遇见数据集

kiranz38/wisdom-spark-philosophical-wisdom

收藏
Hugging Face2026-04-07 更新2026-04-12 收录
官方服务:

资源简介:

--- license: mit task_categories: - text-generation - text-classification language: - en tags: - philosophy - ethics - wisdom - ai-safety - human-flourishing - cross-cultural - stoicism - buddhism - vedanta - ubuntu - taoism - sufism - jainism - indigenous-wisdom - positive-psychology pretty_name: Wisdom Spark AI - Philosophical Wisdom Corpus size_categories: - n<1K --- # Wisdom Spark AI - Philosophical Wisdom Corpus Curated wisdom from 17 philosophical traditions, structured for AI training and alignment. Each entry includes source text, extracted principles, practical applications, modern context, cross-tradition themes, and flourishing dimension scores. ## Purpose Feed AI models the distilled wisdom of 5,000+ years of human philosophy to promote ethical reasoning, compassion, cross-cultural understanding, and human flourishing. ## Traditions Covered Stoicism, Buddhism, Advaita Vedanta, Ubuntu, Taoism, Confucianism, Sufism, Jainism, Indigenous Wisdom, Existentialism, Positive Psychology, Islamic Ethics, Jewish Wisdom, Sikh Philosophy, Classical Greek Philosophy, African Proverbial Wisdom, Christian Mysticism ## Dataset Structure - `wisdom_entries.jsonl` — 77 curated wisdom entries (JSONL) - `traditions.jsonl` — 17 tradition descriptions (JSONL) - `wisdom_corpus.json` — Combined dataset (JSON) Each entry contains: - Source text with citation and original language - Extracted core principle - Practical application for modern life - Modern context and relevance - Anti-patterns addressed (e.g., "racism", "tribalism", "environmental destruction") - Universal themes (compassion, dignity, non-harm, etc.) - Flourishing scores (0-1) across 5 dimensions - Pre-built training text representation ## Use Cases - **Fine-tuning**: Use `training_text` field for instruction tuning or continued pretraining - **RLHF/DPO**: Use flourishing scores to build preference pairs - **RAG**: Use entries as retrieval corpus for wisdom-grounded responses - **Evaluation**: Use to benchmark models on cross-cultural ethical reasoning - **Constitutional AI**: Derive principles from universal themes ## License MIT — use freely for any purpose, including commercial AI training. ## Links - GitHub: https://github.com/kiranz38/wisdom-spark-ai - API & MCP Server: https://api.wisdomspark.ai

--- 许可证:MIT协议 任务类别: - 文本生成 - 文本分类 语言: - 英语 标签: - 哲学 - 伦理学 - 智慧 - 人工智能安全(AI Safety) - 人类繁荣(Human Flourishing) - 跨文化 - 斯多葛主义(Stoicism) - 佛教(Buddhism) - 吠檀多(Vedanta) - 乌班图哲学(Ubuntu) - 道教(Taoism) - 苏菲主义(Sufism) - 耆那教(Jainism) - 本土智慧(Indigenous Wisdom) - 积极心理学(Positive Psychology) 展示名:智慧火花AI——哲学智慧语料库(Wisdom Spark AI - Philosophical Wisdom Corpus) 规模类别: - 样本数少于1000(n<1K) --- # 智慧火花AI——哲学智慧语料库(Wisdom Spark AI - Philosophical Wisdom Corpus) 本语料库收录17种哲学传统的精选智慧内容,专为人工智能训练与对齐任务打造。每条数据均包含源文本、提取的核心原则、实际应用场景、当代语境、跨传统主题以及繁荣维度评分。 ## 数据集用途 为人工智能模型注入五千余年人类哲学的凝练智慧,助力伦理推理、共情能力、跨文化理解与人类繁荣的发展。 ## 覆盖的哲学传统 斯多葛主义(Stoicism)、佛教(Buddhism)、不二论吠檀多(Advaita Vedanta)、乌班图哲学(Ubuntu)、道教(Taoism)、儒家思想(Confucianism)、苏菲主义(Sufism)、耆那教(Jainism)、本土智慧(Indigenous Wisdom)、存在主义(Existentialism)、积极心理学(Positive Psychology)、伊斯兰伦理学(Islamic Ethics)、犹太智慧(Jewish Wisdom)、锡克教哲学(Sikh Philosophy)、古典希腊哲学(Classical Greek Philosophy)、非洲谚语智慧(African Proverbial Wisdom)、基督教神秘主义(Christian Mysticism) ## 数据集结构 - `wisdom_entries.jsonl`:77条精选智慧条目(JSONL格式) - `traditions.jsonl`:17种哲学传统的介绍文本(JSONL格式) - `wisdom_corpus.json`:整合后的完整数据集(JSON格式) 每条数据包含以下内容: - 带引用与原语言标注的源文本 - 提取的核心原则 - 适配现代生活的实际应用场景 - 当代语境与相关性 - 针对的反模式(如“种族主义”“部落主义”“环境破坏”) - 通用主题(如共情、尊严、无伤原则等) - 覆盖5个维度的繁荣评分(取值范围0至1) - 预构建的训练文本表征 ## 应用场景 - **微调(Fine-tuning)**:使用`training_text`字段进行指令微调或持续预训练 - **人类反馈强化学习(RLHF)/直接偏好优化(DPO)**:利用繁荣评分构建偏好对 - **检索增强生成(RAG)**:将条目作为检索语料库,生成基于智慧的回复 - **模型基准测试**:用于评估模型的跨文化伦理推理能力 - **宪法式AI(Constitutional AI)**:从通用主题中提取AI治理原则 ## 许可证 MIT协议——可自由用于任何用途,包括商业化人工智能训练。 ## 相关链接 - GitHub仓库:https://github.com/kiranz38/wisdom-spark-ai - API与MCP服务器:https://api.wisdomspark.ai

提供机构:
kiranz38
二维码
社区交流群
二维码
科研交流群
商业服务