{
  "schemaVersion": 1,
  "study": "reference-access-2026-09-25",
  "measuredDate": "2026-09-25",
  "model": "zai-org/GLM-5.3-Flash",
  "taskCount": 20,
  "agreementsPerTask": 50,
  "repetitions": 5,
  "sourceSegments": 3,
  "conditions": {
    "reference": {
      "attempts": 100,
      "correct": 99,
      "score": 0.99,
      "rowAccuracy": 0.99,
      "meanTurns": 2.03,
      "meanSeconds": 42.72062809954,
      "meanUsd": null,
      "knownUsd": "0.42709952",
      "missingUsageRequests": 1,
      "toolRejections": 0,
      "timeouts": 1,
      "incorrectAnswers": 0
    },
    "sql": {
      "attempts": 100,
      "correct": 99,
      "score": 0.99,
      "rowAccuracy": 0.9998,
      "meanTurns": 3.9,
      "meanSeconds": 86.32905955584,
      "meanUsd": "0.0092795195",
      "knownUsd": "0.92795195",
      "missingUsageRequests": 0,
      "toolRejections": 13,
      "timeouts": 0,
      "incorrectAnswers": 1
    }
  },
  "tasks": [
    {
      "task": 1,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.2,
        "meanSeconds": 38.327311108,
        "meanUsd": "0.00430653",
        "knownUsd": "0.02153265",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 4.8,
        "meanSeconds": 114.4823609168,
        "meanUsd": "0.011750488",
        "knownUsd": "0.05875244",
        "missingUsageRequests": 0,
        "toolRejections": 1,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 2,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 34.4887552084,
        "meanUsd": "0.003738348",
        "knownUsd": "0.01869174",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.4,
        "meanSeconds": 68.0488149586,
        "meanUsd": "0.008012326",
        "knownUsd": "0.04006163",
        "missingUsageRequests": 0,
        "toolRejections": 1,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 3,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.2,
        "meanSeconds": 53.931124658,
        "meanUsd": "0.005354302",
        "knownUsd": "0.02677151",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 4.6,
        "meanSeconds": 83.3000624168,
        "meanUsd": "0.010442368",
        "knownUsd": "0.05221184",
        "missingUsageRequests": 0,
        "toolRejections": 1,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 4,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 42.8268248584,
        "meanUsd": "0.004876444",
        "knownUsd": "0.02438222",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.8,
        "meanSeconds": 89.72767485840001,
        "meanUsd": "0.00899481",
        "knownUsd": "0.04497405",
        "missingUsageRequests": 0,
        "toolRejections": 2,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 5,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 41.9050882748,
        "meanUsd": "0.003936276",
        "knownUsd": "0.01968138",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.2,
        "meanSeconds": 76.7795235084,
        "meanUsd": "0.008089686",
        "knownUsd": "0.04044843",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 6,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 30.8809026166,
        "meanUsd": "0.003731044",
        "knownUsd": "0.01865522",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.8,
        "meanSeconds": 92.49481775,
        "meanUsd": "0.008909814",
        "knownUsd": "0.04454907",
        "missingUsageRequests": 0,
        "toolRejections": 2,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 7,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 39.0626919752,
        "meanUsd": "0.004275966",
        "knownUsd": "0.02137983",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.8,
        "meanSeconds": 67.5194951668,
        "meanUsd": "0.007508184",
        "knownUsd": "0.03754092",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 8,
      "reference": {
        "attempts": 5,
        "correct": 4,
        "score": 0.8,
        "rowAccuracy": 0.8,
        "meanTurns": 2.0,
        "meanSeconds": 86.797205658,
        "meanUsd": null,
        "knownUsd": "0.01993059",
        "missingUsageRequests": 1,
        "toolRejections": 0,
        "timeouts": 1,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.4,
        "meanSeconds": 68.2971895418,
        "meanUsd": "0.007709522",
        "knownUsd": "0.03854761",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 9,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 33.4962429082,
        "meanUsd": "0.003416362",
        "knownUsd": "0.01708181",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 4,
        "score": 0.8,
        "rowAccuracy": 0.996,
        "meanTurns": 4.0,
        "meanSeconds": 86.9433984168,
        "meanUsd": "0.009076082",
        "knownUsd": "0.04538041",
        "missingUsageRequests": 0,
        "toolRejections": 1,
        "timeouts": 0,
        "incorrectAnswers": 1
      }
    },
    {
      "task": 10,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 55.467203058399996,
        "meanUsd": "0.005437884",
        "knownUsd": "0.02718942",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 5.0,
        "meanSeconds": 124.619440575,
        "meanUsd": "0.014031316",
        "knownUsd": "0.07015658",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 11,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 51.527233366800004,
        "meanUsd": "0.004284808",
        "knownUsd": "0.02142404",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.0,
        "meanSeconds": 96.46759979160001,
        "meanUsd": "0.008534684",
        "knownUsd": "0.04267342",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 12,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 36.1677544666,
        "meanUsd": "0.004063918",
        "knownUsd": "0.02031959",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.2,
        "meanSeconds": 76.8066393164,
        "meanUsd": "0.008029152",
        "knownUsd": "0.04014576",
        "missingUsageRequests": 0,
        "toolRejections": 1,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 13,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 47.7420702918,
        "meanUsd": "0.004867722",
        "knownUsd": "0.02433861",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 4.2,
        "meanSeconds": 82.234858625,
        "meanUsd": "0.009425594",
        "knownUsd": "0.04712797",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 14,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 34.592157433400004,
        "meanUsd": "0.004063198",
        "knownUsd": "0.02031599",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 4.4,
        "meanSeconds": 75.1836007584,
        "meanUsd": "0.010267858",
        "knownUsd": "0.05133929",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 15,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 37.7454683332,
        "meanUsd": "0.003836038",
        "knownUsd": "0.01918019",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 5.0,
        "meanSeconds": 82.0651075168,
        "meanUsd": "0.009495098",
        "knownUsd": "0.04747549",
        "missingUsageRequests": 0,
        "toolRejections": 4,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 16,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 36.55686595,
        "meanUsd": "0.003842888",
        "knownUsd": "0.01921444",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.8,
        "meanSeconds": 83.1767726914,
        "meanUsd": "0.009442694",
        "knownUsd": "0.04721347",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 17,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 34.6938161916,
        "meanUsd": "0.003892858",
        "knownUsd": "0.01946429",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.6,
        "meanSeconds": 93.7638201164,
        "meanUsd": "0.008441688",
        "knownUsd": "0.04220844",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 18,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.2,
        "meanSeconds": 44.374911158,
        "meanUsd": "0.00488219",
        "knownUsd": "0.02441095",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 4.2,
        "meanSeconds": 79.2220699084,
        "meanUsd": "0.009382056",
        "knownUsd": "0.04691028",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 19,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 32.361928067,
        "meanUsd": "0.00438851",
        "knownUsd": "0.02194255",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 3.8,
        "meanSeconds": 96.6630610664,
        "meanUsd": "0.009701564",
        "knownUsd": "0.04850782",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    },
    {
      "task": 20,
      "reference": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 2.0,
        "meanSeconds": 41.467006408399996,
        "meanUsd": "0.00423850",
        "knownUsd": "0.02119250",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 5,
        "correct": 5,
        "score": 1.0,
        "rowAccuracy": 1.0,
        "meanTurns": 4.0,
        "meanSeconds": 88.7848832166,
        "meanUsd": "0.008345406",
        "knownUsd": "0.04172703",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 0
      }
    }
  ],
  "sensitivity": {
    "rejectedCalls": 13,
    "affectedPairs": 12,
    "affectedTasks": 8,
    "allPairs": {
      "pairs": 100,
      "reference": {
        "attempts": 100,
        "correct": 99,
        "score": 0.99,
        "rowAccuracy": 0.99,
        "meanTurns": 2.03,
        "meanSeconds": 42.72062809954,
        "meanUsd": null,
        "knownUsd": "0.42709952",
        "missingUsageRequests": 1,
        "toolRejections": 0,
        "timeouts": 1,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 100,
        "correct": 99,
        "score": 0.99,
        "rowAccuracy": 0.9998,
        "meanTurns": 3.9,
        "meanSeconds": 86.32905955584,
        "meanUsd": "0.0092795195",
        "knownUsd": "0.92795195",
        "missingUsageRequests": 0,
        "toolRejections": 13,
        "timeouts": 0,
        "incorrectAnswers": 1
      },
      "turnReductionFraction": 0.47948717948717956
    },
    "withoutRejectedSqlPairs": {
      "pairs": 88,
      "reference": {
        "attempts": 88,
        "correct": 87,
        "score": 0.9886363636363636,
        "rowAccuracy": 0.9886363636363636,
        "meanTurns": 2.034090909090909,
        "meanSeconds": 43.558091711125,
        "meanUsd": null,
        "knownUsd": "0.38031349",
        "missingUsageRequests": 1,
        "toolRejections": 0,
        "timeouts": 1,
        "incorrectAnswers": 0
      },
      "sql": {
        "attempts": 88,
        "correct": 87,
        "score": 0.9886363636363636,
        "rowAccuracy": 0.9997727272727273,
        "meanTurns": 3.6818181818181817,
        "meanSeconds": 83.86553571306818,
        "meanUsd": "0.009017993068181818181818181818",
        "knownUsd": "0.79358339",
        "missingUsageRequests": 0,
        "toolRejections": 0,
        "timeouts": 0,
        "incorrectAnswers": 1
      },
      "turnReductionFraction": 0.44753086419753085
    },
    "postHoc": true,
    "causalEffectOfRemovingGuardEstablished": false
  },
  "pairedAnalysis": {
    "method": "paired_task_cluster_percentile_bootstrap",
    "independent_task_groups": 20,
    "draws": 20000,
    "seed": 1701,
    "difference_direction": "reference_minus_sql",
    "metrics": {
      "model_turns": {
        "reference_minus_sql": -1.8699999999999997,
        "percentile_95_interval": [
          -2.1300000000000003,
          -1.6100000000000003
        ],
        "no_observed_task_variation_in_difference": false
      },
      "latency_seconds": {
        "reference_minus_sql": -43.6084314563,
        "percentile_95_interval": [
          -51.597878808387755,
          -34.45088028819126
        ],
        "no_observed_task_variation_in_difference": false
      },
      "exact_table_accuracy": {
        "reference_minus_sql": 0.0,
        "percentile_95_interval": [
          -0.030000000000000027,
          0.030000000000000027
        ],
        "no_observed_task_variation_in_difference": false
      },
      "row_accuracy": {
        "reference_minus_sql": -0.009800000000000031,
        "percentile_95_interval": [
          -0.030000000000000027,
          0.0006000000000000449
        ],
        "no_observed_task_variation_in_difference": false
      }
    },
    "cost_comparison_available": false,
    "boundary_caution": "A degenerate accuracy interval does not establish equivalence or rule out unseen failures."
  },
  "excludedProviderRejections": 1,
  "protocol": {
    "reasoning": "max",
    "temperature": 0.2,
    "maxModelRequests": 12,
    "maxOutputTokens": 16384,
    "timeoutSeconds": 240,
    "score": "exact_50_row_table",
    "cost": "received_token_usage_at_published_rates"
  }
}
