{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 29.33,
    "wer_norm": 28.95,
    "wer_numcanon": 28.95,
    "space_norm_wer": 21.97,
    "mer": 11.85,
    "cer_norm": 11.36,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.38,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.98,
      "norm_to_mer": 17.1
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 20.1,
    "wer_norm": 19.5,
    "wer_numcanon": 19.5,
    "space_norm_wer": 12.98,
    "mer": 4.41,
    "cer_norm": 4.55,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.6,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.52,
      "norm_to_mer": 15.09
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 27.71,
    "wer_norm": 17.66,
    "wer_numcanon": 17.68,
    "space_norm_wer": 12.17,
    "mer": 11.03,
    "cer_norm": 10.79,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 10.05,
      "norm_to_numcanon": -0.02,
      "norm_to_space_norm": 5.49,
      "norm_to_mer": 6.63
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 18.72,
    "wer_norm": 18.62,
    "wer_numcanon": 18.62,
    "space_norm_wer": 13.75,
    "mer": 5.09,
    "cer_norm": 5.02,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.1,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.87,
      "norm_to_mer": 13.53
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 15.59,
    "wer_norm": 15.58,
    "wer_numcanon": 15.58,
    "space_norm_wer": 11.01,
    "mer": 5.01,
    "cer_norm": 4.8,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.01,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.57,
      "norm_to_mer": 10.57
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 30.43,
    "wer_norm": 30.38,
    "wer_numcanon": 30.38,
    "space_norm_wer": 21.3,
    "mer": 6.41,
    "cer_norm": 6.75,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.05,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.08,
      "norm_to_mer": 23.97
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 42.13,
    "wer_norm": 41.96,
    "wer_numcanon": 41.96,
    "space_norm_wer": 31.61,
    "mer": 7.97,
    "cer_norm": 8.07,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.17,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.35,
      "norm_to_mer": 33.99
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 26.68,
    "wer_norm": 26.68,
    "wer_numcanon": 26.68,
    "space_norm_wer": 19.1,
    "mer": 7.12,
    "cer_norm": 7.27,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.58,
      "norm_to_mer": 19.56
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 25.26,
    "wer_norm": 25.26,
    "wer_numcanon": 25.26,
    "space_norm_wer": 15.06,
    "mer": 4.88,
    "cer_norm": 5.34,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.2,
      "norm_to_mer": 20.38
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 19.24,
    "wer_norm": 19.07,
    "wer_numcanon": 19.07,
    "space_norm_wer": 14.74,
    "mer": 5.85,
    "cer_norm": 5.58,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.17,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.33,
      "norm_to_mer": 13.22
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 35.33,
    "wer_norm": 35.05,
    "wer_numcanon": 35.05,
    "space_norm_wer": 25.94,
    "mer": 7.87,
    "cer_norm": 7.91,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.28,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.11,
      "norm_to_mer": 27.18
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 28.99,
    "wer_norm": 28.88,
    "wer_numcanon": 28.88,
    "space_norm_wer": 18.5,
    "mer": 5.46,
    "cer_norm": 5.86,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.11,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.38,
      "norm_to_mer": 23.42
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 26.56,
    "wer_norm": 25.61,
    "wer_numcanon": 25.62,
    "space_norm_wer": 18.22,
    "mer": 7.02,
    "cer_norm": 7.07
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 26.63,
    "wer_norm": 25.63,
    "wer_numcanon": 25.63,
    "space_norm_wer": 18.18,
    "mer": 6.91,
    "cer_norm": 6.94
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-ckpt-120000",
    "checkpoint_name": "ckpt-120000-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": 321.98,
    "total_audio_sec": 40354.46,
    "rtf": 0.008,
    "timestamp": "2026-03-31T15:31:03Z",
    "gpu": "NVIDIA A100-SXM4-80GB",
    "framework": "transformers",
    "normalization_version": "v1",
    "jiwer_version": "unknown"
  }
}