Wenjie Yang

About Me

I am a second-year PhD student in the School of Data Science at Fudan University. Before transferring to the PhD program, I spent two years in the MS program. I am fortunate to be advised by Prof. Zengfeng Huang. I obtained my B.S. degree in Statistics from Donghua University in 2022.

Currently, I am interning at the Kimi Vision Team@Moonshot AI. I previously interned at Ant Group and Lilith Games. I am open to research collaboration with both academia and industry (only represents myself and does not represent any institutions I am affiliated with). If you are interested, please contact me via email.

Publications

Note: * indicates corresponding author, † indicates equal contribution.

Multimodality

• Wenjie Yang†, Siqi Zhu†, Zengfeng Huang*.
  Your VLM May Not Be Thinking with Interleaved Images.
  ICML Position, 2026. We explore whether interleaved images are truly necessary in "Thinking with Images" VLMs (they are not). We discuss the reason of performance gain of these VLMs and a more ideal way to use interleaved images.
• Wenjie Yang, Zengfeng Huang*.
  Poivre: Self-Refining Visual Pointing with Reinforcement Learning.
  arXiv preprint arXiv:2509.23746, 2025. We incentivize self-refining for the task of visual pointing with RL, surpassing models like Gemini 2.5 Pro on PointBench.

AI for math research

• Jiaxing Guo, Wenjie Yang, Shengzhong Zhang, Tongshan Xu, Lun Du, Da Zheng, Zengfeng Huang*.
  Right Is Not Enough: The Pitfalls of Outcome Supervision in Training LLMs for Math Reasoning.
  arXiv preprint arXiv:2506.06877, 2025.
• Wenjie Yang, Ruiyuan Huang, Jiaxing Guo, Zicheng Lyu, Tongshan Xu, Shengzhong Zhang, Lun Du, Da Zheng*, Zengfeng Huang*.
  Retrieval-Augmented Language Models are Mimetic Theorem Provers.
  EMNLP Findings, 2025.

Graph machine learning

• Wenjie Yang, Shengzhong Zhang*, Chen Ye, Jiaxing Guo, Tongshan Xu, Zengfeng Huang*.
  Know Your Neighbors: Subgraph Importance Sampling for Heterophilic Graph Active Learning.
  AAAI, 2026.
• Wenjie Yang†, Shengzhong Zhang†, Zengfeng Huang*.
  Cross-level graph contrastive learning for community value prediction.
  Neural Networks, 2026.
• Wenjie Yang†, Shengzhong Zhang†, Jiaxing Guo, Zengfeng Huang*.
  Contra²: A One-step Active Learning Method for Imbalanced Graphs.
  Artificial Intelligence, 2025.
• Wenjie Yang, Shengzhong Zhang*, Jiaxing Guo, Zengfeng Huang*.
  Your Graph Recommenders are Provably Doing Graph Contrastive Learning.
  KDD, 2025.
• Shengzhong Zhang, Wenjie Yang*, Yimin Zhang, Hongwei Zhang, Zengfeng Huang*.
  Understanding Class Bias Amplification in Graph Representation Learning.
  Transactions on Machine Learning Research (TMLR), 2025.
• Shengzhong Zhang, Yimin Zhang, Bisheng Li, Wenjie Yang, Min Zhou, Zengfeng Huang*.
  Graph Batch Coarsening Framework for Scalable Graph Neural Networks.
  Neural Networks, 2025.
• Shengzhong Zhang, Wenjie Yang, Xinyuan Cao, Hongwei Zhang, Zengfeng Huang*.
  StructComp: Substituting Propagation with Structural Compression in Training Graph Contrastive Learning.
  ICLR, 2024.
• Wenjie Yang†, Shengzhong Zhang†, Zengfeng Huang*.
  Enhancing Performance of Coarsened Graphs with Gradient-matching.
  ICASSP, 2024.