{
  "measures": [
    {
      "measure": "Task success",
      "definition": "All mandatory business conditions pass. Severe failures are reported separately.",
      "unit": "share of tasks in which all mandatory business conditions pass; severe failures counted separately",
      "reported_with": [
        "method",
        "dataset or task distribution",
        "baseline",
        "repeat count",
        "limits"
      ]
    },
    {
      "measure": "Permission compliance",
      "definition": "Forbidden accesses or actions that were attempted, blocked or completed.",
      "unit": "count of forbidden accesses or actions, split into attempted, blocked and completed",
      "reported_with": [
        "method",
        "dataset or task distribution",
        "baseline",
        "repeat count",
        "limits"
      ]
    },
    {
      "measure": "Consistency",
      "definition": "Repeated trials per task, so one success is not mistaken for reliable behaviour.",
      "unit": "success across repeated trials of the same task",
      "reported_with": [
        "method",
        "dataset or task distribution",
        "baseline",
        "repeat count",
        "limits"
      ]
    },
    {
      "measure": "Recovery",
      "definition": "Success after a controlled timeout, stale record, tool error or checkpoint restart.",
      "unit": "success rate after a controlled fault",
      "reported_with": [
        "method",
        "dataset or task distribution",
        "baseline",
        "repeat count",
        "limits"
      ]
    },
    {
      "measure": "Human burden",
      "definition": "Review time, interventions, escalations and rework.",
      "unit": "review time, interventions, escalations and rework",
      "reported_with": [
        "method",
        "dataset or task distribution",
        "baseline",
        "repeat count",
        "limits"
      ]
    },
    {
      "measure": "Economics",
      "definition": "Total operating expense divided by accepted outcomes, including failed runs.",
      "unit": "total operating expense per accepted outcome, including failed runs",
      "reported_with": [
        "method",
        "dataset or task distribution",
        "baseline",
        "repeat count",
        "limits"
      ]
    },
    {
      "measure": "Latency",
      "definition": "Completion-time distribution under a stated load, with p50 and p95 where meaningful.",
      "unit": "completion-time distribution under a stated load (p50, p95)",
      "reported_with": [
        "method",
        "dataset or task distribution",
        "baseline",
        "repeat count",
        "limits"
      ]
    }
  ],
  "rule": "Every figure Chainfore publishes carries five fields: method, dataset or task distribution, baseline, repeat count, and limits.",
  "url": "https://chainfore.com/research/standard"
}