{
  "model": "openai/gpt-oss-120b",
  "tasks": {
    "clean_deploy": {
      "seed_0": {
        "task": "clean_deploy",
        "seed_offset": 0,
        "steps": 15,
        "reward_sum": -1.48,
        "rewards_per_step": [
          0.04,
          0.01,
          0.01,
          0.01,
          -0.15,
          -0.15,
          -0.1,
          -0.1,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15
        ],
        "roles_used": [
          "ops",
          "sre"
        ],
        "num_roles_used": 2,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 1.0,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_1": {
        "task": "clean_deploy",
        "seed_offset": 1,
        "steps": 15,
        "reward_sum": -1.0426,
        "rewards_per_step": [
          0.04,
          0.01,
          0.01,
          0.0277,
          0.07,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          -0.15,
          0.1497
        ],
        "roles_used": [
          "ops",
          "sre"
        ],
        "num_roles_used": 2,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 1.0,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_2": {
        "task": "clean_deploy",
        "seed_offset": 2,
        "steps": 20,
        "reward_sum": -1.6045,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0058,
          0.0554,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.008,
          0.008,
          0.0275,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      }
    },
    "broken_pipeline": {
      "seed_0": {
        "task": "broken_pipeline",
        "seed_offset": 0,
        "steps": 20,
        "reward_sum": -1.7022,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0552,
          0.0055,
          -0.05,
          0.0058,
          0.0058,
          0.026,
          0.0254,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          -0.05,
          0.008,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_1": {
        "task": "broken_pipeline",
        "seed_offset": 1,
        "steps": 20,
        "reward_sum": -0.2785,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0552,
          0.0055,
          0.0058,
          0.0058,
          0.026,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.05,
          -0.0007,
          0.0035,
          0.008,
          -0.05,
          0.008,
          0.008,
          -0.05
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_2": {
        "task": "broken_pipeline",
        "seed_offset": 2,
        "steps": 20,
        "reward_sum": -1.6625,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0547,
          0.0065,
          0.0071,
          0.0071,
          -0.05,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.0275,
          0.008,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      }
    },
    "judgment_call": {
      "seed_0": {
        "task": "judgment_call",
        "seed_offset": 0,
        "steps": 20,
        "reward_sum": -1.682,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0552,
          0.0055,
          0.0058,
          -0.05,
          0.0058,
          0.0065,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.008,
          0.008,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_1": {
        "task": "judgment_call",
        "seed_offset": 1,
        "steps": 20,
        "reward_sum": -1.74,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0547,
          0.0065,
          -0.05,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.008,
          0.008,
          -0.05,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_2": {
        "task": "judgment_call",
        "seed_offset": 2,
        "steps": 20,
        "reward_sum": -0.182,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0058,
          0.0065,
          0.056,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.008,
          0.008,
          0.008,
          0.008,
          0.008,
          -0.05
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      }
    },
    "cascading_failure": {
      "seed_0": {
        "task": "cascading_failure",
        "seed_offset": 0,
        "steps": 20,
        "reward_sum": -1.6045,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0058,
          0.0065,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          0.0483,
          0.0035,
          0.008,
          0.0275,
          0.008,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_1": {
        "task": "cascading_failure",
        "seed_offset": 1,
        "steps": 20,
        "reward_sum": -1.7205,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0552,
          0.0055,
          0.0058,
          0.0058,
          0.0065,
          0.0071,
          0.0266,
          0.0072,
          -0.05,
          0.007,
          -0.0007,
          -0.05,
          0.0035,
          0.008,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_2": {
        "task": "cascading_failure",
        "seed_offset": 2,
        "steps": 20,
        "reward_sum": -1.74,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0544,
          0.0058,
          0.0058,
          0.0065,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.008,
          -0.05,
          0.008,
          -0.05,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      }
    },
    "capacity_crisis": {
      "seed_0": {
        "task": "capacity_crisis",
        "seed_offset": 0,
        "steps": 20,
        "reward_sum": -1.6931,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0253,
          0.0241,
          -0.05,
          0.0065,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.008,
          0.008,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_1": {
        "task": "capacity_crisis",
        "seed_offset": 1,
        "steps": 20,
        "reward_sum": -1.798,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0552,
          0.0055,
          -0.05,
          0.0058,
          0.0058,
          -0.05,
          0.0065,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.05,
          -0.0007,
          0.0035,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_2": {
        "task": "capacity_crisis",
        "seed_offset": 2,
        "steps": 20,
        "reward_sum": -0.3274,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0058,
          0.026,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.05,
          -0.0007,
          0.0035,
          -0.05,
          0.008,
          0.008,
          0.008,
          -0.05
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      }
    },
    "random_incident": {
      "seed_0": {
        "task": "random_incident",
        "seed_offset": 0,
        "steps": 20,
        "reward_sum": -1.856,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0058,
          0.0554,
          0.0071,
          0.0071,
          0.0072,
          -0.05,
          0.007,
          -0.0007,
          -0.05,
          0.0035,
          -0.05,
          0.008,
          -0.05,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_1": {
        "task": "random_incident",
        "seed_offset": 1,
        "steps": 20,
        "reward_sum": -0.2889,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0055,
          0.0058,
          0.0058,
          0.0065,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          -0.05,
          0.0035,
          0.008,
          0.008,
          0.008,
          0.008,
          -0.05
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      },
      "seed_2": {
        "task": "random_incident",
        "seed_offset": 2,
        "steps": 20,
        "reward_sum": -1.624,
        "rewards_per_step": [
          0.0432,
          0.0048,
          -0.33,
          0.0063,
          0.0544,
          0.0058,
          0.0058,
          0.0065,
          0.0071,
          0.0071,
          0.0072,
          0.007,
          -0.0007,
          0.0035,
          0.008,
          0.008,
          0.008,
          0.008,
          0.008,
          -1.492
        ],
        "roles_used": [
          "sre"
        ],
        "num_roles_used": 1,
        "all_3_modes_used": false,
        "done": true,
        "success": false,
        "healthy_ratio": 0.8,
        "steps_to_recovery": 0,
        "initial_health": 100.0
      }
    }
  },
  "summary": {
    "clean_deploy": {
      "avg_reward": -1.3757,
      "std_reward": 0.241,
      "avg_steps": 16.67,
      "avg_roles_used": 1.67,
      "success_rate": 0,
      "all_3_modes_hit_rate": 0,
      "avg_steps_to_recovery": 0.0,
      "recovery_episodes_count": 0,
      "n": 3
    },
    "broken_pipeline": {
      "avg_reward": -1.2144,
      "std_reward": 0.662,
      "avg_steps": 20,
      "avg_roles_used": 1,
      "success_rate": 0,
      "all_3_modes_hit_rate": 0,
      "avg_steps_to_recovery": 0.0,
      "recovery_episodes_count": 0,
      "n": 3
    },
    "judgment_call": {
      "avg_reward": -1.2013,
      "std_reward": 0.7212,
      "avg_steps": 20,
      "avg_roles_used": 1,
      "success_rate": 0,
      "all_3_modes_hit_rate": 0,
      "avg_steps_to_recovery": 0.0,
      "recovery_episodes_count": 0,
      "n": 3
    },
    "cascading_failure": {
      "avg_reward": -1.6883,
      "std_reward": 0.0598,
      "avg_steps": 20,
      "avg_roles_used": 1,
      "success_rate": 0,
      "all_3_modes_hit_rate": 0,
      "avg_steps_to_recovery": 0.0,
      "recovery_episodes_count": 0,
      "n": 3
    },
    "capacity_crisis": {
      "avg_reward": -1.2728,
      "std_reward": 0.6699,
      "avg_steps": 20,
      "avg_roles_used": 1,
      "success_rate": 0,
      "all_3_modes_hit_rate": 0,
      "avg_steps_to_recovery": 0.0,
      "recovery_episodes_count": 0,
      "n": 3
    },
    "random_incident": {
      "avg_reward": -1.2563,
      "std_reward": 0.6906,
      "avg_steps": 20,
      "avg_roles_used": 1,
      "success_rate": 0,
      "all_3_modes_hit_rate": 0,
      "avg_steps_to_recovery": 0.0,
      "recovery_episodes_count": 0,
      "n": 3
    }
  }
}