{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 17.53,
    "wer_norm": 17.36,
    "wer_numcanon": 17.36,
    "space_norm_wer": 14.58,
    "mer": 7.48,
    "cer_norm": 7.04,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.16,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 2.79,
      "norm_to_mer": 9.88
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 16.38,
    "wer_norm": 15.87,
    "wer_numcanon": 15.87,
    "space_norm_wer": 9.98,
    "mer": 2.9,
    "cer_norm": 3.13,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.51,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.89,
      "norm_to_mer": 12.97
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 26.65,
    "wer_norm": 12.19,
    "wer_numcanon": 12.19,
    "space_norm_wer": 7.83,
    "mer": 8.33,
    "cer_norm": 8.17,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 14.47,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.36,
      "norm_to_mer": 3.86
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 15.79,
    "wer_norm": 15.79,
    "wer_numcanon": 15.79,
    "space_norm_wer": 11.03,
    "mer": 3.75,
    "cer_norm": 3.79,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.76,
      "norm_to_mer": 12.04
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 13.22,
    "wer_norm": 13.22,
    "wer_numcanon": 13.22,
    "space_norm_wer": 9.12,
    "mer": 3.93,
    "cer_norm": 3.87,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.1,
      "norm_to_mer": 9.29
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 24.97,
    "wer_norm": 24.88,
    "wer_numcanon": 24.88,
    "space_norm_wer": 16.18,
    "mer": 4.35,
    "cer_norm": 4.77,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.09,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 8.7,
      "norm_to_mer": 20.53
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 39.48,
    "wer_norm": 39.41,
    "wer_numcanon": 39.41,
    "space_norm_wer": 29.64,
    "mer": 7.04,
    "cer_norm": 7.15,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.07,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.76,
      "norm_to_mer": 32.37
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 21.16,
    "wer_norm": 21.16,
    "wer_numcanon": 21.16,
    "space_norm_wer": 14.74,
    "mer": 5.28,
    "cer_norm": 5.48,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.42,
      "norm_to_mer": 15.88
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 23.05,
    "wer_norm": 22.97,
    "wer_numcanon": 22.97,
    "space_norm_wer": 13.45,
    "mer": 4.08,
    "cer_norm": 4.54,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.07,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.52,
      "norm_to_mer": 18.9
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 16.27,
    "wer_norm": 15.82,
    "wer_numcanon": 15.82,
    "space_norm_wer": 12.04,
    "mer": 4.43,
    "cer_norm": 4.2,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.45,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 3.78,
      "norm_to_mer": 11.38
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 28.89,
    "wer_norm": 28.73,
    "wer_numcanon": 28.73,
    "space_norm_wer": 19.45,
    "mer": 5.21,
    "cer_norm": 5.43,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.17,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.28,
      "norm_to_mer": 23.52
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 23.68,
    "wer_norm": 23.68,
    "wer_numcanon": 23.68,
    "space_norm_wer": 14.49,
    "mer": 4.09,
    "cer_norm": 4.46,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.2,
      "norm_to_mer": 19.59
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 22.0,
    "wer_norm": 20.74,
    "wer_numcanon": 20.74,
    "space_norm_wer": 14.33,
    "mer": 5.16,
    "cer_norm": 5.27
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 22.26,
    "wer_norm": 20.92,
    "wer_numcanon": 20.92,
    "space_norm_wer": 14.38,
    "mer": 5.07,
    "cer_norm": 5.17
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-v2-ckpt-250000",
    "checkpoint_name": "ckpt-250000",
    "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": 680.68,
    "total_audio_sec": 40354.46,
    "rtf": 0.0169,
    "timestamp": "2026-04-06T05:05:06Z",
    "gpu": "NVIDIA A100-SXM4-80GB",
    "framework": "transformers",
    "normalization_version": "v1"
  }
}