{
  "benchmark": "AI2D-Captioning Official Test Split (N=75)",
  "target_sota": {
    "diagrammergpt": {
      "cider": 31.7,
      "bertscore": 90.1,
      "bleu_4": 8.9
    }
  },
  "metrics": {
    "num_samples": 75,
    "cider": 35.67,
    "cider_raw": 0.3567,
    "bertscore": {
      "precision": 94.17,
      "recall": 88.06,
      "f1": 90.92
    },
    "bleu_1": 32.94,
    "bleu_4": 11.51,
    "rouge_1": 50.17,
    "rouge_2": 24.26,
    "rouge_l": 41.58
  },
  "status": "SOTA EXCEEDED (CIDEr 35.67 > 31.7, BERTScore 90.92% > 90.1%, BLEU-4 11.51 > 8.90)",
  "margin_over_sota": {
    "cider_points": 3.97,
    "cider_relative_pct": 12.52,
    "bleu_4_points": 2.61
  }
}