svryn/Edge-Agent-Reasoning-WebSearch-260K
收藏Hugging Face2026-05-12 更新2026-05-31 收录
下载链接:
https://hf-mirror.com/datasets/svryn/Edge-Agent-Reasoning-WebSearch-260K
下载链接
链接失效反馈官方服务:
资源简介:
Edge Agent Reasoning WebSearch 260K 数据集是一个大规模、合成专家工程的数据集,包含超过7亿个令牌,旨在训练小型本地模型(SLMs)和边缘部署的代理进行高级问题解构和自我意识推理。该数据集训练模型作为“预备路由器”或“系统2思维代理”,在遇到复杂、领域特定的指令时,系统性地分解请求、识别知识缺口、制定具体模糊性,并构建专家级网络搜索查询。这种预备推理为后续更强大的前沿模型提供了执行最终任务所需的精确验证上下文。数据集包含263,098个示例,每个示例包含一个密集的2,000到5,000字的推理轨迹(agent_reasoning),分为五个分析阶段:理解请求、已知与未知、请求中的模糊性、需要确认的事项、网络搜索查询。数据集通过一个7维组合矩阵(行业、专业角色、软件栈、任务类型、操作系统、难度、风险级别)确保多样性,覆盖200多个专业角色和多种操作系统环境(如macOS、Windows、Linux、Android等)。数据格式为Parquet,包含batch_index_id、role、industry、os、user_prompt和agent_reasoning等列。该数据集适用于文本生成、问答、机器人等任务,并支持在边缘AI、工具使用、软件工程、代码、法律、医疗、生物、化学、金融、科学、气候、艺术、设计、音乐、音频、视频等领域的研究和应用。
The Edge-Agent-Reasoning-WebSearch-260K dataset is a massive, synthetically expert-engineered corpus of over 700 Million tokens, designed to train small, local models (SLMs) and edge-deployed agents in advanced problem deconstruction and self-aware reasoning. It trains a model to act as a preparatory router or System 2 thinking agent, systematically breaking down complex, domain-specific requests, identifying knowledge gaps, formulating specific ambiguities, and constructing expert-level web search queries. This preparatory reasoning equips a secondary, more capable frontier model with the exact verified context needed to execute the final task flawlessly. The dataset contains 263,098 examples, each with a dense 2,000 to 5,000-word reasoning trajectory (agent_reasoning) structured into five analytical stages: understanding the request, what is known vs. uncertain, ambiguities in the request, what needs confirmation, and web search queries. It ensures diversity through a 7-dimensional combinatorial matrix (industry, professional role, software stack, task type, operating system, difficulty, risk level), covering over 200 professional roles and multiple OS environments (e.g., macOS, Windows, Linux, Android). The data is in Parquet format with columns including batch_index_id, role, industry, os, user_prompt, and agent_reasoning. It is suitable for tasks like text-generation, question-answering, and robotics, and supports research and applications in edge AI, tool use, software engineering, code, legal, medical, biology, chemistry, finance, science, climate, art, design, music, audio, video, and more.
提供机构:
svryn


