Improve `minecraft-motion-action-dataset` card: Add paper, code, project page, tasks, tags, and usage

README.md

@@ -1,5 +1,16 @@
 ---
 license: mit
+task_categories:
+- robotics
+- image-text-to-text
+language:
+- en
+tags:
+- minecraft
+- agent
+- reinforcement-learning
+- multimodal
+- vision-language-model
 dataset_info:
   features:
   - name: id
@@ -42,4 +53,114 @@ configs:
     path: data/valid-*
 ---
 
-<!-- **minecraft-motion-action-dataset** is part of the OpenHA suite, introduced in our paper [OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft](https://arxiv.org/pdf/2509.13347). -->
+# Minecraft Motion Action Dataset
+
+This `minecraft-motion-action-dataset` is a fundamental component of the [OpenHA suite](https://github.com/CraftJarvis/OpenHA), which introduces a series of open-source hierarchical agentic models designed for the Minecraft environment. This specific dataset focuses on **Motion Action** data.
+
+It is used in the research presented in the paper:
+**[Training One Model to Master Cross-Level Agentic Actions via Reinforcement Learning](https://huggingface.co/papers/2512.09706)**
+
+The paper introduces CrossAgent, a unified agentic model capable of mastering heterogeneous action spaces and autonomously selecting the most effective interface for each step of a trajectory. This dataset supports the training of such agents to learn adaptive action switching—balancing high-level efficiency with low-level precision—without human-specified rules.
+
+-   **Project Page**: [CraftJarvis Homepage](https://craftjarvis.github.io/)
+-   **Code**: [CraftJarvis/OpenHA GitHub Repository](https://github.com/CraftJarvis/OpenHA)
+
+## Sample Usage
+
+To utilize this dataset within the OpenHA framework, you can follow the installation and inference procedures outlined below, as found in the associated GitHub repository.
+
+### Installation
+Clone the OpenHA repository and install dependencies:
+
+```sh
+git clone --recurse-submodules https://github.com/CraftJarvis/OpenHA.git
+conda create -n openha python=3.10
+conda activate openha
+pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url https://download.pytorch.org/whl/cu124  # check your CUDA version
+cd OpenHA
+conda install --channel=conda-forge openjdk=8 -y
+pip install -e .
+```
+
+> ⚠️ Note: The script will install **minestudio** automatically. If you have not used MineStudio before, please check [the tutorial](https://craftjarvis.github.io/MineStudio/overview/getting-started.html).
+
+### Inference
+OpenHA supports multiple ways to serve and load models. We recommend **vLLM** for efficient multi-GPU / multi-process rollout. Example:
+
+```sh
+CUDA_VISIBLE_DEVICES=0,1,2,3 vllm serve CraftJarvis/minecraft-openha-qwen2vl-7b-2509  \
+    --served-model-name minecraft-openha-qwen2vl-7b-2509  \
+    --port 11000 \
+    --limit-mm-per-prompt image=25  \
+    --trust-remote-code --gpu-memory-utilization 0.90  \
+    --pipeline-parallel-size 1  \
+    --tensor-parallel-size 4  \
+    --max-num-seqs 16 \
+    --max-logprobs 20 \
+    --max-model-len 32768
+```
+
+Once the model is loaded, run rollout:
+
+```sh
+python examples/rollout_openha.py --output_mode text_action  \
+    --vlm_client_mode online \
+    --system_message_tag text_action \
+    --model_ips localhost --model_ports 11000 \
+    --model_id minecraft-openha-qwen2vl-7b-2509 \
+    --record_path "/DATA/limuyao/evaluate" \
+    --max_steps_num 200 \
+    --num_rollouts 8
+```
+
+OpenHA also supports HuggingFace Transformers (`hf`) or offline `vllm` loading. Just change the `--vlm_client_mode` argument accordingly.
+
+## Interaction Details
+
+You can control the **output format** of OpenHA via `system_message_tag` in `rollout_openha.py`.
+
+| Parameter          | Output Example                                                                                                       | System Prompt                                                                 |
+|--------------------|-----------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------|
+| `text_action`      | `Action: move(dx='4.0', dy='-1.0') and keyDown(keys=(keyboard.left.control, keyboard.w))`                            | [text_action.txt](https://github.com/CraftJarvis/OpenHA/blob/master/openagents/assets/system_prompt/text_action.txt)                      |
+| `grounding_action` | `Grounding: move_camera <\|object_ref_start\|>empty slot<\|object_ref_end\|><\|point_start\|>(881,558)<\|point_end\|>` | [grounding.txt](https://github.com/CraftJarvis/OpenHA/blob/master/openagents/assets/system_prompt/grounding.txt)                          |
+| `motion_action`    | `Motion: cursor move left and down`                                                                                  | [motion.txt](https://github.com/CraftJarvis/OpenHA/blob/master/openagents/assets/system_prompt/motion.txt)                                |
+| `grounding_coa`    | `Grounding: ... (615,505)... 
+, Action: move(19, 0) and press()`                                                    | [grounding_coa.txt](https://github.com/CraftJarvis/OpenHA/blob/master/openagents/assets/system_prompt/grounding_coa.txt)                  |
+| `motion_coa`       | `Motion: cursor move right and up 
+, Action: move(17, 0) and press()`                                               | [motion_coa.txt](https://github.com/CraftJarvis/OpenHA/blob/master/openagents/assets/system_prompt/motion_coa.txt)                        |
+
+Corresponding `output_mode` values:
+
+```python
+MODE_SYSTEM_PROMPT_MAP = {
+    "greedy": {"motion_coa", "grounding_coa"},
+    "text_action": {"text_action"},
+    "grounding": {"grounding_action"},
+    "motion": {"motion_action"},
+}
+```
+
+## Datasets on 🤗 Hugging Face
+
+This dataset is one of several action space datasets released for the OpenHA project:
+
+| Action Space     | Size        | HuggingFace URL                                                                 |
+|------------------|-------------|---------------------------------------------------------------------------------|
+| Motion Action    | 1B Tokens   | https://huggingface.co/datasets/CraftJarvis/minecraft-motion-action-dataset                 |
+| Grounding Action | 0.5B Tokens   | https://huggingface.co/datasets/CraftJarvis/minecraft-grounding-action-dataset              |
+| Text Action      | 2B Tokens   | https://huggingface.co/datasets/CraftJarvis/minecraft-text-action-dataset                   |
+| Motion CoA       | 0.5B Tokens | https://huggingface.co/datasets/CraftJarvis/minecraft-motion-coa-dataset                    |
+| Grounding CoA    | 0.2B Tokens | https://huggingface.co/datasets/CraftJarvis/minecraft-grounding-coa-dataset                 |
+
+## Citation
+If you find **OpenHA** useful, please give us a ⭐ on GitHub or cite us:
+
+```bibtex
+@article{wang2025openha,
+      title={OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft},
+      author={Zihao Wang and Muyao Li and Kaichen He and Xiangyu Wang and Zhancun Mu and Anji Liu and Yitao Liang},
+      journal = {arXiv preprint arXiv:2509.13347},
+      year={2025},
+      url={https://arxiv.org/abs/2509.13347},
+}
+```