GraphWalker-7B
📄 Paper (arXiv:2603.28533) | 💻 GitHub | 🤗 Model
GraphWalker-7B is a specialized large language model fine-tuned from Qwen2.5-7B-Instruct for Agentic Knowledge Graph Question Answering (KGQA). GraphWalker learns to navigate knowledge graphs via synthetic trajectory curriculum — achieving strong generalization with a single, compact 7B model.
🌟 Overview
GraphWalker is an agentic framework for multi-turn Knowledge Graph Question Answering (KGQA) over Global Knowledge Graphs (e.g., Freebase). It transforms LLMs into reasoning agents that autonomously navigate massive KGs through a "Think-Query-Observe" loop, optimized via a synthetic curriculum.
🛠️ Usage
1. Environment Setup
pip install vllm transformers
2. Download the Model
# Via huggingface-cli
huggingface-cli download xushuwen23/GraphWalker-7B --local-dir ./GraphWalker-7B
3. Inference with vLLM (Recommended)
Start the vLLM server:
vllm serve "xushuwen23/GraphWalker-7B" \
--host 0.0.0.0 --port 22240 \
--served-model-name graphwalker-7b \
--gpu-memory-utilization 0.9 \
--dtype auto \
--chat-template "./GraphWalker-7B/chat_template.jinja"
For training and evaluation, see 💻 GitHub for details.
📈 Evaluation Results
| Method | Backbone | CWQ EM | CWQ F1 | WebQSP EM | WebQSP F1 |
|---|---|---|---|---|---|
| GraphWalker | |||||
| †Vanilla Agent | Qwen2.5-7B-Instruct | 40.7 | 33.2 | 68.4 | 66.1 |
| †Vanilla Agent | GPT-4o-mini | 63.4 | 60.3 | 79.6 | 70.6 |
| †Vanilla Agent | DeepSeek-V3.2 | 69.8 | 63.5 | 76.7 | 71.8 |
| GraphWalker-7B-SFT | Qwen2.5-7B-Instruct | 68.3 | 63.2 | 82.0 | 79.1 |
| GraphWalker-3B-SFT-RL | Qwen2.5-3B-Instruct | 70.9 | 65.2 | 83.5 | 81.7 |
| GraphWalker-8B-SFT-RL | LLaMA3.1-8B-Instruct | 78.5 | 69.6 | 88.2 | 84.5 |
| GraphWalker-7B-SFT-RL | Qwen2.5-7B-Instruct | 79.6 | 74.2 | 91.5 | 88.6 |
📝 Citation
If you use GraphWalker-7B or find this work helpful, please cite:
@misc{xu2026graphwalkeragenticknowledgegraph,
title={GraphWalker: Agentic Knowledge Graph Question Answering via Synthetic Trajectory Curriculum},
author={Shuwen Xu and Yao Xu and Jiaxiang Liu and Chenhao Yuan and Wenshuo Peng and Jun Zhao and Kang Liu},
year={2026},
eprint={2603.28533},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2603.28533},
}
📄 License
This model is released under the Apache 2.0 License, consistent with the base model Qwen2.5-7B-Instruct.