About Me
My name is Zheng Wu, and I am currently a Master’s student at the School of Computer Science, Shanghai Jiao Tong University, under the supervision of Prof. Zhuosheng Zhang. I received my Bachelor’s degree from the same institution in 2025.
My research interests include (multimodal) large language models, agent post-training and agent evaluation.
I am currently a Qingyun Plan Intern with the WeLM Foundation Model Team at Tencent WXG.
如果您对我的研究方向感兴趣，欢迎与我联系、探讨合作可能性。尤其欢迎有意向选择Zhuosheng Zhang老师作为导师的本科生或上海交通大学的校内本科生进行科研实习与合作。
Education
Shanghai Jiao Tong University Supervisor: Prof. Zhuosheng Zhang
Shanghai Jiao Tong University Supervisor: Prof. Zhuosheng Zhang
Internship
Tencent WXG 
Meituan 
OPPO
Awards & Honors
National Scholarship, 2026 (top 0.2% nationwide)
Best Bachelor Thesis, 2025 (top 1% in SJTU)
National Scholarship, 2024 (top 0.2% nationwide)
National Scholarship, 2022 (top 0.2% nationwide)
Publications
* Equal contribution · # Corresponding author.
1. GEM: Gaussian Embedding Modeling for Out-of-Distribution Detection for GUI Agents
2. Agent-Dice: Disentangling Knowledge Updates via Geometric Consensus for Agent Continual Learning
3. Mobile-Aptus: Confidence-Driven Proactive and Robust Interaction in MLLM-based Mobile-Using Agents
4. OS-Kairos: Adaptive Interaction for MLLM-Powered GUI Agents
5. Would You Walk to the Car Wash? Revealing the Salience Bias of Large Language Models in Commonsense Reasoning
6. Beyond a Single Judge: Simulating Social Persona Panels for Generative UI Evaluation
7. GUI-CIDER: Mid-training GUI Agents via Causal Internalization and Density-aware Exemplar Reselection
8. OS-SPEAR: A Toolkit for the Safety, Performance, Efficiency, and Robustness Analysis of OS Agents
9. Quick on the Uptake: Eliciting Implicit Intents from Human Demonstrations for Personalized Mobile-Use Agents
10. VeriOS: Query-Driven Proactive Human-Agent-GUI Interaction for Trustworthy OS Agents
11. UrbanGround: From Local Perception to Spatial Agency in a Real-Scale City
12. ColorEcosystem: Powering Personalized, Standardized, and Trustworthy Agentic Service in Massive-agent Ecosystem
13. ColorAgent: Building a Robust, Personalized, and Interactive OS Agent
14. Faithful Mobile GUI Agents with Guided Advantage Estimator
15. Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation
16. Hidden Ghost Hand: Unveiling Backdoor Vulnerabilities in MLLM-powered Mobile GUI Agents
17. See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles
18. LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition
19. Behavior Safety of Autonomous Interactive Agents: Risks, Attacks, Defenses and Evaluation
20. MineExplorer: Evaluating Open-World Exploration of MLLM Agents in Minecraft
21. Causal Probing for Internal Visual Representations in Multimodal Large Language Models
22. Communication Policy Evolution for Proactive LLM Agents
23. Do GUI Agents Know When Not to Act? Enabling Conflict-Aware Termination for Multimodal GUI Agents
24. Agent-ScanKit: Unraveling Memory and Reasoning of MLLM-Based Agents via Sensitivity Perturbations
25. Atomic-to-Compositional Generalization for Mobile Agents with Systematic Scheduling
26. Smoothing Grounding and Reasoning for MLLM-Powered GUI Agents with Query-Oriented Pivot Tasks
27. Domain Adaptation of MLLM-based Computer-Using Agents with Standard Operating Procedure
28. Say One Thing, Do Another? Diagnosing Reasoning-Execution Gaps in VLM-Powered Mobile-Use Agents
Tutorials and Contributions
Participating: Dive into LLMs《动手学大模型》Course Series 
Participating: 《大模型开发全流程》Course Series
Academic Services
Reviewer of conferences:
AAAI, ACL, ICML, EMNLP, ICLR
Reviewer of journals:
TIFS, IJHCI