Portrait of Yan Li

Yan Li 李岩

  • Incoming Ph.D. student, HKUST CSE
  • Research Associate, NTU

About

Hello! I am Yan Li (李岩). In February 2027, I will join the Department of Computer Science and Engineering at HKUST as a Ph.D. student in Prof. Jieyu Zhao’s group. I am currently a Research Associate at Nanyang Technological University (NTU), working with Prof. Sean Du.

Previously, I received my M.Sc. in Machine Learning from the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), advised by Prof. Kun Zhang, and worked closely with Guangyi Chen and Yunlong Deng in the MBZUAI Causality Group. From July to December 2025, I was a visiting student in the Department of Philosophy at Carnegie Mellon University (CMU), working with Prof. Peter Spirtes and Prof. Kun Zhang and collaborating with the CMU CLeaR Group. I received my B.Eng. in Computer Science and Technology from Tsinghua University in 2024, where I worked on HCI and context-aware systems with Prof. Chun Yu at the Research Institute of Human–Computer Interaction and Media Integration.

My research focuses on representation learning, particularly for foundation models. My broader goal is to improve the reliability and robustness of large language models, for example by reducing hallucinations and enabling stronger generalization.

I have published extensively on causal representation learning, much of it revisiting conventional adaptation problems from a causal perspective. My collaborators and I have also built datasets and written position papers that aim to clarify important research directions in this area.

Since 2026, my research has increasingly shifted toward improving the capabilities of LLMs through the lens of representation learning. I developed MPM, a lossless approach for extending multimodal large language models with additional modalities. I am also collaborating with Prof. Sean Du on noise detection in LLMs, particularly in the reasoning trajectories produced by agents.

Another major direction of my current work is improving the long-horizon capabilities of agents. Instead of treating context as a single flat sequence of tokens, we organize it into structured blocks. During generation, the model can emit dedicated editing tokens from its vocabulary to perform atomic operations on specific blocks, including deletion, rewriting, and exact restoration by block identifier. See our recent preprint, Editing as Language: Teaching LLM Agents a Grammar for Native Context Editing.

Research interests

  • Representation learning
  • Reliability and robustness of LLMs
  • Adaptation and generalization
  • Multimodal and vision–language models
  • Long-horizon LLM agents

News

  • 2026.11 I will join Prof. Jieyu Zhao’s group at HKUST CSE as a Ph.D. student in February 2027.
  • 2026.10 New preprint: “Editing as Language: Teaching LLM Agents a Grammar for Native Context Editing” (arXiv, project page).
  • 2026.01 “Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens” accepted to ICLR 2026.
  • 2025.12 Presented two posters at NeurIPS 2025 in San Diego.
  • 2025.09 “CausalVerse: Benchmarking Causal Representation Learning with Configurable High-Fidelity Simulations” accepted to NeurIPS 2025 as a spotlight.
  • 2025.09 “Towards Self-Refinement of Vision-Language Models with Triangular Consistency” accepted to NeurIPS 2025.
  • 2025.07 Visiting student at Carnegie Mellon University, Department of Philosophy (July–December 2025).
  • 2025.04 “A General Representation-Based Approach to Multi-Source Domain Adaptation” accepted to ICML 2025.
Earlier news
  • 2024.10 Started the M.Sc. in Machine Learning at MBZUAI, advised by Prof. Kun Zhang.
  • 2024.06 Graduated from Tsinghua University with a B.Eng. in Computer Science and Technology.
  • 2023.06 Research intern at Mian Bi Intelligent Technology Co., Ltd., working on pre-training data preparation for large language models.
  • 2022 Third Prize, 40th Challenge Cup of Tsinghua University.
  • 2020 Second-Class Scholarship for Freshmen, Tsinghua University.

Publications

* denotes equal contribution.

  1. Editing as Language: Teaching LLM Agents a Grammar for Native Context Editing
    Yunlong Deng*, Yan Li*, Xing Hu, Yida Zhu, Sean Du, Weiyang Liu, Guangyi Chen
    Preprint, 2026 arXivProject pageCode
  2. A Dialogue between Causal and Traditional Representation Learning: Toward Mutual Benefits in a Unified Formulation
    Yan Li, Yuewen Sun, Shaoan Xie, Gongxu Luo, Yunlong Deng, Kun Zhang, Guangyi Chen
    Preprint, 2026 arXiv
  3. Multimodal LLMs under Pairwise Modalities
    Yan Li*, Yunlong Deng*, Yuewen Sun, Gongxu Luo, Kun Zhang, Guangyi Chen
    Preprint, 2026 arXiv
  4. Should Bias be Eliminated? A General Framework to Use Bias for OOD Generalization
    Yan Li, Yunlong Deng, Zijian Li, Anpeng Wu, Zeyu Tang, Kun Zhang, Guangyi Chen
    Preprint, 2025 arXiv
  5. Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens
    Yunlong Deng, Boyang Sun, Yan Li, Lingjing Kong, Zeyu Tang, Kun Zhang, Guangyi Chen
    ICLR 2026 arXiv
  6. Towards Self-Refinement of Vision-Language Models with Triangular Consistency
    Yunlong Deng, Guangyi Chen, Tianpei Gu, Lingjing Kong, Yan Li, Zeyu Tang, Kun Zhang
    NeurIPS 2025 arXiv
  7. A General Representation-Based Approach to Multi-Source Domain Adaptation
    Ignavier Ng*, Yan Li*, Zijian Li, Yujia Zheng, Guangyi Chen, Kun Zhang
    ICML 2025 PMLR
  8. ProtoGS: Efficient and High-Quality Rendering with 3D Gaussian Prototypes
    Zhengqing Gao, Dongting Hu, Jia-Wang Bian, Huan Fu, Yan Li, Tongliang Liu, Mingming Gong, Kun Zhang
    Preprint, 2025 arXiv
  9. Revealing Personalized Positions in Indoor Spaces: A User-Centric Approach to Context-Aware Area Discovery
    Zhaoheng Li, Chun Yu, Yan Li, Yuanchun Shi
    Undergraduate work on human–computer interaction and context-aware area discovery.

Industry experience

Mian Bi Intelligent Technology Co., Ltd. · Jun. 2023 – Jul. 2023
Research intern, supervised by Jie Cai (Algorithm Engineer)

  • Worked on data preparation for pre-training large language models, including data-cleaning algorithms and large-scale batch cleaning.
  • Participated in discussions on classical text LLMs, data pipelines, and preliminary design.

Education

  • Ph.D. in Computer Science and Engineering (incoming), The Hong Kong University of Science and Technology · from Feb. 2027
    Advisor: Prof. Jieyu Zhao
  • M.Sc. in Machine Learning, Mohamed bin Zayed University of Artificial Intelligence · Oct. 2024 – Jun. 2026
    Advisor: Prof. Kun Zhang
  • Visiting Student, Department of Philosophy, Carnegie Mellon University · Jul. 2025 – Dec. 2025
    Hosted by Prof. Peter Spirtes and Prof. Kun Zhang
  • B.Eng. in Computer Science and Technology, Tsinghua University · Sept. 2020 – Jun. 2024

Honors and awards

  • MBZUAI Conference Travel Scholarship, 2025
  • Tsinghua University Academic Progress Scholarship, 2023
  • Social Practice Excellence Scholarship of Tsinghua University, 2022
  • Third Prize, 40th Challenge Cup of Tsinghua University, 2022
  • Second-Class Scholarship for Freshmen, Tsinghua University, 2020

Service

  • Reviewer: ICLR, ICML, NeurIPS

Contact

  • Email: lyan012010 (at) gmail (dot) com