{
  "schema_version": 1,
  "model": "Qwen3-0.6B",
  "revision": "6130ef31402718485ca4d80a6234f70d9a4cf362",
  "experiments": {
    "full": {
      "initial": {
        "all": {
          "count": 240,
          "bce": 2.797879710263199,
          "brier": 0.44017959982352844,
          "accuracy": 0.5125
        },
        "singles": {
          "count": 80,
          "bce": 2.5989240627105574,
          "brier": 0.4524544972244879,
          "accuracy": 0.4875
        },
        "compounds": {
          "count": 160,
          "bce": 2.89735753403952,
          "brier": 0.4340421511230487,
          "accuracy": 0.525
        },
        "consistency_mean_squared_residual": 0.26970594965789746
      },
      "arms": {
        "trained_pairs": [
          {
            "epoch": 1,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6912725896884998,
                "brier": 0.24877889092481087,
                "accuracy": 0.55
              },
              "singles": {
                "count": 80,
                "bce": 0.698501494526863,
                "brier": 0.2525475204182717,
                "accuracy": 0.5125
              },
              "compounds": {
                "count": 160,
                "bce": 0.6876581372693181,
                "brier": 0.2468945761780804,
                "accuracy": 0.56875
              },
              "consistency_mean_squared_residual": 0.01645940415757584
            },
            "online_train_bce": 0.8700355769090893,
            "online_train_consistency": 0.088222957701116,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 2,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6784292731434107,
                "brier": 0.24272720767306039,
                "accuracy": 0.6
              },
              "singles": {
                "count": 80,
                "bce": 0.6728713203221559,
                "brier": 0.24021277695266074,
                "accuracy": 0.5875
              },
              "compounds": {
                "count": 160,
                "bce": 0.681208249554038,
                "brier": 0.24398442303326023,
                "accuracy": 0.60625
              },
              "consistency_mean_squared_residual": 0.01550376633727353
            },
            "online_train_bce": 0.7090116314291954,
            "online_train_consistency": 0.03302131691408704,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 3,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6651852332055569,
                "brier": 0.23659983209814903,
                "accuracy": 0.5916666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.6690574835985899,
                "brier": 0.23830150932401684,
                "accuracy": 0.5875
              },
              "compounds": {
                "count": 160,
                "bce": 0.6632491080090404,
                "brier": 0.2357489934852151,
                "accuracy": 0.59375
              },
              "consistency_mean_squared_residual": 0.020538321759299587
            },
            "online_train_bce": 0.6738235643357039,
            "online_train_consistency": 0.03182658189082972,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 4,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6751832305143277,
                "brier": 0.23994941651292487,
                "accuracy": 0.6083333333333333
              },
              "singles": {
                "count": 80,
                "bce": 0.655123658105731,
                "brier": 0.2322081918452985,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 0.685213016718626,
                "brier": 0.24382002884673803,
                "accuracy": 0.6
              },
              "consistency_mean_squared_residual": 0.027540484558916056
            },
            "online_train_bce": 0.6533930205106735,
            "online_train_consistency": 0.03511735076684033,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 5,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6531197127265235,
                "brier": 0.2287134168793762,
                "accuracy": 0.65
              },
              "singles": {
                "count": 80,
                "bce": 0.635483156889677,
                "brier": 0.22299667810619123,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 0.6619379906449467,
                "brier": 0.2315717862659687,
                "accuracy": 0.6625
              },
              "consistency_mean_squared_residual": 0.041875045605937765
            },
            "online_train_bce": 0.6314454612191767,
            "online_train_consistency": 0.05292515419042748,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 6,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6725992381883165,
                "brier": 0.23722440110093773,
                "accuracy": 0.625
              },
              "singles": {
                "count": 80,
                "bce": 0.7505759821273387,
                "brier": 0.2736695511014346,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 0.6336108662188054,
                "brier": 0.21900182610068927,
                "accuracy": 0.66875
              },
              "consistency_mean_squared_residual": 0.07645798480923348
            },
            "online_train_bce": 0.6047581194564701,
            "online_train_consistency": 0.07274628277885009,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 7,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7048459393903613,
                "brier": 0.23784925716932676,
                "accuracy": 0.6583333333333333
              },
              "singles": {
                "count": 80,
                "bce": 0.7435728173702956,
                "brier": 0.262903201108282,
                "accuracy": 0.5875
              },
              "compounds": {
                "count": 160,
                "bce": 0.6854825004003942,
                "brier": 0.2253222851998491,
                "accuracy": 0.69375
              },
              "consistency_mean_squared_residual": 0.07514777584253617
            },
            "online_train_bce": 0.5180205555278807,
            "online_train_consistency": 0.08576734493623683,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 8,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7606193039294643,
                "brier": 0.24685987639988863,
                "accuracy": 0.6541666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.7999645276577212,
                "brier": 0.27941148536961136,
                "accuracy": 0.5875
              },
              "compounds": {
                "count": 160,
                "bce": 0.7409466920653358,
                "brier": 0.23058407191502725,
                "accuracy": 0.6875
              },
              "consistency_mean_squared_residual": 0.1472353122534201
            },
            "online_train_bce": 0.4365049054492265,
            "online_train_consistency": 0.08533142963776276,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 9,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.8374117475197029,
                "brier": 0.2547511736700749,
                "accuracy": 0.6458333333333334
              },
              "singles": {
                "count": 80,
                "bce": 0.7873681813129224,
                "brier": 0.2599854128793661,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 0.8624335306230932,
                "brier": 0.2521340540654292,
                "accuracy": 0.65625
              },
              "consistency_mean_squared_residual": 0.12587043743369974
            },
            "online_train_bce": 0.37341824287746567,
            "online_train_consistency": 0.11927109681679819,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 10,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.1411160637516635,
                "brier": 0.2954688537069601,
                "accuracy": 0.625
              },
              "singles": {
                "count": 80,
                "bce": 1.280222963484175,
                "brier": 0.34725964297420925,
                "accuracy": 0.5625
              },
              "compounds": {
                "count": 160,
                "bce": 1.0715626138854077,
                "brier": 0.2695734590733355,
                "accuracy": 0.65625
              },
              "consistency_mean_squared_residual": 0.21782895864386997
            },
            "online_train_bce": 0.28744937091447353,
            "online_train_consistency": 0.1045376866994593,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 11,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.539519040103257,
                "brier": 0.3434825518842539,
                "accuracy": 0.5791666666666667
              },
              "singles": {
                "count": 80,
                "bce": 1.7007498369224776,
                "brier": 0.3855036006946173,
                "accuracy": 0.525
              },
              "compounds": {
                "count": 160,
                "bce": 1.4589036416936465,
                "brier": 0.3224720274790722,
                "accuracy": 0.60625
              },
              "consistency_mean_squared_residual": 0.2150597183996397
            },
            "online_train_bce": 0.2556025594144885,
            "online_train_consistency": 0.10757993735091967,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 12,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.3024355706501212,
                "brier": 0.26407574298461106,
                "accuracy": 0.6666666666666666
              },
              "singles": {
                "count": 80,
                "bce": 1.1136987867859716,
                "brier": 0.26471320534121323,
                "accuracy": 0.6625
              },
              "compounds": {
                "count": 160,
                "bce": 1.396803962582196,
                "brier": 0.26375701180630995,
                "accuracy": 0.66875
              },
              "consistency_mean_squared_residual": 0.24028849681657774
            },
            "online_train_bce": 0.18135354359155487,
            "online_train_consistency": 0.09216108997542381,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 13,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 2.011862748199651,
                "brier": 0.33487785634724204,
                "accuracy": 0.625
              },
              "singles": {
                "count": 80,
                "bce": 1.9086057923994304,
                "brier": 0.3868805046599705,
                "accuracy": 0.5625
              },
              "compounds": {
                "count": 160,
                "bce": 2.063491226099761,
                "brier": 0.30887653219087785,
                "accuracy": 0.65625
              },
              "consistency_mean_squared_residual": 0.2722556116199634
            },
            "online_train_bce": 0.15222653928454694,
            "online_train_consistency": 0.07521749778730186,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 14,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.390415902547034,
                "brier": 0.282861645778398,
                "accuracy": 0.65
              },
              "singles": {
                "count": 80,
                "bce": 1.226921251896647,
                "brier": 0.3043845800859416,
                "accuracy": 0.6125
              },
              "compounds": {
                "count": 160,
                "bce": 1.4721632278722274,
                "brier": 0.27210017862462615,
                "accuracy": 0.66875
              },
              "consistency_mean_squared_residual": 0.2392183324627149
            },
            "online_train_bce": 0.1680347861685418,
            "online_train_consistency": 0.07957240797947651,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 15,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.7976772821084377,
                "brier": 0.3056101568516621,
                "accuracy": 0.6583333333333333
              },
              "singles": {
                "count": 80,
                "bce": 1.9261352901449755,
                "brier": 0.3827216131025323,
                "accuracy": 0.575
              },
              "compounds": {
                "count": 160,
                "bce": 1.7334482780901688,
                "brier": 0.267054428726227,
                "accuracy": 0.7
              },
              "consistency_mean_squared_residual": 0.3061342606902169
            },
            "online_train_bce": 0.11935167519251501,
            "online_train_consistency": 0.06227658872920817,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 16,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 2.216978316529371,
                "brier": 0.3263631367244487,
                "accuracy": 0.6416666666666667
              },
              "singles": {
                "count": 80,
                "bce": 2.334272910506131,
                "brier": 0.3833361148851177,
                "accuracy": 0.575
              },
              "compounds": {
                "count": 160,
                "bce": 2.1583310195409906,
                "brier": 0.29787664764411426,
                "accuracy": 0.675
              },
              "consistency_mean_squared_residual": 0.2889504754615689
            },
            "online_train_bce": 0.18937085365203918,
            "online_train_consistency": 0.07923085255648017,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 17,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 2.0337006579400376,
                "brier": 0.2893893477411182,
                "accuracy": 0.6875
              },
              "singles": {
                "count": 80,
                "bce": 1.6776074487620172,
                "brier": 0.32218811872062575,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 2.211747262529048,
                "brier": 0.2729899622513644,
                "accuracy": 0.71875
              },
              "consistency_mean_squared_residual": 0.25932546314791427
            },
            "online_train_bce": 0.12789308705505126,
            "online_train_consistency": 0.07580642961221552,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 18,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 2.1392702062173,
                "brier": 0.30205941002263575,
                "accuracy": 0.6666666666666666
              },
              "singles": {
                "count": 80,
                "bce": 2.734093774735495,
                "brier": 0.4152853826370665,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 1.841858421958203,
                "brier": 0.2454464237154203,
                "accuracy": 0.73125
              },
              "consistency_mean_squared_residual": 0.2658120575522531
            },
            "online_train_bce": 0.12123284367694338,
            "online_train_consistency": 0.06060933452014107,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 19,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 2.5141952279698327,
                "brier": 0.30136402168150983,
                "accuracy": 0.6708333333333333
              },
              "singles": {
                "count": 80,
                "bce": 2.634057778830233,
                "brier": 0.36332109704761273,
                "accuracy": 0.6
              },
              "compounds": {
                "count": 160,
                "bce": 2.4542639525396326,
                "brier": 0.2703854839984584,
                "accuracy": 0.70625
              },
              "consistency_mean_squared_residual": 0.25181865851921714
            },
            "online_train_bce": 0.08209334940898169,
            "online_train_consistency": 0.048548789508771784,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 20,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 2.605278405919127,
                "brier": 0.31971765023482707,
                "accuracy": 0.6583333333333333
              },
              "singles": {
                "count": 80,
                "bce": 3.0607192401844143,
                "brier": 0.4254081346431642,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 2.3775579887864833,
                "brier": 0.26687240803065854,
                "accuracy": 0.7125
              },
              "consistency_mean_squared_residual": 0.3133243185931374
            },
            "online_train_bce": 0.10451867091792386,
            "online_train_consistency": 0.04574096209959903,
            "reload_logit_delta": 0.0
          }
        ],
        "trained_pairs_consistency": [
          {
            "epoch": 1,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6858528750638168,
                "brier": 0.2463882498890775,
                "accuracy": 0.5375
              },
              "singles": {
                "count": 80,
                "bce": 0.6908663302659989,
                "brier": 0.24889129065276636,
                "accuracy": 0.4875
              },
              "compounds": {
                "count": 160,
                "bce": 0.6833461474627256,
                "brier": 0.24513672950723303,
                "accuracy": 0.5625
              },
              "consistency_mean_squared_residual": 0.007210918521293033
            },
            "online_train_bce": 0.8513709255217181,
            "online_train_consistency": 0.07308507667824089,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 2,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6724238649010659,
                "brier": 0.23976203751181738,
                "accuracy": 0.5875
              },
              "singles": {
                "count": 80,
                "bce": 0.6673522826284171,
                "brier": 0.23729190332594968,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 0.6749596560373903,
                "brier": 0.2409971046047512,
                "accuracy": 0.56875
              },
              "consistency_mean_squared_residual": 0.005447118445927236
            },
            "online_train_bce": 0.722751635171473,
            "online_train_consistency": 0.028075928483527406,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 3,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6747750665992498,
                "brier": 0.24113884032507477,
                "accuracy": 0.5666666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.6829064652323723,
                "brier": 0.24507056670088806,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 0.6707093672826886,
                "brier": 0.2391729771371681,
                "accuracy": 0.575
              },
              "consistency_mean_squared_residual": 0.013628334178358391
            },
            "online_train_bce": 0.6851360881850124,
            "online_train_consistency": 0.020495293355528996,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 4,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6489413851251205,
                "brier": 0.22846922880875978,
                "accuracy": 0.625
              },
              "singles": {
                "count": 80,
                "bce": 0.6525195680558682,
                "brier": 0.23021355990804335,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 0.6471522936597467,
                "brier": 0.22759706325911794,
                "accuracy": 0.625
              },
              "consistency_mean_squared_residual": 0.009804644444044138
            },
            "online_train_bce": 0.6565514976158738,
            "online_train_consistency": 0.011086096619681693,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 5,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6460438410441081,
                "brier": 0.22714625560912174,
                "accuracy": 0.6291666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.656113475561142,
                "brier": 0.23219886384296923,
                "accuracy": 0.5875
              },
              "compounds": {
                "count": 160,
                "bce": 0.6410090237855911,
                "brier": 0.224619951492198,
                "accuracy": 0.65
              },
              "consistency_mean_squared_residual": 0.020266167739119937
            },
            "online_train_bce": 0.64965324402228,
            "online_train_consistency": 0.023403367251309534,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 6,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.638258077390492,
                "brier": 0.2233481963810852,
                "accuracy": 0.6375
              },
              "singles": {
                "count": 80,
                "bce": 0.6710412975400686,
                "brier": 0.24032834425364094,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 0.6218664673157036,
                "brier": 0.21485812244480731,
                "accuracy": 0.68125
              },
              "consistency_mean_squared_residual": 0.028045463512721844
            },
            "online_train_bce": 0.6165920368488877,
            "online_train_consistency": 0.02856850684503267,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 7,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6329470505937934,
                "brier": 0.21673087140437353,
                "accuracy": 0.6875
              },
              "singles": {
                "count": 80,
                "bce": 0.68070667386055,
                "brier": 0.24366062243083153,
                "accuracy": 0.575
              },
              "compounds": {
                "count": 160,
                "bce": 0.6090672389604151,
                "brier": 0.20326599589114452,
                "accuracy": 0.74375
              },
              "consistency_mean_squared_residual": 0.038309397476144626
            },
            "online_train_bce": 0.5474252840783447,
            "online_train_consistency": 0.03524779673407367,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 8,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.621567866485566,
                "brier": 0.21212000405131048,
                "accuracy": 0.6791666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.6405022306367755,
                "brier": 0.2253450168998179,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 0.6121006844099611,
                "brier": 0.20550749762705678,
                "accuracy": 0.70625
              },
              "consistency_mean_squared_residual": 0.05360753123809744
            },
            "online_train_bce": 0.4973309539053589,
            "online_train_consistency": 0.04242562274610843,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 9,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6790092518553138,
                "brier": 0.22455367843424912,
                "accuracy": 0.6708333333333333
              },
              "singles": {
                "count": 80,
                "bce": 0.6958753482438624,
                "brier": 0.24937678462009383,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 0.6705762036610394,
                "brier": 0.21214212534132676,
                "accuracy": 0.7375
              },
              "consistency_mean_squared_residual": 0.12768090701636162
            },
            "online_train_bce": 0.4422362478082068,
            "online_train_consistency": 0.05031257512326079,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 10,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.6697997413381623,
                "brier": 0.2192244617018041,
                "accuracy": 0.7
              },
              "singles": {
                "count": 80,
                "bce": 0.7213038931600749,
                "brier": 0.25210429398908873,
                "accuracy": 0.6125
              },
              "compounds": {
                "count": 160,
                "bce": 0.644047665427206,
                "brier": 0.20278454555816178,
                "accuracy": 0.74375
              },
              "consistency_mean_squared_residual": 0.09868803738650252
            },
            "online_train_bce": 0.3797718913895078,
            "online_train_consistency": 0.0495080318712896,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 11,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.8995163983792736,
                "brier": 0.2644899090938223,
                "accuracy": 0.625
              },
              "singles": {
                "count": 80,
                "bce": 1.0569489147106652,
                "brier": 0.31507798998626874,
                "accuracy": 0.5625
              },
              "compounds": {
                "count": 160,
                "bce": 0.8208001402135778,
                "brier": 0.23919586864759906,
                "accuracy": 0.65625
              },
              "consistency_mean_squared_residual": 0.17528139156108002
            },
            "online_train_bce": 0.3171051122047938,
            "online_train_consistency": 0.0626928274463581,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 12,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.8132625088782031,
                "brier": 0.2438452127748381,
                "accuracy": 0.6791666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.8260781587974634,
                "brier": 0.2640426056251968,
                "accuracy": 0.6375
              },
              "compounds": {
                "count": 160,
                "bce": 0.8068546839185728,
                "brier": 0.23374651634965873,
                "accuracy": 0.7
              },
              "consistency_mean_squared_residual": 0.17527264125969097
            },
            "online_train_bce": 0.3168611138663255,
            "online_train_consistency": 0.06573254934138982,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 13,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.117344486264877,
                "brier": 0.2881781153328163,
                "accuracy": 0.6375
              },
              "singles": {
                "count": 80,
                "bce": 1.1464489900310582,
                "brier": 0.3309140776887192,
                "accuracy": 0.575
              },
              "compounds": {
                "count": 160,
                "bce": 1.1027922343817864,
                "brier": 0.2668101341548648,
                "accuracy": 0.66875
              },
              "consistency_mean_squared_residual": 0.21248332473492915
            },
            "online_train_bce": 0.2424991555675515,
            "online_train_consistency": 0.07297411633268412,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 14,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.2727445882821788,
                "brier": 0.27370571440160496,
                "accuracy": 0.65
              },
              "singles": {
                "count": 80,
                "bce": 1.130817635008134,
                "brier": 0.3042492681075169,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 1.3437080649192013,
                "brier": 0.258433937548649,
                "accuracy": 0.7
              },
              "consistency_mean_squared_residual": 0.2218148207245331
            },
            "online_train_bce": 0.18102024189819008,
            "online_train_consistency": 0.047730903066453965,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 15,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.1097636039459,
                "brier": 0.27486663200653116,
                "accuracy": 0.6708333333333333
              },
              "singles": {
                "count": 80,
                "bce": 1.2262635563539335,
                "brier": 0.33361499778477094,
                "accuracy": 0.575
              },
              "compounds": {
                "count": 160,
                "bce": 1.0515136277418833,
                "brier": 0.2454924491174113,
                "accuracy": 0.71875
              },
              "consistency_mean_squared_residual": 0.23615640481660313
            },
            "online_train_bce": 0.17651366597613583,
            "online_train_consistency": 0.06464040100282992,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 16,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.18084035784885,
                "brier": 0.2622163066980739,
                "accuracy": 0.6791666666666667
              },
              "singles": {
                "count": 80,
                "bce": 1.2152634781219604,
                "brier": 0.3140623296561463,
                "accuracy": 0.575
              },
              "compounds": {
                "count": 160,
                "bce": 1.1636287977122948,
                "brier": 0.23629329521903766,
                "accuracy": 0.73125
              },
              "consistency_mean_squared_residual": 0.22513600830525213
            },
            "online_train_bce": 0.13013237538144448,
            "online_train_consistency": 0.060617828063138784,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 17,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.5957437954519642,
                "brier": 0.2712891177227644,
                "accuracy": 0.6833333333333333
              },
              "singles": {
                "count": 80,
                "bce": 1.3804146400202626,
                "brier": 0.29523745341779123,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 1.7034083731678151,
                "brier": 0.259314949875251,
                "accuracy": 0.7125
              },
              "consistency_mean_squared_residual": 0.26284974813264467
            },
            "online_train_bce": 0.10342015212699505,
            "online_train_consistency": 0.029866594306349612,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 18,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.7686559001978392,
                "brier": 0.30682622265836207,
                "accuracy": 0.65
              },
              "singles": {
                "count": 80,
                "bce": 2.2141526985094173,
                "brier": 0.382759161622086,
                "accuracy": 0.5625
              },
              "compounds": {
                "count": 160,
                "bce": 1.5459075010420502,
                "brier": 0.26885975317650007,
                "accuracy": 0.69375
              },
              "consistency_mean_squared_residual": 0.2229922901541698
            },
            "online_train_bce": 0.12822506481158272,
            "online_train_consistency": 0.06613326209064405,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 19,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.8190542272046595,
                "brier": 0.30026328412983344,
                "accuracy": 0.6541666666666667
              },
              "singles": {
                "count": 80,
                "bce": 2.00841132036237,
                "brier": 0.3619352586944317,
                "accuracy": 0.5875
              },
              "compounds": {
                "count": 160,
                "bce": 1.7243756806258042,
                "brier": 0.2694272968475343,
                "accuracy": 0.6875
              },
              "consistency_mean_squared_residual": 0.3212853905745128
            },
            "online_train_bce": 0.14386428100445808,
            "online_train_consistency": 0.08116597757507614,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 20,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.9560574391636658,
                "brier": 0.276058614377741,
                "accuracy": 0.6916666666666667
              },
              "singles": {
                "count": 80,
                "bce": 1.8511595179635691,
                "brier": 0.3117603796372024,
                "accuracy": 0.625
              },
              "compounds": {
                "count": 160,
                "bce": 2.0085063997637143,
                "brier": 0.25820773174801037,
                "accuracy": 0.725
              },
              "consistency_mean_squared_residual": 0.22255777878123456
            },
            "online_train_bce": 0.12104937226158265,
            "online_train_consistency": 0.06513511843489582,
            "reload_logit_delta": 0.0
          }
        ]
      }
    },
    "one": {
      "initial": {
        "all": {
          "count": 240,
          "bce": 2.797879710263199,
          "brier": 0.44017959982352844,
          "accuracy": 0.5125
        },
        "singles": {
          "count": 80,
          "bce": 2.5989240627105574,
          "brier": 0.4524544972244879,
          "accuracy": 0.4875
        },
        "compounds": {
          "count": 160,
          "bce": 2.89735753403952,
          "brier": 0.4340421511230487,
          "accuracy": 0.525
        },
        "consistency_mean_squared_residual": 0.26970594965789746
      },
      "arms": {
        "trained_pairs": [
          {
            "epoch": 1,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7282227860142787,
                "brier": 0.2657061639866628,
                "accuracy": 0.5
              },
              "singles": {
                "count": 80,
                "bce": 0.75650639757514,
                "brier": 0.2777940890114784,
                "accuracy": 0.525
              },
              "compounds": {
                "count": 160,
                "bce": 0.7140809802338481,
                "brier": 0.259662201474255,
                "accuracy": 0.4875
              },
              "consistency_mean_squared_residual": 0.029989932527007854
            },
            "masked_training": {
              "count": 500,
              "bce": 0.7119221907258034,
              "brier": 0.2577294744566241,
              "accuracy": 0.56
            },
            "online_train_bce": 1.0269259251215772,
            "online_train_consistency": 0.1146136517453499,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 2,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7315568106248975,
                "brier": 0.2657594524754393,
                "accuracy": 0.55
              },
              "singles": {
                "count": 80,
                "bce": 0.7879516130313278,
                "brier": 0.2896249304008255,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 0.7033594094216824,
                "brier": 0.2538267135127462,
                "accuracy": 0.55
              },
              "consistency_mean_squared_residual": 0.0803360281452156
            },
            "masked_training": {
              "count": 500,
              "bce": 0.7302600430548191,
              "brier": 0.26462164698075563,
              "accuracy": 0.558
            },
            "online_train_bce": 0.7937408776953816,
            "online_train_consistency": 0.0322341072679567,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 3,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.72397865789632,
                "brier": 0.26325659303341425,
                "accuracy": 0.5333333333333333
              },
              "singles": {
                "count": 80,
                "bce": 0.7737860463559627,
                "brier": 0.28521362078129286,
                "accuracy": 0.525
              },
              "compounds": {
                "count": 160,
                "bce": 0.6990749636664987,
                "brier": 0.25227807915947487,
                "accuracy": 0.5375
              },
              "consistency_mean_squared_residual": 0.030178769831859065
            },
            "masked_training": {
              "count": 500,
              "bce": 0.6950730970799923,
              "brier": 0.2501191895954406,
              "accuracy": 0.548
            },
            "online_train_bce": 0.624194153137505,
            "online_train_consistency": 0.05972870982289351,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 4,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7801739279801647,
                "brier": 0.27877279159225427,
                "accuracy": 0.55
              },
              "singles": {
                "count": 80,
                "bce": 0.8090580098330975,
                "brier": 0.29385801327760985,
                "accuracy": 0.5
              },
              "compounds": {
                "count": 160,
                "bce": 0.7657318870536983,
                "brier": 0.27123018074957644,
                "accuracy": 0.575
              },
              "consistency_mean_squared_residual": 0.0739478893276534
            },
            "masked_training": {
              "count": 500,
              "bce": 0.766945435628295,
              "brier": 0.27323828919261123,
              "accuracy": 0.564
            },
            "online_train_bce": 0.5875598385185004,
            "online_train_consistency": 0.04691442930574283,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 5,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.9292507452579836,
                "brier": 0.31282426307359995,
                "accuracy": 0.525
              },
              "singles": {
                "count": 80,
                "bce": 1.1986268103588373,
                "brier": 0.37674274542409203,
                "accuracy": 0.4875
              },
              "compounds": {
                "count": 160,
                "bce": 0.7945627127075567,
                "brier": 0.28086502189835383,
                "accuracy": 0.54375
              },
              "consistency_mean_squared_residual": 0.16148922410213346
            },
            "masked_training": {
              "count": 500,
              "bce": 1.0097588803563267,
              "brier": 0.32746562704899596,
              "accuracy": 0.554
            },
            "online_train_bce": 0.47036977142794056,
            "online_train_consistency": 0.1107922143466795,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 6,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.8768239394761622,
                "brier": 0.3037314039063046,
                "accuracy": 0.5208333333333334
              },
              "singles": {
                "count": 80,
                "bce": 0.9134990317746997,
                "brier": 0.32262044124808625,
                "accuracy": 0.4375
              },
              "compounds": {
                "count": 160,
                "bce": 0.8584863933268935,
                "brier": 0.2942868852354138,
                "accuracy": 0.5625
              },
              "consistency_mean_squared_residual": 0.11476548073196027
            },
            "masked_training": {
              "count": 500,
              "bce": 0.8557767989039421,
              "brier": 0.29674358422064323,
              "accuracy": 0.542
            },
            "online_train_bce": 0.5113462197757326,
            "online_train_consistency": 0.1494703031620575,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 7,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.4102849007584155,
                "brier": 0.3777781617053449,
                "accuracy": 0.5375
              },
              "singles": {
                "count": 80,
                "bce": 1.1072347325272858,
                "brier": 0.33436356994734917,
                "accuracy": 0.5875
              },
              "compounds": {
                "count": 160,
                "bce": 1.5618099848739804,
                "brier": 0.3994854575843427,
                "accuracy": 0.5125
              },
              "consistency_mean_squared_residual": 0.26606797332728704
            },
            "masked_training": {
              "count": 500,
              "bce": 1.2391046124193816,
              "brier": 0.3648829223850491,
              "accuracy": 0.504
            },
            "online_train_bce": 0.45669589593671844,
            "online_train_consistency": 0.16436309924817213,
            "reload_logit_delta": 0.0
          }
        ],
        "trained_pairs_consistency": [
          {
            "epoch": 1,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7107202423736453,
                "brier": 0.25814376593713245,
                "accuracy": 0.5083333333333333
              },
              "singles": {
                "count": 80,
                "bce": 0.7262479418888688,
                "brier": 0.26492585236903854,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 0.7029563926160336,
                "brier": 0.2547527227211793,
                "accuracy": 0.49375
              },
              "consistency_mean_squared_residual": 0.027181599264313884
            },
            "masked_training": {
              "count": 500,
              "bce": 0.6962558002471924,
              "brier": 0.2510286668733105,
              "accuracy": 0.558
            },
            "online_train_bce": 1.009421896916283,
            "online_train_consistency": 0.06715826545309211,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 2,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7805559674898783,
                "brier": 0.28707026041833844,
                "accuracy": 0.4625
              },
              "singles": {
                "count": 80,
                "bce": 0.7521415695548057,
                "brier": 0.27672019563464817,
                "accuracy": 0.3875
              },
              "compounds": {
                "count": 160,
                "bce": 0.7947631664574146,
                "brier": 0.2922452928101836,
                "accuracy": 0.5
              },
              "consistency_mean_squared_residual": 0.044777818783774105
            },
            "masked_training": {
              "count": 500,
              "bce": 0.7725240466594696,
              "brier": 0.2850269883753478,
              "accuracy": 0.458
            },
            "online_train_bce": 0.69074115306139,
            "online_train_consistency": 0.03284830819258161,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 3,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7103739934042096,
                "brier": 0.2577143170716877,
                "accuracy": 0.5291666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.7357348725199699,
                "brier": 0.26901120422728847,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 0.6976935538463295,
                "brier": 0.2520658734938873,
                "accuracy": 0.51875
              },
              "consistency_mean_squared_residual": 0.019074471635780407
            },
            "masked_training": {
              "count": 500,
              "bce": 0.6881927189528942,
              "brier": 0.2476577166460212,
              "accuracy": 0.544
            },
            "online_train_bce": 0.6496737927570939,
            "online_train_consistency": 0.012579378286293377,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 4,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7167795719578862,
                "brier": 0.2589590317049423,
                "accuracy": 0.5666666666666667
              },
              "singles": {
                "count": 80,
                "bce": 0.7480419300496578,
                "brier": 0.2737711183710509,
                "accuracy": 0.5125
              },
              "compounds": {
                "count": 160,
                "bce": 0.7011483929120004,
                "brier": 0.25155298837188794,
                "accuracy": 0.59375
              },
              "consistency_mean_squared_residual": 0.03299840542573568
            },
            "masked_training": {
              "count": 500,
              "bce": 0.7087644826769829,
              "brier": 0.2553443526388318,
              "accuracy": 0.584
            },
            "online_train_bce": 0.6132138941623271,
            "online_train_consistency": 0.018274644489184765,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 5,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7180139424589773,
                "brier": 0.25988417865884633,
                "accuracy": 0.575
              },
              "singles": {
                "count": 80,
                "bce": 0.7733029527589679,
                "brier": 0.2844266195596313,
                "accuracy": 0.5125
              },
              "compounds": {
                "count": 160,
                "bce": 0.690369437308982,
                "brier": 0.24761295820845386,
                "accuracy": 0.60625
              },
              "consistency_mean_squared_residual": 0.059048227857188176
            },
            "masked_training": {
              "count": 500,
              "bce": 0.6998500396162272,
              "brier": 0.25155228181231853,
              "accuracy": 0.562
            },
            "online_train_bce": 0.5324606975913048,
            "online_train_consistency": 0.02593888111816213,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 6,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.7215013353619725,
                "brier": 0.25835048947608735,
                "accuracy": 0.6041666666666666
              },
              "singles": {
                "count": 80,
                "bce": 0.8004489971324802,
                "brier": 0.2931572948386245,
                "accuracy": 0.525
              },
              "compounds": {
                "count": 160,
                "bce": 0.6820275044767186,
                "brier": 0.24094708679481874,
                "accuracy": 0.64375
              },
              "consistency_mean_squared_residual": 0.06846306707978272
            },
            "masked_training": {
              "count": 500,
              "bce": 0.6861591701284051,
              "brier": 0.2427640531302066,
              "accuracy": 0.602
            },
            "online_train_bce": 0.5142884172452614,
            "online_train_consistency": 0.035828589515401174,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 7,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.788197694416158,
                "brier": 0.27781122428633886,
                "accuracy": 0.575
              },
              "singles": {
                "count": 80,
                "bce": 0.8192545361816883,
                "brier": 0.2931450861582746,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 0.7726692735333927,
                "brier": 0.27014429335037105,
                "accuracy": 0.5875
              },
              "consistency_mean_squared_residual": 0.09659540351682121
            },
            "masked_training": {
              "count": 500,
              "bce": 0.7112594054266811,
              "brier": 0.24792827244358215,
              "accuracy": 0.606
            },
            "online_train_bce": 0.38974821574927776,
            "online_train_consistency": 0.07554774080675088,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 8,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.8181726142463351,
                "brier": 0.27955445623409325,
                "accuracy": 0.6083333333333333
              },
              "singles": {
                "count": 80,
                "bce": 0.9562138020992279,
                "brier": 0.3339641675881435,
                "accuracy": 0.525
              },
              "compounds": {
                "count": 160,
                "bce": 0.7491520203198888,
                "brier": 0.2523496005570681,
                "accuracy": 0.65
              },
              "consistency_mean_squared_residual": 0.12007789851546498
            },
            "masked_training": {
              "count": 500,
              "bce": 0.7156787494653836,
              "brier": 0.2437013476828792,
              "accuracy": 0.636
            },
            "online_train_bce": 0.30587712115142496,
            "online_train_consistency": 0.060739139624225516,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 9,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.8999742304324172,
                "brier": 0.28885066232994683,
                "accuracy": 0.6125
              },
              "singles": {
                "count": 80,
                "bce": 1.1655285601038485,
                "brier": 0.3758261277109626,
                "accuracy": 0.4875
              },
              "compounds": {
                "count": 160,
                "bce": 0.7671970655967015,
                "brier": 0.24536292963943898,
                "accuracy": 0.675
              },
              "consistency_mean_squared_residual": 0.19574424954290537
            },
            "masked_training": {
              "count": 500,
              "bce": 0.7585776502140797,
              "brier": 0.24866379055129934,
              "accuracy": 0.644
            },
            "online_train_bce": 0.2344151316105854,
            "online_train_consistency": 0.08899258340424239,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 10,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 0.9014485469674886,
                "brier": 0.2848464856843408,
                "accuracy": 0.6125
              },
              "singles": {
                "count": 80,
                "bce": 1.0786516824737191,
                "brier": 0.34038882350335553,
                "accuracy": 0.5
              },
              "compounds": {
                "count": 160,
                "bce": 0.8128469792143733,
                "brier": 0.2570753167748335,
                "accuracy": 0.66875
              },
              "consistency_mean_squared_residual": 0.12261563281858123
            },
            "masked_training": {
              "count": 500,
              "bce": 1.0085119497890118,
              "brier": 0.2943725747417964,
              "accuracy": 0.608
            },
            "online_train_bce": 0.22096310122477006,
            "online_train_consistency": 0.07040396038442975,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 11,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.040046605115155,
                "brier": 0.31376920718952367,
                "accuracy": 0.5958333333333333
              },
              "singles": {
                "count": 80,
                "bce": 1.2338341071270407,
                "brier": 0.3842562419358117,
                "accuracy": 0.5
              },
              "compounds": {
                "count": 160,
                "bce": 0.943152854109212,
                "brier": 0.2785256898163796,
                "accuracy": 0.64375
              },
              "consistency_mean_squared_residual": 0.14672082638149478
            },
            "masked_training": {
              "count": 500,
              "bce": 0.8331332823993871,
              "brier": 0.25794591831536684,
              "accuracy": 0.642
            },
            "online_train_bce": 0.39740959772403583,
            "online_train_consistency": 0.09368932328550728,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 12,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.3549420283858733,
                "brier": 0.344807181703685,
                "accuracy": 0.5625
              },
              "singles": {
                "count": 80,
                "bce": 1.689118728655012,
                "brier": 0.39126679756933325,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 1.1878536782513038,
                "brier": 0.3215773737708609,
                "accuracy": 0.56875
              },
              "consistency_mean_squared_residual": 0.1947597416020275
            },
            "masked_training": {
              "count": 500,
              "bce": 1.2620596601419747,
              "brier": 0.32424097947613306,
              "accuracy": 0.596
            },
            "online_train_bce": 0.16317406494625175,
            "online_train_consistency": 0.08114757343256909,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 13,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.2133905063181374,
                "brier": 0.31870322399369816,
                "accuracy": 0.6041666666666666
              },
              "singles": {
                "count": 80,
                "bce": 1.319540628680261,
                "brier": 0.3735331041323055,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 1.1603154451370756,
                "brier": 0.2912882839243945,
                "accuracy": 0.63125
              },
              "consistency_mean_squared_residual": 0.1980687515992789
            },
            "masked_training": {
              "count": 500,
              "bce": 0.9859466492788997,
              "brier": 0.26338458237486806,
              "accuracy": 0.662
            },
            "online_train_bce": 0.15389185555219229,
            "online_train_consistency": 0.07428259644554296,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 14,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.5245554151290739,
                "brier": 0.33587329733530846,
                "accuracy": 0.6041666666666666
              },
              "singles": {
                "count": 80,
                "bce": 1.8687765699149168,
                "brier": 0.39964924684830155,
                "accuracy": 0.55
              },
              "compounds": {
                "count": 160,
                "bce": 1.3524448377361522,
                "brier": 0.30398532257881195,
                "accuracy": 0.63125
              },
              "consistency_mean_squared_residual": 0.2305133335704446
            },
            "masked_training": {
              "count": 500,
              "bce": 1.1717442425510198,
              "brier": 0.27992163607842196,
              "accuracy": 0.656
            },
            "online_train_bce": 0.06258920108410165,
            "online_train_consistency": 0.07861407037868554,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 15,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.3966942172141785,
                "brier": 0.3262418718117441,
                "accuracy": 0.6041666666666666
              },
              "singles": {
                "count": 80,
                "bce": 1.587022002399317,
                "brier": 0.3807795404239699,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 1.3015303246216092,
                "brier": 0.29897303750563115,
                "accuracy": 0.6375
              },
              "consistency_mean_squared_residual": 0.2655190471723328
            },
            "masked_training": {
              "count": 500,
              "bce": 1.1212483402269435,
              "brier": 0.2647117574216445,
              "accuracy": 0.676
            },
            "online_train_bce": 0.0740258354580726,
            "online_train_consistency": 0.07944470788344535,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 16,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.6147496980303306,
                "brier": 0.359158758063724,
                "accuracy": 0.5875
              },
              "singles": {
                "count": 80,
                "bce": 1.549438691372052,
                "brier": 0.37599630849525023,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 1.64740520135947,
                "brier": 0.35073998284796093,
                "accuracy": 0.6125
              },
              "consistency_mean_squared_residual": 0.20317697648947852
            },
            "masked_training": {
              "count": 500,
              "bce": 1.4066114703121129,
              "brier": 0.3145951148759321,
              "accuracy": 0.622
            },
            "online_train_bce": 0.08742677452136377,
            "online_train_consistency": 0.09385904415530882,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 17,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.9598490126623234,
                "brier": 0.3551545006313761,
                "accuracy": 0.6
              },
              "singles": {
                "count": 80,
                "bce": 2.233700388665761,
                "brier": 0.43881505996272485,
                "accuracy": 0.4875
              },
              "compounds": {
                "count": 160,
                "bce": 1.8229233246606042,
                "brier": 0.3133242209657018,
                "accuracy": 0.65625
              },
              "consistency_mean_squared_residual": 0.22654918719977638
            },
            "masked_training": {
              "count": 500,
              "bce": 1.6143440279081196,
              "brier": 0.28752953545736293,
              "accuracy": 0.668
            },
            "online_train_bce": 0.05441721308417414,
            "online_train_consistency": 0.06829261733286393,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 18,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.250802196820905,
                "brier": 0.3195511470109326,
                "accuracy": 0.6041666666666666
              },
              "singles": {
                "count": 80,
                "bce": 1.351317834376823,
                "brier": 0.36359770654916523,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 1.200544378042946,
                "brier": 0.29752786724181635,
                "accuracy": 0.6375
              },
              "consistency_mean_squared_residual": 0.25004737232744056
            },
            "masked_training": {
              "count": 500,
              "bce": 1.0913713516092711,
              "brier": 0.2623462583388506,
              "accuracy": 0.676
            },
            "online_train_bce": 0.10486247220574114,
            "online_train_consistency": 0.11599385921554567,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 19,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 1.4979513462675924,
                "brier": 0.3360776258614249,
                "accuracy": 0.6166666666666667
              },
              "singles": {
                "count": 80,
                "bce": 1.7675331288606686,
                "brier": 0.4104814967143807,
                "accuracy": 0.5375
              },
              "compounds": {
                "count": 160,
                "bce": 1.3631604549710545,
                "brier": 0.29887569043494694,
                "accuracy": 0.65625
              },
              "consistency_mean_squared_residual": 0.2566793765024351
            },
            "masked_training": {
              "count": 500,
              "bce": 1.339972244543538,
              "brier": 0.2765089893821906,
              "accuracy": 0.684
            },
            "online_train_bce": 0.032552987146998474,
            "online_train_consistency": 0.06973058009111092,
            "reload_logit_delta": 0.0
          },
          {
            "epoch": 20,
            "metrics": {
              "all": {
                "count": 240,
                "bce": 2.225523833705834,
                "brier": 0.3732415991983974,
                "accuracy": 0.5833333333333334
              },
              "singles": {
                "count": 80,
                "bce": 2.844045914252669,
                "brier": 0.4349075681429956,
                "accuracy": 0.525
              },
              "compounds": {
                "count": 160,
                "bce": 1.9162627934324168,
                "brier": 0.3424086147260982,
                "accuracy": 0.6125
              },
              "consistency_mean_squared_residual": 0.26645934850564557
            },
            "masked_training": {
              "count": 500,
              "bce": 1.52833389474198,
              "brier": 0.28304756640629336,
              "accuracy": 0.674
            },
            "online_train_bce": 0.17919961594132067,
            "online_train_consistency": 0.09475719073234327,
            "reload_logit_delta": 0.0
          }
        ]
      }
    }
  },
  "sources": {
    "pilot375-full20-complete": "81244c454205c45919bbcf416f37510524119602e4e7b82a0263c9c7e63e9fce",
    "pilot375-one-label-stopped-v1": "7064b26a660beec6e5e4276fcd59aa2d820645cea64fb5ea84b8ed68f904d62a",
    "pilot375-one-label-consistency-complete": "274b1224ef6e30658fe8c3db1501d765f987f5cce57633a92ceb615c6b64d72c"
  },
  "smoke": {
    "source_document": "docs/fixed-kv-smoke.md",
    "source_sha256": "d0662bf77cdb15e49f984b43efb277af2d770b5549f23b8f5fc1cf6324836b7a",
    "train": [
      {
        "arm": "singles",
        "queries": 10,
        "updates": 1,
        "bce_before": 5.13617,
        "bce_after": 3.34181
      },
      {
        "arm": "pairs",
        "queries": 120,
        "updates": 1,
        "bce_before": 3.042312,
        "bce_after": 2.161691,
        "consistency_before": 0.361743,
        "consistency_after": 0.360404
      },
      {
        "arm": "pairs_consistency",
        "queries": 120,
        "updates": 1,
        "bce_before": 3.042312,
        "bce_after": 2.162989,
        "consistency_before": 0.361743,
        "consistency_after": 0.358503
      }
    ],
    "validation": [
      {
        "checkpoint": "initial",
        "singles20_bce": 3.232157,
        "pairs60_bce": 2.908728,
        "pairs60_brier": 0.448511,
        "pairs_consistency": 0.423986
      },
      {
        "checkpoint": "singles",
        "singles20_bce": 2.101527,
        "pairs60_bce": 2.106276,
        "pairs60_brier": 0.455081,
        "pairs_consistency": 0.457002
      },
      {
        "checkpoint": "pairs",
        "singles20_bce": 2.134721,
        "pairs60_bce": 2.006804,
        "pairs60_brier": 0.445607,
        "pairs_consistency": 0.429788
      },
      {
        "checkpoint": "pairs_consistency",
        "singles20_bce": 2.137763,
        "pairs60_bce": 2.011799,
        "pairs60_brier": 0.44593,
        "pairs_consistency": 0.428787
      }
    ]
  }
}
