{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 28.21,
    "wer_norm": 27.72,
    "wer_numcanon": 27.72,
    "space_norm_wer": 21.41,
    "mer": 11.79,
    "cer_norm": 11.27,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.49,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.31,
      "norm_to_mer": 15.93
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 19.82,
    "wer_norm": 19.22,
    "wer_numcanon": 19.22,
    "space_norm_wer": 12.18,
    "mer": 4.14,
    "cer_norm": 4.33,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.6,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.04,
      "norm_to_mer": 15.08
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 26.23,
    "wer_norm": 16.84,
    "wer_numcanon": 16.84,
    "space_norm_wer": 11.1,
    "mer": 10.57,
    "cer_norm": 10.37,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 9.38,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.74,
      "norm_to_mer": 6.27
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 20.02,
    "wer_norm": 19.72,
    "wer_numcanon": 19.72,
    "space_norm_wer": 14.17,
    "mer": 5.43,
    "cer_norm": 5.44,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.31,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.55,
      "norm_to_mer": 14.28
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 15.96,
    "wer_norm": 15.96,
    "wer_numcanon": 15.96,
    "space_norm_wer": 11.3,
    "mer": 5.02,
    "cer_norm": 4.85,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.66,
      "norm_to_mer": 10.94
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 30.96,
    "wer_norm": 30.91,
    "wer_numcanon": 30.91,
    "space_norm_wer": 21.52,
    "mer": 6.43,
    "cer_norm": 6.8,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.05,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.39,
      "norm_to_mer": 24.48
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 42.7,
    "wer_norm": 42.18,
    "wer_numcanon": 42.18,
    "space_norm_wer": 31.5,
    "mer": 8.09,
    "cer_norm": 8.18,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.51,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.68,
      "norm_to_mer": 34.09
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 26.7,
    "wer_norm": 26.7,
    "wer_numcanon": 26.7,
    "space_norm_wer": 18.43,
    "mer": 7.22,
    "cer_norm": 7.4,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 8.27,
      "norm_to_mer": 19.47
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 26.06,
    "wer_norm": 26.04,
    "wer_numcanon": 26.04,
    "space_norm_wer": 15.26,
    "mer": 5.44,
    "cer_norm": 5.93,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.02,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.78,
      "norm_to_mer": 20.6
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 19.8,
    "wer_norm": 19.57,
    "wer_numcanon": 19.57,
    "space_norm_wer": 15.24,
    "mer": 6.02,
    "cer_norm": 5.74,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.23,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.33,
      "norm_to_mer": 13.55
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 36.54,
    "wer_norm": 35.7,
    "wer_numcanon": 35.7,
    "space_norm_wer": 26.15,
    "mer": 7.65,
    "cer_norm": 7.79,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.84,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.56,
      "norm_to_mer": 28.05
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 30.14,
    "wer_norm": 29.88,
    "wer_numcanon": 29.88,
    "space_norm_wer": 18.63,
    "mer": 5.66,
    "cer_norm": 6.11,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.26,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.25,
      "norm_to_mer": 24.22
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 26.83,
    "wer_norm": 25.82,
    "wer_numcanon": 25.82,
    "space_norm_wer": 18.11,
    "mer": 7.06,
    "cer_norm": 7.14
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 26.93,
    "wer_norm": 25.87,
    "wer_numcanon": 25.87,
    "space_norm_wer": 18.07,
    "mer": 6.95,
    "cer_norm": 7.02
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-ckpt-76000",
    "checkpoint_name": "ckpt-76000",
    "model_id": "qwen3-asr-mixed",
    "model_type": "qwen3-asr-1.7B-mixed",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 8,
    "inference_time_sec": 1898.44,
    "total_audio_sec": 40354.46,
    "rtf": 0.047,
    "timestamp": "2026-03-31T03:24:58Z",
    "gpu": "NVIDIA A100-SXM4-80GB",
    "framework": "transformers",
    "normalization_version": "v1"
  }
}