{
  "title": "Unseen tasks in Skild’s internal evaluation",
  "period": "S1 release, 2026",
  "compiled_on": "2026-10-07",
  "scope": "Selected figures compiled for an unpublished report preview; not a full underlying dataset",
  "methodology": "Company-reported results at 100,000 hours of pre-training, using the same data and compute. The metric is cumulative per-step success across tasks. Human intervention was used to recover from failures during rollouts. These percentages are not end-to-end autonomous task completion rates.",
  "rows": [
    {
      "How the task is specified": "Language-prompted baseline",
      "Cumulative per-step success": "9%",
      "source_url": "https://skild.ai/blogs/s1",
      "source_period": "2026 release",
      "verification": "Primary source checked 2026-10-07"
    },
    {
      "How the task is specified": "Video demonstration in context",
      "Cumulative per-step success": "66%",
      "source_url": "https://skild.ai/blogs/s1",
      "source_period": "2026 release",
      "verification": "Primary source checked 2026-10-07"
    }
  ]
}