AgentRVOS

Reasoning Over Object Tracks for Zero-Shot Referring Video Object Segmentation

Arxiv 2026

   

Woojeong Jin*   Jaeho Lee*   Heeseong Shin   Seungho Jang   Junhwan Heo   Seungryong Kim^†

KAIST AI
_{* Equal contribution   ^† Corresponding author}

🗞️ News

Date	Update
2026.03	📄 Paper released on ArXiv
2026.03	🌐 Project page is available!

---

📦 Opensource Progress

Component	Status
Inference Code	🔜 Coming Soon
Evaluation Code	🔜 Coming Soon
Gradio Demo	🔜 Coming Soon

---

✅ Update Checklist

• Project page release
• Paper release
• Code release
• Demo release

---

📊 Main Results

For more results, please see our Project page!

---

📝 Citation

If you find this work useful, please consider citing:

@misc{jin2026agentrvosreasoningobjecttracks,
      title={AgentRVOS: Reasoning over Object Tracks for Zero-Shot Referring Video Object Segmentation}, 
      author={Woojeong Jin and Jaeho Lee and Heeseong Shin and Seungho Jang and Junhwan Heo and Seungryong Kim},
      year={2026},
      eprint={2603.23489},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2603.23489},
}

🙏 Acknowledgements

We thank the authors of SAM3, Qwen3-VL and vLLM for their excellent open-source contributions.