{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 28.74,
    "wer_norm": 26.51,
    "wer_numcanon": 26.51,
    "mer": 10.49,
    "cer_norm": 9.91,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 2.23,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 16.03
    },
    "wer_norm_nonum": 26.0,
    "mer_nonum": 9.97,
    "numeric_samples_dropped": 11,
    "space_norm_wer": 22.11
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 13.53,
    "wer_norm": 12.98,
    "wer_numcanon": 12.98,
    "mer": 3.57,
    "cer_norm": 3.61,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.55,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 9.41
    },
    "wer_norm_nonum": 10.44,
    "mer_nonum": 1.65,
    "numeric_samples_dropped": 69,
    "space_norm_wer": 9.13
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 26.34,
    "wer_norm": 10.68,
    "wer_numcanon": 10.68,
    "mer": 5.68,
    "cer_norm": 5.89,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 15.66,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 4.99
    },
    "wer_norm_nonum": 8.37,
    "mer_nonum": 3.53,
    "numeric_samples_dropped": 97,
    "space_norm_wer": 7.49
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 14.22,
    "wer_norm": 14.22,
    "wer_numcanon": 14.22,
    "mer": 3.93,
    "cer_norm": 3.95,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 10.28
    },
    "wer_norm_nonum": 12.39,
    "mer_nonum": 2.45,
    "numeric_samples_dropped": 33,
    "space_norm_wer": 10.45
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 9.02,
    "wer_norm": 9.02,
    "wer_numcanon": 9.02,
    "mer": 2.94,
    "cer_norm": 2.82,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 6.08
    },
    "wer_norm_nonum": 8.37,
    "mer_nonum": 2.37,
    "numeric_samples_dropped": 11,
    "space_norm_wer": 6.63
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 19.54,
    "wer_norm": 19.52,
    "wer_numcanon": 19.52,
    "mer": 3.63,
    "cer_norm": 3.96,
    "empty_hypotheses": 1,
    "normalization_delta": {
      "raw_to_norm": 0.02,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 15.89
    },
    "wer_norm_nonum": 18.49,
    "mer_nonum": 2.76,
    "numeric_samples_dropped": 13,
    "space_norm_wer": 11.65
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 35.03,
    "wer_norm": 35.03,
    "wer_numcanon": 35.03,
    "mer": 6.1,
    "cer_norm": 6.16,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 28.93
    },
    "wer_norm_nonum": 35.01,
    "mer_nonum": 6.05,
    "numeric_samples_dropped": 3,
    "space_norm_wer": 26.43
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 19.42,
    "wer_norm": 19.42,
    "wer_numcanon": 19.42,
    "mer": 5.77,
    "cer_norm": 5.94,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 13.64
    },
    "wer_norm_nonum": 15.81,
    "mer_nonum": 3.21,
    "numeric_samples_dropped": 62,
    "space_norm_wer": 13.91
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 20.83,
    "wer_norm": 20.83,
    "wer_numcanon": 20.83,
    "mer": 4.24,
    "cer_norm": 4.62,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 16.6
    },
    "wer_norm_nonum": 19.25,
    "mer_nonum": 2.87,
    "numeric_samples_dropped": 45,
    "space_norm_wer": 12.49
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 13.63,
    "wer_norm": 13.63,
    "wer_numcanon": 13.63,
    "mer": 3.81,
    "cer_norm": 3.64,
    "empty_hypotheses": 1,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 9.82
    },
    "wer_norm_nonum": 13.34,
    "mer_nonum": 3.61,
    "numeric_samples_dropped": 4,
    "space_norm_wer": 10.15
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 22.42,
    "wer_norm": 22.42,
    "wer_numcanon": 22.42,
    "mer": 3.6,
    "cer_norm": 3.88,
    "empty_hypotheses": 2,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 18.82
    },
    "wer_norm_nonum": 22.27,
    "mer_nonum": 3.53,
    "numeric_samples_dropped": 2,
    "space_norm_wer": 13.15
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 22.04,
    "wer_norm": 22.04,
    "wer_numcanon": 22.04,
    "mer": 2.88,
    "cer_norm": 3.45,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 19.15
    },
    "wer_norm_nonum": 22.08,
    "mer_nonum": 2.85,
    "numeric_samples_dropped": 2,
    "space_norm_wer": 11.38
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 20.5,
    "wer_norm": 18.98,
    "wer_numcanon": 18.98,
    "mer": 4.8,
    "cer_norm": 4.92,
    "wer_norm_nonum": 18.1,
    "mer_nonum": 3.95,
    "numeric_samples_dropped": 352,
    "space_norm_wer": 13.12
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 20.4,
    "wer_norm": 18.86,
    "wer_numcanon": 18.86,
    "mer": 4.72,
    "cer_norm": 4.82,
    "wer_norm_nonum": 17.65,
    "mer_nonum": 3.74,
    "space_norm_wer": 12.91
  },
  "__meta__": {
    "checkpoint": "api/gemini-3.1-pro",
    "checkpoint_name": "baseline",
    "model_id": "gemini-3.1-pro",
    "model_type": "gemini-3.1-pro-preview",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 1,
    "inference_time_sec": 0,
    "total_audio_sec": 40354.42,
    "rtf": 0,
    "timestamp": "2026-03-27T07:17:46Z",
    "normalization_version": "v1",
    "framework": "api"
  }
}