artifacts/evals/summary.json

{
  "task_count": 15,
  "baseline_mean_reward": 0.5427,
  "improved_mean_reward": 0.9777,
  "mean_reward_delta": 0.435,
  "median_reward_delta": 0.42,
  "baseline_success_rate": 0.3333,
  "improved_success_rate": 1,
  "success_rate_delta": 0.6667,
  "baseline_mean_violations": 0,
  "improved_mean_violations": 0,
  "violation_delta": 0,
  "baseline_mean_steps": 1,
  "improved_mean_steps": 3.5333,
  "step_delta": 2.5333,
  "tasks_with_positive_reward_delta": 15,
  "tasks_with_no_reward_delta": 0,
  "per_difficulty": {
    "easy": {
      "count": 5,
      "baseline_mean_reward": 0.4967,
      "improved_mean_reward": 0.9687,
      "reward_delta": 0.472,
      "baseline_mean_steps": 1,
      "improved_mean_steps": 2.6,
      "step_delta": 1.6
    },
    "medium": {
      "count": 5,
      "baseline_mean_reward": 0.5992,
      "improved_mean_reward": 0.9745,
      "reward_delta": 0.3753,
      "baseline_mean_steps": 1,
      "improved_mean_steps": 3,
      "step_delta": 2
    },
    "hard": {
      "count": 5,
      "baseline_mean_reward": 0.5323,
      "improved_mean_reward": 0.99,
      "reward_delta": 0.4577,
      "baseline_mean_steps": 1,
      "improved_mean_steps": 5,
      "step_delta": 4
    }
  }
}