{
 "cols": [
  "Trio-Spark",
  "Jev 1.14",
  "Clef",
  "Clef-Flash",
  "Laya"
 ],
 "rows": [
  {
   "name": "AG News",
   "n": 300,
   "correct": {
    "Trio-Spark": 272,
    "Jev 1.14": 248,
    "Clef": 261,
    "Clef-Flash": 266,
    "Laya": 275
   },
   "pct": {
    "Trio-Spark": 90.7,
    "Jev 1.14": 82.7,
    "Clef": 87.0,
    "Clef-Flash": 88.7,
    "Laya": 91.7
   }
  },
  {
   "name": "ARC-Easy",
   "n": 300,
   "correct": {
    "Trio-Spark": 291,
    "Jev 1.14": 299,
    "Clef": 299,
    "Clef-Flash": 299,
    "Laya": 144
   },
   "pct": {
    "Trio-Spark": 97.0,
    "Jev 1.14": 99.7,
    "Clef": 99.7,
    "Clef-Flash": 99.7,
    "Laya": 48.0
   }
  },
  {
   "name": "BoolQ",
   "n": 300,
   "correct": {
    "Trio-Spark": 275,
    "Jev 1.14": 271,
    "Clef": 278,
    "Clef-Flash": 275,
    "Laya": 242
   },
   "pct": {
    "Trio-Spark": 91.7,
    "Jev 1.14": 90.3,
    "Clef": 92.7,
    "Clef-Flash": 91.7,
    "Laya": 80.7
   }
  },
  {
   "name": "HellaSwag",
   "n": 300,
   "correct": {
    "Trio-Spark": 284,
    "Jev 1.14": 284,
    "Clef": 294,
    "Clef-Flash": 296,
    "Laya": 85
   },
   "pct": {
    "Trio-Spark": 94.7,
    "Jev 1.14": 94.7,
    "Clef": 98.0,
    "Clef-Flash": 98.7,
    "Laya": 28.3
   }
  },
  {
   "name": "PIQA",
   "n": 300,
   "correct": {
    "Trio-Spark": 279,
    "Jev 1.14": 287,
    "Clef": 273,
    "Clef-Flash": 271,
    "Laya": 161
   },
   "pct": {
    "Trio-Spark": 93.0,
    "Jev 1.14": 95.7,
    "Clef": 91.0,
    "Clef-Flash": 90.3,
    "Laya": 53.7
   }
  },
  {
   "name": "PubMedQA",
   "n": 300,
   "correct": {
    "Trio-Spark": 248,
    "Jev 1.14": 230,
    "Clef": 233,
    "Clef-Flash": 239,
    "Laya": 171
   },
   "pct": {
    "Trio-Spark": 82.7,
    "Jev 1.14": 76.7,
    "Clef": 77.7,
    "Clef-Flash": 79.7,
    "Laya": 57.0
   }
  },
  {
   "name": "RTE",
   "n": 277,
   "correct": {
    "Trio-Spark": 257,
    "Jev 1.14": 253,
    "Clef": 254,
    "Clef-Flash": 252,
    "Laya": 230
   },
   "pct": {
    "Trio-Spark": 92.8,
    "Jev 1.14": 91.3,
    "Clef": 91.7,
    "Clef-Flash": 91.0,
    "Laya": 83.0
   }
  },
  {
   "name": "SST-2",
   "n": 300,
   "correct": {
    "Trio-Spark": 289,
    "Jev 1.14": 286,
    "Clef": 283,
    "Clef-Flash": 289,
    "Laya": 273
   },
   "pct": {
    "Trio-Spark": 96.3,
    "Jev 1.14": 95.3,
    "Clef": 94.3,
    "Clef-Flash": 96.3,
    "Laya": 91.0
   }
  },
  {
   "name": "TweetEval offensive",
   "n": 300,
   "correct": {
    "Trio-Spark": 261,
    "Jev 1.14": 253,
    "Clef": 248,
    "Clef-Flash": 253,
    "Laya": 235
   },
   "pct": {
    "Trio-Spark": 87.0,
    "Jev 1.14": 84.3,
    "Clef": 82.7,
    "Clef-Flash": 84.3,
    "Laya": 78.3
   }
  },
  {
   "name": "WinoGrande",
   "n": 300,
   "correct": {
    "Trio-Spark": 269,
    "Jev 1.14": 268,
    "Clef": 281,
    "Clef-Flash": 294,
    "Laya": 151
   },
   "pct": {
    "Trio-Spark": 89.7,
    "Jev 1.14": 89.3,
    "Clef": 93.7,
    "Clef-Flash": 98.0,
    "Laya": 50.3
   }
  },
  {
   "name": "SNLI",
   "n": 150,
   "correct": {
    "Trio-Spark": 133,
    "Jev 1.14": 114,
    "Clef": 124,
    "Clef-Flash": 124,
    "Laya": 69
   },
   "pct": {
    "Trio-Spark": 88.7,
    "Jev 1.14": 76.0,
    "Clef": 82.7,
    "Clef-Flash": 82.7,
    "Laya": 46.0
   }
  },
  {
   "name": "HelpSteer2 (0–4 score)",
   "n": 300,
   "pct": {
    "Trio-Spark": 44.3,
    "Jev 1.14": 38.9,
    "Clef": 41.3,
    "Clef-Flash": 42.7,
    "Laya": 35.3
   },
   "note": "peers: mean over 5 option orders (1,500 decisions); ours one ask per item, over-limit requests count wrong"
  },
  {
   "name": "Banking77 (77-way)†",
   "n": 300,
   "pct": {
    "Trio-Spark": 88.2,
    "Jev 1.14": 80.1,
    "Clef": 94.6,
    "Clef-Flash": 94.7,
    "Laya": 38.7
   },
   "note": "77 options exceed the API's 8-choice limit: ours read out offline (v1.3 served Q4, v1.2 BF16 research adapter), 5 option orders; v1.0 not measured"
  }
 ],
 "summary": {
  "panel_total_correct": {
   "Trio-Spark": 2725,
   "Jev 1.14": 2679,
   "Clef": 2704,
   "Clef-Flash": 2734,
   "Laya": 1967
  },
  "panel_total_n": 2977,
  "latency_ms": {
   "n": 4566,
   "p50": 118.4,
   "p90": 229.0
  },
  "model_version": "trio-spark-2026-10-28",
  "best_in_row": {
   "Trio-Spark": 6,
   "Jev 1.14": 2,
   "Clef": 2,
   "Clef-Flash": 5,
   "Laya": 1
  }
 }
}