OpenManus-RL

# Dataset Card for OpenManusRL

## Dataset Description

### Overview

<p align="center">
💻 <a href="https://github.com/OpenManus/OpenManus-RL" target="_blank">[Github Repo]</a>
</p>

**OpenManusRL** combines agent trajectories from [AgentInstruct](https://huggingface.co/datasets/THUDM/AgentInstruct), [Agent-FLAN](https://huggingface.co/datasets/internlm/Agent-FLAN) and [AgentTraj-L(AgentGym)](https://huggingface.co/datasets/AgentGym/AgentTraj-L) with features:

- 🔍 **ReAct Framework** - <a href="https://react-lm.github.io/" target="_blank">Reasoning-Acting integration</a>
- 🧠 **Structured Training** - Separate format/reasoning learning
- 🚫 **Anti-Hallucination** - Negative samples + environment grounding
- 🌐 **6 Domains** - OS, DB, Web, KG, Household, E-commerce

### Dataset Overview

| Source | Trajectories | Avg Turns | Key Features |
|--------|--------------|-----------|--------------|
| [Agent-FLAN](https://huggingface.co/datasets/internlm/Agent-FLAN) | 34442 | 3-35 | Error recovery patterns, CoT reasoning|
[AgentTraj-L](https://huggingface.co/datasets/AgentGym/AgentTraj-L) | 14485 | 3-35 | interactive environments and tasks
| **Combined** | 48927 | 4-20 | Enhanced generalization, uniform format with broader task coverage |

### Supported Tasks
- **text-generation**: ReAct-style instruction following
- **conversational-ai**: Tool-augmented dialogues

### Languages
English

## Dataset Structure

### Data Instances

**ReAct Pattern Example**:
```json
{
"id": "os_0",
"conversations": [
{"role": "user", "content": "Count files in /etc"},
{"role": "assistant", "content": "Think: Need reliable counting method\nAct: bash\n```bash\nls -1 /etc | wc -l\n```"},
{"role": "user", "content": "OS Output: 220"},
{"role": "assistant", "content": "Think: Verified through execution\nAct: answer(220)"}
]
}
```

## Licensing
- **License:** [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
- **Sources:**
- [AgentInstruct](https://huggingface.co/datasets/THUDM/AgentInstruct) (CC-BY-NC-4.0)
- [Agent-FLAN](https://huggingface.co/datasets/internlm/Agent-FLAN) (Apache 2.0)

## Citation
```bibtex
@misc{zeng2023agenttuning,
title={AgentTuning: Enabling Generalized Agent Abilities for LLMs},
author={Aohan Zeng and Mingdao Liu and Rui Lu and Bowen Wang and Xiao Liu and Yuxiao Dong and Jie Tang},
year={2023},
eprint={2310.12823},
archivePrefix={arXiv},
primaryClass={cs.CL}
}

@article{chen2024agent,
title={Agent-FLAN: Designing Data and Methods of Effective Agent Tuning for Large Language Models},
author={Chen, Zehui and Liu, Kuikun and Wang, Qiuchen and Zhang, Wenwei and Liu, Jiangning and Lin, Dahua and Chen, Kai and Zhao, Feng},
journal={arXiv preprint arXiv:2403.12881},
year={2024}
}

@misc{xi2024agentgym,
title={AgentGym: Evolving Large Language Model-based Agents across Diverse Environments},
author={Zhiheng Xi and Yiwen Ding and Wenxiang Chen and Boyang Hong and Honglin Guo and Junzhe Wang and Dingwen Yang and Chenyang Liao and Xin Guo and Wei He and Songyang Gao and Lu Chen and Rui Zheng and Yicheng Zou and Tao Gui and Qi Zhang and Xipeng Qiu and Xuanjing Huang and Zuxuan Wu and Yu-Gang Jiang},
year={2024},
eprint={2406.04151},
archivePrefix={arXiv},
primaryClass={cs.AI}
}
```

## Contact
[OpenManus Team](https://github.com/OpenManus/OpenManus-RL)

# OpenManusRL 数据集卡片

## 数据集描述

### 概览
<p align="center">
💻 <a href="https://github.com/OpenManus/OpenManus-RL" target="_blank">[GitHub 仓库]</a>
</p>

**OpenManusRL** 整合了来自 [AgentInstruct](https://huggingface.co/datasets/THUDM/AgentInstruct)、[Agent-FLAN](https://huggingface.co/datasets/internlm/Agent-FLAN) 以及 [AgentTraj-L(AgentGym)](https://huggingface.co/datasets/AgentGym/AgentTraj-L) 的智能体轨迹数据，具备以下特性：
- 🔍 **ReAct 框架（ReAct Framework）** - <a href="https://react-lm.github.io/" target="_blank">推理-行动一体化</a>
- 🧠 **结构化训练** - 分离格式与推理学习流程
- 🚫 **抗幻觉** - 负样本与环境锚定机制
- 🌐 **六大领域** - 操作系统（OS）、数据库（DB）、网络（Web）、知识库（KG）、智能家居（Household）、电子商务（E-commerce）

### 数据集概览
| 数据来源 | 轨迹数量 | 平均交互轮次 | 核心特性 |
|--------|--------------|-----------|--------------|
| [Agent-FLAN](https://huggingface.co/datasets/internlm/Agent-FLAN) | 34442 | 3-35 | 错误恢复模式、思维链（Chain of Thought，CoT）推理 |
| [AgentTraj-L](https://huggingface.co/datasets/AgentGym/AgentTraj-L) | 14485 | 3-35 | 交互环境与任务 |
| **合并后数据集** | 48927 | 4-20 | 增强泛化能力，格式统一且覆盖更广泛的任务场景 |

### 支持任务
- **文本生成**：ReAct 风格指令遵循
- **对话式AI**：工具增强型对话

### 语言
英语

## 数据集结构

### 数据实例
**ReAct 模式示例**：
json
{
"id": "os_0",
"conversations": [
{"role": "user", "content": "统计 /etc 目录下的文件数量"},
{"role": "assistant", "content": "思考：需要可靠的统计方法
行动：bash
bash
ls -1 /etc | wc -l
"},
{"role": "user", "content": "操作系统输出：220"},
{"role": "assistant", "content": "思考：已通过执行结果验证
行动：回答(220)"}
]
}

## 许可证
- **许可证**：[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
- **来源许可证**：
- [AgentInstruct](https://huggingface.co/datasets/THUDM/AgentInstruct) (CC-BY-NC-4.0)
- [Agent-FLAN](https://huggingface.co/datasets/internlm/Agent-FLAN) (Apache 2.0)

## 引用
bibtex
@misc{zeng2023agenttuning,
title={AgentTuning: Enabling Generalized Agent Abilities for LLMs},
author={Aohan Zeng and Mingdao Liu and Rui Lu and Bowen Wang and Xiao Liu and Yuxiao Dong and Jie Tang},
year={2023},
eprint={2310.12823},
archivePrefix={arXiv},
primaryClass={cs.CL}
}

@article{chen2024agent,
title={Agent-FLAN: Designing Data and Methods of Effective Agent Tuning for Large Language Models},
author={Chen, Zehui and Liu, Kuikun and Wang, Qiuchen and Zhang, Wenwei and Liu, Jiangning and Lin, Dahua and Chen, Kai and Zhao, Feng},
journal={arXiv preprint arXiv:2403.12881},
year={2024}
}

@misc{xi2024agentgym,
title={AgentGym: Evolving Large Language Model-based Agents across Diverse Environments},
author={Zhiheng Xi and Yiwen Ding and Wenxiang Chen and Boyang Hong and Honglin Guo and Junzhe Wang and Dingwen Yang and Chenyang Liao and Xin Guo and Wei He and Songyang Gao and Lu Chen and Rui Zheng and Yicheng Zou and Tao Gui and Qi Zhang and Xipeng Qiu and Xuanjing Huang and Zuxuan Wu and Yu-Gang Jiang},
year={2024},
eprint={2406.04151},
archivePrefix={arXiv},
primaryClass={cs.AI}
}

## 联系方式
[OpenManus 团队](https://github.com/OpenManus/OpenManus-RL)