davanstrien/aud-qwen3-4b-instruct-20260428

该数据集是由classify-and-augment工具生成的LLM标注数据集，用于分类任务。数据集使用Qwen/Qwen3-4B-Instruct-2507模型进行标注，包含positive和negative两类标签。原始输入180行，输出233行。标签分布中negative标签149条（均为真实数据），positive标签84条（31条真实数据，53条合成数据）。合成审计显示positive类别合成接受率为93.2%，即同一模型将合成候选重新分类为目标类别的比例。

LLM-annotated dataset produced by classify-and-augment tool for classification tasks. The dataset is annotated using Qwen/Qwen3-4B-Instruct-2507 model with labels positive and negative. Original input rows: 180, output rows: 233. Label distribution: 149 negative (all real data), 84 positive (31 real, 53 synthetic). Synthesis audit shows 93.2% acceptance rate for positive class, indicating the proportion of synthetic candidates re-classified as target class by the same model.