|
Yiming Huang 黄奕铭 My name is Yiming Huang (黄奕铭 in Chinese, Tim is my nickname). I am an M.Phil. student at Data Science and Analytics Thrust of The Hong Kong University of Science and Technology, Guangzhou, under the supervision of Prof. Xinlei He. Before that, I received my B.Eng. degree in Software Engineering at School of Computer of Beijing University of Technology, under the supervision of Prof. Xiaodan Zhang. I'm also a high school graduate of Tsinghua University High School. My research interest focuses on Trustworthy AI. The long-term research goal of my study is to analyze and control the most advanced AI algorithms, models, and systems. To be specific, I am exploring the interpretability, robustness, granularity, and generalizability of AI, to cope with the explainability (XAI), safety, privacy, fairness, more refined controllability and human-interactivity problems in AI applications. I think this research goal can also be called as "Analytics and Control of Massive Parameter System". Currently, I am focusing on mechanistic interpretability for safe AI and societal AI. I have wonderful collaborations with Ziyu Zhou@HKUST(GZ) CityMind Lab, Zihao Wang@HKUST EMIA and Biquan Bie. I am also looking forward to more interdisciplinary collaboration in the future.
🎉News 2025.8.23: Our collaboration work (with Junyan Zhang) is accepted by Findings of EMNLP '25.
✔️News 2025.5.28: Three papers are submitted these days, GOOD! I am coming up with some new idea~
🎉News 2024.12.10: Congratulations to Zihao - Our collaboration work MSV-PCT is accepted by AAAI '25!.
✔️News 2024.8.16: A paper I participated in has been submitted to AAAI '25.
🤩News 2024.5.27: I received official offer of Red Bird Master of Philosophy Program at HKUST(GZ) and I make my final decision to accept it. See you guys in Guangzhou!
🎉News 2024.4.17: Congratulations to Ziyu - His SDformer is accepted by IJCAI '24!. It's really a interesting work!
|
|
Publications(In chronological order, * is equal contribution, [.] is the topic, my newest publication is updating at my Google Scholar) |
|
Do BERT-Like Bidirectional Models Still Perform Better on Text Classification in the Era of LLMs?
Junyan Zhang*, Yiming Huang*, Shuliang Liu, Yubo Gao, Xuming Hu Findings of EMNLP '25 paper [Text Classification, BERT, LLM] TL;DR: We comprehensively compare Bert-like bidirectional model and LLM-based methods (direct query & internal probe) on various text classification tasks. The result shows bert models are still more robust on some sort of tasks than LLMs. |
|
MSV-PCT: Multi-Sparse-View Enhanced Transformer Framework for Salient Object Detection in Point Clouds
Zihao Wang, Yiming Huang, Gengyu Lyu, Yucheng Zhao, Ziyu Zhou, Bochen Xie, Zhen Yang, Yongjian Deng AAAI '25 paper [Point Cloud, Multi-View] TL;DR: We propose a novel framework for Point Cloud Salient Object Detection (PCSOD) task, named MSV-PCT. Which leverage efficient multi-sparse-view branches to enhance the performance of transformer-based point cloud backbone on PCSOD task. |
|
SDformer: Transformer with Spectral Filter and Dynamic Attention for Multivariate Time Series Long-term Forecasting
Ziyu Zhou, Gengyu Lyu, Yiming Huang, Zihao Wang, Ziyu Jia, Zhen Yang IJCAI '24 paper / code [Time-Series, Attention] TL;DR: We propose a novel Transformer architecture (named SDformer) for long-term time series forecasting. It is the first time to address the problem of smooth attention distribution when modeling time series data with a large number of variates. |
|
Generic Attention-model Explainability by Weighted Relevance Accumulation
Yiming Huang, Aozhe Jia, Xiaodan Zhang, Jiawei Zhang MMAsia '23 (reprentative work of my undergrad research) paper / sildes / video / code [XAI, Attention] TL;DR: This work explores the explainability of multimodal attention by layer-wise relevancy accumulation with attention map, especially stressing the importance of relevancy in residual connection for better explainability. |
|
TimesNet-PM2.5: Interpretable TimesNet for Disentangling Intraperiod and Interperiod Variations in PM2.5 Prediction
Yiming Huang*, Ziyu Zhou*, Zihao Wang*, Xiaoying Zhi, Xiliang Liu (*equal contribution) Atmosphere Journal paper [XAI, Time-Series] TL;DR: In this paper, we accomplish task-specific adaption of TimesNet (ICLR '23) named TimesNet-PM2.5. This specialized version improved the performance and interpretability of the PM2.5 prediction of Haikou, Hainan Province. |
|
STFM: Enhancing Autism Spectrum Disorder Classification Through Ensemble Learning-Based Fusion of Temporal and Spatial fMRI Patterns
Ziyu Zhou*, Yiming Huang*, Yining Wang*, Yin Liang PRICAI '23 paper [Medical AI, Attention] TL;DR: We propose a Spatio and Temporal framework based on cross-attention for better autism spectrum disorder classification. |
Miscellanea |
|
I love biking, although all of my bikes are broken now. My favorite animation: Rebuild of EVANGELION series. I love miHoYo games. My favorite TV series: House of the Dragon. My favorite poem: Sailing to Byzantium. My favorite book:《文学的邀约》. My favorite restaurants: 宜宾招待所，东发道. I have lived in Beijing for more than 10 years and my hometown is Jilin, China. My motto: 缘自尘心起，逆旅天地间。 As its name goes, my WeChat Friend Cycle is only open to friends. But don't worry, maybe you will see it in the future after a long time of acquaintance. In addition, I don't open it to teachers and elders, because some contents may not be serious. |
Gallery |
<
>
|
|
Last update: 2025.08.25. This page is powered by Jon Barron's Template, ChyoD1811's CSDN blog and ChatGPT. |