My name is Yue Li (李越). I am currently a master’s student at the School of Computer Science and Technology, East China Normal University, under the supervision of Professor Linlin Wang. My primary collaborator, Dr. Xin Yi, provides me with invaluable guidance and support throughout my studies. Prior to this, I earned my BEng degree from Xiangtan University, where I was mentored by Associate Professor Xuan Lin.

In 2026, I started my industry internships. I first joined the Shanghai Artificial Intelligence Laboratory on the Xuhui West Bund in Shanghai, where I spent a rewarding few months with my supportive mentor and colleagues. I then moved to Ant Group in Hangzhou, where I investigated the inherent safety of large language models, particularly in reinforcement learning (RL) and on-policy distillation (OPD) for agents.

My research interests mainly lie in Trustworthy AI and Model Post-Training (current focus). I have published 5+ papers at the top international AI conferences and journals such as ACL, KDD, KBS and ESWA.

🔥 News

• 2026.06: 💼 I joined Ant Group as a research intern in Hangzhou.
• 2026.05: 🎉 My first-authored paper has been accepted to KDD 2026!
• 2026.04: 💼 I joined Shanghai AI Lab (Pjlab) as a research intern in Shanghai.
• 2025.05: 🎉 My first-authored paper has been accepted to ACL 2025!

📝 Publications

🎯 Safety and Trustworthy AI

Jailbreak Attacks and Defenses

ACL 2025 Findings

Hierarchical Safety Realignment: Lightweight Restoration of Safety in Pruned Large Vision-Language Models \ Yue Li^*, Xin Yi^*, Dongsheng Shi, Gerard de Melo, Xiaoling Wang and Linlin Wang^†.

Arxiv | Project | ACL Anthology

• The current pruning methods will lead to a significant degradation of the model’s safety at a higher sparsity.
• The HSR (Hierarchical Safety Realignment) method we proposed can achieve safety realignment for the pruned model by restoring only a very small number of neurons. HSR is effective for both LLM and LVLM.

• ESWA 2026 Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks, Xin Yi, Yue Li, Dongsheng Shi, Linlin Wang^†, Xiaoling Wang and Liang He.

• Preprint Unified defense for large language models against jailbreak and fine-tuning attacks in education, Xin Yi, Yue Li, Dongsheng Shi, Linlin Wang^†, Xiaoling Wang and Liang He.

Intellectual Property Security

KDD 2026

AGMark: Attention-Guided Dynamic Watermarking for Large Vision-Language Models \ Yue Li^*, Xin Yi^*, Dongsheng Shi, Yongyi Cui, Gerard de Melo and Linlin Wang^†.

Arxiv | Project

• We propose AGmark, a watermarking method for LVLMs that follows the red–green token partitioning paradigm.
• At each generation step, AGmark identifies candidate token weights and determines the protected token set size, effectively mitigating the trade-off between text quality and watermark detectability.

• Preprint From Construction to Injection: Edit-Based Fingerprints for Large Language Models, Yue Li^*, Xin Yi^*, Dongsheng Shi, Yongyi Cui, Gerard de Melo and Linlin Wang^†.

• KBS 2025 Unified Attacks to Large Language Model Watermarks: Spoofing and Scrubbing in Unauthorized Knowledge Distillation, Xin Yi, Yue Li, Shunfan Zheng, Linlin Wang^†, Xiaoling Wang and Liang He.

⚙️ Model Post-Training

• Preprint Reinforcement Learning for Large Language Model Selective Evidence Adoption from Contaminated Retrieval Results, Yanyu Chen^*, Yue Li^*, Yongyi Cui, Dongsheng Shi, Lichang Dai^†.

📦 Others

Medical Agent Systems

• Preprint EMR: Self-Evolving Medical Multi-Agent System via Experience Mining and Reuse, Dongsheng Shi, Xin Yi, Yue Li, Linlin Wang^†.

• Preprint SURGENT: A Surgical Multi-Agent Assistance System Across the Perioperative Workflow, Dongsheng Shi, Yue Li, Xin Yi, Huawei Feng, Linlin Wang^†.

Benchmarks

• ESWA 2026 Benchmarking Large Language Models for End-to-End Clinical Support in Traditional Chinese Medicine, Dongsheng Shi, Xin Yi, Yue Li, Linlin Wang^†.

• Preprint Evaluating LLMs in Database Scenarios: A Lifecycle Benchmark for Assessing Their Potential in Core Database Tasks, Shunfan Zheng, Dongsheng Shi, Yue Li, Xin Yi, Linlin Wang^†, Gerard de Melo.

💼 Internships

Ant Group, Security and Risk Management | Hangzhou

• Duration: June 2026 – Present
• Mentors: Feng Wen and Qiu Zhi
• Focus: Intrinsic safety of LLMs, with a particular emphasis on agent tool calling, including agentic reinforcement learning and on-policy distillation.

Shanghai AI Laboratory, Center for Safe and Trustworthy AI | Shanghai

• Duration: April 2026 – June 2026
• Mentor: Jie Li
• Focus: LLM/Agent Safety, including participation in the construction of the OpenClaw evaluation benchmark and support for safety testing of the Intern series models.

🎖 Honors and Awards

• 2025.10, East China Normal University Outstanding Academic Scholarship (First Prize)
• 2023.05, The 2023 China College Student Programming Competition (CCPC) National Invitational (Hunan), Silver Medal
• 2022.04, The 46th International Collegiate Programming Contest (ICPC) Asian Regional Competition (Kunming), Bronze Medal

🎓 Education

• 2024.09 - Present, Master of Engineering, East China Normal University, Shanghai.
• 2020.09 - 2024.06, Bachelor of Engineering, Xiangtan University, Xiangtan.

📋 Academic Services

• The ACM Web Conference (The International World Wide Web Conference, WWW)
• Association for the Advancement of Artificial Intelligence (AAAI)
• IEEE/INNS International Joint Conference on Neural Networks (IJCNN)
• Association for Computational Linguistics Rolling Review (ARR)