davanstrien/aud-llama-3.2-3b-instruct-20260428

这是一个由classify-and-augment工具生成的LLM标注数据集，使用了meta-llama/Llama-3.2-3B-Instruct模型进行标注。数据集包含180条输入行和221条输出行，标签分为positive和negative两类。标签分布显示，negative标签有154条真实数据和0条合成数据，positive标签有26条真实数据和41条合成数据。合成审计部分显示，positive类需要74条合成数据，生成了86条，验证了74条，保留了41条，接受率为86.0%。

This is an LLM-annotated dataset produced by classify-and-augment, using the meta-llama/Llama-3.2-3B-Instruct model. The dataset contains 180 input rows and 221 output rows, with labels divided into positive and negative. The label distribution shows 154 real and 0 synthetic data for negative, and 26 real and 41 synthetic data for positive. The synthesis audit indicates that 74 synthetic data were needed for the positive class, 86 were generated, 74 were validated, 41 were kept, with an acceptance rate of 86.0%.