austindixson/agent-dataset-hybrid-v2
收藏资源简介:
--- license: apache-2.0 task_categories: - text-generation - question-answering language: - en tags: - agents - tool-use - coding - reasoning - claude - anthropic - tool-calling - function-calling - code-generation - attribution-required size_categories: - 100K<n<1M dataset_info: - attribution: - Contains data from nohurry/Opus-4.6-Reasoning-3000x-filtered (Apache 2.0) - Contains data from Jackrong/Qwen3.5-reasoning-700x (Apache 2.0) - license: apache-2.0 --- # Agent Dataset Hybrid v2 Fine-tuning dataset for training AI agents with tool-use capabilities and reasoning skills. ## ⚠️ Attribution Notice This dataset incorporates data from the following sources: ### Original Sources 1. **[Opus-4.6-Reasoning-3000x-filtered](https://huggingface.co/datasets/nohurry/Opus-4.6-Reasoning-3000x-filtered)** by nohurry - License: [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) - Contribution: Mathematical and logical reasoning samples with explicit thinking tags - Samples in this dataset: ~2,093 (5.2%) 2. **[Qwen3.5-reasoning-700x](https://huggingface.co/datasets/Jackrong/Qwen3.5-reasoning-700x)** by Jackrong - License: [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) - Contribution: Natural language Chain-of-Thought reasoning - Samples in this dataset: ~569 (1.4%) 3. **Agent conversations (primary dataset)** by austindixson - License: Apache 2.0 - Contribution: Claude Code agent workflows with tool-use - Samples in this dataset: ~39,971 (93.3%) ### License Notice This derivative work is licensed under the **Apache 2.0 License**. Original datasets are used in accordance with their Apache 2.0 licenses. This dataset: - ✅ Includes attribution to original creators - ✅ States this is a derivative work - ✅ Uses the same Apache 2.0 license - ✅ Includes license notice ## Dataset Description Multi-purpose fine-tuning dataset combining: - **Tool-Use Workflows**: Real agent conversations with file operations, code execution, search - **Mathematical Reasoning**: Step-by-step problem solving with `<thinking>` tags - **Natural Reasoning**: Conversational Chain-of-Thought processes ## Dataset Statistics - **Training**: 38,369 conversations (965,819 turns) - **Validation**: 4,264 conversations (44,644 turns) - **Total**: 42,633 conversations (1,010,463 turns) - **Average turns**: 25.3 per conversation - **Tool-use conversations**: ~3,867 (9.7%) ## Data Format Each example is a conversation with multiple turns: ```json { "conversations": [ {"from": "human", "value": "User message"}, {"from": "gpt", "value": "Assistant response"}, {"from": "tool", "value": "Tool execution result"}, {"from": "gpt", "value": "Assistant response to tool result"} ] } ``` ## Use Cases Perfect for fine-tuning models on: - Tool-use and function calling - Code generation and debugging - Multi-step reasoning - Agent workflows - Software development tasks - Mathematical problem solving ## Training Recommendations - Models: Claude, GPT-4, LLaMA, Gemma, GLM - Method: QLoRA, full fine-tuning - Sequence length: 8192+ tokens recommended - Batch size: 16-32 effective ## Citation If you use this dataset, please cite the original sources: ```bibtex @dataset{opus_reasoning, title = {Opus-4.6-Reasoning-3000x-filtered}, author = {nohurry}, year = {2024}, publisher = {HuggingFace} } @dataset{qwen_reasoning, title = {Qwen3.5-reasoning-700x}, author = {Jackrong}, year = {2024}, publisher = {HuggingFace} } @dataset{agent_hybrid, title = {Agent Dataset Hybrid v2}, author = {austindixson}, year = {2025}, publisher = {HuggingFace} } ``` ## License Copyright 2025 austindixson Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ## Changes from Original Datasets This dataset: - Combines multiple datasets into a unified format - Converts all datasets to a conversational format - Adds tool-use execution results from Claude Code conversations - Filters and curates samples for quality - Splits into train/validation sets All modifications comply with Apache 2.0 license requirements.
--- 许可证:Apache 2.0许可证(Apache 2.0) 任务类别: - 文本生成 - 问答 语言: - 英语 标签: - 智能体(AI Agent) - 工具使用(tool-use) - 编程 - 推理 - Claude - Anthropic - 工具调用(tool-calling) - 函数调用(function-calling) - 代码生成(code-generation) - 需注明来源(attribution-required) 规模类别:10万 < 样本数 < 100万 数据集信息: - 来源说明: - 包含来自nohurry/Opus-4.6-Reasoning-3000x-filtered(Apache 2.0)的数据 - 包含来自Jackrong/Qwen3.5-reasoning-700x(Apache 2.0)的数据 - 许可证:Apache 2.0许可证(Apache 2.0) --- # 混合智能体数据集v2(Agent Dataset Hybrid v2) 用于训练具备工具使用能力与推理技能的AI智能体(AI Agent)的微调数据集。 ## ⚠️ 来源注明须知 本数据集整合了以下原始数据源的数据: ### 原始数据源 1. **[Opus-4.6-Reasoning-3000x-filtered](https://huggingface.co/datasets/nohurry/Opus-4.6-Reasoning-3000x-filtered)**,作者:nohurry - 许可证:[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) - 贡献内容:带有显式`<thinking>`思考标记的数学与逻辑推理样本 - 本数据集占比:约2093条(5.2%) 2. **[Qwen3.5-reasoning-700x](https://huggingface.co/datasets/Jackrong/Qwen3.5-reasoning-700x)**,作者:Jackrong - 许可证:[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) - 贡献内容:自然语言思维链(Chain-of-Thought)推理样本 - 本数据集占比:约569条(1.4%) 3. **智能体对话(核心数据集)**,作者:austindixson - 许可证:Apache 2.0 - 贡献内容:带有工具使用流程的Claude Code智能体工作流 - 本数据集占比:约39971条(93.3%) ### 许可证声明 本衍生作品采用Apache 2.0许可证(Apache 2.0)进行授权。 原始数据集均按照其自身的Apache 2.0许可证要求进行使用。本数据集: - ✅ 包含对原始创作者的署名声明 - ✅ 明确标注本作品为衍生作品 - ✅ 采用与原始数据集一致的Apache 2.0许可证 - ✅ 附带完整的许可证声明 ## 数据集描述 多功能微调数据集,整合了以下三类内容: - **工具使用工作流**:包含文件操作、代码执行、搜索等操作的真实智能体对话 - **数学推理样本**:带有`<thinking>`标记的分步问题求解内容 - **自然语言推理**:对话式思维链(Chain-of-Thought)流程 ## 数据集统计信息 - **训练集**:38369轮对话(总计965819个交互回合) - **验证集**:4264轮对话(总计44644个交互回合) - **总样本**:42633轮对话(总计1010463个交互回合) - **平均回合数**:每轮对话平均包含25.3个交互回合 - **工具使用对话**:约3867轮(占总样本的9.7%) ## 数据格式 每个样本均为多轮对话结构: json { "conversations": [ {"from": "human", "value": "User message"}, {"from": "gpt", "value": "Assistant response"}, {"from": "tool", "value": "Tool execution result"}, {"from": "gpt", "value": "Assistant response to tool result"} ] } ## 应用场景 适用于以下场景的模型微调: - 工具使用与函数调用 - 代码生成与调试 - 多步推理任务 - 智能体工作流构建 - 软件开发相关任务 - 数学问题求解 ## 训练建议 - 支持模型:Claude、GPT-4、LLaMA、Gemma、GLM - 微调方法:QLoRA、全量微调 - 序列长度:建议使用8192+ Token(Token) - 有效批次大小:16-32 ## 引用方式 若使用本数据集,请引用以下原始数据源: bibtex @dataset{opus_reasoning, title = {Opus-4.6-Reasoning-3000x-filtered}, author = {nohurry}, year = {2024}, publisher = {HuggingFace} } @dataset{qwen_reasoning, title = {Qwen3.5-reasoning-700x}, author = {Jackrong}, year = {2024}, publisher = {HuggingFace} } @dataset{agent_hybrid, title = {Agent Dataset Hybrid v2}, author = {austindixson}, year = {2025}, publisher = {HuggingFace} } ## 许可证 版权所有 2025 austindixson 根据Apache许可证2.0版("许可证")获得许可;除非符合许可证要求,否则不得使用本作品。您可以通过以下链接获取许可证副本: http://www.apache.org/licenses/LICENSE-2.0 除非适用法律要求或书面同意,根据许可证分发的软件均按"按现状"提供,不附带任何明示或暗示的担保或条件。有关许可证下特定语言的权限和限制,请参阅许可证。 ## 与原始数据集的差异 本数据集相较于原始数据源作出了以下修改: - 将多个独立数据集整合为统一格式 - 将所有原始数据集转换为标准对话格式 - 从Claude Code对话中补充工具执行结果字段 - 对样本进行筛选与整理以提升整体质量 - 划分为训练集与验证集子集 所有修改均符合Apache 2.0许可证的相关要求。




