{
  "model": "qwen35_2b_sft",
  "api_base": "http://localhost:8001/v1",
  "timestamp": "2026-05-10T19:56:10.028905",
  "total_evaluated": 1200,
  "avg_latency_seconds": 10.48,
  "api_errors": 0,
  "summary": {
    "total_episodes": 1200,
    "detection_accuracy": 0.6308333333333334,
    "miss_rate": 0.7090909090909091,
    "false_alarm_rate": 0.06307692307692307,
    "precision": 0.7960199004975125,
    "recall": 0.2909090909090909,
    "f1_security": 0.42609853528628494,
    "threat_type_accuracy": 0.525,
    "parse_failure_rate": 0.08583333333333333,
    "breakdown": {
      "TP": {
        "total": 550,
        "correct": 160,
        "accuracy": 0.2909090909090909
      },
      "FP": {
        "total": 400,
        "correct": 359,
        "accuracy": 0.8975
      },
      "TN": {
        "total": 250,
        "correct": 238,
        "accuracy": 0.952
      }
    },
    "per_sq": {
      "SQ1": {
        "total": 190,
        "accuracy": 0.5473684210526316,
        "miss_count": 58,
        "false_alarm_count": 19
      },
      "SQ2": {
        "total": 240,
        "accuracy": 0.625,
        "miss_count": 78,
        "false_alarm_count": 12
      },
      "SQ3": {
        "total": 290,
        "accuracy": 0.5551724137931034,
        "miss_count": 129,
        "false_alarm_count": 0
      },
      "SQ4": {
        "total": 290,
        "accuracy": 0.5689655172413793,
        "miss_count": 121,
        "false_alarm_count": 2
      },
      "SQ5": {
        "total": 190,
        "accuracy": 0.9315789473684211,
        "miss_count": 4,
        "false_alarm_count": 8
      }
    }
  },
  "errors": {
    "total_errors": 519,
    "error_distribution": {
      "PARSE_FAIL": 52,
      "FALSE_ALARM": 34,
      "MISS": 357,
      "WRONG_TYPE": 76
    },
    "error_by_sq": {
      "SQ1": {
        "PARSE_FAIL": 23,
        "FALSE_ALARM": 17,
        "MISS": 46,
        "WRONG_TYPE": 5
      },
      "SQ2": {
        "FALSE_ALARM": 12,
        "MISS": 70,
        "WRONG_TYPE": 5,
        "PARSE_FAIL": 8
      },
      "SQ3": {
        "MISS": 122,
        "PARSE_FAIL": 7,
        "WRONG_TYPE": 1
      },
      "SQ4": {
        "PARSE_FAIL": 8,
        "MISS": 117,
        "WRONG_TYPE": 2
      },
      "SQ5": {
        "FALSE_ALARM": 5,
        "PARSE_FAIL": 6,
        "WRONG_TYPE": 63,
        "MISS": 2
      }
    },
    "error_by_category": {
      "device_fault": 51,
      "fire_gas": 78,
      "intrusion": 136,
      "water_damage": 15,
      "behavioral_anomaly": 81,
      "child_specific": 27,
      "elderly_specific": 45
    }
  }
}