About Me

Tianhao Ma is currently a second-year M.E. student at the University of Tokyo, under the supervision of Prof. Masashi Sugiyama and Prof. Takashi Ishida, and collaborates with Dr. Gang Niu. He received his B.E. degree from Jilin University under the supervision of Prof. Ximing Li.

His primary research interests lie in statistical machine learning, learning from label proportions, and diffusion LLMs. Outside of research, he is an amateur football player, a China National Second-Class Athlete, and a National Level-3 Football Referee.

If you are interested in discussing with me, feel free to drop me an email (matianhao2120 at g.ecc.u-tokyo.ac.jp).

Furthermore, I am currently seeking PhD positions and am open to discussing potential research opportunities.

Ongoing Projects

• Submitted Learning from Label Proportions with High-Order Moment Information (First Author)

I am currently extending my research to diffusion large language model (DLLM) safety, applying the theoretical analysis and algorithm design expertise developed through weakly supervised learning to this rapidly evolving frontier.

• 95% Complete Designing Watermarks for Diffusion Large Language Models (Project Lead)
  Designing robust watermarking methods to overcome the challenges posed by any-order generation and highly parallel decoding.

Publications

During my master’s study, I focus on weakly supervised learning and design algorithms with theoretical guarantees, with an emphasis on principled consistency and risk analysis.

• ICLR 2026 Learning from Label Proportions via Proportional Value Classification
  Tianhao Ma, Wei Wang, Ximing Li, Gang Niu, Masashi Sugiyama
  International Conference on Learning Representations (ICLR), 2026
  PDF

• ICLR 2026 Rethinking Consistent Multi-Label Classification under Inexact Supervision
  Wei Wang*, Tianhao Ma*, Ming-Kun Xie, Gang Niu, Masashi Sugiyama
  International Conference on Learning Representations (ICLR), 2026
  PDF

My earlier undergraduate research explored heuristic strategies to improve empirical performance of learning from label proportions, focusing on more effective optimization objectives and practical gains.

• CVPR 2025 Forming Auxiliary High-confident Instance-level Loss to Promote Learning from Label Proportions
  Tianhao Ma*, Han Chen*, Juncheng Hu, Yungang Zhu, Ximing Li
  IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025
  PDF

• CVPR 2026 Findings Learning from Label Proportion with Dual-proportion Constraints
  Tianhao Ma, Ximing Li, Changchun Li, Renchu Guan
  IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026
  PDF

*Equal contribution (co-first authors)

Honors and Awards

I did not perform particularly well academically in the past and received few academic awards.

• National Second Prize Scholarship, 2023
• Yunnan Provincial Campus Football Tournament — 1st Place, 2019
• Jilin Provincial Collegiate Football Championship Super Division — 5th Place, 2021

Educations

• 2025.04 - 2027.03 (expected), Master, University of Tokyo, Tokyo.
• 2020.09 - 2024.07, Bachelor, Jilin University, Jilin.

Invited Talks

• Top Conference Session on Learning Theory and Its Applications, Forum on Information Technology (FIT 2026), Kitakyushu, Japan, Sep. 2026.

Academic Services

• Conference Reviewer: ICLR 2027
• Conference Reviewer: NeurIPS 2026