NHK Recipe Dataset
收藏资源简介:
NHK Recipe Dataset是一个高质量的日本食谱数据集,用于评估大型语言模型(LLMs)对烹饪过程中食材状态变化的识别能力。该数据集由日本广播公司(NHK)的专业厨师和编辑创建的食谱组成,每个食谱都附有清晰的食材状态变化注释。数据集包含大约25,000个食谱条目,每个食谱都经过精心编辑,以消除不一致性和错误。数据集在Hugging Face Hub上公开可用,旨在帮助LLMs更好地理解烹饪过程,并支持人类在现实世界中的烹饪活动。
NHK Recipe Dataset is a high-quality Japanese recipe dataset designed for evaluating the ability of large language models (LLMs) to recognize changes in ingredient states during cooking processes. This dataset comprises recipes created by professional chefs and editors from the Japan Broadcasting Corporation (NHK), with each recipe accompanied by clear annotations of ingredient state changes. It contains approximately 25,000 recipe entries, and every recipe has been meticulously edited to eliminate inconsistencies and errors. The dataset is publicly available on the Hugging Face Hub, aiming to help LLMs better understand cooking processes and support human cooking activities in real-world scenarios.

- 1A Highly Clean Recipe Dataset with Ingredient States Annotation for State Probing Task东京大学 · 2025年



