Runpeng Yu

Runpeng Yu 于润芃 Ph.D. Student @ xML Lab, Institute of Data Science, National University of Singapore. [ Scholar ] [ GitHub ] Contact Email: r.yu [at] u (dot) nus (dot) edu Address: 4 ENGINEERING DRIVE 3, 117583, Singapore

Actively Seeking Intern and Visiting Opportunities

Research Interests

• Large Language Model. Large Language Model (LLM) reasoning; Chain-of-Thought (CoT).

• Multimodal Large Language Model. Multimodal Chain-of-Thought; perception process of Multimodal Large Language Model (MLLM).

• Machine Learning. Lineage relationships among pretrained models; pretrained model reuse.

Publications

• Introducing Visual Perception Token into Multimodal Large Language Model [Paper]
  Runpeng Yu , Xinyin Ma and Xinchao Wang.
  Preprint (2025).

• CoT-Valve: Length-Compressible Chain-of-Thought Tuning [Paper]
  Xinyin Ma, Guangnian Wan, Runpeng Yu , Gongfan Fang, and Xinchao Wang
  Preprint (2025).

• Neural Phylogeny: Fine-Tuning Relationship Detection among Neural Networks [Paper]
  Runpeng Yu and Xinchao Wang.
  Accepted by ICLR (2025).

• Attention Prompting on Image for Large Vision-Language Models [Paper]
  Runpeng Yu , Weihao Yu and Xinchao Wang.
  Accepted by ECCV (2024).

• Neural Lineage [Paper]
  Runpeng Yu ,and Xinchao Wang.
  Accepted as CVPR (2024) Oral.

• Generator Born from Classifier [Paper]
  Runpeng Yu ,and Xinchao Wang.
  Accepted by NeurIPS (2023).

• Distribution Shift Inversion for Out-of-Distribution Prediction [Paper]
  Runpeng Yu , Songhua Liu, Xingyi Yang, and Xinchao Wang.
  Accepted by CVPR (2023).

• Regularization Penalty Optimization for Addressing Data Quality Variance in OoD Algorithms [Paper]
  Runpeng Yu , Hong Zhu, Kaican Li, Lanqing Hong, Rui Zhang, Nanyang Ye, Shao-Lun Huang and Xiuqiang He.
  Accepted by AAAI (2022).

• A Spectral Perspective On Graph Data Augmentation For Graph Classification [Paper]
  Yutong Xia*, Runpeng Yu* , Yuxuan Liang, Xavier Bresson, Xinchao Wang and Roger Zimmermannand
  Accepted by AAAI (2025).