---
library_name: transformers
pipeline_tag: image-text-to-text
---

This repository contains the RuleMaze model checkpoint, a LoRA-fine-tuned version of Qwen2.5-VL-3B for rule-compliant visual spatial planning in multimodal large language models. The model is introduced in the paper [RuleMaze: Rule-Compliant Visual Spatial Planning for Multimodal Large Language Models](https://huggingface.co/papers/2608.20237).

For more details about the benchmark, dataset, and training pipeline, please refer to the [project page](https://fish-03.github.io/RULEMAZE/) and the [GitHub repository](https://github.com/oceanflowlab/RuleMaze).