Chunzheng Zhu

Chunzheng Zhu

Ph.D. Candidate

Hunan University

Biography

I am a Ph.D. candidate in Computer Science and Technology at Hunan University and a Research Intern at ByteDance. My research focuses on multimodal large language models, agentic reinforcement learning, representation learning, and medical AI. I aim to develop efficient and trustworthy reasoning systems through post-training and rigorous evaluation.

Expected PhD graduation: June 2027.

Interests
  • Multimodal Large Language Models
  • Agentic Reinforcement Learning
  • Representation Learning
  • Medical AI
  • Computer Vision
Education

  • Ph.D. in Computer Science and Technology, 2022.09 - 2027.06

    Hunan University

  • B.Eng. in Data Science and Big Data Technology, 2018.09 - 2022.06

    Minzu University of China

Experience

Research Intern

ByteDance
Research on post-training and biomedical mid-training for native multimodal large language models, with a focus on capability optimization and evaluation. Built pipelines for heterogeneous biomedical data integration, mid-training, and post-training to strengthen biomedical understanding and multimodal reasoning capabilities.

Student Researcher

National Supercomputing Center in Changsha
Research on AI assisted fetal ultrasound diagnosis and post training of medical multimodal large language models. Developed self supervised and continual learning methods for ultrasound and explored agentic reinforcement learning, reflection, and latent reasoning for medical reasoning.

Publications

*
All Featured Publications
CRAFT: Causal Responsibility and Failure Tracing in Medical Vision Language Models
Chunzheng Zhu, Jiaqi Zeng, Hongbo Zhao, Yihang Chen, Yijun Wang, Jianxin Lin
Advances in Neural Information Processing Systems (NeurIPS), 2026 (Spotlight)
Causal tracing and targeted intervention for arbitration and brake failures in medical vision language models.
PDF Code Paper
CRAFT: Causal Responsibility and Failure Tracing in Medical Vision Language Models
PACE: Pareto-Adaptive Compression for Efficient Native MLLMs
Lei Tian¹, Chunzheng Zhu†, Kai WU¹, YiFan Zhang², Haihua Yang¹
Advances in Neural Information Processing Systems (NeurIPS), 2026
Pareto-adaptive visual token compression for efficient native multimodal language models.
PDF Paper
PACE: Pareto-Adaptive Compression for Efficient Native MLLMs
Trustworthy Medical Concept Recalibration through Logic-Aware Multimodal Evidence Fusion
Lei Zhao, Chunzheng Zhu†, Zhan Gao, Pengchen Liang, Xiaoyun Liang, Alex Chit-Yuen Cheng, Chi-Man Pun
Information Fusion, 2026
ConChecker autonomously recalibrates medical concepts through logic-aware multimodal evidence fusion and verifiable policy optimization.
Trustworthy Medical Concept Recalibration through Logic-Aware Multimodal Evidence Fusion
The Path to Self-Evolving Clinical Systems: Scaling Medical Agents from Assistance to Autonomy
Chunzheng Zhu, Lei Tian, Bohan Tan, Ziqi Zhou, Yuxuan Sun, Yijun Wang, Chengchao Lv, Yilin Wen, Yijun He, Jinghao Lin, Yihang Chen, Cheewei Tan, Qianshan Wei, Lei Zhao, Bin Pu, Kenli Li, Yuan Xue, Jianxin Lin
arXiv preprint arXiv:2607.11175, 2026
A survey and roadmap for scaling medical agents from clinical assistance toward autonomous, self-evolving clinical systems.
PDF Project DOI Paper Project
The Path to Self-Evolving Clinical Systems: Scaling Medical Agents from Assistance to Autonomy
MedSynapse-V: Bridging Visual Perception and Clinical Intuition via Latent Memory Evolution
Chunzheng Zhu, Jiaqi Zeng, Junyu Jiang, Jianxin Lin, Yijun Wang
European Conference on Computer Vision (ECCV), 2026
A latent diagnostic memory evolution framework that bridges visual perception and clinical intuition in medical VLMs.
Code DOI Paper
MedSynapse-V: Bridging Visual Perception and Clinical Intuition via Latent Memory Evolution
Anatomy-Anchored Self-Supervision: Distilling Vision Foundation Models for Invariant Ultrasound Representation
Chunzheng Zhu, Yijun Wang, Jianxin Lin, Feng Wang, Hongwei Wang, Lei Zhao, Shengli Li, Kenli Li
International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2026)
Anatomy-anchored self-supervised learning for invariant ultrasound representation.
Code Paper
Anatomy-Anchored Self-Supervision: Distilling Vision Foundation Models for Invariant Ultrasound Representation
When Models Learn to Ask Why: Adaptive Causal Reasoning for Trustworthy Medical Vision-Language Models
Jianxin Lin†, Chunzheng Zhu†, Peter J. Kneuertz, Yunfei Bai, Yuan Xue
IEEE/CVF Conference on Computer Vision and Pattern Recognition Findings (CVPR Findings), 2026
Adaptive causal reasoning with self-reflection for trustworthy medical vision-language models.
Code DOI Paper
When Models Learn to Ask Why: Adaptive Causal Reasoning for Trustworthy Medical Vision-Language Models
MedEyes: Learning Dynamic Visual Focus for Medical Progressive Diagnosis
Chunzheng Zhu, Yangfang Lin, Shen Chen, Yijun Wang, Jianxin Lin
Proceedings of the AAAI Conference on Artificial Intelligence, 2026
A reinforcement learning framework that models clinician-style dynamic visual focusing for medical progressive diagnosis.
Code DOI Paper
MedEyes: Learning Dynamic Visual Focus for Medical Progressive Diagnosis
fMRI2GES: Co-Speech Gesture Reconstruction From fMRI Signal With Dual Brain Decoding Alignment
Chunzheng Zhu, Jialin Shao, Jianxin Lin, Yijun Wang, Jing Wang, Jinhui Tang, Kenli Li
IEEE Transactions on Circuits and Systems for Video Technology, 2025
Co-speech gesture reconstruction from fMRI signals with dual brain decoding alignment.
Paper
fMRI2GES: Co-Speech Gesture Reconstruction From fMRI Signal With Dual Brain Decoding Alignment
Pathology-Aware Prototype Evolution via LLM-Driven Semantic Disambiguation for Multicenter Diabetic Retinopathy Diagnosis
Chunzheng Zhu, Yangfang Lin, Jialin Shao, Jianxin Lin, Yijun Wang
Proceedings of the 33rd ACM International Conference on Multimedia (ACM MM), 2025 (Oral)
LLM-driven pathology-aware prototype evolution for multicenter diabetic retinopathy diagnosis.
Code DOI Paper
Pathology-Aware Prototype Evolution via LLM-Driven Semantic Disambiguation for Multicenter Diabetic Retinopathy Diagnosis
Advancing Ultrasound Medical Continuous Learning with Task-Specific Generalization and Adaptability
Chunzheng Zhu, Jianxin Lin, Guanghua Tan, Ningbo Zhu, Kenli Li, Chunlian Wang, Shengli Li
IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2024 (Oral)
Task-specific continual learning for ultrasound medical image analysis.
Paper
Advancing Ultrasound Medical Continuous Learning with Task-Specific Generalization and Adaptability

Awards

National Scholarship, 2026
Hunan University Sep 2026
Graduate Student Role Model, Hunan University, 2026
College of Computer Science and Electronic Engineering, Hunan University Sep 2026
Recognized as a Graduate Student Role Model by the College of Computer Science and Electronic Engineering, Hunan University.
First Prize, 19th Challenge Cup Artificial Intelligence+ Special Competition
Challenge Cup Jan 2025
Global-leading AI + ultrasound technology solution.
Outstanding Graduate Student, Hunan University, 2024-2025
Hunan University Jan 2025
Outstanding Graduate Student, Hunan University, 2023-2024
Hunan University Jan 2024
Second Prize, MathorCup Big Data Challenge for Graduate Students
MathorCup Jan 2023
Third Prize, China Graduate Mathematical Contest in Modeling
China Graduate Mathematical Contest in Modeling Jan 2023
First-Class Academic Scholarship, Minzu University of China, 2021-2022
Minzu University of China Jan 2022
Second-Class Academic Scholarship, Minzu University of China, 2020-2021
Minzu University of China Jan 2021
Meritorious Winner, Mathematical Contest in Modeling
Mathematical Contest in Modeling Jan 2020
National Second Prize, China Computer Design Competition
China Computer Design Competition Jan 2020
First Prize, Beijing Division, CUMCM
Contemporary Undergraduate Mathematical Contest in Modeling Jan 2020
Second Prize, Beijing Undergraduate Animation Design Competition
Beijing Undergraduate Animation Design Competition Jan 2019