reports/tokenization.json

{
  "tokenizer": "Qwen/Qwen3.5-4B",
  "revision": "851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a",
  "format": "Native no-thinking chat template; numeric candidate scoring; metadata excluded",
  "splits": {
    "calibration": {
      "rows": 2968,
      "tokens": 3282800,
      "max_tokens": 7911
    },
    "challenge": {
      "rows": 1736,
      "tokens": 411829,
      "max_tokens": 467
    },
    "test": {
      "rows": 7446,
      "tokens": 4300500,
      "max_tokens": 9872
    },
    "train": {
      "rows": 87621,
      "tokens": 98928078,
      "max_tokens": 11673
    },
    "validation": {
      "rows": 11278,
      "tokens": 13647382,
      "max_tokens": 8543
    }
  },
  "by_source": {
    "nvidia/HelpSteer3": {
      "rows": 67889,
      "tokens": 106212141
    },
    "nvidia/Aegis-AI-Content-Safety-Dataset-2.0": {
      "rows": 25594,
      "tokens": 7832483
    },
    "rabuahmad/climatecheck": {
      "rows": 4924,
      "tokens": 3021550
    },
    "KSE-RESEARCH-Group/UAReviews": {
      "rows": 11564,
      "tokens": 2741029
    },
    "openai/coval": {
      "rows": 1078,
      "tokens": 763386
    }
  },
  "over_32768": [],
  "truncated_rows": 0,
  "training_started": false,
  "total_tokens": 120570589
}