vamboai/fikira

Fikira（斯瓦希里语意为“思考/推理”）是一个针对非洲语言的多语言推理数据集，由Vambo AI开发。该数据集包含50,000个推理示例，覆盖10种非洲语言，每种语言有5,000个示例。这些示例分为六个推理类别：数学（算术、代数和数值推理）、常识（日常知识和实际推理）、逻辑（演绎和归纳推理）、阅读理解（文本理解和分析）、因果（因果关系）和类比（模式识别和比较）。每个示例包含问题陈述、逐步推理和最终答案，均以目标语言呈现。数据集可用于微调大型语言模型、多语言评估、教育应用、研究和聊天机器人开发。

Fikira (Swahili for "thinking/reasoning") is a multilingual reasoning dataset for African languages, developed by Vambo AI. This dataset contains 50,000 reasoning examples across 10 African languages, with 5,000 examples per language. The examples are categorized into six reasoning types: Mathematical (arithmetic, algebra, and numerical reasoning), Commonsense (everyday knowledge and practical reasoning), Logical (deductive and inductive reasoning), Reading Comprehension (understanding and analyzing text), Causal (cause and effect relationships), and Analogical (pattern recognition and comparisons). Each example includes a problem statement, step-by-step reasoning, and the final answer, all in the target language. The dataset is suitable for fine-tuning LLMs, multilingual evaluation, educational applications, research, and chatbot development.