Xiaoce Wang

Senior undergraduate at Tsinghua University, Xinya College / Dept. of Computer Science and Technology (B.S. 2023–2028). I work on Multimodal AI (vision-language models, diffusion & flow matching, multimodal agents) and AI Safety & Alignment (safe LLMs, safe multimodal AI, multimodal alignment, interpretable AI).

📧 wangxc23@mails.tsinghua.edu.cn 🐙 github.com/ZephinueCode 🎓 Google Scholar Beijing, China

Education

Tsinghua University
B.S. in Computer Science and Technology, Xinya College
2023 – 2028 (expected)

Publications

Why Do DiT Editors Drift? Plug-and-Play Low Frequency Alignment in VAE Latent SpaceNeurIPS 2026
Xiaoce Wang*, Sifan Zhou*, Kaifei Wang*, Leli Xu, Tao He, Ming Li  (*co-first author)
ToolTok: Tool Tokenization for Efficient and Generalizable GUI AgentsPreprint
Xiaoce Wang, Guibin Zhang, Junzhe Li, Jinzhe Tu, Chun Li, Ming Li
It Is Not the Phrase; It Is the Expectation: Understanding User Perceptions of Contrastive Negation in AI-Generated ResponsesPreprint
Guobin Wang*, Xiaoce Wang*, Minqian Yang, Pei-Luen Patrick Rau  (*co-first author)
BARREL: Boundary-Aware Reasoning for Factual and Reliable LRMsICLR 2026
Junxiao Yang, Jinzhe Tu, Haoran Liu, Xiaoce Wang, Chujie Zheng, Zhexin Zhang, Shiyao Cui, Caishun Chen, Tiantian He, Hongning Wang, Yew-Soon Ong, Minlie Huang
StyleTailor: Towards Personalized Fashion Styling via Hierarchical Negative FeedbackAAAI 2026 Oral
Hongbo Ma, Fei Shen, Hongbin Xu, Xiaoce Wang, Gang Xu, Jinkai Zheng, Liangqiong Qu, Ming Li
SynCred-Bench: Benchmarking Synthetic Credibility in AI-Generated Visual MisinformationPreprint
Junxiao Yang, Minghao Zhang, Xiaoce Wang, Haoran Liu, Shiyao Cui, Hongning Wang, Minlie Huang

Research & Industry Experience

Research Intern
MemoraX
May 2026 – Present
  • Multimodal Memory: long-context & agentic memory engineering from a multimodal perspective.
Algorithm Engineer Intern
Xiaomi Corporation
Feb 2026 – Jul 2026
  • VLM Pretraining & Alignment: pretraining of edge-deployable vision-language models.
Research Intern
Guangming Laboratory, Shenzhen
Jul 2025 – May 2026
  • Multimodal Agentic Systems: multimodal agent design & engineering, multi-agent systems, cross-modal learning, representation learning.
  • Generative Models: mechanistic interpretability of DiT-based generative models.
Research Intern
Conversational AI (CoAI) Lab @ CST, Tsinghua University
Mar 2025 – Present
  • Safe & Interpretable LLMs: truthful LLMs, reasoning LLMs, supervised fine-tuning, reinforcement learning, CoT.
  • Developing new thinking paradigms for reasoning LLMs to admit their ignorance on certain topics via SFT and RL techniques.

Honors & Leadership

  • President of Student Union, Xinya College, Tsinghua University
  • Outstanding Student Leader, Tsinghua University (Top 6 among 400+ students, Xinya College), 2023–2024
  • Scholarship for Society Work Contributions, Tsinghua University, 2024–2025
  • Second Grade Prize, NOIP 2021 (Beijing)

Languages

  • Chinese (native); English (TOEFL 105, IELTS 8.5); Japanese (JLPT N1, 154/180)