aisingapore/kalahi

Kalahi是一个由菲律宾本地人协作创建的高质量、手动制作的文化数据集，旨在评估大型语言模型在处理菲律宾人日常生活中面临的文化特定情境时提供相关回应的能力。该数据集包含150个情境丰富的提示以及与菲律宾文化相关和不相关的响应，覆盖了共享的菲律宾文化知识和价值观。数据集涉及的文化主题包括美容与服饰、信仰与习俗、职业与生计、交流与肢体语言、约会与求爱、家庭与婚姻、食物与聚会、友谊、健康与福祉、本地知识以及社交礼仪等。

Kalahi is a high-quality, manually-crafted cultural dataset created collaboratively by native Filipino speakers, designed to evaluate the ability of LLMs to provide relevant responses to culturally-specific situations that Filipinos encounter in their daily lives. The dataset consists of 150 situationally-enriched prompts and culturally relevant and irrelevant responses, covering shared Filipino cultural knowledge and values, including topics such as beauty and clothing, beliefs and practices, career and livelihood, communication and body language, dating and courtship, family and marriage, food and gatherings, friendship, health and wellness, local know-how, and social etiquette.