Baixuan Xu Tony 徐柏轩

Third-year Ph.D. student at HKUST

prof_pic.jpg

Third-year Ph.D. student at HKUST

I'm a Tennis Fan

Bibliography

Hi, welcome to my personal webpage! I’m Baixuan, a third-year Ph.D. student from Computer Science and Engineering Department at the Hong Kong University of Science and Techonology supervised by Professor Yangqiu Song. I received my bachelor degree in Computer Science and Mathematics from the same institution in 2024, and I am honored to be funded by the prestigious HKPFS award. I was an intern at Xiaomi MiMo building frontier foundation models (2026-02 - 2026-09). I mainly focus on:

Pre-train data engineering and curation, particularly code-related data and web data preparation.

SFT/RL data construction via scaling agentic environments for post-training.

I am a core contributor to the MiMo-V2.6 series.

This page is updated on 2026/09/22

Research Interests

My research interests lie in the following areas:

User-intent understanding, particularly for e-commerce applications. [MIND]

Pre-training data recipes and evaluation, including data mixture design, quality control, and capability-oriented assessment. [Work conducted during my internship at Xiaomi MiMo]

Agentic RL, with a focus on synthesizing scalable environments and constructing high-quality trajectories for mid-training and SFT. [Work conducted during my internship at Xiaomi MiMo]

I also have done some research on understanding long-horizon agents. [Cognitive Bandwidth Bottleneck]

If you are interested in these areas, please feel free to contact me at bxuan at connect dot ust dot hk.

Education

The Hong Kong University of Science and Technology (HKUST)

  • Doctor of Philosophy in Computer Science and Engineering supervised by Professor Yangqiu Song, Sep. 2024 – Jul. 2028 (Expected)
  • Bachelor of Science in Computer Science and Mathematics, Sep. 2020 – Jun. 2024, First Class Honors, Academic Achievement Medal (~top 1% graduate)

Academic Service

  • Conference Reviewer: ACL Rolling Review

Awards

  • Hong Kong PhD Fellowship (2024 - 2028)
  • Dean’s List for the School of Engineering (8 Semesters)
  • University’s Scholarship Scheme for Continuing Undergraduate Students (2021 - 2023)

news

Sep 22, 2026 I’m thrilled to announce the release of our latest frontier foundation model, MiMo-V2.6-Pro, and proud to have been a core contributor to the MiMo-V2.6 series!
Aug 25, 2026 I am delighted to share that two of our papers have been accepted to EMNLP 2026. Many thanks to my amazing collaborators! The code and public version would be ready soon.
May 16, 2025 Our paper on multi-agent reasoning system design arxiv has been arxived! We are open for discussion through email!:smile:
Sep 20, 2024 Our new multimodal purchase intention understanding framework MIND has been accepted to EMNLP2024 main conference! :sparkles: :smile:

Selected Publications

  1. Tech Report
    MiMo-V2.6-Pro Technical Report
    Xiaomi MiMo Team (Core Contributor)
    2026
    PDF Website
  2. EMNLP
    The Cognitive Bandwidth Bottleneck: Shifting Long-Horizon Agents from Planning with Actions to Planning with Schemas
    Baixuan Xu, Tianshi Zheng, Zhaowei Wang, and 4 more authors
    In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, 2026
    arXiv Code
  3. EMNLP
    Finding the Right Evidence: Factor-Guided Coarse-to-Fine Reasoning for Long Videos
    Baixuan Xu, Yinyui Xu, Tianshi Zheng, and 9 more authors
    In Findings of the Association for Computational Linguistics: EMNLP 2026, 2026
    arXiv PDF Code
  4. EMNLP
    MIND: Multimodal Shopping Intention Distillation from Large Vision-language Models for E-commerce Purchase Understanding
    Baixuan Xu, Weiqi Wang, Haochen Shi, and 12 more authors
    In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024, Miami, FL, USA, November 12-16, 2024, 2024