{
  "seed": 1267,
  "math_check": {
    "stationarity_gradient_norm": 6.938893903907227e-13,
    "hessian_error_norm": 1.5476311098377395e-07,
    "analytic_u": [
      0.14427515036590355,
      -0.24895276547974993
    ]
  },
  "toy_predictions": {
    "lambdas": [
      0.0,
      0.1,
      0.3,
      1.0,
      3.0,
      10.0,
      30.0,
      100.0
    ],
    "solutions": [
      [
        2.0,
        2.0
      ],
      [
        1.4285714285714286,
        1.9512195121951221
      ],
      [
        0.9090909090909091,
        1.8604651162790697
      ],
      [
        0.4,
        1.6
      ],
      [
        0.15384615384615385,
        1.1428571428571428
      ],
      [
        0.04878048780487805,
        0.5714285714285714
      ],
      [
        0.01652892561983471,
        0.23529411764705882
      ],
      [
        0.004987531172069825,
        0.07692307692307693
      ]
    ],
    "trajectory_kl": [
      8.5,
      4.557539851157596,
      2.0855588680948824,
      0.6400000000000001,
      0.21060258422895783,
      0.04557539851157595,
      0.0074668259892055515,
      0.0007893959047989377
    ],
    "prediction_1_kl_monotone": true,
    "prediction_2_expected_lambda_u_tail": [
      0.5,
      8.0
    ],
    "prediction_2_observed_lambda_u_tail": [
      0.4941419212836014,
      6.821805645335057
    ],
    "prediction_2_relative_error": 0.1472742943330995,
    "prediction_3_predicted_curvature_ratio": 4.0,
    "prediction_3_observed_curvature_ratio": 4.0,
    "prediction_3_relative_error": 0.0
  },
  "policy_experiment": {
    "device": "cuda",
    "lambda": 2.0,
    "action_regularization": {
      "task_mse_half": 0.3949859142303467,
      "action_penalty": 0.049373235553503036,
      "trajectory_kl": 0.1613738238811493,
      "W": [
        [
          0.20000000298023224,
          -0.07999999821186066
        ],
        [
          0.03999998793005943,
          0.1599999964237213
        ],
        [
          -0.13999998569488525,
          0.06000000238418579
        ]
      ]
    },
    "noise_whitened_drift_regularization": {
      "task_mse_half": 0.411968857049942,
      "action_penalty": 0.11426855623722076,
      "trajectory_kl": 0.037015583366155624,
      "W": [
        [
          0.0476190522313118,
          -0.20000000298023224
        ],
        [
          0.00952380895614624,
          0.4000000059604645
        ],
        [
          -0.03333333507180214,
          0.15000000596046448
        ]
      ]
    }
  }
}