About me

I am currently a postdoctoral researcher at the National University of Singapore, working with Prof. Wee Sun Lee and Prof. Yee Whye Teh on AI reliability. Previously, I received my Ph.D. from Huazhong University of Science and Technology (HUST) in 2024 (advisor: Prof. Ruixuan Li) and my B.S. from HUST in 2019.

Contact: weiliumg@gmail.com

I’m currently looking for opportunities at industry AI labs.

Research

My research focuses on making AI systems reliable throughout the lifecycle of their knowledge. In particular, I work on three closely related directions: diagnosing model failures through interpretability and causal analysis; correcting outdated or harmful knowledge through model editing; and preserving important capabilities during continual updates through efficient replay and model merging. More broadly, my goal is to develop AI systems that are debuggable, patchable, and upgradable.

Preprint

• Are We Evaluating the Edit Locality of LLM Model Editing Properly? paper code
  Wei Liu, Haomei Xu, Hongkai Liu, Zhiying Deng, Ruixuan Li, Heng Huang, Yee Whye Teh, Wee Sun Lee
  Preprint

Publications

* denotes correspondence.

• Editing Large Language Models with Geometry-Aware Regularization
  Bingqing Liu, Wei Liu*, Jun Wang, Xiaobo sun, Yuhua Li, Ruixuan Li
  In NeurIPS 2026.

• Rethinking Sequential Locate-Then-Edit: Optimality and Stability
  Bingqing Liu, Zhiying Deng, Wei Liu*, Jun Wang, Yuhua Li, Ruixuan Li
  In NeurIPS 2026

• NORMA: Norm-Guided Explanation Subgraph Discovery
  Xiangyu Fu, Wei Liu*, Jun Wang, Yang Qiu, Yuhua Li, Ruixuan Li
  In NeurIPS 2026

• From Backward Spreading to Forward Replay: Revisiting Target Construction in LLM Parameter Editing. paper code
  Wei Liu, Hongkai Liu, Zhiying Deng, Yee Whye Teh, Wee Sun Lee
  In ICML 2026

• Escaping Graphical Causal Priors: Multi-Generator Cooperation for Rationalization
  Wei Liu, Zhiying Deng, Lang Gao, Yuankai Zhang, Haozhao Wang, Jun Wang, Rui Zhang, Heng Huang, Ruixuan Li
  In Artificial Intelligence (AIJ)

• The Cylindrical Representation Hypothesis for Language Model Steering.
  Lang Gao, Jinghui Zhang, Wei Liu, Fengxian Ji, Chenxi Wang, Zirui Song, Akash Ghosh, Youssef Mohamed, Preslav Nakov, Xiuying Chen
  In ICML 2026

• Confounder Balance in Next Basket Prediction.
  Zhiying Deng, Jianjun Li; Wei Liu*; Juan Zhao
  In TCYB 2026.

• When Personalization Tricks Detectors: The Feature-Inversion Trap in Machine-Generated Text Detection
  Lang Gao, Xuhui Li, Chenxi Wang, Mingzhe Li, Wei Liu, Zirui Song, Jinghui Zhang, Rui Yan, Preslav Nakov, Xiuying Chen
  In ACL 2026.

• Time-Interval-Aware Disentangled Expert Modeling for Next-Basket Recommendation.
  Zhiying Deng, Yuan Fu, Usman Farooq, Ziwei Tian, Wei Liu*, Jianjun Li
  In SIGIR 2026.

• Behavior-Aware Global-Enhanced Neural Modeling for Sequential Set Recommendation.
  Zhiying Deng, Wei Liu*, Jianjun Li, Zhiqiang Guo, Qian Chen, Juan Zhao
  In TAI 2026.

• Unbiased Interest Modeling in Sequential Basket Analysis: Addressing Repetition Bias with Multi-Factor Estimation.
  Zhiying Deng, Jianjun Li, Wei Liu*, Juan Zhao
  In TORS 2025.

• Re-fed+: A better replay strategy for federated incremental learning.
  Yichen Li, Haozhao Wang, Yining Qi, Wei Liu, Ruixuan Li
  In TPAMI 2025.

• Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets. paper code
  Wei Liu, Zhongyu Niu, Lang Gao, Zhiying Deng, Jun Wang, Haozhao Wang, Ruixuan Li
  Conference: ICML 2025.

• Exploring Practical Gaps in Using Cross Entropy to Implement Maximum Mutual Information Criterion for Rationalization. paper code
  Wei Liu, Zhiying Deng, Zhongyu Niu, Jun Wang, Haozhao Wang, Ruixuan Li
  In TACL 2025.

• Breaking Free from MMI: A New Frontier in Rationalization by Probing Input Utilization. paper code
  Wei Liu, Zhiying Deng, Zhongyu Niu, Jun Wang, Haozhao Wang, Zhigang Zeng, Ruixuan Li
  Conference: ICLR 2025.

• Is the MMI Criterion Necessary for Interpretability? Degenerating Non-causal Features to Plain Noise for Self-Rationalization. paper code
  Wei Liu, Zhiying Deng, Zhongyu Niu, Jun Wang, Haozhao Wang, YuanKai Zhang, Ruixuan Li
  Conference: NeurIPS 2024.

• Decoupling Representation and Knowledge for Few-Shot Intent Classification and Slot Filling.
  Jie Han, Yixiong Zou, Haozhao Wang, Jun Wang, Wei Liu, Yao Wu, Tao Zhang, Ruixuan Li.
  Conference: AAAI 2024.

• Enhancing the Rationale-Input Alignment for Self-explaining Rationalization. paper code
  Wei Liu, Haozhao Wang, Jun Wang, Zhiying Deng, YuanKai Zhang, Cheng Wang, Ruixuan Li
  Conference: ICDE 2024.

• D-Separation for Causal Self-Explanation. (NeurIPS 2023) paper code
  Wei Liu, Jun Wang, Haozhao Wang, Ruixuan Li, Zhiying Deng, Yuankai Zhang, Yang Qiu

• Decoupled Rationalization with Asymmetric Learning Rates: A Flexible Lipschitz Restraint. (KDD 2023) paper code
  Wei Liu, Jun Wang, Haozhao Wang, Ruixuan Li, Yang Qiu, YuanKai Zhang, Jie Han, Yixiong Zou

• MGR: Multi-generator Based Rationalization. (ACL 2023, Oral, Top 4%) paper code
  Wei Liu, Haozhao Wang, Jun Wang, Ruixuan Li, Xinyang Li, Yuankai Zhang, Yang Qiu

• Multi-view Multi-aspect Neural Networks for Next-Basket Recommendation. (SIGIR 2023)
  Zhiying Deng, Jianjun Li, Zhiqiang Guo, Wei Liu, Li Zou, Guohui Li

• FR: Folded Rationalization with a Unified Encoder (NeurIPS 2022) paper code video
  Wei Liu, Haozhao Wang, Jun Wang, Ruixuan Li, Chao Yue, Yuankai Zhang

Service

Program co-chair of AAAI AIR-FM workshop, 2026.
Great reviewer of ACL ARR December.