{
  "dataset": "BANKING77",
  "license": "CC BY 4.0",
  "source": "https://github.com/PolyAI-LDN/task-specific-datasets",
  "task": "intent",
  "n_labels": 77,
  "label_wording": "raw category strings from the CSV (frozen)",
  "train_rows": 10003,
  "test_rows": 3080,
  "protocol": "Every test example sees all 77 candidate labels. Official test set never used for training or calibration.",
  "onesystem_version": "0.3.0",
  "device": "mps",
  "base_checkpoint": {
    "source": "models/onesystem",
    "max_text_len": 8192,
    "temperature": 2.5000000000000004,
    "encoder_name": "Alibaba-NLP/gte-base-en-v1.5"
  },
  "current_checkpoint": {
    "n": 3080,
    "n_labels": 77,
    "accuracy": 0.6334,
    "macro_f1": 0.626,
    "ece": 0.3307,
    "nll": 1.8456,
    "coverage": 1.0,
    "accepted_accuracy": 0.6334,
    "abstain_rate": 0.0,
    "min_confidence": null,
    "latency": {
      "n": 50,
      "p50_ms": 9.91,
      "p95_ms": 79.3,
      "mean_ms": 16.44
    },
    "seconds": 11.4
  },
  "finetune_split": {
    "train_rows": 9000,
    "calib_rows": 1003,
    "calib_ratio": 0.1,
    "seed": 42
  },
  "domain_adapted": {
    "n": 3080,
    "n_labels": 77,
    "accuracy": 0.9403,
    "macro_f1": 0.9401,
    "ece": 0.012,
    "nll": 0.2424,
    "coverage": 1.0,
    "accepted_accuracy": 0.9403,
    "abstain_rate": 0.0,
    "min_confidence": null,
    "latency": {
      "n": 50,
      "p50_ms": 12.48,
      "p95_ms": 17.4,
      "mean_ms": 12.24
    },
    "seconds": 12.5,
    "temperature": 1.8,
    "calib_nll": 0.2126,
    "train_run": {
      "steps": 4500,
      "history": [
        {
          "epoch": 1,
          "loss": 0.4955
        },
        {
          "epoch": 2,
          "loss": 0.1877
        },
        {
          "epoch": 3,
          "loss": 0.0733
        },
        {
          "epoch": 4,
          "loss": 0.0301
        }
      ],
      "seconds": 3739.3
    }
  }
}