Tengjun Huang

Tengjun Huang

HKUST(GZ)

I am Tengjun Huang, a first-year MPhil student at The Hong Kong University of Science and Technology (Guangzhou). I received my bachelor’s degree from the School of Computer Science, Shandong University. My passion lies in uncovering the intricate relationships between diverse perceptual modalities and behavioral paradigms. My research interests are primarily focused on multimodal learning, MLLM, and representation learning.

My previous work has centered around computer vision and multimodal learning, delving into the intrinsic connections between a single modality and multiple other modalities. I also possess a background in video understanding, medical image segmentation, and self-supervised learning. Meanwhile, I am also interested in exploring the application of LLM-based agents in multimodal learning.

In my research, I aim to go beyond the surface-level understanding of data and instead seek to uncover deeper. Currently, an increasing number of approaches focus on scaling up data and models to address the gaps between different modalities, with the aim of learning holistic semantic connections. However, it may also be important to pay attention to underlying issues that cannot be resolved by simply enlarging the task scale. For instance, current multimodal models exhibit hallucinations during complex cross-modal reasoning and suffer significant performance drops when transferred to different domains with substantial gaps. On the other hand, we should perhaps explore more diverse learning paradigms, such as more fundamental task design, more efficient feature storage structures, more generalized optimization methods, and tighter representation coupling. This approach ensures that when conventional development paths encounter challenging dilemmas, we still have various alternative strategies to explore. This perspective drives me to explore innovative methods and frameworks that can effectively integrate and leverage these interactions.

If you are interested in discussing my research or exploring potential collaborations, please feel free to contact me. [📧Email]

🔥 News

  • 2025.06:  🎓 Graduated from Shandong University. Completed my bachelor thesis with Excellent.
  • 2025.01:  🎉 Thrilled to receive an offer from HKUST(GZ) for the MPhil program.
  • 2024.06:  📝 Submitted one paper to NeurIPS 2024 (Datasets and Benchmarks).
  • 2024.03:  🎉 My independent work on “Multimodal Representation Alignment and Efficient Transfer Learning” was accepted to the workshop of The Twelfth International Conference on Learning Representations (ICLR 2024).
  • 2023.11:  🏆 Achieved 2nd place (Runner-up) in all tracks of the NeurIPS 2023 Road Event Detection with Requirements Challenge and gave an oral presentation at NeurIPS 2023. (Competed solo: Track 1 — ranked 2nd out of 318 submissions by different teams, Track 2 — ranked 2nd out of 149 submissions by different teams).
  • 2023.06:  🎉 Ranked in the Top 0.6% (long-term competition, currently in the top 0.2%) in the Alibaba Cloud Tianchi Tmall Repurchase Prediction Challenge.
  • 2022.11:  🥈 Awarded Second Prize in the National College Mathematical Contest.
  • 2022.07:  🥇 Secured a National Gold Medal in the Chinese Collegiate Algorithm Design & Programming Challenge.

📝 Selected Papers

Efficient Remote Sensing with Harmonized Transfer Learning and Modality Alignment

Tengjun Huang
The Twelfth International Conference on Learning Representations (ICLR 2024), ML4RS.
[Paper] [Arxiv] [Stars]

🎖 Honors and Awards

  • 2nd Place (Runner-up) in All Tracks (Oral Presentation & Solo), NeurIPS 2023 ROAD-R Challenge, 2023
  • National Gold Medal, Chinese Collegiate Algorithm Design & Programming Challenge Contest, 2022
  • Asian Second Prize (Leader), Asia and Pacific Mathematical Contest in Modeling, 2022
  • Top 0.2% Worldwide (Solo), Alibaba Cloud Tianchi Tmall Repurchase Prediction Challenge, 2023
  • Second Prize, National College Mathematical Contest (CMC, organized by the Chinese Mathematical Society), 2022
  • Second Prize (Leader), China Contemporary Undergraduate Mathematical Contest in Modeling (CUMCM), 2022
  • First Prize (Top 5% & Leader), Mathematical Modeling Contest of Shandong University, 2022
  • National Third Prize, National Finals of New Engineering National Innovation Competition, 2023
  • National Third Prize (Top 3%), The 17th iCAN Innovation Contest (National Finals), 2023
  • First-Class (Top 1.2%), Research and Innovation Scholarship of Shandong University, 2022

💡 Patents

Old people's behavioral safety monitored control system

Jiahui Cui, Tengjun Huang, Dongxu Liu, Xiaohuan Zhao, Siyuan Niu, Jiayi Lu, Yunhan Hu, Zhengshuai Jiang, Guangrun Wu
Invention Patent Grant, China National Intellectual Property Administration.
Publication of CN116246424B.

💼 Academic Service

  • Journal Reviewer: IEEE Transactions on Artificial Intelligence (TAI).

📖 Educations

  • 2025 - (expected 2027): MPhil Student, The Hong Kong University of Science and Technology (Guangzhou).
  • 2021 - 2025: B.Eng. in Artificial Intelligence, Shandong University.

💬 Oral Presentations

  • Dec 2023, the Thirty-seventh Annual Conference on Neural Information Processing Systems (NeurIPS 2023), Ernest N. Morial Convention Center, New Orleans, US, Room 353.