{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 50.48,
    "wer_norm": 50.32,
    "wer_numcanon": 50.32,
    "space_norm_wer": 42.53,
    "mer": 24.32,
    "cer_norm": 22.8,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.16,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.79,
      "norm_to_mer": 26.0
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 31.82,
    "wer_norm": 31.18,
    "wer_numcanon": 31.18,
    "space_norm_wer": 23.13,
    "mer": 10.32,
    "cer_norm": 10.07,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.64,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 8.05,
      "norm_to_mer": 20.86
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 27.03,
    "wer_norm": 14.18,
    "wer_numcanon": 13.69,
    "space_norm_wer": 8.35,
    "mer": 8.63,
    "cer_norm": 8.63,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 12.85,
      "norm_to_numcanon": 0.49,
      "norm_to_space_norm": 5.83,
      "norm_to_mer": 5.55
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 26.77,
    "wer_norm": 26.4,
    "wer_numcanon": 26.4,
    "space_norm_wer": 20.62,
    "mer": 8.97,
    "cer_norm": 8.72,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.37,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.78,
      "norm_to_mer": 17.43
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 14.09,
    "wer_norm": 13.93,
    "wer_numcanon": 13.91,
    "space_norm_wer": 10.85,
    "mer": 5.35,
    "cer_norm": 4.95,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.16,
      "norm_to_numcanon": 0.02,
      "norm_to_space_norm": 3.08,
      "norm_to_mer": 8.58
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 46.7,
    "wer_norm": 45.98,
    "wer_numcanon": 45.98,
    "space_norm_wer": 34.65,
    "mer": 12.05,
    "cer_norm": 12.32,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.72,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.33,
      "norm_to_mer": 33.93
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 52.54,
    "wer_norm": 51.29,
    "wer_numcanon": 51.29,
    "space_norm_wer": 40.92,
    "mer": 12.6,
    "cer_norm": 12.39,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.25,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.37,
      "norm_to_mer": 38.69
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 40.71,
    "wer_norm": 40.47,
    "wer_numcanon": 40.47,
    "space_norm_wer": 30.97,
    "mer": 13.95,
    "cer_norm": 14.0,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.24,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.5,
      "norm_to_mer": 26.52
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 41.26,
    "wer_norm": 40.87,
    "wer_numcanon": 40.87,
    "space_norm_wer": 30.8,
    "mer": 13.27,
    "cer_norm": 13.04,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.39,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.07,
      "norm_to_mer": 27.6
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 27.44,
    "wer_norm": 27.07,
    "wer_numcanon": 27.07,
    "space_norm_wer": 20.87,
    "mer": 11.11,
    "cer_norm": 10.52,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.37,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.2,
      "norm_to_mer": 15.96
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 49.12,
    "wer_norm": 47.52,
    "wer_numcanon": 47.52,
    "space_norm_wer": 36.7,
    "mer": 12.47,
    "cer_norm": 12.49,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.6,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.82,
      "norm_to_mer": 35.05
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 41.17,
    "wer_norm": 39.59,
    "wer_numcanon": 39.59,
    "space_norm_wer": 27.6,
    "mer": 9.65,
    "cer_norm": 9.93,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.58,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.99,
      "norm_to_mer": 29.94
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 37.59,
    "wer_norm": 35.98,
    "wer_numcanon": 35.94,
    "space_norm_wer": 27.64,
    "mer": 12.24,
    "cer_norm": 12.02
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 37.43,
    "wer_norm": 35.73,
    "wer_numcanon": 35.69,
    "space_norm_wer": 27.33,
    "mer": 11.89,
    "cer_norm": 11.66
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-ckpt-100000",
    "checkpoint_name": "ckpt-100000",
    "model_id": "qwen3-asr",
    "model_type": "Qwen3-ASR-1.7B",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 64,
    "inference_time_sec": 271.21,
    "total_audio_sec": 40354.46,
    "rtf": 0.0067,
    "timestamp": "2026-03-28T06:50:20Z",
    "gpu": "NVIDIA H200 80GB",
    "framework": "vllm",
    "normalization_version": "v1",
    "jiwer_version": "3.1.0"
  }
}