56 lines
3.1 KiB
Markdown
56 lines
3.1 KiB
Markdown
---
|
|
license: apache-2.0
|
|
datasets:
|
|
- Agent-Ark/Toucan-1.5M
|
|
language:
|
|
- en
|
|
base_model:
|
|
- Qwen/Qwen2.5-7B-Instruct
|
|
tags:
|
|
- agent
|
|
---
|
|
|
|
# 🦤 Toucan-1.5M:
|
|
|
|
Toucan-1.5M is the largest fully synthetic tool-agent dataset to date, designed to advance tool use in agentic LLMs. It comprises over 1.5 million trajectories synthesized from 495 real-world Model Context Protocols (MCPs) spanning 2,000+ tools. By leveraging authentic MCP environments, Toucan-1.5M generates diverse, realistic, and challenging tasks requires using multiple tools, with trajectories involving real tool executions across multi-round, multi-turn, sequential, and parallel tool calls. Models fine-tuned on Toucan-1.5M outperform much larger closed-source counterparts on the BFCL V3 benchmark and extend the Pareto frontier on the MCP-Universe benchmark.
|
|
|
|
- 📄 [Technical Report](https://arxiv.org/abs/2510.01179) - Discover the methodology and technical details behind Toucan-1.5M
|
|
- 💾 [Github Repo](https://github.com/TheAgentArk/Toucan) - Access the complete pipeline used to produce Toucan-1.5M
|
|
- 🤗 [HF Dataset](https://huggingface.co/datasets/Agent-Ark/Toucan-1.5M) - Full dataset (You are here!)
|
|
- 🤖 Model Checkpoints - [Qwen2.5-7B](https://huggingface.co/Agent-Ark/Toucan-Qwen2.5-7B-Instruct-v0.1) | [Qwen2.5-14B](https://huggingface.co/Agent-Ark/Toucan-Qwen2.5-7B-Instruct-v0.1) | [Qwen2.5-32B](https://huggingface.co/Agent-Ark/Toucan-Qwen2.5-32B-Instruct-v0.1)
|
|
|
|
|
|
|
|
## About This Model
|
|
|
|
This model is a fine-tuned variant of **Qwen2.5-7B-Instruct**, trained on a curated subset of the [Toucan-1.5M](https://huggingface.co/datasets/Agent-Ark/Toucan-1.5M) dataset. The supervised fine-tuning (SFT) subset consists of **119.3K instances** in total, including:
|
|
|
|
- **28.3K** from the original pipeline
|
|
- **40K** from Extension 1 (*Irrelevance*)
|
|
- **15.8K** from Extension 2 (*Diversify*)
|
|
- **35.2K** from Extension 3 (*Multi-Turn*)
|
|
|
|
We adopt the `Hermes` prompt template for fine-tuning. For a detailed description of the training setup and hyperparameters, please refer to our [technical report](https://arxiv.org/abs/2510.01179).
|
|
|
|
## Model Performance
|
|
|
|
Toucan-1.5M remarkably improves baseline model performance through SFT and enables smaller models to outperform larger models across different evaluation aspects, as evidenced in BFCL-V3 and MCP Universe benchmarks.
|
|
|
|
|
|
|
|
## 📚 Citation
|
|
|
|
If you find the data or code useful, please cite:
|
|
```
|
|
@misc{xu2025toucan,
|
|
title={TOUCAN: Synthesizing 1.5M Tool-Agentic Data from Real-World MCP Environments},
|
|
author={Zhangchen Xu and Adriana Meza Soria and Shawn Tan and Anurag Roy and Ashish Sunil Agrawal and Radha Poovendran and Rameswar Panda},
|
|
year={2025},
|
|
eprint={2510.01179},
|
|
archivePrefix={arXiv},
|
|
primaryClass={cs.LG},
|
|
url={https://arxiv.org/abs/2510.01179},
|
|
}
|
|
```
|
|
|
|
**Contact**: For questions, please contact [Zhangchen](mailto:zxu9@uw.edu) by email. |