{
  "acceptance_MPS": {
    "bootstrap_draws": 5000,
    "bootstrap_seed": 42,
    "cluster_bootstrap_95_percent": [
      0.465,
      0.56
    ],
    "cluster_count": 300,
    "improvement_count": 312,
    "models": {
      "c-1213": {
        "document_field_error_counts": {
          "all_invented": 3,
          "all_omitted": 2,
          "invalid_document_plan": 1,
          "limit_changed": 1,
          "order_changed": 15,
          "scope_changed": 1,
          "types_omitted": 1
        },
        "empty_n": 60,
        "empty_passed": 58,
        "false_acceptances": 0,
        "false_refusals": 0,
        "generation_ms_median_p95": [
          583.7723544682376,
          983.8346101169006
        ],
        "n": 600,
        "outcomes": 575,
        "plans": 575,
        "strata": {
          "boolean": 99,
          "capability": 100,
          "credentials": 83,
          "filename": 96,
          "scope_format": 98,
          "time_order": 99
        },
        "synthetic_fixture_execution_ms_median_p95": [
          3.5388120159041137,
          35.2817041130038
        ]
      },
      "incumbent-p0": {
        "document_field_error_counts": {
          "all_invented": 23,
          "all_omitted": 6,
          "any_invented": 6,
          "any_omitted": 6,
          "credential_changed": 22,
          "exclude_invented": 5,
          "exclude_omitted": 1,
          "exclude_paths_invented": 2,
          "exclude_paths_omitted": 8,
          "invalid_document_plan": 35,
          "order_changed": 1,
          "scope_changed": 41,
          "time_changed": 5,
          "types_invented": 34,
          "types_omitted": 39,
          "wrong_output_mode": 166
        },
        "empty_n": 60,
        "empty_passed": 26,
        "false_acceptances": 11,
        "false_refusals": 104,
        "generation_ms_median_p95": [
          540.1951874955557,
          1000.8711083355593
        ],
        "n": 600,
        "outcomes": 267,
        "plans": 242,
        "strata": {
          "boolean": 43,
          "capability": 63,
          "credentials": 5,
          "filename": 88,
          "scope_format": 44,
          "time_order": 24
        },
        "synthetic_fixture_execution_ms_median_p95": [
          3.4563539957161993,
          39.59561458905227
        ]
      },
      "incumbent-p1": {
        "document_field_error_counts": {
          "all_invented": 39,
          "all_omitted": 10,
          "any_invented": 5,
          "any_omitted": 6,
          "credential_changed": 33,
          "exclude_invented": 6,
          "exclude_omitted": 1,
          "exclude_paths_invented": 3,
          "exclude_paths_omitted": 9,
          "invalid_document_plan": 22,
          "limit_changed": 1,
          "order_changed": 5,
          "scope_changed": 49,
          "time_changed": 5,
          "types_invented": 52,
          "types_omitted": 38,
          "wrong_output_mode": 132
        },
        "empty_n": 60,
        "empty_passed": 32,
        "false_acceptances": 6,
        "false_refusals": 79,
        "generation_ms_median_p95": [
          555.6130625191145,
          1038.996921115904
        ],
        "n": 600,
        "outcomes": 305,
        "plans": 275,
        "strata": {
          "boolean": 52,
          "capability": 74,
          "credentials": 10,
          "filename": 90,
          "scope_format": 42,
          "time_order": 37
        },
        "synthetic_fixture_execution_ms_median_p95": [
          3.452375007327646,
          39.26825032394844
        ]
      }
    },
    "paired_gain": 0.5133333333333333,
    "regression_count": 4
  },
  "acceptance_board_sha256": "fbb5e57c06cab2dbeb9b33b395a640b2c0f0c05329b12d0308435c1a9dd94cb1",
  "acceptance_failure_analysis": {
    "boolean": {
      "field_error_counts": {
        "scope_changed": 1
      },
      "outcome_failures": 1
    },
    "capability": {
      "field_error_counts": {},
      "outcome_failures": 0
    },
    "credentials": {
      "field_error_counts": {
        "all_invented": 1,
        "limit_changed": 1,
        "order_changed": 15,
        "types_omitted": 1
      },
      "outcome_failures": 17
    },
    "filename": {
      "field_error_counts": {},
      "outcome_failures": 4
    },
    "scope_format": {
      "field_error_counts": {
        "all_invented": 1,
        "all_omitted": 1,
        "invalid_document_plan": 1
      },
      "outcome_failures": 2
    },
    "time_order": {
      "field_error_counts": {
        "all_invented": 1,
        "all_omitted": 1
      },
      "outcome_failures": 1
    }
  },
  "acceptance_gate_failures": [
    "credentials: 83/100, requires 85/100"
  ],
  "acceptance_opened": true,
  "acceptance_thresholds": {
    "cases": 600,
    "maximum_false_acceptances": 0,
    "outcomes": 540,
    "per_stratum": 85,
    "positive_paired_gain_required": true
  },
  "application_deployed": false,
  "c_checkpoint_timing_tradeoff": "C middle scores 1183/1280 and C final 1183/1280. Their retention scores are 139/160 and 132/160; development scores are 572/600 and 576/600. The operator explicitly preferred C final as the default. Equal totals do not mean identical capabilities.",
  "checkpoint_sha256": {
    "a-1213": "2b70ca3f912822ffb8a976fe6fb1e2d093b67e03b5048457f6b1f779da2b110c",
    "a-303": "d3bcd9aa0a5273e36aff0267f9493fc4398e25ffbe225f8430e00796a233edb0",
    "a-607": "1b1853a956694289616c5861073b0d285fadf25d9419b407ec4cd06b5adc9e36",
    "b-1213": "615cfa626204f4ea0180ed680ef0789e28d89fdd4862319a8e54f2f8aec19271",
    "b-303": "31d11d248992cd6e7b180c5b60940696324592f606a319d4f3f990e379b473e6",
    "b-607": "f578dfd56e6eeeb1bdafec674065d52c66351b9edc4ad2fdcf3599ff68e97214",
    "c-1213": "1a398f8880b23a18da9e524ceb0136851ec2215eb6a066c25d425fdab5cd7169",
    "c-303": "2e9d6be577484b102501f12953324c726fcaed9cff773a1ae589b9eba4c6b1e9",
    "c-607": "93ca32d330fe00ea550b530606491fcab1afba6f0479be33edea560c66ee3eba",
    "d-1213": "a7f94f4e27f12667d692f48546bec69b46f3aca661a4a3fd70e17c978e49d253",
    "d-303": "d16942be26032270da5760346d962a968e9d33aa4ba78e00427fcc9a650fea51",
    "d-607": "fed47a0415e70f332f72f38e79c18ada4da2a0f7564c940371ac64a1c1c3aa2f"
  },
  "conclusion": "c-1213 is jeremiah_v1, the operator-selected research incumbent: 1183/1280 shared diagnostic outcomes and 575/600 sealed acceptance outcomes, versus 267/600 for incumbent P0 and 305/600 for incumbent P1. The strict synthetic acceptance gate failed: credentials: 83/100, requires 85/100. Historical regressions remain tradeoffs; application release is pending.",
  "context_sha256": {
    "p0": "78145836417d8c16b3651739f212a2a68dcf9faadaeb1231fe65ce95fb3c4e1f",
    "p1": "529717c6555d64164e307e7e2bed79d8e40d1920541f6dca1b18921df1085b53"
  },
  "credential_followup": "Credentials is the first follow-up priority: 15 of its 17 failed acceptance cases involve a changed ordering field. Use contrastive examples that preserve requested ordering while varying provider and credential wording, and separately exercise limit, all-results and file-type constraints. Field counts can overlap; this is an observed output-error pattern, not a claim about the model's internal reasoning.",
  "diagnostic_provenance_sha256": {
    "board_sha256": "c65b444e276ac9e1fc4eb81b58ce01d48b2bbb3b6c5329102f3e71058bdf390c",
    "broker_sha256": "635f64ccc6d7c288de50e87656da2685dbd87c6607bf9335dbe537ec5aa4af70",
    "dev_sha256": "ccc6d0e9cc3f7c9dc163e11311f893c8d0c6234d2726286b6818df94afe83929",
    "gpu_board_sha256": "20adfe8fd3dd6abb5baa0fdb1fb59f5ac02caa84dd08fc62b3aea5b6b437049f"
  },
  "eligibility": {
    "a-1213": {
      "dev_strata": {
        "boolean": 77,
        "capability": 80,
        "credentials": 21,
        "filename": 81,
        "scope_format": 68,
        "time_order": 65
      },
      "eligible": false,
      "regression_count": 24,
      "transfer_families": {
        "boolean": 40,
        "credentials": 40,
        "dates": 40,
        "dotenv": 40,
        "formats": 39,
        "refusal": 40
      }
    },
    "a-303": {
      "dev_strata": {
        "boolean": 72,
        "capability": 79,
        "credentials": 18,
        "filename": 81,
        "scope_format": 55,
        "time_order": 52
      },
      "eligible": false,
      "regression_count": 37,
      "transfer_families": {
        "boolean": 39,
        "credentials": 39,
        "dates": 40,
        "dotenv": 39,
        "formats": 38,
        "refusal": 40
      }
    },
    "a-607": {
      "dev_strata": {
        "boolean": 77,
        "capability": 80,
        "credentials": 19,
        "filename": 81,
        "scope_format": 69,
        "time_order": 65
      },
      "eligible": false,
      "regression_count": 24,
      "transfer_families": {
        "boolean": 40,
        "credentials": 40,
        "dates": 40,
        "dotenv": 40,
        "formats": 39,
        "refusal": 40
      }
    },
    "b-1213": {
      "dev_strata": {
        "boolean": 46,
        "capability": 75,
        "credentials": 12,
        "filename": 80,
        "scope_format": 33,
        "time_order": 29
      },
      "eligible": false,
      "regression_count": 57,
      "transfer_families": {
        "boolean": 30,
        "credentials": 40,
        "dates": 40,
        "dotenv": 30,
        "formats": 30,
        "refusal": 40
      }
    },
    "b-303": {
      "dev_strata": {
        "boolean": 39,
        "capability": 76,
        "credentials": 7,
        "filename": 80,
        "scope_format": 23,
        "time_order": 29
      },
      "eligible": false,
      "regression_count": 67,
      "transfer_families": {
        "boolean": 27,
        "credentials": 40,
        "dates": 40,
        "dotenv": 23,
        "formats": 29,
        "refusal": 37
      }
    },
    "b-607": {
      "dev_strata": {
        "boolean": 46,
        "capability": 74,
        "credentials": 12,
        "filename": 79,
        "scope_format": 29,
        "time_order": 28
      },
      "eligible": false,
      "regression_count": 60,
      "transfer_families": {
        "boolean": 30,
        "credentials": 40,
        "dates": 40,
        "dotenv": 26,
        "formats": 31,
        "refusal": 39
      }
    },
    "c-1213": {
      "dev_strata": {
        "boolean": 96,
        "capability": 99,
        "credentials": 91,
        "filename": 97,
        "scope_format": 94,
        "time_order": 99
      },
      "eligible": false,
      "regression_count": 45,
      "transfer_families": {
        "boolean": 40,
        "credentials": 39,
        "dates": 40,
        "dotenv": 40,
        "formats": 39,
        "refusal": 40
      }
    },
    "c-303": {
      "dev_strata": {
        "boolean": 96,
        "capability": 95,
        "credentials": 89,
        "filename": 93,
        "scope_format": 95,
        "time_order": 95
      },
      "eligible": false,
      "regression_count": 48,
      "transfer_families": {
        "boolean": 39,
        "credentials": 40,
        "dates": 40,
        "dotenv": 40,
        "formats": 38,
        "refusal": 40
      }
    },
    "c-607": {
      "dev_strata": {
        "boolean": 96,
        "capability": 98,
        "credentials": 87,
        "filename": 97,
        "scope_format": 95,
        "time_order": 99
      },
      "eligible": false,
      "regression_count": 38,
      "transfer_families": {
        "boolean": 40,
        "credentials": 40,
        "dates": 40,
        "dotenv": 40,
        "formats": 39,
        "refusal": 40
      }
    },
    "d-1213": {
      "dev_strata": {
        "boolean": 83,
        "capability": 88,
        "credentials": 62,
        "filename": 87,
        "scope_format": 88,
        "time_order": 82
      },
      "eligible": false,
      "regression_count": 78,
      "transfer_families": {
        "boolean": 31,
        "credentials": 40,
        "dates": 40,
        "dotenv": 20,
        "formats": 30,
        "refusal": 38
      }
    },
    "d-303": {
      "dev_strata": {
        "boolean": 82,
        "capability": 84,
        "credentials": 46,
        "filename": 83,
        "scope_format": 85,
        "time_order": 79
      },
      "eligible": false,
      "regression_count": 108,
      "transfer_families": {
        "boolean": 29,
        "credentials": 39,
        "dates": 40,
        "dotenv": 20,
        "formats": 29,
        "refusal": 36
      }
    },
    "d-607": {
      "dev_strata": {
        "boolean": 84,
        "capability": 86,
        "credentials": 59,
        "filename": 88,
        "scope_format": 88,
        "time_order": 81
      },
      "eligible": false,
      "regression_count": 80,
      "transfer_families": {
        "boolean": 30,
        "credentials": 40,
        "dates": 40,
        "dotenv": 20,
        "formats": 31,
        "refusal": 38
      }
    }
  },
  "estimated_compute_usd": 7.4286,
  "filename-retention_MPS": {
    "bootstrap_draws": 5000,
    "bootstrap_seed": 42,
    "cluster_bootstrap_95_percent": [
      -0.125,
      0.0
    ],
    "cluster_count": 40,
    "improvement_count": 0,
    "models": {
      "c-1213": {
        "document_field_error_counts": {},
        "empty_n": 0,
        "empty_passed": 0,
        "false_acceptances": 0,
        "false_refusals": 0,
        "generation_ms_median_p95": [
          520.725770998979,
          1492.830949954805
        ],
        "n": 40,
        "outcomes": 37,
        "plans": 34,
        "strata": {
          "find_bounded_depth": 5,
          "find_directories": 4,
          "find_empty_result": 2,
          "find_exact_name": 7,
          "find_iname": 5,
          "find_name_patterns_with_pruning": 6,
          "find_recent_mtime": 4,
          "find_size_threshold": 4
        },
        "synthetic_fixture_execution_ms_median_p95": [
          57.40989602054469,
          68.65132645471022
        ]
      },
      "incumbent-p0": {
        "document_field_error_counts": {},
        "empty_n": 0,
        "empty_passed": 0,
        "false_acceptances": 0,
        "false_refusals": 0,
        "generation_ms_median_p95": [
          507.4872710101772,
          1492.2620211495087
        ],
        "n": 40,
        "outcomes": 39,
        "plans": 38,
        "strata": {
          "find_bounded_depth": 6,
          "find_directories": 4,
          "find_empty_result": 3,
          "find_exact_name": 7,
          "find_iname": 5,
          "find_name_patterns_with_pruning": 6,
          "find_recent_mtime": 4,
          "find_size_threshold": 4
        },
        "synthetic_fixture_execution_ms_median_p95": [
          57.90695798350498,
          67.37462944875004
        ]
      },
      "incumbent-p1": {
        "document_field_error_counts": {},
        "empty_n": 0,
        "empty_passed": 0,
        "false_acceptances": 0,
        "false_refusals": 0,
        "generation_ms_median_p95": [
          518.6489379848354,
          1493.8425895787077
        ],
        "n": 40,
        "outcomes": 39,
        "plans": 37,
        "strata": {
          "find_bounded_depth": 6,
          "find_directories": 4,
          "find_empty_result": 3,
          "find_exact_name": 7,
          "find_iname": 5,
          "find_name_patterns_with_pruning": 6,
          "find_recent_mtime": 4,
          "find_size_threshold": 4
        },
        "synthetic_fixture_execution_ms_median_p95": [
          58.91058349516243,
          68.93927645578515
        ]
      }
    },
    "paired_gain": -0.05,
    "regression_count": 2
  },
  "gpu_deleted": true,
  "historical_MPS": {
    "bootstrap_draws": 5000,
    "bootstrap_seed": 42,
    "cluster_bootstrap_95_percent": [
      0.21652279769809651,
      0.317913366839273
    ],
    "cluster_count": 451,
    "improvement_count": 212,
    "models": {
      "c-1213": {
        "document_field_error_counts": {
          "all_invented": 2,
          "all_omitted": 14,
          "any_invented": 6,
          "any_omitted": 2,
          "credential_changed": 1,
          "exclude_invented": 9,
          "exclude_omitted": 7,
          "exclude_paths_invented": 2,
          "exclude_paths_omitted": 8,
          "invalid_document_plan": 8,
          "scope_changed": 2,
          "types_invented": 15,
          "types_omitted": 16,
          "wrong_output_mode": 30
        },
        "empty_n": 0,
        "empty_passed": 0,
        "false_acceptances": 0,
        "false_refusals": 27,
        "generation_ms_median_p95": [
          610.906917048851,
          1193.3272520982428
        ],
        "n": 680,
        "outcomes": 606,
        "plans": 560,
        "strata": {
          "all_any": 10,
          "boolean": 40,
          "capability": 8,
          "challenge-00": 0,
          "challenge-01": 0,
          "challenge-02": 6,
          "challenge-03": 0,
          "challenge-04": 6,
          "challenge-05": 0,
          "challenge-06": 4,
          "challenge-07": 6,
          "challenge-08": 6,
          "challenge-09": 6,
          "challenge-10": 6,
          "challenge-11": 6,
          "compound": 10,
          "conversational-filename": 12,
          "credential": 10,
          "credential-association": 0,
          "credentials": 39,
          "dates": 40,
          "document-00": 6,
          "document-01": 5,
          "document-02": 10,
          "document-03": 9,
          "document-04": 9,
          "document-05": 8,
          "document-06": 8,
          "document-07": 10,
          "document-08": 6,
          "document-09": 11,
          "document-10": 10,
          "document-11": 8,
          "dotenv": 40,
          "filename": 10,
          "find_bounded_depth": 5,
          "find_directories": 4,
          "find_empty_result": 2,
          "find_exact_name": 7,
          "find_iname": 5,
          "find_name_patterns_with_pruning": 6,
          "find_recent_mtime": 4,
          "find_size_threshold": 4,
          "formats": 39,
          "fresh_all": 33,
          "fresh_any": 30,
          "needs-clarification": 2,
          "needs-document-format": 9,
          "needs-email-access": 6,
          "needs-image-content": 6,
          "needs-provenance-metadata": 3,
          "not-a-file-search": 9,
          "or_recall": 2,
          "refusal": 40,
          "rg_files_candidates": 5,
          "scope_format": 10,
          "window_topk": 10
        },
        "synthetic_fixture_execution_ms_median_p95": [
          3.1800630094949156,
          55.681664272560724
        ]
      },
      "incumbent-p0": {
        "document_field_error_counts": {
          "all_invented": 29,
          "all_omitted": 10,
          "any_omitted": 2,
          "credential_changed": 16,
          "exclude_invented": 9,
          "exclude_paths_invented": 1,
          "exclude_paths_omitted": 15,
          "invalid_document_plan": 99,
          "scope_changed": 9,
          "types_invented": 44,
          "types_omitted": 47,
          "wrong_output_mode": 38
        },
        "empty_n": 0,
        "empty_passed": 0,
        "false_acceptances": 30,
        "false_refusals": 29,
        "generation_ms_median_p95": [
          603.3937915053684,
          1193.7494145764504
        ],
        "n": 680,
        "outcomes": 426,
        "plans": 384,
        "strata": {
          "all_any": 10,
          "boolean": 8,
          "capability": 9,
          "challenge-00": 1,
          "challenge-01": 0,
          "challenge-02": 6,
          "challenge-03": 0,
          "challenge-04": 6,
          "challenge-05": 6,
          "challenge-06": 0,
          "challenge-07": 6,
          "challenge-08": 6,
          "challenge-09": 3,
          "challenge-10": 6,
          "challenge-11": 6,
          "compound": 10,
          "conversational-filename": 11,
          "credential": 10,
          "credential-association": 1,
          "credentials": 0,
          "dates": 25,
          "document-00": 7,
          "document-01": 5,
          "document-02": 9,
          "document-03": 8,
          "document-04": 9,
          "document-05": 7,
          "document-06": 7,
          "document-07": 10,
          "document-08": 5,
          "document-09": 11,
          "document-10": 10,
          "document-11": 7,
          "dotenv": 0,
          "filename": 10,
          "find_bounded_depth": 6,
          "find_directories": 4,
          "find_empty_result": 3,
          "find_exact_name": 7,
          "find_iname": 5,
          "find_name_patterns_with_pruning": 6,
          "find_recent_mtime": 4,
          "find_size_threshold": 4,
          "formats": 4,
          "fresh_all": 37,
          "fresh_any": 34,
          "needs-clarification": 1,
          "needs-document-format": 8,
          "needs-email-access": 6,
          "needs-image-content": 6,
          "needs-provenance-metadata": 3,
          "not-a-file-search": 9,
          "or_recall": 6,
          "refusal": 13,
          "rg_files_candidates": 5,
          "scope_format": 10,
          "window_topk": 10
        },
        "synthetic_fixture_execution_ms_median_p95": [
          2.524666488170624,
          57.11858787108212
        ]
      },
      "incumbent-p1": {
        "document_field_error_counts": {
          "all_invented": 66,
          "all_omitted": 9,
          "any_omitted": 4,
          "credential_changed": 51,
          "exclude_invented": 19,
          "exclude_omitted": 1,
          "exclude_paths_invented": 1,
          "exclude_paths_omitted": 23,
          "invalid_document_plan": 28,
          "scope_changed": 12,
          "types_invented": 41,
          "types_omitted": 40,
          "wrong_output_mode": 36
        },
        "empty_n": 0,
        "empty_passed": 0,
        "false_acceptances": 29,
        "false_refusals": 29,
        "generation_ms_median_p95": [
          605.5316664860584,
          1213.36485275533
        ],
        "n": 680,
        "outcomes": 449,
        "plans": 409,
        "strata": {
          "all_any": 10,
          "boolean": 17,
          "capability": 9,
          "challenge-00": 1,
          "challenge-01": 0,
          "challenge-02": 6,
          "challenge-03": 0,
          "challenge-04": 6,
          "challenge-05": 6,
          "challenge-06": 0,
          "challenge-07": 6,
          "challenge-08": 6,
          "challenge-09": 5,
          "challenge-10": 6,
          "challenge-11": 6,
          "compound": 10,
          "conversational-filename": 11,
          "credential": 10,
          "credential-association": 1,
          "credentials": 0,
          "dates": 25,
          "document-00": 7,
          "document-01": 5,
          "document-02": 9,
          "document-03": 9,
          "document-04": 10,
          "document-05": 7,
          "document-06": 9,
          "document-07": 9,
          "document-08": 5,
          "document-09": 10,
          "document-10": 9,
          "document-11": 7,
          "dotenv": 4,
          "filename": 10,
          "find_bounded_depth": 6,
          "find_directories": 4,
          "find_empty_result": 3,
          "find_exact_name": 7,
          "find_iname": 5,
          "find_name_patterns_with_pruning": 6,
          "find_recent_mtime": 4,
          "find_size_threshold": 4,
          "formats": 16,
          "fresh_all": 36,
          "fresh_any": 33,
          "needs-clarification": 1,
          "needs-document-format": 9,
          "needs-email-access": 6,
          "needs-image-content": 6,
          "needs-provenance-metadata": 3,
          "not-a-file-search": 9,
          "or_recall": 4,
          "refusal": 13,
          "rg_files_candidates": 5,
          "scope_format": 8,
          "window_topk": 10
        },
        "synthetic_fixture_execution_ms_median_p95": [
          3.178562503308058,
          57.588802100508474
        ]
      }
    },
    "paired_gain": 0.2647058823529412,
    "regression_count": 32
  },
  "historical_MPS_gate": {
    "applied_gate_passed": true,
    "operator_policy": {
      "basis": "sum of executed outcomes across the same 1280 diagnostic cases",
      "default_checkpoint": "c-1213",
      "historical_regressions_are_reported_tradeoffs": true,
      "sha256": "23da0c3ceb3094e3cbe5d7dbc287b1c17fca218a197e4011bed74191270b6a61"
    },
    "original_zero_regression_gate_passed": false,
    "paired_outcome_gain": 0.2647058823529412,
    "regressed_cases": 49,
    "transfer_families": {
      "boolean": 40,
      "credentials": 39,
      "dates": 40,
      "dotenv": 40,
      "formats": 39,
      "refusal": 40
    }
  },
  "historical_MPS_passed": true,
  "human_canary_collected": false,
  "initialization_sha256": "984ff1aaed71652d9f53c91074e0787b3386bc3ae41900b4b830e6211805fc64",
  "learning_rate_finding": "At the final checkpoint, lowering the learning rate from 1e-5 to 3e-6 changes M0 from 1009 to 836 outcomes and M1 from 1183 to 1023 outcomes, out of 1,280. This is a matched one-epoch comparison. It does not establish how the lower rate would perform with more training.",
  "limitations": [
    "Synthetic language uses the same teacher family for authoring and review.",
    "Forty filename compatibility cases are already included in the wider board.",
    "Historical and development boards are selection diagnostics, not independent acceptance.",
    "Independent human requests, application smoke checks and production latency remain outstanding.",
    "One training seed per arm; matched differences are not estimates of variability across retraining seeds.",
    "Known corpora and boards were audited for exact overlap and conflicting labels; this does not establish semantic independence.",
    "Overall ranking weights each of the 1,280 cases equally: 600 development, 240 transfer, 160 retention and 280 wider. This benchmark mix does not estimate real-world request frequency.",
    "The operator changed selection after viewing A/B/C diagnostics: C final is the default; D must strictly exceed its overall outcome count. This is a disclosed post-hoc research choice, not a pass of the original frozen gate."
  ],
  "matched_final_checkpoint_deltas": {
    "context_only": {
      "after": "incumbent-p1",
      "before": "incumbent-p0",
      "outcome_delta": {
        "dev": 17,
        "retention": -6,
        "transfer": 24,
        "wider": 4
      }
    },
    "lower_lr_m0": {
      "after": "b-1213",
      "before": "a-1213",
      "outcome_delta": {
        "dev": -117,
        "retention": -19,
        "transfer": -28,
        "wider": -9
      }
    },
    "lower_lr_m1": {
      "after": "d-1213",
      "before": "c-1213",
      "outcome_delta": {
        "dev": -86,
        "retention": -4,
        "transfer": -39,
        "wider": -31
      }
    },
    "new_mixture_lr_1e-5": {
      "after": "c-1213",
      "before": "a-1213",
      "outcome_delta": {
        "dev": 184,
        "retention": -21,
        "transfer": -1,
        "wider": 12
      }
    },
    "new_mixture_lr_3e-6": {
      "after": "d-1213",
      "before": "b-1213",
      "outcome_delta": {
        "dev": 215,
        "retention": -6,
        "transfer": -12,
        "wider": -10
      }
    }
  },
  "metrics": {
    "a-1213": {
      "dev": {
        "n": 600,
        "outcomes": 392,
        "plans": 368
      },
      "retention": {
        "n": 160,
        "outcomes": 153,
        "plans": 143
      },
      "transfer": {
        "n": 240,
        "outcomes": 239,
        "plans": 239
      },
      "wider": {
        "n": 280,
        "outcomes": 225,
        "plans": 203
      }
    },
    "a-303": {
      "dev": {
        "n": 600,
        "outcomes": 357,
        "plans": 333
      },
      "retention": {
        "n": 160,
        "outcomes": 146,
        "plans": 136
      },
      "transfer": {
        "n": 240,
        "outcomes": 235,
        "plans": 235
      },
      "wider": {
        "n": 280,
        "outcomes": 218,
        "plans": 193
      }
    },
    "a-607": {
      "dev": {
        "n": 600,
        "outcomes": 391,
        "plans": 368
      },
      "retention": {
        "n": 160,
        "outcomes": 150,
        "plans": 140
      },
      "transfer": {
        "n": 240,
        "outcomes": 239,
        "plans": 239
      },
      "wider": {
        "n": 280,
        "outcomes": 227,
        "plans": 205
      }
    },
    "b-1213": {
      "dev": {
        "n": 600,
        "outcomes": 275,
        "plans": 251
      },
      "retention": {
        "n": 160,
        "outcomes": 134,
        "plans": 114
      },
      "transfer": {
        "n": 240,
        "outcomes": 211,
        "plans": 210
      },
      "wider": {
        "n": 280,
        "outcomes": 216,
        "plans": 192
      }
    },
    "b-303": {
      "dev": {
        "n": 600,
        "outcomes": 254,
        "plans": 230
      },
      "retention": {
        "n": 160,
        "outcomes": 124,
        "plans": 107
      },
      "transfer": {
        "n": 240,
        "outcomes": 196,
        "plans": 196
      },
      "wider": {
        "n": 280,
        "outcomes": 213,
        "plans": 190
      }
    },
    "b-607": {
      "dev": {
        "n": 600,
        "outcomes": 268,
        "plans": 244
      },
      "retention": {
        "n": 160,
        "outcomes": 131,
        "plans": 111
      },
      "transfer": {
        "n": 240,
        "outcomes": 206,
        "plans": 206
      },
      "wider": {
        "n": 280,
        "outcomes": 215,
        "plans": 191
      }
    },
    "c-1213": {
      "dev": {
        "n": 600,
        "outcomes": 576,
        "plans": 576
      },
      "retention": {
        "n": 160,
        "outcomes": 132,
        "plans": 114
      },
      "transfer": {
        "n": 240,
        "outcomes": 238,
        "plans": 238
      },
      "wider": {
        "n": 280,
        "outcomes": 237,
        "plans": 210
      }
    },
    "c-303": {
      "dev": {
        "n": 600,
        "outcomes": 563,
        "plans": 561
      },
      "retention": {
        "n": 160,
        "outcomes": 120,
        "plans": 101
      },
      "transfer": {
        "n": 240,
        "outcomes": 237,
        "plans": 237
      },
      "wider": {
        "n": 280,
        "outcomes": 235,
        "plans": 213
      }
    },
    "c-607": {
      "dev": {
        "n": 600,
        "outcomes": 572,
        "plans": 572
      },
      "retention": {
        "n": 160,
        "outcomes": 139,
        "plans": 124
      },
      "transfer": {
        "n": 240,
        "outcomes": 239,
        "plans": 239
      },
      "wider": {
        "n": 280,
        "outcomes": 233,
        "plans": 206
      }
    },
    "d-1213": {
      "dev": {
        "n": 600,
        "outcomes": 490,
        "plans": 466
      },
      "retention": {
        "n": 160,
        "outcomes": 128,
        "plans": 111
      },
      "transfer": {
        "n": 240,
        "outcomes": 199,
        "plans": 199
      },
      "wider": {
        "n": 280,
        "outcomes": 206,
        "plans": 168
      }
    },
    "d-303": {
      "dev": {
        "n": 600,
        "outcomes": 459,
        "plans": 434
      },
      "retention": {
        "n": 160,
        "outcomes": 128,
        "plans": 112
      },
      "transfer": {
        "n": 240,
        "outcomes": 193,
        "plans": 193
      },
      "wider": {
        "n": 280,
        "outcomes": 178,
        "plans": 138
      }
    },
    "d-607": {
      "dev": {
        "n": 600,
        "outcomes": 486,
        "plans": 462
      },
      "retention": {
        "n": 160,
        "outcomes": 126,
        "plans": 111
      },
      "transfer": {
        "n": 240,
        "outcomes": 199,
        "plans": 199
      },
      "wider": {
        "n": 280,
        "outcomes": 204,
        "plans": 165
      }
    },
    "incumbent-p0": {
      "dev": {
        "n": 600,
        "outcomes": 246,
        "plans": 224
      },
      "retention": {
        "n": 160,
        "outcomes": 145,
        "plans": 129
      },
      "transfer": {
        "n": 240,
        "outcomes": 49,
        "plans": 40
      },
      "wider": {
        "n": 280,
        "outcomes": 229,
        "plans": 213
      }
    },
    "incumbent-p1": {
      "dev": {
        "n": 600,
        "outcomes": 263,
        "plans": 238
      },
      "retention": {
        "n": 160,
        "outcomes": 139,
        "plans": 127
      },
      "transfer": {
        "n": 240,
        "outcomes": 73,
        "plans": 64
      },
      "wider": {
        "n": 280,
        "outcomes": 233,
        "plans": 215
      }
    },
    "sept10-p1": {
      "dev": {
        "n": 600,
        "outcomes": 356,
        "plans": 327
      },
      "retention": {
        "n": 160,
        "outcomes": 139,
        "plans": 126
      },
      "transfer": {
        "n": 240,
        "outcomes": 240,
        "plans": 240
      },
      "wider": {
        "n": 280,
        "outcomes": 238,
        "plans": 218
      }
    }
  },
  "model_name_status": "research_incumbent_selected",
  "next_step": "Use development diagnostics to design the next intervention. Retire the opened acceptance set from independent testing and author a fresh sealed set before the next experiment.",
  "nominee": "c-1213",
  "operator_selection_policy": {
    "basis": "sum of executed outcomes across the same 1280 diagnostic cases",
    "default_checkpoint": "c-1213",
    "historical_regressions_are_reported_tradeoffs": true,
    "sha256": "23da0c3ceb3094e3cbe5d7dbc287b1c17fca218a197e4011bed74191270b6a61"
  },
  "original_frozen_nominee": null,
  "overall_outcomes": {
    "a-1213": 1009,
    "a-303": 956,
    "a-607": 1007,
    "b-1213": 836,
    "b-303": 787,
    "b-607": 820,
    "c-1213": 1183,
    "c-303": 1155,
    "c-607": 1183,
    "d-1213": 1023,
    "d-303": 958,
    "d-607": 1015,
    "incumbent-p0": 669,
    "incumbent-p1": 708,
    "sept10-p1": 973
  },
  "paired_historical": {
    "a-1213": {
      "retention": {
        "outcome_improvements": 11,
        "outcome_or_plan_regressions": 3,
        "outcome_regressions": 3
      },
      "transfer": {
        "outcome_improvements": 190,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 12,
        "outcome_or_plan_regressions": 21,
        "outcome_regressions": 16
      }
    },
    "a-303": {
      "retention": {
        "outcome_improvements": 10,
        "outcome_or_plan_regressions": 9,
        "outcome_regressions": 9
      },
      "transfer": {
        "outcome_improvements": 187,
        "outcome_or_plan_regressions": 1,
        "outcome_regressions": 1
      },
      "wider": {
        "outcome_improvements": 8,
        "outcome_or_plan_regressions": 27,
        "outcome_regressions": 19
      }
    },
    "a-607": {
      "retention": {
        "outcome_improvements": 10,
        "outcome_or_plan_regressions": 5,
        "outcome_regressions": 5
      },
      "transfer": {
        "outcome_improvements": 190,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 12,
        "outcome_or_plan_regressions": 19,
        "outcome_regressions": 14
      }
    },
    "b-1213": {
      "retention": {
        "outcome_improvements": 6,
        "outcome_or_plan_regressions": 28,
        "outcome_regressions": 17
      },
      "transfer": {
        "outcome_improvements": 164,
        "outcome_or_plan_regressions": 2,
        "outcome_regressions": 2
      },
      "wider": {
        "outcome_improvements": 8,
        "outcome_or_plan_regressions": 27,
        "outcome_regressions": 21
      }
    },
    "b-303": {
      "retention": {
        "outcome_improvements": 5,
        "outcome_or_plan_regressions": 34,
        "outcome_regressions": 26
      },
      "transfer": {
        "outcome_improvements": 150,
        "outcome_or_plan_regressions": 3,
        "outcome_regressions": 3
      },
      "wider": {
        "outcome_improvements": 7,
        "outcome_or_plan_regressions": 30,
        "outcome_regressions": 23
      }
    },
    "b-607": {
      "retention": {
        "outcome_improvements": 5,
        "outcome_or_plan_regressions": 30,
        "outcome_regressions": 19
      },
      "transfer": {
        "outcome_improvements": 159,
        "outcome_or_plan_regressions": 2,
        "outcome_regressions": 2
      },
      "wider": {
        "outcome_improvements": 8,
        "outcome_or_plan_regressions": 28,
        "outcome_regressions": 22
      }
    },
    "c-1213": {
      "retention": {
        "outcome_improvements": 5,
        "outcome_or_plan_regressions": 24,
        "outcome_regressions": 18
      },
      "transfer": {
        "outcome_improvements": 189,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 20,
        "outcome_or_plan_regressions": 21,
        "outcome_regressions": 12
      }
    },
    "c-303": {
      "retention": {
        "outcome_improvements": 3,
        "outcome_or_plan_regressions": 37,
        "outcome_regressions": 28
      },
      "transfer": {
        "outcome_improvements": 188,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 12,
        "outcome_or_plan_regressions": 11,
        "outcome_regressions": 6
      }
    },
    "c-607": {
      "retention": {
        "outcome_improvements": 7,
        "outcome_or_plan_regressions": 18,
        "outcome_regressions": 13
      },
      "transfer": {
        "outcome_improvements": 190,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 15,
        "outcome_or_plan_regressions": 20,
        "outcome_regressions": 11
      }
    },
    "d-1213": {
      "retention": {
        "outcome_improvements": 4,
        "outcome_or_plan_regressions": 27,
        "outcome_regressions": 21
      },
      "transfer": {
        "outcome_improvements": 151,
        "outcome_or_plan_regressions": 1,
        "outcome_regressions": 1
      },
      "wider": {
        "outcome_improvements": 8,
        "outcome_or_plan_regressions": 50,
        "outcome_regressions": 31
      }
    },
    "d-303": {
      "retention": {
        "outcome_improvements": 6,
        "outcome_or_plan_regressions": 28,
        "outcome_regressions": 23
      },
      "transfer": {
        "outcome_improvements": 145,
        "outcome_or_plan_regressions": 1,
        "outcome_regressions": 1
      },
      "wider": {
        "outcome_improvements": 6,
        "outcome_or_plan_regressions": 79,
        "outcome_regressions": 57
      }
    },
    "d-607": {
      "retention": {
        "outcome_improvements": 4,
        "outcome_or_plan_regressions": 27,
        "outcome_regressions": 23
      },
      "transfer": {
        "outcome_improvements": 150,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 8,
        "outcome_or_plan_regressions": 53,
        "outcome_regressions": 33
      }
    },
    "incumbent-p0": {
      "retention": {
        "outcome_improvements": 0,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "transfer": {
        "outcome_improvements": 0,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 0,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      }
    },
    "incumbent-p1": {
      "retention": {
        "outcome_improvements": 2,
        "outcome_or_plan_regressions": 9,
        "outcome_regressions": 8
      },
      "transfer": {
        "outcome_improvements": 27,
        "outcome_or_plan_regressions": 3,
        "outcome_regressions": 3
      },
      "wider": {
        "outcome_improvements": 8,
        "outcome_or_plan_regressions": 5,
        "outcome_regressions": 4
      }
    },
    "sept10-p1": {
      "retention": {
        "outcome_improvements": 6,
        "outcome_or_plan_regressions": 14,
        "outcome_regressions": 12
      },
      "transfer": {
        "outcome_improvements": 191,
        "outcome_or_plan_regressions": 0,
        "outcome_regressions": 0
      },
      "wider": {
        "outcome_improvements": 17,
        "outcome_or_plan_regressions": 11,
        "outcome_regressions": 8
      }
    }
  },
  "requested_model_name": "jeremiah_v1",
  "research_incumbent_selected": true,
  "scheduling": "Mac diagnostic replay overlapped the fixed GPU queue; final nomination followed verified recovery and GPU deletion.",
  "schema": "shlex-retention-matrix-public-v1",
  "selected_paired_outcome_tradeoffs_vs_incumbent_p0": {
    "dev": {
      "improvements": 330,
      "regressions": 0
    },
    "retention": {
      "improvements": 5,
      "regressions": 18
    },
    "transfer": {
      "improvements": 189,
      "regressions": 0
    },
    "wider": {
      "improvements": 20,
      "regressions": 12
    }
  },
  "selected_regression_field_error_counts": {
    "all_omitted": 1,
    "invalid_document_plan": 6,
    "scope_changed": 2,
    "types_invented": 11,
    "types_omitted": 11,
    "wrong_output_mode": 7
  },
  "synthetic_acceptance_passed": false,
  "training": {
    "arms": {
      "a": [
        "m0",
        1e-05
      ],
      "b": [
        "m0",
        3e-06
      ],
      "c": [
        "m1",
        1e-05
      ],
      "d": [
        "m1",
        3e-06
      ]
    },
    "base_model_family": "Qwen2.5-Coder-1.5B-Instruct",
    "maximum_new_rows_per_spec": 4,
    "new_rows_in_m1": 8000,
    "new_unique_semantic_specs": 7290,
    "rows_per_arm": 19408,
    "snapshot_steps": [
      303,
      607,
      1213
    ]
  },
  "verified_checkpoint_count": 12
}
