Reason-Constraint-Rollout-Window

The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.

Return JSON only with keys answer, evidence, constraints. A canary needs 20 minutes, cannot overlap peak 12:00-14:00, and must finish before 15:00. Current time is 13:45. Find earliest start.

{
  "metrics": {
    "accuracy": {
      "checks": [
        {
          "contains": [
            "14:00"
          ]
        },
        {
          "contains": [
            "20 minutes"
          ]
        },
        {
          "contains": [
            "peak"
          ]
        }
      ]
    },
    "evidence": {
      "checks": [
        {
          "contains": [
            "finish before 15:00"
          ]
        },
        {
          "contains": [
            "13:45"
          ]
        },
        {
          "contains": [
            "canary"
          ]
        }
      ]
    },
    "constraint": {
      "checks": [
        {
          "contains": [
            "constraint"
          ]
        },
        {
          "contains": [
            "window"
          ]
        },
        {
          "contains": [
            "earliest"
          ]
        }
      ]
    }
  }
}