davanstrien/imdb-classify-augment-v4-prompt-fix

这是一个由LLM标注的IMDB分类增强数据集，使用classify-and-augment工具生成。数据集采用HuggingFaceTB/SmolLM3-3B模型进行标注，包含positive和negative两类情感标签。原始输入20条数据，通过增强生成30条数据。标签分布均衡，正负标签各15条（真实数据11负/9正，合成数据4负/6正）。合成数据经过自我一致性验证，接受率为50%。

An LLM-annotated IMDB classification augmentation dataset produced by classify-and-augment. The dataset was annotated using HuggingFaceTB/SmolLM3-3B model with positive and negative sentiment labels. Original 20 input rows were augmented to 30 output rows. Balanced label distribution with 15 each (11 real negative/9 real positive, 4 synthetic negative/6 synthetic positive). Synthetic data underwent self-consistency check with 50% acceptance rate.