{
  "schema": "quantum-toricgt-benchmark-v1",
  "backend": "PyTorch CUDA exact statevector/autograd simulation",
  "quantum_hardware": false,
  "quantum_advantage_established": false,
  "dataset_sha256": "4147088938a0742d956aceb7e50a3c7206d349b8f1196cb7a3b6fefb1a774156",
  "seeds": [
    17,
    29,
    43
  ],
  "epochs": 80,
  "batch_size": 32,
  "learning_rate": 0.0015,
  "selection": "Lowest validation node MSE within each predeclared run; test and OOD are scored once after selection.",
  "gpu_name": "NVIDIA GeForce RTX 4090",
  "aggregate": {
    "quantum": {
      "test": {
        "node_rmse_cost": {
          "mean": 0.26522473321613077,
          "sample_std": 0.00291078108278665,
          "values": [
            0.2629056889382799,
            0.2642773211510682,
            0.26849118955904405
          ]
        },
        "source_cost_rmse": {
          "mean": 0.36378012634630846,
          "sample_std": 0.014275008377616076,
          "values": [
            0.34793252476867664,
            0.367777604764446,
            0.3756302495058028
          ]
        },
        "mean_proposal_regret_cost": {
          "mean": 0.03857421875,
          "sample_std": 0.0067658234670659265,
          "values": [
            0.03466796875,
            0.04638671875,
            0.03466796875
          ]
        },
        "optimal_proposal_fraction": {
          "mean": 0.8385416666666666,
          "sample_std": 0.023867581744561665,
          "values": [
            0.859375,
            0.8125,
            0.84375
          ]
        },
        "initial_certificate_fraction": {
          "mean": 0.24739583333333334,
          "sample_std": 0.004510548978043952,
          "values": [
            0.25,
            0.2421875,
            0.25
          ]
        }
      },
      "ood": {
        "node_rmse_cost": {
          "mean": 0.6836920015957456,
          "sample_std": 0.14987351794320858,
          "values": [
            0.8370406812856083,
            0.67648138040095,
            0.5375539431006787
          ]
        },
        "source_cost_rmse": {
          "mean": 0.8571273711945175,
          "sample_std": 0.20891432973345744,
          "values": [
            1.0865851307990195,
            0.8068736772077547,
            0.6779233055767785
          ]
        },
        "mean_proposal_regret_cost": {
          "mean": 0.224609375,
          "sample_std": 0.01982205383807074,
          "values": [
            0.20703125,
            0.24609375,
            0.220703125
          ]
        },
        "optimal_proposal_fraction": {
          "mean": 0.546875,
          "sample_std": 0.034053897996411515,
          "values": [
            0.5703125,
            0.5078125,
            0.5625
          ]
        },
        "initial_certificate_fraction": {
          "mean": 0.06510416666666667,
          "sample_std": 0.03853814736184569,
          "values": [
            0.0390625,
            0.046875,
            0.109375
          ]
        }
      }
    },
    "classical": {
      "test": {
        "node_rmse_cost": {
          "mean": 0.2664683993425742,
          "sample_std": 0.008655906078727533,
          "values": [
            0.27445310267430145,
            0.26768259111472964,
            0.25726950423869144
          ]
        },
        "source_cost_rmse": {
          "mean": 0.3590641055380546,
          "sample_std": 0.008346595198053587,
          "values": [
            0.3682024341378148,
            0.35714708456052274,
            0.35184279791582646
          ]
        },
        "mean_proposal_regret_cost": {
          "mean": 0.03564453125,
          "sample_std": 0.00634765625,
          "values": [
            0.03564453125,
            0.0419921875,
            0.029296875
          ]
        },
        "optimal_proposal_fraction": {
          "mean": 0.84375,
          "sample_std": 0.0078125,
          "values": [
            0.84375,
            0.8359375,
            0.8515625
          ]
        },
        "initial_certificate_fraction": {
          "mean": 0.3046875,
          "sample_std": 0.05633673867912483,
          "values": [
            0.2890625,
            0.2578125,
            0.3671875
          ]
        }
      },
      "ood": {
        "node_rmse_cost": {
          "mean": 0.7896133690284488,
          "sample_std": 0.1165931765500807,
          "values": [
            0.8498458007181783,
            0.8637708768516053,
            0.6552234295155629
          ]
        },
        "source_cost_rmse": {
          "mean": 0.8903589023375479,
          "sample_std": 0.08632146599098542,
          "values": [
            0.8688072584104902,
            0.9854142619947771,
            0.8168551866073767
          ]
        },
        "mean_proposal_regret_cost": {
          "mean": 0.2625325520833333,
          "sample_std": 0.07086865909295953,
          "values": [
            0.31591796875,
            0.28955078125,
            0.18212890625
          ]
        },
        "optimal_proposal_fraction": {
          "mean": 0.5,
          "sample_std": 0.04752158226795484,
          "values": [
            0.46875,
            0.4765625,
            0.5546875
          ]
        },
        "initial_certificate_fraction": {
          "mean": 0.0546875,
          "sample_std": 0.034053897996411515,
          "values": [
            0.03125,
            0.0390625,
            0.09375
          ]
        }
      }
    },
    "frozen": {
      "test": {
        "node_rmse_cost": {
          "mean": 0.2837077085769666,
          "sample_std": 0.011348188307108301,
          "values": [
            0.2884472020419697,
            0.291917861613602,
            0.2707580620753281
          ]
        },
        "source_cost_rmse": {
          "mean": 0.3915300037211487,
          "sample_std": 0.021777995436216082,
          "values": [
            0.40300344469704946,
            0.4051724266443309,
            0.3664141398220659
          ]
        },
        "mean_proposal_regret_cost": {
          "mean": 0.039876302083333336,
          "sample_std": 0.0048909437504286194,
          "values": [
            0.03955078125,
            0.044921875,
            0.03515625
          ]
        },
        "optimal_proposal_fraction": {
          "mean": 0.8307291666666666,
          "sample_std": 0.019661027175184247,
          "values": [
            0.828125,
            0.8125,
            0.8515625
          ]
        },
        "initial_certificate_fraction": {
          "mean": 0.2578125,
          "sample_std": 0.05122998847110938,
          "values": [
            0.3125,
            0.2109375,
            0.25
          ]
        }
      },
      "ood": {
        "node_rmse_cost": {
          "mean": 0.6757861724631721,
          "sample_std": 0.06855302716045324,
          "values": [
            0.6784199701589657,
            0.7429843439657833,
            0.6059542032647672
          ]
        },
        "source_cost_rmse": {
          "mean": 0.8976164564067535,
          "sample_std": 0.10718238143885964,
          "values": [
            0.9956706375597918,
            0.913988876988274,
            0.7831898546721946
          ]
        },
        "mean_proposal_regret_cost": {
          "mean": 0.232421875,
          "sample_std": 0.02627660946694084,
          "values": [
            0.224609375,
            0.26171875,
            0.2109375
          ]
        },
        "optimal_proposal_fraction": {
          "mean": 0.5286458333333334,
          "sample_std": 0.009021097956087904,
          "values": [
            0.5390625,
            0.5234375,
            0.5234375
          ]
        },
        "initial_certificate_fraction": {
          "mean": 0.09895833333333333,
          "sample_std": 0.022552744890219755,
          "values": [
            0.0859375,
            0.0859375,
            0.125
          ]
        }
      }
    }
  },
  "representative_trace": "quantum-seed17/trace.json",
  "runs": [
    {
      "name": "quantum-seed17",
      "mode": "quantum",
      "seed": 17,
      "trainable_parameters": 25515,
      "selected_epoch": 80,
      "theta_change_norm": 0.14594265818595886,
      "training_wall_seconds": 14.288623139029369,
      "inference_batch_ms": 2.5625604088418186,
      "cuda_peak_allocated_bytes": 183420928,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.22584017344411716,
          "source_cost_rmse": 0.32708175265825307,
          "mean_proposal_regret_cost": 0.037353515625,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.857421875,
          "initial_certificate_fraction": 0.197265625,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.23473067317009833,
          "source_cost_rmse": 0.34115330195623556,
          "mean_proposal_regret_cost": 0.0419921875,
          "max_proposal_regret_cost": 1.3125,
          "optimal_proposal_fraction": 0.8828125,
          "initial_certificate_fraction": 0.265625,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.2629056889382799,
          "source_cost_rmse": 0.34793252476867664,
          "mean_proposal_regret_cost": 0.03466796875,
          "max_proposal_regret_cost": 0.6875,
          "optimal_proposal_fraction": 0.859375,
          "initial_certificate_fraction": 0.25,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.8370406812856083,
          "source_cost_rmse": 1.0865851307990195,
          "mean_proposal_regret_cost": 0.20703125,
          "max_proposal_regret_cost": 1.625,
          "optimal_proposal_fraction": 0.5703125,
          "initial_certificate_fraction": 0.0390625,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "classical-seed17",
      "mode": "classical",
      "seed": 17,
      "trainable_parameters": 25515,
      "selected_epoch": 80,
      "theta_change_norm": 0.1159605011343956,
      "training_wall_seconds": 10.511146526085213,
      "inference_batch_ms": 1.8024774035438895,
      "cuda_peak_allocated_bytes": 182443008,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.22889581663030328,
          "source_cost_rmse": 0.3358077576961649,
          "mean_proposal_regret_cost": 0.0382080078125,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.833984375,
          "initial_certificate_fraction": 0.220703125,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.24480186941650087,
          "source_cost_rmse": 0.367352916802833,
          "mean_proposal_regret_cost": 0.0380859375,
          "max_proposal_regret_cost": 1.1875,
          "optimal_proposal_fraction": 0.875,
          "initial_certificate_fraction": 0.2421875,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.27445310267430145,
          "source_cost_rmse": 0.3682024341378148,
          "mean_proposal_regret_cost": 0.03564453125,
          "max_proposal_regret_cost": 0.6875,
          "optimal_proposal_fraction": 0.84375,
          "initial_certificate_fraction": 0.2890625,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.8498458007181783,
          "source_cost_rmse": 0.8688072584104902,
          "mean_proposal_regret_cost": 0.31591796875,
          "max_proposal_regret_cost": 1.875,
          "optimal_proposal_fraction": 0.46875,
          "initial_certificate_fraction": 0.03125,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "frozen-seed17",
      "mode": "frozen",
      "seed": 17,
      "trainable_parameters": 25449,
      "selected_epoch": 80,
      "theta_change_norm": 0.0,
      "training_wall_seconds": 11.116672069067135,
      "inference_batch_ms": 2.579692006111145,
      "cuda_peak_allocated_bytes": 183421440,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.23646842323039674,
          "source_cost_rmse": 0.34292517441384734,
          "mean_proposal_regret_cost": 0.038330078125,
          "max_proposal_regret_cost": 1.125,
          "optimal_proposal_fraction": 0.857421875,
          "initial_certificate_fraction": 0.2734375,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.2586307401218303,
          "source_cost_rmse": 0.3594988910820111,
          "mean_proposal_regret_cost": 0.06103515625,
          "max_proposal_regret_cost": 2.0625,
          "optimal_proposal_fraction": 0.8359375,
          "initial_certificate_fraction": 0.265625,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.2884472020419697,
          "source_cost_rmse": 0.40300344469704946,
          "mean_proposal_regret_cost": 0.03955078125,
          "max_proposal_regret_cost": 0.6875,
          "optimal_proposal_fraction": 0.828125,
          "initial_certificate_fraction": 0.3125,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.6784199701589657,
          "source_cost_rmse": 0.9956706375597918,
          "mean_proposal_regret_cost": 0.224609375,
          "max_proposal_regret_cost": 1.875,
          "optimal_proposal_fraction": 0.5390625,
          "initial_certificate_fraction": 0.0859375,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "quantum-seed29",
      "mode": "quantum",
      "seed": 29,
      "trainable_parameters": 25515,
      "selected_epoch": 80,
      "theta_change_norm": 0.15102992951869965,
      "training_wall_seconds": 14.14329302101396,
      "inference_batch_ms": 2.5679314509034157,
      "cuda_peak_allocated_bytes": 183426048,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.22127390790259743,
          "source_cost_rmse": 0.3277488657482276,
          "mean_proposal_regret_cost": 0.0445556640625,
          "max_proposal_regret_cost": 0.875,
          "optimal_proposal_fraction": 0.8359375,
          "initial_certificate_fraction": 0.21484375,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.23970194485841773,
          "source_cost_rmse": 0.36576271969977425,
          "mean_proposal_regret_cost": 0.05029296875,
          "max_proposal_regret_cost": 1.1875,
          "optimal_proposal_fraction": 0.859375,
          "initial_certificate_fraction": 0.2265625,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.2642773211510682,
          "source_cost_rmse": 0.367777604764446,
          "mean_proposal_regret_cost": 0.04638671875,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.8125,
          "initial_certificate_fraction": 0.2421875,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.67648138040095,
          "source_cost_rmse": 0.8068736772077547,
          "mean_proposal_regret_cost": 0.24609375,
          "max_proposal_regret_cost": 1.75,
          "optimal_proposal_fraction": 0.5078125,
          "initial_certificate_fraction": 0.046875,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "classical-seed29",
      "mode": "classical",
      "seed": 29,
      "trainable_parameters": 25515,
      "selected_epoch": 80,
      "theta_change_norm": 0.11134495586156845,
      "training_wall_seconds": 9.880282870028168,
      "inference_batch_ms": 1.8074609455652535,
      "cuda_peak_allocated_bytes": 182443008,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.23723664897830582,
          "source_cost_rmse": 0.34307516666510707,
          "mean_proposal_regret_cost": 0.03662109375,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.845703125,
          "initial_certificate_fraction": 0.201171875,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.2513624408691692,
          "source_cost_rmse": 0.3491684142583363,
          "mean_proposal_regret_cost": 0.0390625,
          "max_proposal_regret_cost": 1.3125,
          "optimal_proposal_fraction": 0.8984375,
          "initial_certificate_fraction": 0.2265625,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.26768259111472964,
          "source_cost_rmse": 0.35714708456052274,
          "mean_proposal_regret_cost": 0.0419921875,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.8359375,
          "initial_certificate_fraction": 0.2578125,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.8637708768516053,
          "source_cost_rmse": 0.9854142619947771,
          "mean_proposal_regret_cost": 0.28955078125,
          "max_proposal_regret_cost": 1.875,
          "optimal_proposal_fraction": 0.4765625,
          "initial_certificate_fraction": 0.0390625,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "frozen-seed29",
      "mode": "frozen",
      "seed": 29,
      "trainable_parameters": 25449,
      "selected_epoch": 78,
      "theta_change_norm": 0.0,
      "training_wall_seconds": 11.221429720986634,
      "inference_batch_ms": 2.5649628485552967,
      "cuda_peak_allocated_bytes": 183421440,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.24597454211365638,
          "source_cost_rmse": 0.35610038382857256,
          "mean_proposal_regret_cost": 0.044189453125,
          "max_proposal_regret_cost": 1.125,
          "optimal_proposal_fraction": 0.841796875,
          "initial_certificate_fraction": 0.212890625,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.2549261882018437,
          "source_cost_rmse": 0.3613161816450009,
          "mean_proposal_regret_cost": 0.03466796875,
          "max_proposal_regret_cost": 1.1875,
          "optimal_proposal_fraction": 0.8984375,
          "initial_certificate_fraction": 0.2578125,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.291917861613602,
          "source_cost_rmse": 0.4051724266443309,
          "mean_proposal_regret_cost": 0.044921875,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.8125,
          "initial_certificate_fraction": 0.2109375,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.7429843439657833,
          "source_cost_rmse": 0.913988876988274,
          "mean_proposal_regret_cost": 0.26171875,
          "max_proposal_regret_cost": 1.875,
          "optimal_proposal_fraction": 0.5234375,
          "initial_certificate_fraction": 0.0859375,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "quantum-seed43",
      "mode": "quantum",
      "seed": 43,
      "trainable_parameters": 25515,
      "selected_epoch": 77,
      "theta_change_norm": 0.15764325857162476,
      "training_wall_seconds": 14.13036054908298,
      "inference_batch_ms": 2.5576373445801437,
      "cuda_peak_allocated_bytes": 183426048,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.22389563895383,
          "source_cost_rmse": 0.3361150948195806,
          "mean_proposal_regret_cost": 0.033203125,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.86328125,
          "initial_certificate_fraction": 0.240234375,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.2444983722290313,
          "source_cost_rmse": 0.3680172593018272,
          "mean_proposal_regret_cost": 0.04736328125,
          "max_proposal_regret_cost": 1.3125,
          "optimal_proposal_fraction": 0.875,
          "initial_certificate_fraction": 0.2421875,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.26849118955904405,
          "source_cost_rmse": 0.3756302495058028,
          "mean_proposal_regret_cost": 0.03466796875,
          "max_proposal_regret_cost": 0.6875,
          "optimal_proposal_fraction": 0.84375,
          "initial_certificate_fraction": 0.25,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.5375539431006787,
          "source_cost_rmse": 0.6779233055767785,
          "mean_proposal_regret_cost": 0.220703125,
          "max_proposal_regret_cost": 1.875,
          "optimal_proposal_fraction": 0.5625,
          "initial_certificate_fraction": 0.109375,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "classical-seed43",
      "mode": "classical",
      "seed": 43,
      "trainable_parameters": 25515,
      "selected_epoch": 80,
      "theta_change_norm": 0.10721113532781601,
      "training_wall_seconds": 9.998661946039647,
      "inference_batch_ms": 1.8094555009156466,
      "cuda_peak_allocated_bytes": 182443008,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.22315672626185454,
          "source_cost_rmse": 0.3305933218221348,
          "mean_proposal_regret_cost": 0.0360107421875,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.857421875,
          "initial_certificate_fraction": 0.306640625,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.23779149613130063,
          "source_cost_rmse": 0.3571432739699932,
          "mean_proposal_regret_cost": 0.0244140625,
          "max_proposal_regret_cost": 0.5,
          "optimal_proposal_fraction": 0.8828125,
          "initial_certificate_fraction": 0.3671875,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.25726950423869144,
          "source_cost_rmse": 0.35184279791582646,
          "mean_proposal_regret_cost": 0.029296875,
          "max_proposal_regret_cost": 0.5625,
          "optimal_proposal_fraction": 0.8515625,
          "initial_certificate_fraction": 0.3671875,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.6552234295155629,
          "source_cost_rmse": 0.8168551866073767,
          "mean_proposal_regret_cost": 0.18212890625,
          "max_proposal_regret_cost": 1.75,
          "optimal_proposal_fraction": 0.5546875,
          "initial_certificate_fraction": 0.09375,
          "final_certified_fraction": 1.0
        }
      }
    },
    {
      "name": "frozen-seed43",
      "mode": "frozen",
      "seed": 43,
      "trainable_parameters": 25449,
      "selected_epoch": 79,
      "theta_change_norm": 0.0,
      "training_wall_seconds": 11.319593856111169,
      "inference_batch_ms": 2.5727179599925876,
      "cuda_peak_allocated_bytes": 183421440,
      "metrics": {
        "train": {
          "node_rmse_cost": 0.22658661187875903,
          "source_cost_rmse": 0.32876963627598443,
          "mean_proposal_regret_cost": 0.041748046875,
          "max_proposal_regret_cost": 0.875,
          "optimal_proposal_fraction": 0.8359375,
          "initial_certificate_fraction": 0.224609375,
          "final_certified_fraction": 1.0
        },
        "validation": {
          "node_rmse_cost": 0.24575807748337586,
          "source_cost_rmse": 0.3554108692197271,
          "mean_proposal_regret_cost": 0.04248046875,
          "max_proposal_regret_cost": 1.1875,
          "optimal_proposal_fraction": 0.8671875,
          "initial_certificate_fraction": 0.2421875,
          "final_certified_fraction": 1.0
        },
        "test": {
          "node_rmse_cost": 0.2707580620753281,
          "source_cost_rmse": 0.3664141398220659,
          "mean_proposal_regret_cost": 0.03515625,
          "max_proposal_regret_cost": 0.8125,
          "optimal_proposal_fraction": 0.8515625,
          "initial_certificate_fraction": 0.25,
          "final_certified_fraction": 1.0
        },
        "ood": {
          "node_rmse_cost": 0.6059542032647672,
          "source_cost_rmse": 0.7831898546721946,
          "mean_proposal_regret_cost": 0.2109375,
          "max_proposal_regret_cost": 1.75,
          "optimal_proposal_fraction": 0.5234375,
          "initial_certificate_fraction": 0.125,
          "final_certified_fraction": 1.0
        }
      }
    }
  ],
  "scope": "Original bounded3-qubit-per-token graph learning experiment; exact DP repair is separate from learned proposal success; sample standard deviation over3seeds."
}
