Boxuan Zhang

Boxuan Zhang

PhD student @Rutgers CS

🧐 About Me

I am a 1st year CS PhD student at Rutgers University, advised by Prof. Ruixiang (Ryan) Tang. I have completed my master’s degree at Sensing IntelliGence and MAchine learning(SIGMA) Lab in Wuhan University, under the supervision of Prof. Zengmao Wang and Prof. Bo Du.

Previously, I was a research intern at RZ-Lab, Purdue University, advised by Prof. Ruqi Zhang. I was also fortunate to work with Prof. Lu Cheng, Prof. Bo Han and Dr. Jianing Zhu.

🧩 Research Interests:
My current research centers on building trustworthy agentic and generative AI systems, with an emphasis on Reinforcement Learning methods for Agentic LLMs and Diffusion Models, focusing on safety, reliability, and robustness. I am also working on reliable detection for AI-generated visual content (e.g., images, videos), and interdisciplinary applications of machine learning in healthcare (e.g., Immunology, Cell Morphology).

If you share the same research interests with me and are interested in these areas or my previous works, feel free to drop me an Email or add my Wechat . I am always delighted for potential collaborations!

🔥 News

💻 Recent Projects

(Click the images for more details)

    Agent-Foresight    MDMF    Agent-Matrix

    CoT-UQ    CoVer    HCL-Spark


📝 Publications

(* indicates equal contribution)

Selected Publication

ACL 2025
sym

CoT-UQ: Improving Response-wise Uncertainty Quantification in LLMs with Chain-of-Thought

[ACL 2025 Findings]

Boxuan Zhang and Ruqi Zhang

TL;DR: Propose to quantify response-wise uncertainty by integrating LLMs’ inherent reasoning capabilities through Chain-of-Thought (CoT) into the UQ process.

[PDF]   [Code]   [BibTeX]

@article{zhang2025cot,
    title={CoT-UQ: Improving Response-wise Uncertainty Quantification in LLMs with Chain-of-Thought},
    author={Zhang, Boxuan and Zhang, Ruqi},
    journal={arXiv preprint arXiv:2502.17214},
    year={2025}
}
NeurIPS 2024
sym

What If the Input is Expanded in OOD Detection?

[Neural Information Processing Systems (NeurIPS), 2024]

Boxuan Zhang*, Jianing Zhu*, Zengmao Wang, Tongliang Liu, Bo Du, and Bo Han

TL;DR: Propose a novel perspective to employ different common corruptions on the input space to expand the representation dimension for OOD detection.

[PDF]   [Project Page]   [Code]   [BibTeX]

@inproceedings{zhang2024what,
    title={What If the Input is Expanded in OOD Detection?},
    author={Zhang, Boxuan and Zhu, Jianing and Wang, Zengmao and Liu, Tongliang and Du, Bo and Han, Bo},
    booktitle={The Thirty-Eighth Annual Conference on Neural Information Processing Systems},
    year={2024},
}

Working Preprint

Preprint
sym

AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems

[Preprint. Under Review]

Boxuan Zhang*, Jianing Zhu*, Zeru Shi, Dongfang Liu, and Ruixiang Tang

TL;DR: We reframe agentic failure analysis from post-hoc attribution on completed trajectories to online auditing on unfolding prefixes, where an auditor commits a continue-or-alarm verdict at every step.

[PDF]   [Code]   [Project Page]   [BibTeX]

@article{zhang2026agentforesight,
  title={AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems},
  author={Zhang, Boxuan and Zhu, Jianing and Shi, Zeru and Liu, Dongfang and Tang, Ruixiang},
  journal={arXiv preprint arXiv:2605.08715},
  year={2026}
}
Preprint
sym

Micro-Defects Expose Macro-Fakes: Detecting AI-Generated Images via Local Distributional Shifts

[Preprint. Under Review]

Boxuan Zhang, Jianing Zhu, Qifan Wang, Jiang Liu, and Ruixiang Tang

TL;DR: We propose Micro-Defects expose Macro-Fakes (MDMF), a local distribution-aware detection framework that amplifies micro-scale statistical irregularities into macro-level distributional discrepancies.

[PDF]   [Code]   [Project Page]   [BibTeX]

@article{zhang2026mdmf,
  title={Micro-Defects Expose Macro-Fakes: Detecting AI-Generated Images via Local Distributional Shifts},
  author={Zhang, Boxuan and Zhu, Jianing and Wang, Qifan and Liu, Jiang and Tang, Ruixiang},
  journal={arXiv preprint arXiv:2605.09296},
  year={2026}
}
Preprint
sym

Dive into the Agent Matrix: A Realistic Evaluation of Self-Replication Risk in LLM Agents

[Preprint. Under Review]

Boxuan Zhang*, Yi Yu*, Jiaxuan Guo, and Jing Shao

TL;DR: We present a comprehensive evaluation framework for quantifying self-replication risks. Our framework establishes authentic production environments and realistic tasks to enable scenario-driven assessment of agent behaviors.

[PDF]   [BLOG]   [BibTeX]

@article{zhang2025dive,
  title={Dive into the Agent Matrix: A Realistic Evaluation of Self-Replication Risk in LLM Agents},
  author={Zhang, Boxuan and Yu, Yi and Guo, Jiaxuan and Shao, Jing},
  journal={arXiv preprint arXiv:2509.25302},
  year={2025}
}

📖 Education

  • 2025.09 - present, PhD, Department of Computer Science, Rutgers University, United States.
  • 2022.09 - 2024.06, Master, School of Computer Science, Wuhan University, China.
  • 2018.09 - 2022.06, Undergraduate, School of Computer Science, Wuhan University, China.

👨🏻‍💻 Research Experience

💼 Work Experience

💁🏻‍♂️ Academic Service

  • Journal Reviewer: ISPRS Journal of Photogrammetry and Remote Sensing
  • Conference Reviewer: NeurIPS (2025, 2026) ICLR (2026) ICML (2026)