{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 49.16,
    "wer_norm": 48.93,
    "wer_numcanon": 48.93,
    "space_norm_wer": 41.2,
    "mer": 24.04,
    "cer_norm": 22.53,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.23,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.73,
      "norm_to_mer": 24.89
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 31.95,
    "wer_norm": 31.31,
    "wer_numcanon": 31.31,
    "space_norm_wer": 23.33,
    "mer": 10.5,
    "cer_norm": 10.22,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.64,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.98,
      "norm_to_mer": 20.81
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 26.41,
    "wer_norm": 14.12,
    "wer_numcanon": 13.64,
    "space_norm_wer": 8.35,
    "mer": 8.57,
    "cer_norm": 8.6,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 12.29,
      "norm_to_numcanon": 0.48,
      "norm_to_space_norm": 5.77,
      "norm_to_mer": 5.55
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 26.69,
    "wer_norm": 26.26,
    "wer_numcanon": 26.26,
    "space_norm_wer": 20.54,
    "mer": 8.98,
    "cer_norm": 8.77,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.43,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.72,
      "norm_to_mer": 17.28
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 14.04,
    "wer_norm": 13.97,
    "wer_numcanon": 13.95,
    "space_norm_wer": 10.97,
    "mer": 5.45,
    "cer_norm": 5.02,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.07,
      "norm_to_numcanon": 0.02,
      "norm_to_space_norm": 3.0,
      "norm_to_mer": 8.52
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 46.52,
    "wer_norm": 45.65,
    "wer_numcanon": 45.65,
    "space_norm_wer": 34.08,
    "mer": 12.02,
    "cer_norm": 12.31,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.87,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.57,
      "norm_to_mer": 33.63
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 52.48,
    "wer_norm": 50.88,
    "wer_numcanon": 50.88,
    "space_norm_wer": 40.62,
    "mer": 12.38,
    "cer_norm": 12.19,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.6,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.26,
      "norm_to_mer": 38.5
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 39.03,
    "wer_norm": 38.57,
    "wer_numcanon": 38.57,
    "space_norm_wer": 30.58,
    "mer": 12.72,
    "cer_norm": 12.72,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.46,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.99,
      "norm_to_mer": 25.85
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 40.8,
    "wer_norm": 40.43,
    "wer_numcanon": 40.43,
    "space_norm_wer": 29.91,
    "mer": 13.34,
    "cer_norm": 13.12,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.37,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.52,
      "norm_to_mer": 27.09
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 28.13,
    "wer_norm": 27.87,
    "wer_numcanon": 27.87,
    "space_norm_wer": 20.02,
    "mer": 11.86,
    "cer_norm": 11.37,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.26,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.85,
      "norm_to_mer": 16.01
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 49.71,
    "wer_norm": 47.97,
    "wer_numcanon": 47.97,
    "space_norm_wer": 36.85,
    "mer": 12.73,
    "cer_norm": 12.76,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.74,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.12,
      "norm_to_mer": 35.24
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 40.91,
    "wer_norm": 39.15,
    "wer_numcanon": 39.15,
    "space_norm_wer": 27.27,
    "mer": 9.47,
    "cer_norm": 9.78,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.76,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.88,
      "norm_to_mer": 29.68
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 37.28,
    "wer_norm": 35.64,
    "wer_numcanon": 35.6,
    "space_norm_wer": 27.26,
    "mer": 12.16,
    "cer_norm": 11.96
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 37.15,
    "wer_norm": 35.43,
    "wer_numcanon": 35.38,
    "space_norm_wer": 26.98,
    "mer": 11.84,
    "cer_norm": 11.62
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-ckpt-200000",
    "checkpoint_name": "ckpt-200000",
    "model_id": "qwen3-asr",
    "model_type": "Qwen3-ASR-1.7B",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 64,
    "inference_time_sec": 265.92,
    "total_audio_sec": 40354.46,
    "rtf": 0.0066,
    "timestamp": "2026-03-28T06:50:53Z",
    "gpu": "NVIDIA H200 80GB",
    "framework": "vllm",
    "normalization_version": "v1",
    "jiwer_version": "3.1.0"
  }
}