leosrepo/MotionHalluc

📌 Overview

MotionHalluc studies kinematic hallucinations in fine-grained motion reasoning by comparing query and reference videos. The dataset includes benchmark questions for directional, attributional, and temporal hallucinations, together with the original video-pair annotations, corrective instructions, and frame annotations.

[arXiv]

---

📁 Dataset Structure

Contains all annotation files, including:

• Three MotionHalluc tasks
• Original curated annotations used to construct the benchmark

Each file is in JSON format.

---

🎥 Video Data

We distribute annotations and benchmark only.

Since we do not own the original videos, users are required to download them separately.

• Fit3D dataset: https://fit3d.imar.ro/

To respect data copyright and protect privacy, we do not redistribute the motion data in this repository. Users can obtain the source data from Fit3D or reconstruct motion from videos using 4D-Humans.

---

💻 Preprocessing & Evaluation Code

We provide video processing and evaluation scripts in the official code repository:

👉 GitHub repository: https://github.com/Leocoding0971/MotionHalluc

---

📜 License

This dataset is released for non-commercial scientific research purposes only.

• The annotation data and benchmark design are released under the CC BY 4.0 License.
• Motion representations are derived from a pretrained human motion reconstruction model.
• Video data is not redistributed due to licensing restrictions and must be obtained from the original source.

Users must comply with the license terms of all underlying datasets used in this benchmark, including the Fit3D dataset.

---

📌 Citation

If you find this code useful for your research, please consider citing the following paper:

@misc{guo2026motionhalluc,
  title={{MotionHalluc}: Diagnosing Kinematic Hallucinations in Fine-Grained Motion Reasoning},
  author={Weile Guo and Shenghong He and Danying Mo and Chengdong Xu and Xuexun Liu and Chao Yu},
  year={2026},
  eprint={2606.23061},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2606.23061}
}