About me

I received my PhD from the School of Computer Science, The University of Sydney. I study overfitting in machine learning and work on LLM pre- and post-training under imperfect supervision. I was named an Outstanding Reviewer at ICLR, ICML, NeurIPS, and CVPR.

Recent Work

• Decoupling Token Roles in Autoregressive Pretraining.

  Suqin Yuan, Runqi Lin, Kevin Qinghong Lin, Junchi Yu, Lei Feng, Chris Russell, Tongliang Liu.

  arXiv preprint:2609.33405. [Paper]

• Are Human-Aligned LLMs Models of Humans? A Turing-Test Gap in Preference Alignment.

  Suqin Yuan, Runqi Lin, Muyang Li, Guanzhe Hong, Jindong Gu, Lei Feng, Chris Russell, Tongliang Liu.

  arXiv preprint:2609.23640. [Paper]

• Understanding Diversity Collapse in RLVR via the Lens of Overtraining.

  Suqin Yuan, Jinkun Chen, Jiyang Zheng, Muyang Li, Lei Feng, Dadong Wang, Tao Xiang, Tongliang Liu, Bo An.

  arXiv preprint:2606.15455. [Paper]

Publications

• Mitigating Mismatch within Reference-based Preference Optimization.

  Suqin Yuan, Xingrui Yu, Jiyang Zheng, Lei Feng, Dadong Wang, Ivor Tsang, Tongliang Liu.

  International Conference on Learning Representations (ICLR), 2026. [Paper] [Code]

• Instance-dependent Early Stopping for Adaptive Data Pruning.

  Suqin Yuan, Runqi Lin, Felix Azian, Lei Feng, Bo Han, Gang Niu, Masashi Sugiyama, Tongliang Liu.

  IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). [Paper]

  International Conference on Learning Representations (ICLR), 2025. (Spotlight) [Paper] [Code]

• Early Stopping Without Validation Data in Weakly Supervised Learning.

  Suqin Yuan, Muyang Li, Lei Feng, Bo Han, Gang Niu, Masashi Sugiyama, Tongliang Liu.

  IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). [Paper]

  International Conference on Learning Representations (ICLR), 2024. [Paper] [Code]

• Enhancing Sample Selection Against Label Noise by Cutting Mislabeled Easy Examples.

  Suqin Yuan, Lei Feng, Bo Han, Tongliang Liu.

  Advances in Neural Information Processing Systems (NeurIPS), 2025. [Paper] [Code]

• Late Stopping: Avoiding Confidently Learning from Mislabeled Examples.

  Suqin Yuan, Lei Feng, Tongliang Liu.

  International Conference on Computer Vision (ICCV), 2023. [Paper] [Code]

Education and Experience

• The University of Sydney, 2022 – 2026
  Ph.D., Computer Science, Advisor: Prof. Tongliang Liu

• Chongqing University, 2018 – 2022
  B.Eng.,Computer Science, Advisor: Prof. Lei Feng

• University of Oxford, Mar. – Sep. 2026
  Student Visitor, Torr Vision Group

• Nanyang Technological University, Dec. 2025 – Feb. 2026
  Visiting Student, AI-for-X Centre

• Agency for Science, Technology and Research (A*STAR), Aug. – Dec. 2025
  Research Officer, Centre for Frontier AI Research

Honors and Awards

• USYD Faculty of Engineering Career Advancement Award, 2025
• 12th China Youth Science and Technology Innovation Award, 100 recipients biennially, 2020
• 26th China High School Biology Olympiad, First Prize, ranked 12th, 2017