{
  "schema": "biosim.solvent_extension_comparison_v4",
  "status": "completed_comparison_mixed_results_not_promoted",
  "incumbent_changed": false,
  "counts": {
    "new_solvent_training_configurations": 2056,
    "new_solvent_validation_configurations": 512,
    "new_validation_solute_groups": 64,
    "validation_records_each_arm": 4508,
    "updates_each_arm": 14464
  },
  "comparison": {
    "status": "completed_paired_solvent_force_and_energy_comparison",
    "records": 512,
    "solute_roots": 64,
    "metrics": {
      "incumbent": {
        "all_force_component_rmse_ev_a": 0.08908521735741955,
        "solute_force_component_rmse_ev_a": 0.10236063275253979,
        "water_force_component_rmse_ev_a": 0.0784206465137683,
        "within_solute_energy_rmse_ev": 0.26077537972182163,
        "total_residual_energy_rmse_ev": 0.8913260828107529
      },
      "control": {
        "all_force_component_rmse_ev_a": 0.08762495364301598,
        "solute_force_component_rmse_ev_a": 0.10286493632844647,
        "water_force_component_rmse_ev_a": 0.07508419672879856,
        "within_solute_energy_rmse_ev": 0.25742204004578106,
        "total_residual_energy_rmse_ev": 0.6902507522930265
      },
      "solvent": {
        "all_force_component_rmse_ev_a": 0.08286563346493488,
        "solute_force_component_rmse_ev_a": 0.10292668183290211,
        "water_force_component_rmse_ev_a": 0.06516868964868044,
        "within_solute_energy_rmse_ev": 0.21705315263808472,
        "total_residual_energy_rmse_ev": 0.7941702881320793
      }
    },
    "treatment_vs_control_change_percent": {
      "all_force_component_rmse_ev_a": -5.431466700080179,
      "solute_force_component_rmse_ev_a": 0.06002580340738195,
      "water_force_component_rmse_ev_a": -13.20585091418447,
      "within_solute_energy_rmse_ev": -15.681985660791497,
      "total_residual_energy_rmse_ev": 15.055331050901444
    },
    "conditional_group_bootstrap_95_percent_intervals": {
      "all_force_component_rmse_ev_a": [
        -7.368178521989282,
        -3.383492225237167
      ],
      "solute_force_component_rmse_ev_a": [
        -2.037285407091424,
        1.664649177399905
      ],
      "water_force_component_rmse_ev_a": [
        -17.65765271430176,
        -8.478017683390767
      ],
      "within_solute_energy_rmse_ev": [
        -20.804587113263775,
        -3.6335985589127104
      ],
      "total_residual_energy_rmse_ev": [
        6.941697208907583,
        24.301031927011604
      ]
    },
    "bootstrap_replicates": 2000,
    "strata": {
      "stratum": {
        "0": {
          "records": 256,
          "roots": 32,
          "metrics": {
            "incumbent": {
              "all_force_component_rmse_ev_a": 0.09325988261313803,
              "solute_force_component_rmse_ev_a": 0.06999954465422742,
              "water_force_component_rmse_ev_a": 0.1554483268876889,
              "within_solute_energy_rmse_ev": 0.29531034340904533,
              "total_residual_energy_rmse_ev": 0.3594422771379637
            },
            "control": {
              "all_force_component_rmse_ev_a": 0.09060604280731101,
              "solute_force_component_rmse_ev_a": 0.06858541688189089,
              "water_force_component_rmse_ev_a": 0.14995681506212036,
              "within_solute_energy_rmse_ev": 0.2857445103644449,
              "total_residual_energy_rmse_ev": 0.3583181268876271
            },
            "solvent": {
              "all_force_component_rmse_ev_a": 0.08213589578861967,
              "solute_force_component_rmse_ev_a": 0.06559569390687178,
              "water_force_component_rmse_ev_a": 0.12922275061243646,
              "within_solute_energy_rmse_ev": 0.2198084655793107,
              "total_residual_energy_rmse_ev": 0.3209922851416607
            }
          }
        },
        "2": {
          "records": 256,
          "roots": 32,
          "metrics": {
            "incumbent": {
              "all_force_component_rmse_ev_a": 0.08837611613694114,
              "solute_force_component_rmse_ev_a": 0.11226458845095744,
              "water_force_component_rmse_ev_a": 0.07245052500161873,
              "within_solute_energy_rmse_ev": 0.22090585871305504,
              "total_residual_energy_rmse_ev": 1.2081910532707143
            },
            "control": {
              "all_force_component_rmse_ev_a": 0.08712232875502625,
              "solute_force_component_rmse_ev_a": 0.11322967982919459,
              "water_force_component_rmse_ev_a": 0.06924713791306196,
              "within_solute_energy_rmse_ev": 0.2255710269499322,
              "total_residual_energy_rmse_ev": 0.9080200008953109
            },
            "solvent": {
              "all_force_component_rmse_ev_a": 0.08298564752474719,
              "solute_force_component_rmse_ev_a": 0.11397977843561446,
              "water_force_component_rmse_ev_a": 0.06020282428806919,
              "within_solute_energy_rmse_ev": 0.2142624106088895,
              "total_residual_energy_rmse_ev": 1.0762791673088887
            }
          }
        }
      },
      "charge": {
        "-1": {
          "records": 48,
          "roots": 6,
          "metrics": {
            "incumbent": {
              "all_force_component_rmse_ev_a": 0.1978968594970059,
              "solute_force_component_rmse_ev_a": 0.2510997016612552,
              "water_force_component_rmse_ev_a": 0.16820684780661171,
              "within_solute_energy_rmse_ev": 0.7041956270427395,
              "total_residual_energy_rmse_ev": 1.029222897253798
            },
            "control": {
              "all_force_component_rmse_ev_a": 0.19939747993095497,
              "solute_force_component_rmse_ev_a": 0.2567577726742257,
              "water_force_component_rmse_ev_a": 0.1668936707489028,
              "within_solute_energy_rmse_ev": 0.684848065093125,
              "total_residual_energy_rmse_ev": 0.9773582529673698
            },
            "solvent": {
              "all_force_component_rmse_ev_a": 0.18484660990230795,
              "solute_force_component_rmse_ev_a": 0.2576375816531802,
              "water_force_component_rmse_ev_a": 0.13968675100155248,
              "within_solute_energy_rmse_ev": 0.5344836160221994,
              "total_residual_energy_rmse_ev": 0.7653272436017273
            }
          }
        },
        "0": {
          "records": 432,
          "roots": 54,
          "metrics": {
            "incumbent": {
              "all_force_component_rmse_ev_a": 0.07227291495344955,
              "solute_force_component_rmse_ev_a": 0.08423189623961198,
              "water_force_component_rmse_ev_a": 0.0621156461671883,
              "within_solute_energy_rmse_ev": 0.15054459139011617,
              "total_residual_energy_rmse_ev": 0.8779112790119616
            },
            "control": {
              "all_force_component_rmse_ev_a": 0.07006688675463565,
              "solute_force_component_rmse_ev_a": 0.08401924665555352,
              "water_force_component_rmse_ev_a": 0.057805117937872415,
              "within_solute_energy_rmse_ev": 0.15347199975792372,
              "total_residual_energy_rmse_ev": 0.6587105457601523
            },
            "solvent": {
              "all_force_component_rmse_ev_a": 0.06720789065226154,
              "solute_force_component_rmse_ev_a": 0.083893393866176,
              "water_force_component_rmse_ev_a": 0.05175076294558826,
              "within_solute_energy_rmse_ev": 0.1464766195667576,
              "total_residual_energy_rmse_ev": 0.79596092740467
            }
          }
        },
        "1": {
          "records": 32,
          "roots": 4,
          "metrics": {
            "incumbent": {
              "all_force_component_rmse_ev_a": 0.07182359705686342,
              "solute_force_component_rmse_ev_a": 0.08337523350752037,
              "water_force_component_rmse_ev_a": 0.062172836056593773,
              "within_solute_energy_rmse_ev": 0.19561175329912775,
              "total_residual_energy_rmse_ev": 0.8471212602167898
            },
            "control": {
              "all_force_component_rmse_ev_a": 0.06826352449878391,
              "solute_force_component_rmse_ev_a": 0.08071745181370808,
              "water_force_component_rmse_ev_a": 0.057633225899629756,
              "within_solute_energy_rmse_ev": 0.1968706688354413,
              "total_residual_energy_rmse_ev": 0.5767577592899569
            },
            "solvent": {
              "all_force_component_rmse_ev_a": 0.065115382610367,
              "solute_force_component_rmse_ev_a": 0.08092470024708998,
              "water_force_component_rmse_ev_a": 0.0507410426026016,
              "within_solute_energy_rmse_ev": 0.18877534889875303,
              "total_residual_energy_rmse_ev": 0.8122426901726679
            }
          }
        }
      }
    },
    "prediction_calls": 0,
    "bindings": {
      "incumbent": {
        "forces.npz": "38ba52cc2aa882f308ff3ccac08ef017581a26575bcd4ef6b07a5e84f36ac522",
        "records.json": "21b48e28bc3711654e756a7637bb6f09936d5c8545f3e316fd66dd0b48e93bcb"
      },
      "control": {
        "final-solvent-forces.npz": "7ff01d174cededf11d63269269f6e087d3a9f300a70938ae75ea1b02d914b91c",
        "final-solvent-records.json": "55e9c515fe72766176aea075da3b5397888ed4254e011edb8c32a7b2a2748e94"
      },
      "solvent": {
        "final-solvent-forces.npz": "51a19c428feea923169a6ba30b46b5f1b1bf1efa0ae2298b7117cf18ab7a9480",
        "final-solvent-records.json": "50ad7535399d5e4d2d36efb549ff4d29020c5d044751a73e7d6ea48243bf202c"
      }
    },
    "operator_sha256": "8c9d59a5462a669d8208b6b1dc217354e7acfd2931bf3f4d80c1765298e5629b",
    "scope": "Fixed reserved project-adaptation groups; original pretraining exposure possible. Atom partition is source-topology solute/water, not instantaneous species. Matched update schedules, one run per arm; bootstrap does not cover training-seed uncertainty. No barriers, rates or universal model qualification."
  },
  "old_source_change_percent": {
    "bms25": {
      "qm_force_component_rmse_ev_a": -0.003968853466418398,
      "mm_response_component_rmse_ev_a": -0.017071661867074095,
      "within_molecule_energy_rmse_ev": -0.192373918729849
    },
    "omol25": {
      "total_energy_rmse_ev": 21.479615694098953,
      "force_component_rmse_ev_a": 0.1660917742017931
    },
    "spice2": {
      "total_energy_rmse_ev": 52.60305738823994,
      "force_component_rmse_ev_a": 0.2346144056119659
    },
    "spice2_solvent": {
      "force_component_rmse_ev_a": -5.4314667000802,
      "total_residual_energy_rmse_ev": 15.055331050901444,
      "within_solute_energy_rmse_ev": -15.681985660791486
    }
  },
  "energy_error_decomposition": {
    "status": "saved_prediction_energy_error_decomposition",
    "results": {
      "control": {
        "omol25": {
          "records": 825,
          "groups": 566,
          "mean_signed_error_ev": -0.08312843316215025,
          "total_rmse_ev": 0.27523540670414476,
          "group_mean_error_rms_ev": 0.2727862141756864,
          "within_group_centered_error_rms_ev": 0.03663619056742308,
          "multiframe_centered_error_rms_ev": 0.04623516414281892,
          "multiframe_records": 518
        },
        "spice2": {
          "records": 825,
          "groups": 566,
          "mean_signed_error_ev": -0.03272440647422958,
          "total_rmse_ev": 0.25788495708775455,
          "group_mean_error_rms_ev": 0.255252256925395,
          "within_group_centered_error_rms_ev": 0.03675508708526212,
          "multiframe_centered_error_rms_ev": 0.046385212494821575,
          "multiframe_records": 518
        }
      },
      "treatment": {
        "omol25": {
          "records": 825,
          "groups": 566,
          "mean_signed_error_ev": 0.17951204248667196,
          "total_rmse_ev": 0.3343549143182853,
          "group_mean_error_rms_ev": 0.33233423349125385,
          "within_group_centered_error_rms_ev": 0.03670375973341953,
          "multiframe_centered_error_rms_ev": 0.04632043696819876,
          "multiframe_records": 518
        },
        "spice2": {
          "records": 825,
          "groups": 566,
          "mean_signed_error_ev": 0.27478997898035284,
          "total_rmse_ev": 0.39354032906026404,
          "group_mean_error_rms_ev": 0.3918132406668784,
          "within_group_centered_error_rms_ev": 0.03682899720301571,
          "multiframe_centered_error_rms_ev": 0.04647848765179686,
          "multiframe_records": 518
        }
      }
    },
    "model_calls": 0,
    "fitting": false,
    "scope": "Descriptive group-mean versus within-group decomposition of exposed validation residuals; no offsets fitted or applied, and not a correction or an independent test."
  },
  "runtime": {
    "control_seconds": 2648.2930473592132,
    "treatment_seconds": 2595.083467330318,
    "gpu": "RTX 2000 Ada",
    "gpu_count_per_run": 1,
    "peak_allocated_bytes": 1618644992,
    "both_pods_independently_stopped": true,
    "worker_runtime_cost_estimate_usd": 0.3495584343126354
  },
  "hard_training_case": {
    "id": 1988908,
    "initial_force_rmse_ev_a": 54036.270163782305,
    "control_force_rmse_ev_a": 52992.387374583006,
    "treatment_force_rmse_ev_a": 6.5625099744893,
    "evaluation_status": "Exposed training configuration; still inaccurate, not generalisation evidence."
  },
  "decision": "Keep incumbent. Water-force and within-solute-energy benefits are supported for these fixed validation roots; solute-force improvement is not demonstrated. Total-energy regressions remain. No automatic third training run or reaction-rate claim. Next investigate training-only source energy calibration with network/forces fixed.",
  "source_checkpoint_sha256": {
    "initial": "68dbadf704df0fbe0c073963057bcb216a73ff225ce1395d9892f07b5fc2b2c4",
    "control": "bed3a30baa48d313c4d213cb31a78f8acbbcbb4ddedcf0c5c538feb1e92f8513",
    "treatment": "d88a65e03baf291f3e60355c0d6b85e50b56d44a4101995f195e6e4767fbbb34"
  },
  "independent_audits": {
    "control": {
      "status": "passed_complete_continuation_prediction_and_metric_reopening",
      "job": "joint-solvent-control-v4",
      "method_specific_validation_records": 4508,
      "native_bms_fingerprints_reproduced": 2346,
      "cached_initial_metrics_unchanged": true,
      "final_metrics": {
        "bms25": {
          "records": 2346,
          "qm_force_component_rmse_ev_a": 0.14571267938111188,
          "mm_response_component_rmse_ev_a": 0.004875705660824478,
          "within_molecule_energy_rmse_ev": 0.18914573514277166,
          "nonanchor_energy_records": 2205
        },
        "reactions": {
          "records": 1000,
          "qm_force_component_rmse_ev_a": 0.09660017574608351,
          "mm_response_component_rmse_ev_a": 0.001734217979792932,
          "within_molecule_energy_rmse_ev": 0.13293551451214577,
          "nonanchor_energy_records": 952
        },
        "small-molecules-0": {
          "records": 1000,
          "qm_force_component_rmse_ev_a": 0.1596782861041863,
          "mm_response_component_rmse_ev_a": 0.004618311379742581,
          "within_molecule_energy_rmse_ev": 0.20559210302424075,
          "nonanchor_energy_records": 966
        },
        "tripeptides-0": {
          "records": 346,
          "qm_force_component_rmse_ev_a": 0.14779110875977117,
          "mm_response_component_rmse_ev_a": 0.007970714982227135,
          "within_molecule_energy_rmse_ev": 0.2719891974693648,
          "nonanchor_energy_records": 287
        },
        "omol25": {
          "records": 825,
          "total_energy_rmse_ev": 0.27523540670455543,
          "force_component_rmse_ev_a": 0.08443343444255134
        },
        "spice2": {
          "records": 825,
          "total_energy_rmse_ev": 0.25788495708776593,
          "force_component_rmse_ev_a": 0.0833015958885203
        },
        "spice2_solvent": {
          "records": 512,
          "solute_groups": 64,
          "force_component_rmse_ev_a": 0.08762495364301602,
          "total_residual_energy_rmse_ev": 0.6902507522930265,
          "within_solute_energy_rmse_ev": 0.25742204004578106,
          "nonanchor_energy_records": 448
        }
      },
      "relative_error_changes_percent": {
        "bms25": {
          "qm_force_component_rmse_ev_a": 0.203062354782646,
          "mm_response_component_rmse_ev_a": -0.1553388607083539,
          "within_molecule_energy_rmse_ev": 0.15310246211928114
        },
        "omol25": {
          "total_energy_rmse_ev": 0.4100883769207142,
          "force_component_rmse_ev_a": -2.256979820908811
        },
        "spice2": {
          "total_energy_rmse_ev": -2.469854098955504,
          "force_component_rmse_ev_a": -0.5601147236275139
        },
        "spice2_solvent": {
          "force_component_rmse_ev_a": -1.6391762378990649,
          "total_residual_energy_rmse_ev": -22.559121111282366,
          "within_solute_energy_rmse_ev": -1.285911146833607
        }
      },
      "source_exposures": {
        "bms25": 89201,
        "omol25": 11560,
        "spice2": 11564
      },
      "source_quotas": {
        "bms25": 11570,
        "omol25": 1447,
        "spice2": 1447
      },
      "checkpoint_sha256": "bed3a30baa48d313c4d213cb31a78f8acbbcbb4ddedcf0c5c538feb1e92f8513",
      "result_archive_sha256": "48f20e15cd8e76b911ef19b4390e5941a0049cbe39935d3a0844f334783f3553",
      "review_sha256": "503cd1e30f92f938111927b9d6253d998d15adab97f73a117053ddddc26adc73",
      "operator_sha256": "f4e6491da4f50376686593b76f1d373d28d35a8cf78349d4255cfe17758d5a3d",
      "elapsed_seconds": 15.575894142501056,
      "model_qualified": false,
      "scope": "Every saved native/solvent evaluation reconciled to source targets and ownership. No inference or fitting. Existing source/pretraining exposure and numerical exclusions remain."
    },
    "treatment": {
      "status": "passed_complete_continuation_prediction_and_metric_reopening",
      "job": "joint-solvent-treatment-v4",
      "method_specific_validation_records": 4508,
      "native_bms_fingerprints_reproduced": 2346,
      "cached_initial_metrics_unchanged": true,
      "final_metrics": {
        "bms25": {
          "records": 2346,
          "qm_force_component_rmse_ev_a": 0.14570689613984128,
          "mm_response_component_rmse_ev_a": 0.004874873296157822,
          "within_molecule_energy_rmse_ev": 0.18878186807996714,
          "nonanchor_energy_records": 2205
        },
        "reactions": {
          "records": 1000,
          "qm_force_component_rmse_ev_a": 0.0966542821948417,
          "mm_response_component_rmse_ev_a": 0.0017333194869683906,
          "within_molecule_energy_rmse_ev": 0.13284226852328787,
          "nonanchor_energy_records": 952
        },
        "small-molecules-0": {
          "records": 1000,
          "qm_force_component_rmse_ev_a": 0.1597143383861059,
          "mm_response_component_rmse_ev_a": 0.004615635497794346,
          "within_molecule_energy_rmse_ev": 0.20447370635688697,
          "nonanchor_energy_records": 966
        },
        "tripeptides-0": {
          "records": 346,
          "qm_force_component_rmse_ev_a": 0.14766912882305788,
          "mm_response_component_rmse_ev_a": 0.0079723465788964,
          "within_molecule_energy_rmse_ev": 0.27303377800461265,
          "nonanchor_energy_records": 287
        },
        "omol25": {
          "records": 825,
          "total_energy_rmse_ev": 0.3343549143186967,
          "force_component_rmse_ev_a": 0.08457367143183647
        },
        "spice2": {
          "records": 825,
          "total_energy_rmse_ev": 0.3935403290604366,
          "force_component_rmse_ev_a": 0.08349703313694098
        },
        "spice2_solvent": {
          "records": 512,
          "solute_groups": 64,
          "force_component_rmse_ev_a": 0.08286563346493489,
          "total_residual_energy_rmse_ev": 0.7941702881320793,
          "within_solute_energy_rmse_ev": 0.21705315263808475,
          "nonanchor_energy_records": 448
        }
      },
      "relative_error_changes_percent": {
        "bms25": {
          "qm_force_component_rmse_ev_a": 0.19908536054913384,
          "mm_response_component_rmse_ev_a": -0.17238401762877942,
          "within_molecule_energy_rmse_ev": -0.03956598581662529
        },
        "omol25": {
          "total_energy_rmse_ev": 21.97778947835649,
          "force_component_rmse_ev_a": -2.094636704534958
        },
        "spice2": {
          "total_energy_rmse_ev": 48.8339845202638,
          "force_component_rmse_ev_a": -0.3268147807585886
        },
        "spice2_solvent": {
          "force_component_rmse_ev_a": -6.981611626462147,
          "total_residual_energy_rmse_ev": -10.900140425858273,
          "within_solute_energy_rmse_ev": -16.766240405968137
        }
      },
      "source_exposures": {
        "bms25": 89201,
        "omol25": 11560,
        "spice2": 11576
      },
      "source_quotas": {
        "bms25": 11570,
        "omol25": 1447,
        "spice2": 1447
      },
      "checkpoint_sha256": "d88a65e03baf291f3e60355c0d6b85e50b56d44a4101995f195e6e4767fbbb34",
      "result_archive_sha256": "44cf768e175443cdd3f481dab31e101f6d00fd9d37fe8b9cdc0cfcf34af6b489",
      "review_sha256": "a3a6adc98360fdf99d611081e4b79430b4d14f08e52ad6f1c178b9f1d171d772",
      "operator_sha256": "f4e6491da4f50376686593b76f1d373d28d35a8cf78349d4255cfe17758d5a3d",
      "elapsed_seconds": 14.542222686111927,
      "model_qualified": false,
      "scope": "Every saved native/solvent evaluation reconciled to source targets and ownership. No inference or fitting. Existing source/pretraining exposure and numerical exclusions remain."
    }
  }
}
