Yifan Han

I'm Yifan Han, a second-year Master's student at the National Laboratory of Pattern Recognition (NLPR), Institute of Automation, Chinese Academy of Sciences (CASIA), under the supervision of Prof. Wenzhao Lian. My research interests include robot manipulation, dexterous manipulation, and vision-language-action (VLA), with a focus on multimodal perception and embodied intelligence for real-world robotic systems.

Dexterous Manipulation Affordance Learning VLA Systems Active Perception
hanyifan2024@ia.ac.cn Beijing / Shanghai

Research Focus

  • Policy learning and RL-based post-training for dexterous manipulation in real-world deployment.
  • Active perception and grounding for embodied manipulation.
  • Affordance learning for robot manipulation.
Homepage Google Scholar OpenReview

Education

Academic training

Institute of Automation, Chinese Academy of Sciences

M.S. Student, National Laboratory of Pattern Recognition

GPA: 3.85/4.00 | 2023 - 2025 | Beijing, China

Supervisor: Prof. Wenzhao Lian

Shanghai Jiao Tong University

Research Assistant, School of Artificial Intelligence

2025 - Present | Shanghai, China

Supervisors: Prof. Wenzhao Lian, Prof. Yao Mu

Rutgers University

Research Intern, Department of Computer Science

2026 - Present | New Brunswick, NJ, USA

Supervisor: Prof. Jingjin Yu

Huazhong University of Science and Technology

B.Eng., School of Artificial Intelligence and Automation

GPA: 3.78/4.00 | 2020 - 2024 | Wuhan, China

Research Experience

Selected roles

Research Assistant | SJTU

School of Artificial Intelligence, Shanghai, China (2025 - Present)

Advisors: Prof. Wenzhao Lian, Prof. Yao Mu

  • Conduct research on dexterous manipulation and affordance learning, including policy learning and RL-based post-training.
  • Contribute to active perception, VLA-based robotic systems, and MLLM grounding modules for embodied manipulation tasks.

Research Intern | Rutgers University

Department of Computer Science, New Brunswick, NJ, USA (2026 - Present)

Advisor: Prof. Jingjin Yu

  • Worked on task-conditioned scene-level affordance grounding for real-time robotic manipulation in complex multi-object scenes.
  • Contributed to affordance-conditioned policy learning.

Research Student | CASIA

NLPR, Institute of Automation, Beijing, China (2023 - 2025)

Advisor: Prof. Wenzhao Lian

  • Worked on learning-based robot manipulation, including multi-task imitation learning and long-horizon manipulation.
  • Contributed to human-to-robot skill transfer, affordance-centric manipulation, and active robotic perception.

Publications

First-Author or Co-First

BORA: Bridging Offline Reinforcement Learning and Online Residual Adaptation for Real-World Dexterous VLA Models
Zhongxi Chen*, Yifan Han*, Yanming Shao, Huanming Liu, Congsheng Xu, Yao Mu, Wenzhao Lian
Under Review Equal contribution Project leader

Bridges offline token-action reinforcement learning with online residual adaptation for robust real-world dexterous VLA deployment.

BridgeACT: Bridging Human Demonstrations to Robot Actions via Unified Tool-Target Affordances
Yifan Han, Jianxiang Liu, Haoyu Zhang, Yuqi Gu, Hanqing Wang, Zhongxi Chen, Litao Liu, Wenzhao Lian
Under Review Project leader

Introduces role-conditioned tool-target affordances as an embodiment-agnostic representation for transferring manipulation knowledge from human videos to robots.

Affordance2Action: A Benchmark and Framework for Task-Conditioned Part Affordance Grounding in Real-Time Robot Manipulation
Litao Liu*, Yifan Han*, Pengfei Yi, Wenbo Yu, Hanqing Wang, Haoran Du, Enze Yuan, Zilin Yuan, Ruiding Feng, Michael Liu, Qi Zhang, Jingjin Yu
Under Review Equal contribution Project leader

Builds a benchmark and real-time framework for task-conditioned part affordance grounding in robot manipulation.

DexHiL: A Human-in-the-Loop Framework for Vision-Language-Action Model Post-Training in Dexterous Manipulation
Yifan Han*, Zhongxi Chen*, Yuxuan Zhao, Congsheng Xu, Yanming Shao, Yichuan Peng, Yao Mu, Wenzhao Lian
IROS 2026 Equal contribution

Builds a human-in-the-loop post-training framework for dexterous-manipulation VLA models, combining human feedback and policy optimization.

FSAG: Enhancing Human-to-Dexterous-Hand Finger-Specific Affordance Grounding via Diffusion Models
Yifan Han, Yichuan Peng, Pengfei Yi, Junyan Li, Hanqing Wang, Gaojing Zhang, Qi Peng Liu, Wenzhao Lian
IROS 2026

Studies finger-specific affordance grounding from human hands to dexterous hands using diffusion models.

Main Contributor

FOAM: Foresight-Augmented Multi-Task Imitation Policy for Robotic Manipulation
Litao Liu, Wentao Wang, Yifan Han, Zhuoli Xie, Pengfei Yi, Junyan Li, Yi Qin, Wenzhao Lian
AAAI 2026 Oral Presentation

Contributed to foresight-augmented multi-task imitation learning, introducing task-progress prediction and visual representations to improve generalization in robotic manipulation.

Splatter Joint: 3D Gaussian Splatting for Articulated Objects
Junyan Li, Yifan Han, Pengfei Yi, Wenzhao Lian
IROS 2025

Contributed to 3D Gaussian Splatting for articulated-object modeling, focusing on movable-part representation, structural understanding, and view-consistent reconstruction.

SAGE: Scene Graph-Aware Guidance and Execution for Long-Horizon Manipulation Tasks
Jialiang Li, Wenzheng Wu, Gaojing Zhang, Yifan Han, Wenzhao Lian
ICRA 2026

Contributed to a scene-graph-guided long-horizon manipulation framework that integrates object-relation modeling with execution planning for complex multi-stage tasks.

Affordance-R1: Reinforcement Learning for Generalizable Affordance Reasoning in Multimodal Large Language Models
Hanging Wang, Shaoyang Wang, Yiming Zhong, Zemin Yang, Jiamin Wang, Zhiqing Cui, Jiahao Yuan, Yifan Han, Mingyu Liu, Yuexin Ma
AAAI 2026 Oral Presentation

Contributed to a reinforcement-learning framework for affordance reasoning in MLLMs, improving generalizable operation-region understanding across objects and scenes.

Look, Zoom, Understand: The Robotic Eyeball for Embodied Perception
Jiashu Yang, Yifan Han, Yucheng Xie, Ning Guo, Wenzhao Lian
Preprint 2025

Contributed to an active embodied perception system that improves robot understanding of task-relevant regions and object details through viewpoint adjustment and local observation.

Viewpoint Matters: Dynamically Optimizing Viewpoints with Masked Autoencoder for Visual Manipulation
Pengfei Yi, Yifan Han, Junyan Li, Litao Liu, Wenzhao Lian
Preprint 2026

Contributed to MAE-based dynamic viewpoint optimization, analyzing and improving the robustness of visual manipulation policies under changing viewpoints.

Academic Service

Reviewer
  • IROS
  • ICRA
  • IEEE RA-L
  • AAAI
  • CVPR
  • ECCV

Skills

Technical strengths

Programming

Proficient: Python, MATLAB, C, LaTeX

Advanced: C++, JavaScript

Software

MuJoCo, Isaac Sim, SolidWorks, Adobe Photoshop, MS Office

Hardware

Robots: Franka, xArm robots, Universal Robots

Other: 8051 MCUs, STM32, RealSense, Orbbec cameras

Honors and Awards

Selected

National Robotics Competition (ROBOCON)

National First Prize and Third Place Overall (2022)

National First Prize, Robot Equestrian Event (2022)

National First Prize (2023)

National Third Prize, Robot Equestrian Event (2023)

China University Intelligent Robot Creativity Competition

National Second Prize (2022)

Huazhong University of Science and Technology

Self-Improvement Scholarship (2022)