{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 27.3,
    "wer_norm": 18.0,
    "wer_numcanon": 18.0,
    "mer": 8.07,
    "cer_norm": 7.59,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 9.3,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 9.93
    },
    "wer_norm_nonum": 15.76,
    "mer_nonum": 6.18,
    "numeric_samples_dropped": 60,
    "space_norm_wer": 15.4
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 23.33,
    "wer_norm": 15.65,
    "wer_numcanon": 15.65,
    "mer": 5.02,
    "cer_norm": 4.91,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 7.69,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 10.63
    },
    "wer_norm_nonum": 11.15,
    "mer_nonum": 1.82,
    "numeric_samples_dropped": 136,
    "space_norm_wer": 12.16
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 18.75,
    "wer_norm": 6.82,
    "wer_numcanon": 6.82,
    "mer": 3.08,
    "cer_norm": 3.06,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 11.94,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 3.74
    },
    "wer_norm_nonum": 5.78,
    "mer_nonum": 2.58,
    "numeric_samples_dropped": 97,
    "space_norm_wer": 5.55
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 23.06,
    "wer_norm": 14.71,
    "wer_numcanon": 14.71,
    "mer": 4.26,
    "cer_norm": 4.2,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 8.34,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 10.46
    },
    "wer_norm_nonum": 12.33,
    "mer_nonum": 2.51,
    "numeric_samples_dropped": 51,
    "space_norm_wer": 11.29
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 15.73,
    "wer_norm": 9.8,
    "wer_numcanon": 9.8,
    "mer": 3.46,
    "cer_norm": 3.22,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 5.94,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 6.33
    },
    "wer_norm_nonum": 8.86,
    "mer_nonum": 2.66,
    "numeric_samples_dropped": 29,
    "space_norm_wer": 7.67
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 27.17,
    "wer_norm": 19.81,
    "wer_numcanon": 19.81,
    "mer": 4.4,
    "cer_norm": 4.6,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 7.35,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 15.41
    },
    "wer_norm_nonum": 17.54,
    "mer_nonum": 2.48,
    "numeric_samples_dropped": 55,
    "space_norm_wer": 13.4
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 42.5,
    "wer_norm": 34.74,
    "wer_numcanon": 34.74,
    "mer": 6.9,
    "cer_norm": 6.77,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 7.76,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 27.84
    },
    "wer_norm_nonum": 33.52,
    "mer_nonum": 6.19,
    "numeric_samples_dropped": 52,
    "space_norm_wer": 27.77
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 27.28,
    "wer_norm": 19.66,
    "wer_numcanon": 19.66,
    "mer": 6.68,
    "cer_norm": 6.59,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 7.62,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 12.98
    },
    "wer_norm_nonum": 15.23,
    "mer_nonum": 3.46,
    "numeric_samples_dropped": 85,
    "space_norm_wer": 16.37
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 30.43,
    "wer_norm": 21.57,
    "wer_numcanon": 21.57,
    "mer": 5.64,
    "cer_norm": 5.88,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 8.86,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 15.93
    },
    "wer_norm_nonum": 18.45,
    "mer_nonum": 2.9,
    "numeric_samples_dropped": 114,
    "space_norm_wer": 14.34
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 21.6,
    "wer_norm": 13.58,
    "wer_numcanon": 13.58,
    "mer": 3.9,
    "cer_norm": 3.64,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 8.02,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 9.68
    },
    "wer_norm_nonum": 12.87,
    "mer_nonum": 3.43,
    "numeric_samples_dropped": 28,
    "space_norm_wer": 10.69
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 31.49,
    "wer_norm": 24.31,
    "wer_numcanon": 24.31,
    "mer": 4.03,
    "cer_norm": 4.24,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 7.18,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 20.27
    },
    "wer_norm_nonum": 23.89,
    "mer_nonum": 3.7,
    "numeric_samples_dropped": 26,
    "space_norm_wer": 15.92
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 28.42,
    "wer_norm": 21.44,
    "wer_numcanon": 21.44,
    "mer": 3.26,
    "cer_norm": 3.63,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 6.98,
      "norm_to_numcanon": 0.0,
      "norm_to_mer": 18.18
    },
    "wer_norm_nonum": 21.24,
    "mer_nonum": 2.98,
    "numeric_samples_dropped": 17,
    "space_norm_wer": 12.65
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 26.36,
    "wer_norm": 18.26,
    "wer_numcanon": 18.26,
    "mer": 5.04,
    "cer_norm": 5.01,
    "wer_norm_nonum": 16.52,
    "mer_nonum": 3.61,
    "numeric_samples_dropped": 750,
    "space_norm_wer": 13.61
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 26.42,
    "wer_norm": 18.34,
    "wer_numcanon": 18.34,
    "mer": 4.89,
    "cer_norm": 4.86,
    "wer_norm_nonum": 16.39,
    "mer_nonum": 3.41,
    "space_norm_wer": 13.6
  },
  "__meta__": {
    "checkpoint": "api/sarvam-saaras-v3",
    "checkpoint_name": "baseline",
    "model_id": "sarvam-saaras-v3",
    "model_type": "saaras:v3",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 1,
    "inference_time_sec": 0,
    "total_audio_sec": 40354.42,
    "rtf": 0,
    "timestamp": "2026-03-27T07:17:49Z",
    "normalization_version": "v1",
    "framework": "api"
  }
}