{
  "args": {
    "model_seed": 1,
    "outdir": "results/bci_v2_calibrated_confirmation",
    "r1_gate": "results/bci_v2_calibrated_dev_gate.json",
    "task_seed": 32
  },
  "assays": {
    "challenge": {
      "active_state_episode_steps_by_mode": {
        "acute_critic_lesion": 13197,
        "acute_outcome_lesion": 13197,
        "intact": 13197
      },
      "calibration": {
        "active_state_episode_steps": 14336,
        "episodes": 512,
        "maximum_cursor_summary": {
          "maximum": 1.839465856552124,
          "median": 1.7447079420089722,
          "minimum": 1.5286614894866943
        },
        "quantiles": [
          0.2,
          0.35,
          0.5,
          0.65,
          0.8
        ],
        "seed": 600032,
        "targets": [
          1.7064756870269775,
          1.7277086198329925,
          1.7448158860206604,
          1.7594623565673828,
          1.7777812480926514
        ],
        "uses_outcome_labels": false
      },
      "episodes_per_target": 128,
      "maximum_steps_per_episode": 28,
      "selection_over_evaluation_outcomes": false,
      "targets": [
        1.7064756870269775,
        1.7277086198329925,
        1.7448158860206604,
        1.7594623565673828,
        1.7777812480926514
      ],
      "trajectory_seeds": {
        "1": 642000,
        "2": 642001,
        "3": 642002,
        "4": 642003,
        "5": 642004
      }
    },
    "performance_evaluation_episodes": 256,
    "performance_evaluation_seed": 590032
  },
  "conditions": {
    "critic_training_lesion": {
      "cost": {
        "active_state_episode_steps": 8309,
        "cursor_scalar_observations": 5052,
        "maximum_state_episode_steps": 25088,
        "reverse_mode_calls": 0,
        "role_probe_examples": 2526,
        "task_loss_queries": 0,
        "terminal_outcome_observations": 896
      },
      "critic_l2_after_training": 0.0,
      "daily_success": [
        0.0,
        0.015625,
        0.0625,
        0.359375,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0
      ],
      "early_success": 0.026041666666666668,
      "evaluation_active_state_episode_steps": 256,
      "final_success": 1.0,
      "late_success": 1.0,
      "learning_gain": 0.9739583333333334,
      "role_cosine_after_training": 0.9838115572929382,
      "training_wall_s": 0.14876241609454155
    },
    "fixed_role": {
      "cost": {
        "active_state_episode_steps": 25088,
        "cursor_scalar_observations": 12544,
        "maximum_state_episode_steps": 25088,
        "reverse_mode_calls": 0,
        "role_probe_examples": 6272,
        "task_loss_queries": 0,
        "terminal_outcome_observations": 896
      },
      "critic_l2_after_training": 0.06551171839237213,
      "daily_success": [
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0
      ],
      "early_success": 0.0,
      "evaluation_active_state_episode_steps": 7168,
      "final_success": 0.0,
      "late_success": 0.0,
      "learning_gain": 0.0,
      "role_cosine_after_training": 0.006917059421539307,
      "training_wall_s": 0.20988380163908005
    },
    "intact": {
      "cost": {
        "active_state_episode_steps": 8176,
        "cursor_scalar_observations": 4970,
        "maximum_state_episode_steps": 25088,
        "reverse_mode_calls": 0,
        "role_probe_examples": 2485,
        "task_loss_queries": 0,
        "terminal_outcome_observations": 896
      },
      "critic_l2_after_training": 0.5582403540611267,
      "daily_success": [
        0.0,
        0.015625,
        0.078125,
        0.515625,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0
      ],
      "early_success": 0.03125,
      "evaluation_active_state_episode_steps": 256,
      "final_success": 1.0,
      "late_success": 1.0,
      "learning_gain": 0.96875,
      "role_cosine_after_training": 0.9938780665397644,
      "training_wall_s": 0.23583389073610306
    },
    "oracle_role": {
      "cost": {
        "active_state_episode_steps": 8098,
        "cursor_scalar_observations": 4942,
        "maximum_state_episode_steps": 25088,
        "reverse_mode_calls": 0,
        "role_probe_examples": 2471,
        "task_loss_queries": 0,
        "terminal_outcome_observations": 896
      },
      "critic_l2_after_training": 0.5624566078186035,
      "daily_success": [
        0.0,
        0.015625,
        0.078125,
        0.53125,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0
      ],
      "early_success": 0.03125,
      "evaluation_active_state_episode_steps": 257,
      "final_success": 1.0,
      "late_success": 1.0,
      "learning_gain": 0.96875,
      "role_cosine_after_training": 1.0000001192092896,
      "training_wall_s": 0.16362693533301353
    },
    "outcome_training_lesion": {
      "cost": {
        "active_state_episode_steps": 14370,
        "cursor_scalar_observations": 7642,
        "maximum_state_episode_steps": 25088,
        "reverse_mode_calls": 0,
        "role_probe_examples": 3821,
        "task_loss_queries": 0,
        "terminal_outcome_observations": 896
      },
      "critic_l2_after_training": 0.3587868809700012,
      "daily_success": [
        0.0,
        0.015625,
        0.046875,
        0.125,
        0.34375,
        0.515625,
        0.625,
        0.84375,
        0.96875,
        0.96875,
        0.984375,
        0.984375,
        1.0,
        1.0
      ],
      "early_success": 0.020833333333333332,
      "evaluation_active_state_episode_steps": 552,
      "final_success": 1.0,
      "late_success": 0.9947916666666666,
      "learning_gain": 0.9739583333333334,
      "role_cosine_after_training": 0.8972746729850769,
      "training_wall_s": 0.20119396969676018
    },
    "plasticity_lesion": {
      "cost": {
        "active_state_episode_steps": 25088,
        "cursor_scalar_observations": 12544,
        "maximum_state_episode_steps": 25088,
        "reverse_mode_calls": 0,
        "role_probe_examples": 6272,
        "task_loss_queries": 0,
        "terminal_outcome_observations": 896
      },
      "critic_l2_after_training": 0.06581009179353714,
      "daily_success": [
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0
      ],
      "early_success": 0.0,
      "evaluation_active_state_episode_steps": 7168,
      "final_success": 0.0,
      "late_success": 0.0,
      "learning_gain": 0.0,
      "role_cosine_after_training": 0.9962126016616821,
      "training_wall_s": 0.1819687932729721
    }
  },
  "config": {
    "context_ar": 0.8,
    "context_dim": 16,
    "coupling_scale": 1.0,
    "critic_eta": 0.03,
    "days": 14,
    "eligibility_decay": 0.8,
    "episodes_per_day": 64,
    "feedback": "performance_velocity",
    "forward_eta": 0.1,
    "gamma": 0.8,
    "inertia": 0.65,
    "kappa": 0.0,
    "n_background": 30,
    "n_minus": 5,
    "n_plus": 5,
    "perturb_every": 4,
    "perturb_sigma": 0.03,
    "predictor_eta": 0.2,
    "process_noise": 0.12,
    "steps_per_episode": 28,
    "target": 0.8,
    "terminal_reward": 1.0,
    "vectorizer_eta": 0.03,
    "velocity_reward_scale": 1.0
  },
  "finite": true,
  "hardware": {
    "device": "cpu",
    "platform": "Linux-5.15.0-161-generic-x86_64-with-glibc2.35",
    "threads": 1,
    "torch_version": "2.10.0+cu128"
  },
  "peak_rss_mib": 936.0859375,
  "protocol": {
    "calibration_uses_outcome_labels": false,
    "confirmation_grid_size": 30,
    "fixed_config": {
      "critic_eta": 0.03,
      "forward_eta": 0.1,
      "gamma": 0.8,
      "velocity_reward_scale": 1.0
    },
    "model_seed": 1,
    "name": "oral_b_v2_calibrated_recovery_confirmation_v1",
    "no_further_selection": true,
    "protocol_sha256": "a175dc80fdb7218c010953d5d7e3c77fee0690d5f10fabaf9d700bb8505acde8",
    "r1_gate_sha256": "907648edaea040b0d7e6fbf74f96b9c5ff8fc19c0dc1063b054829d5f2aa9fbc",
    "split": "untouched_confirmation",
    "training_task_seed": 32
  },
  "provenance": {
    "git_commit": "70e180c5ef5f78679f2e163ed3ee30873ee523bf",
    "git_tracked_dirty": false,
    "input_sha256": {
      "base_dynamics": "d5a373314562af0daedf55baa04b975fe643236c7821be1560d2cfb4359688fc",
      "common_runner": "4157a57806dedfc2eb70ec84f7a7ca1f73ab684e5d8f0b227cd80465e8e97aa5",
      "confirmation_analyzer": "2536fa01f62c3b2d87e782b08075cb5948aa26bfdce71a1d8c2446d5b68bb34c",
      "confirmation_runner": "5e9beba106b4b29c4ee1568052d6e8c15690dbde92917a2497a52e43e7f557d1",
      "d4_gate": "636c587e47287338cdca7d9558dc3bfb99a2cec615badb23337025d85b8128ef",
      "development_analyzer": "dde4a2e4d62f5f9da94357b388139c50251b6863f9c991f1a006e71eeb7953e4",
      "development_runner": "446ad26fc015eb79d28e0bb442c8aef71bdd1d0320910494cac20f92e3c7dc18",
      "failed_target_gate": "1df37c0b639542bbee3017da54306b0d572cbf96230428f142fafed139ac92f1",
      "failed_v2_gate": "e53f46cf456d60ce4d33586ace4f4619fb1a31a58082b2b72943d4ed10cd25fd",
      "old_r2_gate": "4f6f969ceae88afa2523e3472a3373522991f5ecaab69440830c07479d2d3597",
      "protocol": "a175dc80fdb7218c010953d5d7e3c77fee0690d5f10fabaf9d700bb8505acde8",
      "r1_gate": "907648edaea040b0d7e6fbf74f96b9c5ff8fc19c0dc1063b054829d5f2aa9fbc",
      "recovery_confirmation_analyzer": "29a168143266bc7078e5229183bb7ea45283973592fecd1fcd715488f5d6242b",
      "recovery_development_analyzer": "2768b877132ece6d612d2a5f1e483c495b277e26fdf54fb8a716aa8de6fd67e2",
      "recovery_metrics": "371cb6a7e2192d25ddbfad69c0ce7a1150aa1c21ca98e141bedb2afc2acbdbaa",
      "recovery_runner": "cea0ef658c698c376b1fda0cf2f7f910ee44f7bcde153f2fbaf18e66a2d3661a",
      "runner": "5e9beba106b4b29c4ee1568052d6e8c15690dbde92917a2497a52e43e7f557d1",
      "v2_dynamics": "786c5aed3091644272ae0a740913ac3cc38cdb02a8cacb6dfc8496e36836cdae",
      "v2_metrics": "421da9287b92e62c742f9eeefbfdccba582c6a4ad5cf92ade4709689120fa1ee"
    },
    "tracked_inputs": {
      "base_dynamics": true,
      "common_runner": true,
      "confirmation_analyzer": true,
      "confirmation_runner": true,
      "d4_gate": true,
      "development_analyzer": true,
      "development_runner": true,
      "failed_target_gate": true,
      "failed_v2_gate": true,
      "old_r2_gate": true,
      "protocol": true,
      "r1_gate": true,
      "recovery_confirmation_analyzer": true,
      "recovery_development_analyzer": true,
      "recovery_metrics": true,
      "recovery_runner": true,
      "runner": true,
      "v2_dynamics": true,
      "v2_metrics": true
    }
  },
  "schema_version": 4,
  "signatures": {
    "acute_outcome_lesion_outcome_balanced_acc": 0.5727208784338244,
    "acute_outcome_lesion_role_aligned_separation": -0.0009131268129543013,
    "causal_role_sign_inversion_index": 0.0448768071145361,
    "challenge_episodes": 640,
    "challenge_failure_count": 326,
    "challenge_success_count": 314,
    "challenge_success_fraction": 0.490625,
    "critic_contribution_value_prediction_corr": 0.9999999999987023,
    "decoder_distance_residual_corr": 0.11275966315486538,
    "mean_abs_raw_soma_corr": 0.9987214042329219,
    "mean_abs_residual_soma_corr": 0.07483543237113474,
    "mean_critic_expectedness_contribution": 0.36799634712323687,
    "nonterminal_training_events": 7280,
    "raw_minus_residual_abs_soma_corr": 0.9238859718617871,
    "role_aligned_error_cv_corr": 0.3891806866805502,
    "role_aligned_velocity_cv_corr": 0.9981286331323391,
    "surrounding_event_decoder_balanced_acc": 0.5415856723757944,
    "target_success_fraction": {
      "1.7064756870269775": 0.7578125,
      "1.7277086198329925": 0.6484375,
      "1.7448158860206604": 0.453125,
      "1.7594623565673828": 0.390625,
      "1.7777812480926514": 0.203125
    },
    "terminal_outcome_separation_drop_under_acute_lesion": 0.39622390380398786,
    "terminal_previous_soma_outcome_balanced_acc": 0.7454769254816147,
    "terminal_residual_minus_previous_soma_acc": 0.2545230745183853,
    "terminal_residual_outcome_balanced_acc": 1.0,
    "terminal_role_aligned_outcome_separation": 0.39531077699103356,
    "terminal_training_events": 896,
    "velocity_minus_error_abs_cv_corr": 0.6089479464517888
  },
  "split": "untouched_confirmation",
  "wall_s": 2.1287473514676094,
  "warmup": {
    "critic_training_lesion": {
      "batches": 100,
      "examples": 6400,
      "instruction_present": false,
      "predictor_max_abs_error": 2.384185791015625e-07,
      "rng_seed": 603201,
      "role_cosine_after_warmup": 0.9919762015342712,
      "role_cursor_scalar_observations": 12800,
      "role_update_enabled": true
    },
    "fixed_role": {
      "batches": 100,
      "examples": 6400,
      "instruction_present": false,
      "predictor_max_abs_error": 2.384185791015625e-07,
      "rng_seed": 603201,
      "role_cosine_after_warmup": 0.006917059421539307,
      "role_cursor_scalar_observations": 12800,
      "role_update_enabled": false
    },
    "intact": {
      "batches": 100,
      "examples": 6400,
      "instruction_present": false,
      "predictor_max_abs_error": 2.384185791015625e-07,
      "rng_seed": 603201,
      "role_cosine_after_warmup": 0.9919762015342712,
      "role_cursor_scalar_observations": 12800,
      "role_update_enabled": true
    },
    "oracle_role": {
      "batches": 100,
      "examples": 6400,
      "instruction_present": false,
      "predictor_max_abs_error": 2.384185791015625e-07,
      "rng_seed": 603201,
      "role_cosine_after_warmup": 1.0000001192092896,
      "role_cursor_scalar_observations": 12800,
      "role_update_enabled": false
    },
    "outcome_training_lesion": {
      "batches": 100,
      "examples": 6400,
      "instruction_present": false,
      "predictor_max_abs_error": 2.384185791015625e-07,
      "rng_seed": 603201,
      "role_cosine_after_warmup": 0.9919762015342712,
      "role_cursor_scalar_observations": 12800,
      "role_update_enabled": true
    },
    "plasticity_lesion": {
      "batches": 100,
      "examples": 6400,
      "instruction_present": false,
      "predictor_max_abs_error": 2.384185791015625e-07,
      "rng_seed": 603201,
      "role_cosine_after_warmup": 0.9919762015342712,
      "role_cursor_scalar_observations": 12800,
      "role_update_enabled": true
    }
  }
}