{
  "scope": "All ten CALVIN sequences for all ten models: 100 adopted single-run episodes. Prefixes and full-sequence success reproduce the retained CALVIN results.",
  "definition": "An execution-error episode contains a failed code-execution round, excluding transport failures, timeouts, and budget guards. Resumed execution requires a later execution_ok=true round. Recovered completion requires a later native-success round.",
  "models": [
    {
      "model": "astra",
      "n": 10,
      "mean_prefix": 3.1,
      "prefix_observed": 10,
      "reached": [
        8,
        8,
        7,
        5,
        3
      ],
      "error_episodes": 3,
      "resumed": 3,
      "completed_after_error": 1,
      "completed": 3
    },
    {
      "model": "sol",
      "n": 10,
      "mean_prefix": 1.4,
      "prefix_observed": 10,
      "reached": [
        8,
        4,
        1,
        1,
        0
      ],
      "error_episodes": 3,
      "resumed": 3,
      "completed_after_error": 0,
      "completed": 0
    },
    {
      "model": "fable",
      "n": 10,
      "mean_prefix": 0.9,
      "prefix_observed": 10,
      "reached": [
        7,
        2,
        0,
        0,
        0
      ],
      "error_episodes": 9,
      "resumed": 9,
      "completed_after_error": 0,
      "completed": 0
    },
    {
      "model": "qwen",
      "n": 10,
      "mean_prefix": 0.8,
      "prefix_observed": 9,
      "reached": [
        5,
        1,
        1,
        1,
        0
      ],
      "error_episodes": 10,
      "resumed": 9,
      "completed_after_error": 0,
      "completed": 0
    },
    {
      "model": "gemini",
      "n": 10,
      "mean_prefix": 1.8,
      "prefix_observed": 10,
      "reached": [
        8,
        5,
        4,
        1,
        0
      ],
      "error_episodes": 8,
      "resumed": 8,
      "completed_after_error": 0,
      "completed": 0
    },
    {
      "model": "qwen35-397b",
      "n": 10,
      "mean_prefix": 0,
      "prefix_observed": 5,
      "reached": [
        0,
        0,
        0,
        0,
        0
      ],
      "error_episodes": 2,
      "resumed": 2,
      "completed_after_error": 0,
      "completed": 0
    },
    {
      "model": "qwen35-27b",
      "n": 10,
      "mean_prefix": 0.1,
      "prefix_observed": 10,
      "reached": [
        1,
        0,
        0,
        0,
        0
      ],
      "error_episodes": 7,
      "resumed": 6,
      "completed_after_error": 0,
      "completed": 0
    },
    {
      "model": "opus55",
      "n": 10,
      "mean_prefix": 2.4,
      "prefix_observed": 10,
      "reached": [
        8,
        7,
        4,
        4,
        1
      ],
      "error_episodes": 9,
      "resumed": 9,
      "completed_after_error": 1,
      "completed": 1
    },
    {
      "model": "deepseek",
      "n": 10,
      "mean_prefix": 1.4,
      "prefix_observed": 10,
      "reached": [
        5,
        4,
        3,
        1,
        1
      ],
      "error_episodes": 10,
      "resumed": 10,
      "completed_after_error": 1,
      "completed": 1
    },
    {
      "model": "gemini4argon",
      "n": 10,
      "mean_prefix": 3.3,
      "prefix_observed": 10,
      "reached": [
        10,
        9,
        5,
        5,
        4
      ],
      "error_episodes": 4,
      "resumed": 3,
      "completed_after_error": 0,
      "completed": 4
    }
  ],
  "total": {
    "n": 100,
    "error_episodes": 65,
    "resumed": 62,
    "completed_after_error": 3,
    "completed": 9
  }
}
