{
  "checked_on": "2026-10-24",
  "python": "3.12.14",
  "numpy": "2.3.5",
  "validation": "synthetic finite-state experiment; no physical robot",
  "result": {
    "seed": 11,
    "numpy": "2.3.5",
    "simulation": "finite-1D-integrator",
    "gradient_checks": {
      "softmax_cross_entropy_max_gradient_error": 1.0411654871589349e-10,
      "expected_policy_score_max_abs": 2.7755575615628914e-17
    },
    "bc_training": {
      "synthetic_episodes": 120,
      "training_samples": 296,
      "validation_samples": 64,
      "validation_cross_entropy": 0.027740822053011344
    },
    "rl_training": {
      "episodes": 9600,
      "environment_steps": 50381,
      "training_curve": [
        {
          "batch": 0,
          "mean_training_return": -0.1275
        },
        {
          "batch": 50,
          "mean_training_return": 0.835
        },
        {
          "batch": 100,
          "mean_training_return": 0.86625
        },
        {
          "batch": 150,
          "mean_training_return": 0.8362499999999999
        },
        {
          "batch": 200,
          "mean_training_return": 0.875
        },
        {
          "batch": 250,
          "mean_training_return": 0.8625
        },
        {
          "batch": 299,
          "mean_training_return": 0.88375
        }
      ]
    },
    "evaluation": {
      "classical": {
        "episodes": 200,
        "successes": 200,
        "success_rate": 1,
        "mean_steps": 3.45,
        "mean_return": 0.9020000000000001,
        "actuation_dropout": 0.15
      },
      "random": {
        "episodes": 200,
        "successes": 100,
        "success_rate": 0.5,
        "mean_steps": 14.89,
        "mean_return": -0.07560000000000008,
        "actuation_dropout": 0.15
      },
      "bc": {
        "episodes": 200,
        "successes": 200,
        "success_rate": 1,
        "mean_steps": 3.45,
        "mean_return": 0.9020000000000001,
        "actuation_dropout": 0.15
      },
      "reinforce": {
        "episodes": 200,
        "successes": 200,
        "success_rate": 1,
        "mean_steps": 3.45,
        "mean_return": 0.9020000000000001,
        "actuation_dropout": 0.15
      }
    },
    "elapsed_seconds": 2.6322139000112657
  }
}