcac3b33097"
  },
  "independent_metadata_reconciliation": {
    "root_replay_receipt_sha256": "0b2b570fbc3fa95cfb19b7b646f8a5f541eea75ae00f388f2bba6a58d768f4c6",
    "independent_result_sha256": "9174779d1d800c7b2938e41843df18f85432a4bf9861288c1d31d446cc2a4de1",
    "scope": "source/metadata/1604journal/rate arithmetic; not second raw-prediction replay or external peer review"
  },
  "native_barrier_slots": 800,
  "ordered_stage_events_verified": 1604,
  "endpoint_conditions": {
    "exact": "Strict JSON on exact stored output with surrounding whitespace stripped only.",
    "terminal_eos": "Remove exactly one final literal <eos> only when recorded stop ID is 1; otherwise retain exact stored output. No fences/interior/duplicate markers or commentary repaired."
  },
  "argument_comparison": "Recursive typed equality, bool distinct from number; finite int/float numerical equivalence; legacy comparator reported separately.",
  "audit": {
    "schema_version": 1,
    "plan_sha256": "45b1fe0db19db6c1cf22c82eb237832126a1275a2c58cbe3b74b82a11bc6ea6c",
    "native_replay_slots": 800,
    "groups": {
      "base": {
        "call": {
          "denominator": 360,
          "native_correct": 228,
          "exact": {
            "valid": 0,
            "typed_correct": 0,
            "legacy_correct": 0,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 0,
              "json_only": 0,
              "native_only": 228,
              "both_incorrect": 132
            },
            "typed_accuracy": 0.0,
            "legacy_accuracy": 0.0
          },
          "terminal_eos": {
            "valid": 0,
            "typed_correct": 0,
            "legacy_correct": 0,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 0,
              "json_only": 0,
              "native_only": 228,
              "both_incorrect": 132
            },
            "typed_accuracy": 0.0,
            "legacy_accuracy": 0.0
          },
          "failures": 0,
          "unrun": 0,
          "status_counts": {
            "completed": 360
          },
          "truncations": 0,
          "stop_mismatches": 0,
          "native_accuracy": 0.6333333333333333
        },
        "no_call": {
          "denominator": 40,
          "native_correct": 0,
          "exact": {
            "valid": 0,
            "typed_correct": 0,
            "legacy_correct": 0,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 0,
              "json_only": 0,
              "native_only": 0,
              "both_incorrect": 40
            },
            "typed_accuracy": 0.0,
            "legacy_accuracy": 0.0
          },
          "terminal_eos": {
            "valid": 0,
            "typed_correct": 0,
            "legacy_correct": 0,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 0,
              "json_only": 0,
              "native_only": 0,
              "both_incorrect": 40
            },
            "typed_accuracy": 0.0,
            "legacy_accuracy": 0.0
          },
          "failures": 0,
          "unrun": 0,
          "status_counts": {
            "completed": 40
          },
          "truncations": 0,
          "stop_mismatches": 0,
          "native_accuracy": 0.0
        }
      },
      "checkpoint600": {
        "call": {
          "denominator": 360,
          "native_correct": 0,
          "exact": {
            "valid": 0,
            "typed_correct": 0,
            "legacy_correct": 0,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 0,
              "json_only": 0,
              "native_only": 0,
              "both_incorrect": 360
            },
            "typed_accuracy": 0.0,
            "legacy_accuracy": 0.0
          },
          "terminal_eos": {
            "valid": 250,
            "typed_correct": 159,
            "legacy_correct": 159,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 0,
              "json_only": 159,
              "native_only": 0,
              "both_incorrect": 201
            },
            "typed_accuracy": 0.44166666666666665,
            "legacy_accuracy": 0.44166666666666665
          },
          "failures": 0,
          "unrun": 0,
          "status_counts": {
            "completed": 360
          },
          "truncations": 8,
          "stop_mismatches": 0,
          "native_accuracy": 0.0
        },
        "no_call": {
          "denominator": 40,
          "native_correct": 40,
          "exact": {
            "valid": 0,
            "typed_correct": 0,
            "legacy_correct": 0,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 0,
              "json_only": 0,
              "native_only": 40,
              "both_incorrect": 0
            },
            "typed_accuracy": 0.0,
            "legacy_accuracy": 0.0
          },
          "terminal_eos": {
            "valid": 39,
            "typed_correct": 36,
            "legacy_correct": 36,
            "disagreements": 0,
            "paired_native": {
              "both_correct": 36,
              "json_only": 0,
              "native_only": 4,
              "both_incorrect": 0
            },
            "typed_accuracy": 0.9,
            "legacy_accuracy": 0.9
          },
          "failures": 0,
          "unrun": 0,
          "status_counts": {
            "completed": 40
          },
          "truncations": 0,
          "stop_mismatches": 0,
          "native_accuracy": 1.0
        }
      }
    },
    "rates": {
      "base": {
        "native": {
          "macro_accuracy": 0.31666666666666665,
          "pooled_accuracy": 0.57
        },
        "exact": {
          "typed": {
            "macro_accuracy": 0.0,
            "pooled_accuracy": 0.0
          },
          "legacy": {
            "macro_accuracy": 0.0,
            "pooled_accuracy": 0.0
          }
        },
        "terminal_eos": {
          "typed": {
            "macro_accuracy": 0.0,
            "pooled_accuracy": 0.0
          },
          "legacy": {
            "macro_accuracy": 0.0,
            "pooled_accuracy": 0.0
          }
        }
      },
      "checkpoint600": {
        "native": {
          "macro_accuracy": 0.5,
          "pooled_accuracy": 0.1
        },
        "exact": {
          "typed": {
            "macro_accuracy": 0.0,
            "pooled_accuracy": 0.0
          },
          "legacy": {
            "macro_accuracy": 0.0,
            "pooled_accuracy": 0.0
          }
        },
        "terminal_eos": {
          "typed": {
            "macro_accuracy": 0.6708333333333334,
            "pooled_accuracy": 0.4875
          },
          "legacy": {
            "macro_accuracy": 0.6708333333333334,
            "pooled_accuracy": 0.4875
          }
        }
      }
    },
    "commitments": {
      "base": {
        "launch_plan": "08edd38fdc79d861c3fef2b1c7d666901b73ec22e0b8fd458eeaa1a39f5c9dda",
        "summary": "9243f75d44049c7d52c51a93c2a618d80e7afa954b0f51dcd74f348e9a225250",
        "terminal": "e201412006673b073a64772dadafe5407eeb3c3e4600090d92a10831ec5685db",
        "events": "ecb861128670585f5801bf4d947fb609953a3321d6ba20a0a10cb64e3e494686",
        "root_reconciliation": "ba9ab668533273718526bfd5ea04bc2f5897075e8f4f027555e71a995763006f"
      },
      "checkpoint600": {
        "launch_plan": "9afe07869ecbc4fb298454f3a81fef4cab23b03d993e62414ce836c9670396dd",
        "summary": "7df807d373665777e3d9e1df611a424a3dfe61b7b051231ce412072197524a1d",
        "terminal": "139236fe627ea9528f27ea58bd942218f110da0452124139f52ff027a8104a75",
        "events": "abe4b58225be5a1ad1adbf570925a2f6d263864634d17399d7534a1021aefb75",
        "root_reconciliation": "c23f4780ce047948f5905fa69d6cc610d7a949a2dc4c6aee4ce1560f51c1669c"
      }
    },
    "raw_text_published": false,
    "historical_native_labels_unchanged": true,
    "primary_argument_comparison": "typed",
    "sensitivity": "legacy permissive numeric/bool equality",
    "external_usd": 0
  },
  "limits": [
    "Original native labels and tables unchanged; original selected SFT 1.7B checkpoint remains unidentified and unrelated to this FunctionGemma endpoint audit.",
    "Historical exposed prefix, five known training-content overlaps, eight earlier probe exposures and lossy native schemas remain.",
    "Post-hoc alternate scoring endpoint; no new generation, causal improvement, equivalent full-schema comparison or general superiority inference.",
    "No-call absence/empty list is not semantic refusal quality; the historical cause remains unproved.",
    "Same-owner AI-assisted source/arithmetic checks are not outside peer review or independent reproduction.",
    "Aggregates/hashes do not permit publicly available item-level correctness rescoring; raw responses/references remain private."
  ],
  "raw_outputs_published": false,
  "additional_paid_usd": 0
}
