{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 54.17,
    "wer_norm": 49.99,
    "wer_numcanon": 49.99,
    "mer": 33.01,
    "cer_norm": 29.97,
    "empty_hypotheses": 3,
    "normalization_delta": {
      "raw_to_norm": 4.17,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 16.98
    },
    "wer_norm_nonum": 49.12,
    "mer_nonum": 22.58,
    "numeric_samples_dropped": 22,
    "space_norm_wer": 43.96
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 18.8,
    "wer_norm": 16.15,
    "wer_numcanon": 16.15,
    "mer": 5.04,
    "cer_norm": 4.99,
    "empty_hypotheses": 1,
    "normalization_delta": {
      "raw_to_norm": 2.66,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 11.11
    },
    "wer_norm_nonum": 12.95,
    "mer_nonum": 2.81,
    "numeric_samples_dropped": 81,
    "space_norm_wer": 11.73
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 21.86,
    "wer_norm": 12.04,
    "wer_numcanon": 12.04,
    "mer": 7.09,
    "cer_norm": 6.99,
    "empty_hypotheses": 7,
    "normalization_delta": {
      "raw_to_norm": 9.82,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 4.94
    },
    "wer_norm_nonum": 11.99,
    "mer_nonum": 7.15,
    "numeric_samples_dropped": 100,
    "space_norm_wer": 9.09
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 23.07,
    "wer_norm": 16.41,
    "wer_numcanon": 16.41,
    "mer": 5.09,
    "cer_norm": 5.08,
    "empty_hypotheses": 1,
    "normalization_delta": {
      "raw_to_norm": 6.66,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 11.32
    },
    "wer_norm_nonum": 14.67,
    "mer_nonum": 3.61,
    "numeric_samples_dropped": 36,
    "space_norm_wer": 11.95
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 14.09,
    "wer_norm": 10.83,
    "wer_numcanon": 10.83,
    "mer": 4.25,
    "cer_norm": 4.05,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 3.25,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 6.59
    },
    "wer_norm_nonum": 10.2,
    "mer_nonum": 3.71,
    "numeric_samples_dropped": 16,
    "space_norm_wer": 7.58
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 35.88,
    "wer_norm": 29.16,
    "wer_numcanon": 29.16,
    "mer": 9.72,
    "cer_norm": 10.15,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 6.72,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 19.45
    },
    "wer_norm_nonum": 22.6,
    "mer_nonum": 4.49,
    "numeric_samples_dropped": 30,
    "space_norm_wer": 16.74
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 452.54,
    "wer_norm": 445.02,
    "wer_numcanon": 445.02,
    "mer": 469.23,
    "cer_norm": 467.93,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 7.52,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": -24.22
    },
    "wer_norm_nonum": 456.79,
    "mer_nonum": 482.11,
    "numeric_samples_dropped": 14,
    "space_norm_wer": 30.58
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 29.12,
    "wer_norm": 23.5,
    "wer_numcanon": 23.5,
    "mer": 7.8,
    "cer_norm": 7.87,
    "empty_hypotheses": 1,
    "normalization_delta": {
      "raw_to_norm": 5.63,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 15.69
    },
    "wer_norm_nonum": 18.74,
    "mer_nonum": 4.46,
    "numeric_samples_dropped": 81,
    "space_norm_wer": 17.89
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 33.64,
    "wer_norm": 26.87,
    "wer_numcanon": 26.87,
    "mer": 6.77,
    "cer_norm": 7.0,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 6.77,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 20.09
    },
    "wer_norm_nonum": 25.13,
    "mer_nonum": 5.0,
    "numeric_samples_dropped": 72,
    "space_norm_wer": 17.84
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 21.17,
    "wer_norm": 14.79,
    "wer_numcanon": 14.79,
    "mer": 4.38,
    "cer_norm": 4.13,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 6.37,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 10.41
    },
    "wer_norm_nonum": 14.46,
    "mer_nonum": 4.13,
    "numeric_samples_dropped": 6,
    "space_norm_wer": 11.71
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 32.31,
    "wer_norm": 26.72,
    "wer_numcanon": 26.72,
    "mer": 4.82,
    "cer_norm": 5.04,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 5.59,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 21.9
    },
    "wer_norm_nonum": 26.55,
    "mer_nonum": 4.74,
    "numeric_samples_dropped": 4,
    "space_norm_wer": 17.63
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 41.76,
    "wer_norm": 32.84,
    "wer_numcanon": 32.84,
    "mer": 10.86,
    "cer_norm": 11.57,
    "empty_hypotheses": 1,
    "normalization_delta": {
      "raw_to_norm": 8.92,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 21.98
    },
    "wer_norm_nonum": 33.0,
    "mer_nonum": 10.85,
    "numeric_samples_dropped": 6,
    "space_norm_wer": 15.12
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 63.8,
    "wer_norm": 57.71,
    "wer_numcanon": 57.71,
    "mer": 66.67,
    "cer_norm": 63.57,
    "wer_norm_nonum": 59.61,
    "mer_nonum": 69.68,
    "numeric_samples_dropped": 468,
    "space_norm_wer": 18.29
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 64.87,
    "wer_norm": 58.69,
    "wer_numcanon": 58.69,
    "mer": 47.34,
    "cer_norm": 47.06,
    "wer_norm_nonum": 58.02,
    "mer_nonum": 46.3,
    "space_norm_wer": 17.65
  },
  "__meta__": {
    "checkpoint": "api/gemini-2.5-flash",
    "checkpoint_name": "baseline",
    "model_id": "gemini-2.5-flash",
    "model_type": "gemini-2.5-flash",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 1,
    "inference_time_sec": 0,
    "total_audio_sec": 40354.42,
    "rtf": 0,
    "timestamp": "2026-03-27T07:17:42Z",
    "normalization_version": "v1",
    "framework": "api"
  }
}