News

• Jun. 2026 Paper accepted at ICML 2026 on faithful LVLM decoding through causal shortcut dismantling.
• Jan. 2026 Paper accepted at AAAI 2026 on object hallucination mitigation in LVLMs.
• Sep. 2025 Started a CSC-funded one-year joint PhD program at the University of Auckland, New Zealand.
• 2025 Papers accepted at ACM MM 2025, ICASSP 2025, and ICME 2025.

About

I am a PhD candidate in Software Engineering at UESTC, advised by Prof. Ping Kuang and Prof. Fan Zhou, and a visiting PhD student in Computer Science at the University of Auckland, New Zealand, working with Prof. Gillian Dobbie. My research focuses on responsible and trustworthy AI, especially bias mitigation and hallucination reduction in large language models, vision-language models, and text-to-image generation systems.

Publications

• Dismantling Pathological Shortcuts: A Causal Framework for Faithful LVLM Decoding ICML 2026
  Liu Yu, Can Chen, Ping Kuang, Zhikun Feng, Fan Zhou, Gillian Dobbie · First Author
• Causally-Grounded Dual-Path Attention Intervention for Object Hallucination Mitigation in LVLMs AAAI 2026
  Liu Yu, Zhonghao Chen, Ping Kuang, Zhikun Feng, Fan Zhou, Lan Wang, Gillian Dobbie · First Author
• Enhancing Decision Boundaries in Continual Learning through a Decoupled Gaussian Framework IP&M 2026
  Zhikun Feng, Cong Chen, Zhen Yang, Liu Yu, Ping Kuang, Mingli Zhou, Kui Dang, Zhiqiang Gao, Yimin Ju · Co-author
• Bimodal Debiasing for Text-to-Image Diffusion: Adaptive Guidance in Textual and Visual Spaces ACM MM 2025
  Liu Yu, Jun Sun, Ping Kuang, Rui Zhou, Fan Zhou, Zhikun Feng · First Author
• Bridging the Fairness Gap: Enhancing Pre-trained Models with LLM-Generated Sentences ICASSP 2025
  Liu Yu, Ludie Guo, Ping Kuang, Fan Zhou · First Author
• Amplifying Commonsense Knowledge via Bi-directional Relation Integrated Graph-based Contrastive Pre-training from Large Language Models IP&M 2025
  Liu Yu, Fan Tian, Ping Kuang, Fan Zhou · First Author
• Knowledge Graphs Acquisition via Forward-Reverse Relation Enhanced Contrastive Pretraining from Large-scale Models ICME 2025
  Liu Yu, Fan Tian, Ping Kuang, Zhikun Feng, Fan Zhou · First Author
• TIPS: Two-level Prompt Selection for More Stability-Plasticity Balance in Continual Learning Pattern Recognition 2025
  Zhikun Feng, Lei Peng, Dapeng Kang, Mingli Zhou, Ping Kuang, Minyi Wu, Liu Yu, Junxian Su · Co-author
• Causal-Debias: Unifying Debiasing in Pretrained Language Models and Fine-tuning via Causal Invariant Learning ACL 2023
  Fan Zhou, Yuzhuo Mao, Liu Yu, Yuming Yang, Tao Zhong · Corresponding Author
• Mixup-based Unified Framework to Overcome Gender Bias Resurgence SIGIR 2023
  Liu Yu, Yuzhuo Mao, Jundong Wu, Fan Zhou · First Author
• Transformer-Enhanced Hawkes Process with Decoupling Training for Information Cascade Prediction KBS 2022
  Liu Yu, Xovee Xu, Goce Trajcevski, Fan Zhou · First Author
• Decoupling Representation and Regressor for Long-Tailed Information Cascade Prediction SIGIR 2021
  Fan Zhou, Liu Yu, Xovee Xu, Goce Trajcevski · Student-First Author

Education & Visiting Experience

Selected Honors & Awards

• Second Prize, Sichuan Science and Technology Progress Award 2024
• Young Elite Scientists Sponsorship Program for PhD Students, China Association for Science and Technology 2025
• National Scholarship 2021, 2022, 2025
• Outstanding Graduate, Sichuan Province 2020, 2023
• President's Scholarship, China Petroleum Scholarship, Shenzhen Stock Exchange Scholarship 2021-2025
• First-Class Academic Scholarship 2021, 2022, 2025
• Outstanding Graduate Student, Outstanding Teaching Assistant, Academic Rising Star, Academic Young Talent 2021-2025

Academic Exchange

• Academic exchange at ICML, Seoul, Korea 2026
• Student Travel Award and academic exchange at AAAI, Singapore 2026
• Student Travel Award and academic exchange at ACM MM, Dublin, Ireland 2025
• Academic exchange at ICME, Nantes, France 2025

Conference Moments

AAAI Singapore · 2026

MM Dublin, Ireland · 2025

ICME Nantes, France · 2025

Academic Service

• Reviewer, ICML, NeurIPS, AAAI, ACL; On-site Volunteer, AAAI 2026
• Reviewer, IEEE/CVF CVPR, IEEE ICME; On-site Volunteer, ACM MM 2025
• Reviewer, IEEE TKDE, Information Processing & Management 2024
• Student Volunteer, SIGIR Online Conference 2021

Open Collaboration

I welcome research conversations and collaborations on responsible AI, trustworthy multimodal models, social bias and hallucination mitigation, evaluation benchmarks, and related problems. I am also open to exploring adjacent or entirely new directions when there is an interesting idea to try.

• Collaboration I am happy to discuss joint projects across LLM/LVLM faithfulness, text-to-image fairness, causal analysis, benchmark construction, and other emerging AI topics.
• Related Work If you have citation-related needs, feel free to share your paper or draft; I am open to reciprocal, appropriate citations when the research connection is genuine and useful for readers.
• Contact Please reach me at liu.yu [at] std.uestc.edu.cn with a short description of the research idea or paper.