{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 56.9,
    "wer_norm": 56.71,
    "wer_numcanon": 56.71,
    "space_norm_wer": 47.4,
    "mer": 28.35,
    "cer_norm": 26.6,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.19,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.31,
      "norm_to_mer": 28.36
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 34.7,
    "wer_norm": 34.07,
    "wer_numcanon": 34.07,
    "space_norm_wer": 25.68,
    "mer": 11.84,
    "cer_norm": 11.52,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.63,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 8.39,
      "norm_to_mer": 22.23
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 27.62,
    "wer_norm": 14.55,
    "wer_numcanon": 14.07,
    "space_norm_wer": 8.7,
    "mer": 8.89,
    "cer_norm": 8.9,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 13.07,
      "norm_to_numcanon": 0.48,
      "norm_to_space_norm": 5.85,
      "norm_to_mer": 5.66
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 28.71,
    "wer_norm": 28.19,
    "wer_numcanon": 28.19,
    "space_norm_wer": 21.69,
    "mer": 9.62,
    "cer_norm": 9.37,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.52,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.5,
      "norm_to_mer": 18.57
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 14.26,
    "wer_norm": 14.12,
    "wer_numcanon": 14.11,
    "space_norm_wer": 10.94,
    "mer": 5.58,
    "cer_norm": 5.13,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.14,
      "norm_to_numcanon": 0.01,
      "norm_to_space_norm": 3.18,
      "norm_to_mer": 8.54
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 50.21,
    "wer_norm": 49.57,
    "wer_numcanon": 49.57,
    "space_norm_wer": 37.46,
    "mer": 13.76,
    "cer_norm": 13.97,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.64,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 12.11,
      "norm_to_mer": 35.81
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 58.59,
    "wer_norm": 56.77,
    "wer_numcanon": 56.77,
    "space_norm_wer": 45.2,
    "mer": 15.42,
    "cer_norm": 15.13,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.82,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.57,
      "norm_to_mer": 41.35
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 41.77,
    "wer_norm": 41.49,
    "wer_numcanon": 41.49,
    "space_norm_wer": 32.52,
    "mer": 13.58,
    "cer_norm": 13.58,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.28,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 8.97,
      "norm_to_mer": 27.91
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 44.16,
    "wer_norm": 43.83,
    "wer_numcanon": 43.83,
    "space_norm_wer": 33.77,
    "mer": 15.13,
    "cer_norm": 14.72,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.33,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.06,
      "norm_to_mer": 28.7
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 29.62,
    "wer_norm": 29.43,
    "wer_numcanon": 29.43,
    "space_norm_wer": 22.15,
    "mer": 11.83,
    "cer_norm": 11.26,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.19,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.28,
      "norm_to_mer": 17.6
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 54.54,
    "wer_norm": 52.9,
    "wer_numcanon": 52.9,
    "space_norm_wer": 41.77,
    "mer": 15.46,
    "cer_norm": 15.37,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.64,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 11.13,
      "norm_to_mer": 37.44
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 46.05,
    "wer_norm": 44.16,
    "wer_numcanon": 44.16,
    "space_norm_wer": 30.68,
    "mer": 11.49,
    "cer_norm": 11.79,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 1.89,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 13.48,
      "norm_to_mer": 32.67
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 40.81,
    "wer_norm": 39.12,
    "wer_numcanon": 39.08,
    "space_norm_wer": 30.17,
    "mer": 13.96,
    "cer_norm": 13.64
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 40.59,
    "wer_norm": 38.82,
    "wer_numcanon": 38.78,
    "space_norm_wer": 29.83,
    "mer": 13.41,
    "cer_norm": 13.11
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-ckpt-24000",
    "checkpoint_name": "ckpt-24000",
    "model_id": "qwen3-asr",
    "model_type": "Qwen3-ASR-1.7B",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 64,
    "inference_time_sec": 147.85,
    "total_audio_sec": 40354.46,
    "rtf": 0.0037,
    "timestamp": "2026-03-28T06:49:47Z",
    "gpu": "NVIDIA H200 80GB",
    "framework": "vllm",
    "normalization_version": "v1",
    "jiwer_version": "3.1.0"
  }
}