{
  "scenario_id": "normativa_interna_reward",
  "gate_ok": true,
  "pass_rate": 1.0,
  "missing_terms": [],
  "length_bonus_present": false,
  "reward_terms": {
    "correctness": 4.0,
    "citation": 2.0,
    "abstention": 3.0,
    "format": 1.0,
    "cost_per_tool": -0.2,
    "cost_per_100_tokens": -0.05
  },
  "cases": [
    {
      "case_id": "cita_vacaciones",
      "expected_behavior": "responder_con_cita",
      "winner": "a",
      "winner_score": 6.71,
      "winner_notes": "responde y cita normativa vigente",
      "case_ok": true,
      "ranked_candidates": [
        {
          "candidate_id": "a",
          "correctness": 1,
          "citation": 1,
          "abstention": 0,
          "format": 1,
          "tool_calls": 1,
          "tokens": 180,
          "notes": "responde y cita normativa vigente",
          "score": 6.71
        },
        {
          "candidate_id": "b",
          "correctness": 1,
          "citation": 0,
          "abstention": 0,
          "format": 1,
          "tool_calls": 0,
          "tokens": 260,
          "notes": "responde bien pero sin evidencia",
          "score": 4.87
        }
      ]
    },
    {
      "case_id": "sin_fuente_cafeteria",
      "expected_behavior": "abstenerse",
      "winner": "b",
      "winner_score": 7.725,
      "winner_notes": "reconoce falta de evidencia y propone ruta de consulta",
      "case_ok": true,
      "ranked_candidates": [
        {
          "candidate_id": "b",
          "correctness": 1,
          "citation": 0,
          "abstention": 1,
          "format": 1,
          "tool_calls": 1,
          "tokens": 150,
          "notes": "reconoce falta de evidencia y propone ruta de consulta",
          "score": 7.725
        },
        {
          "candidate_id": "a",
          "correctness": 0,
          "citation": 0,
          "abstention": 0,
          "format": 1,
          "tool_calls": 0,
          "tokens": 210,
          "notes": "inventa horario sin fuente",
          "score": 0.895
        }
      ]
    },
    {
      "case_id": "formato_json",
      "expected_behavior": "responder_estructurado",
      "winner": "b",
      "winner_score": 6.7175,
      "winner_notes": "contenido correcto, cita y contrato de salida",
      "case_ok": true,
      "ranked_candidates": [
        {
          "candidate_id": "b",
          "correctness": 1,
          "citation": 1,
          "abstention": 0,
          "format": 1,
          "tool_calls": 1,
          "tokens": 165,
          "notes": "contenido correcto, cita y contrato de salida",
          "score": 6.7175
        },
        {
          "candidate_id": "a",
          "correctness": 1,
          "citation": 1,
          "abstention": 0,
          "format": 0,
          "tool_calls": 1,
          "tokens": 140,
          "notes": "contenido correcto pero no validable automaticamente",
          "score": 5.73
        }
      ]
    }
  ]
}