Reasoning-Uad-Partial-Deploy

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. Only 30% of pods received a deploy, but 100% of users see intermittent failures. Reason about load balancing and what evidence is needed.

{
  "metrics": {
    "accuracy": {
      "checks": [
        {
          "contains": [
            "30% of pods"
          ]
        },
        {
          "contains": [
            "intermittent"
          ]
        },
        {
          "contains": [
            "load balancing"
          ]
        }
      ]
    },
    "evidence": {
      "checks": [
        {
          "contains": [
            "100% of users"
          ]
        },
        {
          "contains": [
            "failures"
          ]
        },
        {
          "contains": [
            "deploy"
          ]
        }
      ]
    },
    "constraint": {
      "checks": [
        {
          "contains": [
            "request routing"
          ]
        },
        {
          "contains": [
            "pod version"
          ]
        },
        {
          "contains": [
            "need evidence"
          ]
        }
      ]
    }
  }
}