{
  "seed": 1448,
  "math_checks": {
    "odds_slope_observed": 1.2999999999999996,
    "odds_slope_predicted": 1.3,
    "odds_max_abs_identity_error": 8.881784197001252e-16,
    "boundary_beta_observed": 1.5,
    "boundary_beta_predicted": 1.5,
    "boundary_abs_error": 0.0,
    "beta0_max_transition_difference": 0.0,
    "beta0_sweep_scales": [
      0.0,
      0.5,
      1.0,
      3.0,
      10.0
    ],
    "beta0_sweep_max_errors": [
      0.0,
      0.0,
      0.0,
      0.0,
      0.0
    ],
    "predictions_confirmed": true
  },
  "mini_experiment": {
    "n_episodes": 160,
    "one_step_mse": {
      "baseline": 0.025127682580169907,
      "stl": 0.0252477885589946
    },
    "rollout_50_mse": {
      "baseline": 3.115207491006956,
      "stl": 3.5306199564239877
    },
    "mode2_selection_rate": {
      "baseline": 0.43125,
      "stl": 0.20625
    },
    "beta_sweep_same_protocol": {
      "0.0": {
        "one_step_mse": 0.027298106786175846,
        "mode2_selection_rate": 0.41875
      },
      "0.5": {
        "one_step_mse": 0.025657513167276536,
        "mode2_selection_rate": 0.40625
      },
      "1.0": {
        "one_step_mse": 0.02570714521150827,
        "mode2_selection_rate": 0.39375
      },
      "2.8": {
        "one_step_mse": 0.02604671855802617,
        "mode2_selection_rate": 0.175
      },
      "5.0": {
        "one_step_mse": 0.025344349841651525,
        "mode2_selection_rate": 0.0375
      }
    },
    "robustness_beta": 2.8,
    "note": "Synthetic fixed expert dynamics; no learned parameters or training loop."
  }
}