I am a first-year Ph.D. student at The University of Hong Kong, advised by Prof. Tao Yu in the Executable Language Grounding (XLANG) Lab, part of the HKU NLP Group.

My research focuses on robot-use agents that control robots to accomplish tasks in the real world. I also work on vision-language-action (VLA) models for robot learning, including developing model frameworks and data scaling strategies to ground language in physical actions.

Previously, I studied computer science at Shanghai Jiao Tong University and worked with Prof. Lu Chen and Prof. Kai Yu in the X-LANCE Lab.

🔥 News

• 2026.09: 🎉 FineVLA is accepted at NeurIPS 2026!
• 2026.05: Qwen-VLA and FineVLA are now on arXiv, exploring generalist robot policies and fine-grained instruction following.
• 2025.08: 🎉 When Long Helps Short is accepted at EMNLP 2025 (Main Conference).

📝 Publications

* Equal contribution

• FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies

  Xintong Hu^*, Xuhong Huang^*, Jinyu Zhang, …, Yingming Zheng, Shuai Bai, Tao Yu

  NeurIPS 2026  ·  [Paper]   [Project]   [Code]

  Fine-grained, action-aligned supervision for robot policies that follow instructions about how to act, with a human-verified dataset, a video-understanding benchmark, and a scalable annotation pipeline.

• Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments

  Qiuyue Wang, Mingsheng Li, Jian Guan, …, Yingming Zheng, …, Tao Yu, Xionghui Chen

  Technical report · arXiv 2026  ·  [Paper]   [Repository]   [Blog]

  A generalist VLA model that brings manipulation, navigation, and trajectory prediction into one framework, learning from heterogeneous embodied data across robot platforms.

• When Long Helps Short: How Context Length in Supervised Fine-tuning Affects Behavior of Large Language Models

  Yingming Zheng^*, Hanqi Li^*, Kai Yu, Lu Chen

  EMNLP 2025 · Main Conference  ·  [Paper]   [PDF]   [arXiv]

  Why can long-context fine-tuning help short-context tasks? An analysis of attention, feed-forward networks, and contextual versus parametric knowledge, with guidance for hybrid training.

• CRAVE: A Conflicting Reasoning Approach for Explainable Claim Verification Using LLMs

  Yingming Zheng, Xiaoliang Liu, Peng Wu, Li Pan

  IJCNN 2025  ·  [Paper]   [Code]   [Proceedings]

  Explainable claim verification through evidence retrieval, reasoning from conflicting perspectives, and a small language model that judges the resulting rationales.

🎖 Honors and Awards

• 2024.10: The Tung Scholarship
• 2023.10: National Scholarship
• 2023.10: First Prize Scholarship, Shanghai Jiao Tong University (top student in the major)

📖 Education

• The University of Hong Kong — First-year Ph.D. student, XLANG Lab. Advisor: Prof. Tao Yu.
• Shanghai Jiao Tong University — Undergraduate studies in Computer Science, from September 2022. Research with Prof. Lu Chen and Prof. Kai Yu, X-LANCE Lab.