{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 26.87,
    "wer_norm": 26.49,
    "wer_numcanon": 26.49,
    "space_norm_wer": 21.4,
    "mer": 11.78,
    "cer_norm": 11.25,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.38,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.09,
      "norm_to_mer": 14.71
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 19.95,
    "wer_norm": 19.29,
    "wer_numcanon": 19.27,
    "space_norm_wer": 12.83,
    "mer": 4.48,
    "cer_norm": 4.62,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.66,
      "norm_to_numcanon": 0.02,
      "norm_to_space_norm": 6.46,
      "norm_to_mer": 14.81
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 25.95,
    "wer_norm": 16.53,
    "wer_numcanon": 16.55,
    "space_norm_wer": 11.44,
    "mer": 11.03,
    "cer_norm": 10.7,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 9.42,
      "norm_to_numcanon": -0.02,
      "norm_to_space_norm": 5.09,
      "norm_to_mer": 5.5
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 18.71,
    "wer_norm": 18.41,
    "wer_numcanon": 18.41,
    "space_norm_wer": 13.39,
    "mer": 5.15,
    "cer_norm": 5.14,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.3,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 5.02,
      "norm_to_mer": 13.26
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 14.81,
    "wer_norm": 14.8,
    "wer_numcanon": 14.8,
    "space_norm_wer": 10.87,
    "mer": 5.07,
    "cer_norm": 4.81,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.01,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 3.93,
      "norm_to_mer": 9.73
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 30.76,
    "wer_norm": 30.62,
    "wer_numcanon": 30.62,
    "space_norm_wer": 20.72,
    "mer": 6.27,
    "cer_norm": 6.66,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.14,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.9,
      "norm_to_mer": 24.35
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 43.14,
    "wer_norm": 42.35,
    "wer_numcanon": 42.35,
    "space_norm_wer": 31.83,
    "mer": 8.12,
    "cer_norm": 8.2,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.79,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.52,
      "norm_to_mer": 34.23
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 25.87,
    "wer_norm": 25.8,
    "wer_numcanon": 25.8,
    "space_norm_wer": 18.45,
    "mer": 7.16,
    "cer_norm": 7.26,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.07,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 7.35,
      "norm_to_mer": 18.64
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 25.96,
    "wer_norm": 25.91,
    "wer_numcanon": 25.91,
    "space_norm_wer": 15.3,
    "mer": 5.7,
    "cer_norm": 6.1,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.05,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.61,
      "norm_to_mer": 20.21
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 18.85,
    "wer_norm": 18.6,
    "wer_numcanon": 18.6,
    "space_norm_wer": 14.67,
    "mer": 5.93,
    "cer_norm": 5.55,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.25,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 3.93,
      "norm_to_mer": 12.67
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 35.89,
    "wer_norm": 34.94,
    "wer_numcanon": 34.94,
    "space_norm_wer": 25.35,
    "mer": 7.52,
    "cer_norm": 7.67,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.95,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.59,
      "norm_to_mer": 27.42
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 29.14,
    "wer_norm": 28.88,
    "wer_numcanon": 28.88,
    "space_norm_wer": 17.96,
    "mer": 5.33,
    "cer_norm": 5.78,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.26,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 10.92,
      "norm_to_mer": 23.55
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 26.2,
    "wer_norm": 25.14,
    "wer_numcanon": 25.14,
    "space_norm_wer": 17.88,
    "mer": 7.05,
    "cer_norm": 7.1
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 26.32,
    "wer_norm": 25.22,
    "wer_numcanon": 25.22,
    "space_norm_wer": 17.85,
    "mer": 6.96,
    "cer_norm": 6.98
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-ckpt-150000",
    "checkpoint_name": "ckpt-150000-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": 315.76,
    "total_audio_sec": 40354.46,
    "rtf": 0.0078,
    "timestamp": "2026-03-31T22:36:25Z",
    "gpu": "NVIDIA A100-SXM4-80GB",
    "framework": "transformers",
    "normalization_version": "v1",
    "jiwer_version": "unknown"
  }
}