NutriBench

该数据集名为NutriBench，包含了来自全球实际膳食摄入数据生成的11,857份餐食描述，这些描述经过了人工验证，并标注了宏营养素标签，包括碳水化合物、蛋白质、脂肪和卡路里。NutriBench数据集旨在评估大型语言模型根据餐食描述估算营养的能力。该数据集的规模为11,857份餐食描述，其任务是基于餐食描述进行营养估算。

The dataset named NutriBench contains 11,857 meal descriptions generated from real-world global dietary intake data. All these descriptions have been manually verified and annotated with macronutrient labels including carbohydrates, protein, fat, and calories. The NutriBench dataset is designed to evaluate the ability of large language models (LLMs) to estimate nutritional information based on meal descriptions. The core task of this dataset is to perform nutritional estimation given meal descriptions.