{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 17.66,
    "wer_norm": 17.48,
    "wer_numcanon": 17.48,
    "space_norm_wer": 14.64,
    "mer": 7.52,
    "cer_norm": 7.08,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.18,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 2.84,
      "norm_to_mer": 9.96
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 16.31,
    "wer_norm": 15.81,
    "wer_numcanon": 15.81,
    "space_norm_wer": 9.74,
    "mer": 2.89,
    "cer_norm": 3.13,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.49,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.07,
      "norm_to_mer": 12.92
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 26.54,
    "wer_norm": 12.41,
    "wer_numcanon": 12.41,
    "space_norm_wer": 7.73,
    "mer": 8.28,
    "cer_norm": 8.15,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 14.13,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.68,
      "norm_to_mer": 4.12
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 15.5,
    "wer_norm": 15.5,
    "wer_numcanon": 15.5,
    "space_norm_wer": 10.86,
    "mer": 3.71,
    "cer_norm": 3.74,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.64,
      "norm_to_mer": 11.8
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 13.29,
    "wer_norm": 13.29,
    "wer_numcanon": 13.29,
    "space_norm_wer": 9.22,
    "mer": 3.89,
    "cer_norm": 3.78,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.07,
      "norm_to_mer": 9.4
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 25.0,
    "wer_norm": 24.93,
    "wer_numcanon": 24.93,
    "space_norm_wer": 16.02,
    "mer": 4.33,
    "cer_norm": 4.75,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.07,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 8.92,
      "norm_to_mer": 20.61
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 39.0,
    "wer_norm": 39.0,
    "wer_numcanon": 39.0,
    "space_norm_wer": 29.44,
    "mer": 6.91,
    "cer_norm": 6.99,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.56,
      "norm_to_mer": 32.1
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 20.81,
    "wer_norm": 20.81,
    "wer_numcanon": 20.81,
    "space_norm_wer": 14.46,
    "mer": 5.21,
    "cer_norm": 5.37,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.35,
      "norm_to_mer": 15.6
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 22.86,
    "wer_norm": 22.81,
    "wer_numcanon": 22.81,
    "space_norm_wer": 13.32,
    "mer": 4.04,
    "cer_norm": 4.51,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.06,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.49,
      "norm_to_mer": 18.77
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 16.07,
    "wer_norm": 15.59,
    "wer_numcanon": 15.59,
    "space_norm_wer": 11.99,
    "mer": 4.31,
    "cer_norm": 4.08,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.49,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 3.6,
      "norm_to_mer": 11.28
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 28.67,
    "wer_norm": 28.63,
    "wer_numcanon": 28.63,
    "space_norm_wer": 19.02,
    "mer": 5.09,
    "cer_norm": 5.32,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.04,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.61,
      "norm_to_mer": 23.55
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 23.2,
    "wer_norm": 23.18,
    "wer_numcanon": 23.18,
    "space_norm_wer": 13.88,
    "mer": 3.81,
    "cer_norm": 4.22,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.02,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.31,
      "norm_to_mer": 19.37
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 21.83,
    "wer_norm": 20.61,
    "wer_numcanon": 20.61,
    "space_norm_wer": 14.16,
    "mer": 5.08,
    "cer_norm": 5.19
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 22.08,
    "wer_norm": 20.79,
    "wer_numcanon": 20.79,
    "space_norm_wer": 14.19,
    "mer": 5.0,
    "cer_norm": 5.09
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-v2-ckpt-300000",
    "checkpoint_name": "ckpt-300000",
    "model_id": "qwen3-asr-mixed-v2",
    "model_type": "qwen3-asr-1.7B-mixed-v2",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 128,
    "inference_time_sec": 680.8,
    "total_audio_sec": 40354.46,
    "rtf": 0.0169,
    "timestamp": "2026-04-06T20:52:36Z",
    "gpu": "NVIDIA A100-SXM4-80GB",
    "framework": "transformers",
    "normalization_version": "v1"
  }
}