{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 26.99,
    "wer_norm": 26.63,
    "wer_numcanon": 26.63,
    "space_norm_wer": 20.35,
    "mer": 11.75,
    "cer_norm": 11.29,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.36,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.28,
      "norm_to_mer": 14.88
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 19.63,
    "wer_norm": 18.94,
    "wer_numcanon": 18.94,
    "space_norm_wer": 12.44,
    "mer": 4.27,
    "cer_norm": 4.43,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.69,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.5,
      "norm_to_mer": 14.67
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 25.86,
    "wer_norm": 16.42,
    "wer_numcanon": 16.43,
    "space_norm_wer": 11.16,
    "mer": 10.33,
    "cer_norm": 10.07,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 9.44,
      "norm_to_numcanon": -0.01,
      "norm_to_space_norm": 5.26,
      "norm_to_mer": 6.09
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 18.79,
    "wer_norm": 18.35,
    "wer_numcanon": 18.35,
    "space_norm_wer": 13.24,
    "mer": 4.92,
    "cer_norm": 4.94,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.44,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.11,
      "norm_to_mer": 13.43
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 15.18,
    "wer_norm": 15.11,
    "wer_numcanon": 15.11,
    "space_norm_wer": 10.68,
    "mer": 4.7,
    "cer_norm": 4.56,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.07,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.43,
      "norm_to_mer": 10.41
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 30.14,
    "wer_norm": 29.96,
    "wer_numcanon": 29.96,
    "space_norm_wer": 20.54,
    "mer": 6.27,
    "cer_norm": 6.67,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.18,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.42,
      "norm_to_mer": 23.69
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 41.84,
    "wer_norm": 41.12,
    "wer_numcanon": 41.12,
    "space_norm_wer": 31.11,
    "mer": 7.82,
    "cer_norm": 7.92,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.72,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.01,
      "norm_to_mer": 33.3
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 25.15,
    "wer_norm": 25.1,
    "wer_numcanon": 25.1,
    "space_norm_wer": 18.1,
    "mer": 6.8,
    "cer_norm": 6.93,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.05,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.0,
      "norm_to_mer": 18.3
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 24.88,
    "wer_norm": 24.82,
    "wer_numcanon": 24.82,
    "space_norm_wer": 14.63,
    "mer": 4.91,
    "cer_norm": 5.35,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.06,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.19,
      "norm_to_mer": 19.91
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 19.33,
    "wer_norm": 19.09,
    "wer_numcanon": 19.09,
    "space_norm_wer": 14.71,
    "mer": 5.81,
    "cer_norm": 5.49,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.24,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.38,
      "norm_to_mer": 13.28
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 34.18,
    "wer_norm": 33.09,
    "wer_numcanon": 33.09,
    "space_norm_wer": 24.01,
    "mer": 7.22,
    "cer_norm": 7.34,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.09,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.08,
      "norm_to_mer": 25.87
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 28.77,
    "wer_norm": 28.47,
    "wer_numcanon": 28.47,
    "space_norm_wer": 17.91,
    "mer": 5.4,
    "cer_norm": 5.83,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.3,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.56,
      "norm_to_mer": 23.07
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 25.8,
    "wer_norm": 24.71,
    "wer_numcanon": 24.71,
    "space_norm_wer": 17.43,
    "mer": 6.79,
    "cer_norm": 6.86
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 25.89,
    "wer_norm": 24.76,
    "wer_numcanon": 24.76,
    "space_norm_wer": 17.41,
    "mer": 6.68,
    "cer_norm": 6.74
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-ckpt-246000",
    "checkpoint_name": "ckpt-246000-mixed",
    "model_id": "qwen3-asr-mixed",
    "model_type": "qwen3-asr-1.7B-mixed",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 32,
    "inference_time_sec": 432.42,
    "total_audio_sec": 40354.46,
    "rtf": 0.0107,
    "timestamp": "2026-04-02T01:05:41Z",
    "gpu": "NVIDIA A100-SXM4-80GB",
    "framework": "transformers",
    "normalization_version": "v1",
    "jiwer_version": "unknown"
  }
}