{
  "metric": "full_task_success_percent",
  "tasks": 40,
  "difficulty_tasks": {
    "simple": 21,
    "moderate": 13,
    "complex": 6
  },
  "methods": [
    {
      "method": "RobotUse",
      "overall": 45.0,
      "simple": 53.97,
      "moderate": 35.9,
      "complex": 33.33,
      "episodes": 120,
      "successes": 54,
      "successes_by_difficulty": {
        "simple": 34,
        "moderate": 14,
        "complex": 6
      }
    },
    {
      "method": "CaP-X",
      "overall": 38.33,
      "simple": 42.86,
      "moderate": 35.9,
      "complex": 27.78,
      "episodes": 120,
      "successes": 46,
      "successes_by_difficulty": {
        "simple": 27,
        "moderate": 14,
        "complex": 5
      }
    },
    {
      "method": "Open Robot Skill",
      "overall": 20.0,
      "simple": 31.75,
      "moderate": 10.26,
      "complex": 0,
      "episodes": 120,
      "successes": 24,
      "successes_by_difficulty": {
        "simple": 20,
        "moderate": 4,
        "complex": 0
      }
    },
    {
      "method": "mini-SWE",
      "overall": 20.83,
      "simple": 25.4,
      "moderate": 17.95,
      "complex": 11.11,
      "episodes": 120,
      "successes": 25,
      "successes_by_difficulty": {
        "simple": 16,
        "moderate": 7,
        "complex": 2
      }
    },
    {
      "method": "\u03c0\u2080.\u2085",
      "overall": 30.75,
      "simple": 30.48,
      "moderate": 33.08,
      "complex": 26.67,
      "episodes": 400,
      "successes": 123,
      "successes_by_difficulty": {
        "simple": 64,
        "moderate": 43,
        "complex": 16
      }
    },
    {
      "method": "Cosmos 3",
      "overall": 42.25,
      "simple": 44.76,
      "moderate": 45.38,
      "complex": 26.67,
      "episodes": 400,
      "successes": 169,
      "successes_by_difficulty": {
        "simple": 94,
        "moderate": 59,
        "complex": 16
      }
    }
  ],
  "continual_success_counts": [
    19,
    20,
    22,
    22
  ],
  "continual_episodes_per_revision": 40,
  "sources": [
    "Completed 40-task campaign aggregates, report 142",
    "Continual-harness revision aggregates, report 154",
    "Native episode audits for pi0.5 and mini-SWE",
    "Author-supplied per-task Cosmos 3 report"
  ],
  "source_notes": {
    "pi05": "Recomputed from 400 native episode results.",
    "mini_SWE": "Recomputed from 120 native verifiers.",
    "Cosmos_3": "User-supplied per-task report; raw server logs not independently inspected.",
    "language_agents": "Completed selected-episode campaign aggregates; comparison conditions documented in this file. Original campaign reports are archived by the authors."
  },
  "conditions": "Method-specific execution limits, rerun selection, and episode counts differ. Continual revisions include implementation and execution changes alongside playbook updates."
}