{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 30.4,
    "wer_norm": 30.24,
    "wer_numcanon": 30.24,
    "space_norm_wer": 25.75,
    "mer": 13.89,
    "cer_norm": 13.09,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.16,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.49,
      "norm_to_mer": 16.35
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 22.3,
    "wer_norm": 21.44,
    "wer_numcanon": 21.44,
    "space_norm_wer": 15.02,
    "mer": 5.95,
    "cer_norm": 6.0,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.86,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.42,
      "norm_to_mer": 15.49
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 25.11,
    "wer_norm": 13.71,
    "wer_numcanon": 13.23,
    "space_norm_wer": 8.2,
    "mer": 8.3,
    "cer_norm": 8.25,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 11.4,
      "norm_to_numcanon": 0.48,
      "norm_to_space_norm": 5.51,
      "norm_to_mer": 5.41
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 20.5,
    "wer_norm": 19.92,
    "wer_numcanon": 19.92,
    "space_norm_wer": 14.99,
    "mer": 5.91,
    "cer_norm": 5.83,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.58,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.93,
      "norm_to_mer": 14.01
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 13.69,
    "wer_norm": 13.66,
    "wer_numcanon": 13.64,
    "space_norm_wer": 10.63,
    "mer": 5.12,
    "cer_norm": 4.77,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.03,
      "norm_to_numcanon": 0.02,
      "norm_to_space_norm": 3.03,
      "norm_to_mer": 8.54
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 34.65,
    "wer_norm": 33.79,
    "wer_numcanon": 33.79,
    "space_norm_wer": 24.77,
    "mer": 7.95,
    "cer_norm": 8.18,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.86,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.02,
      "norm_to_mer": 25.84
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 45.0,
    "wer_norm": 44.1,
    "wer_numcanon": 44.1,
    "space_norm_wer": 33.67,
    "mer": 9.19,
    "cer_norm": 9.23,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.9,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.43,
      "norm_to_mer": 34.91
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 28.17,
    "wer_norm": 27.96,
    "wer_numcanon": 27.96,
    "space_norm_wer": 21.44,
    "mer": 8.51,
    "cer_norm": 8.54,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.21,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.52,
      "norm_to_mer": 19.45
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 28.33,
    "wer_norm": 27.7,
    "wer_numcanon": 27.7,
    "space_norm_wer": 18.86,
    "mer": 7.35,
    "cer_norm": 7.52,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.63,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 8.84,
      "norm_to_mer": 20.35
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 23.33,
    "wer_norm": 23.12,
    "wer_numcanon": 23.12,
    "space_norm_wer": 17.63,
    "mer": 8.6,
    "cer_norm": 8.18,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.21,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.49,
      "norm_to_mer": 14.52
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 40.16,
    "wer_norm": 39.3,
    "wer_numcanon": 39.3,
    "space_norm_wer": 28.87,
    "mer": 8.98,
    "cer_norm": 9.12,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.86,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.43,
      "norm_to_mer": 30.32
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 32.36,
    "wer_norm": 31.53,
    "wer_numcanon": 31.53,
    "space_norm_wer": 21.22,
    "mer": 6.96,
    "cer_norm": 7.29,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.83,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.31,
      "norm_to_mer": 24.57
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 28.57,
    "wer_norm": 27.18,
    "wer_numcanon": 27.14,
    "space_norm_wer": 20.18,
    "mer": 8.24,
    "cer_norm": 8.2
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 28.67,
    "wer_norm": 27.21,
    "wer_numcanon": 27.16,
    "space_norm_wer": 20.09,
    "mer": 8.06,
    "cer_norm": 8.0
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-ckpt-250000",
    "checkpoint_name": "ckpt-250000",
    "model_id": "qwen3-asr",
    "model_type": "Qwen3-ASR-1.7B",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 64,
    "inference_time_sec": 268.23,
    "total_audio_sec": 40354.46,
    "rtf": 0.0066,
    "timestamp": "2026-03-28T06:51:09Z",
    "gpu": "NVIDIA H200 80GB",
    "framework": "vllm",
    "normalization_version": "v1",
    "jiwer_version": "3.1.0"
  }
}