Speed Perf Ttft Factual

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

Answer with a single word only (no punctuation, no explanation): What is the capital city of Portugal?

{
  "expected_contains": [
    "Lisbon",
    "lisbon",
    "LISBON"
  ]
}