Reasoning-Mas-Cache-Rollout

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. Compare a changelog saying cache key now includes locale with logs showing en-US and de-DE responses share one cache entry. Identify the inconsistency.

{
  "metrics": {
    "accuracy": {
      "checks": [
        {
          "contains": [
            "locale"
          ]
        },
        {
          "contains": [
            "same cache entry"
          ]
        },
        {
          "contains": [
            "inconsistency"
          ]
        }
      ]
    },
    "evidence": {
      "checks": [
        {
          "contains": [
            "changelog"
          ]
        },
        {
          "contains": [
            "logs"
          ]
        },
        {
          "contains": [
            "en-US"
          ]
        }
      ]
    },
    "constraint": {
      "checks": [
        {
          "contains": [
            "multi-artifact"
          ]
        },
        {
          "contains": [
            "evidence"
          ]
        },
        {
          "contains": [
            "regression"
          ]
        }
      ]
    }
  }
}