Xiangming Gu

I am a fourth-year Ph.D. candidate from NUS Sound and Music Computing Lab, where I am supervised by Prof. Ye Wang. I am affiliated to Integrative Sciences and Engineering Programme and School of Computing at National University of Singapore. Before that, I obtained my B.E. degree of Electronic Engineering and B.S. degree of Finance at Tsinghua University in 2021.

I am also a research intern at Sea AI lab (SAIL). I am mentored by Dr. Tianyu Pang and Dr. Chao Du, and working closely with Dr. Qian Liu, and Dr. Min Lin.

My research topics include: (i) generative models: demystifying attention sink in LLMs [ICLR'2025 Spotlight], and demystifying memorization in diffusion models [TMLR'2025]; (ii) AI safety: introducing infectious jailbreak for (multimodal) LLM multi-agents [ICML'2024], and elucidating confidence calibration on LLM guard models [ICLR'2025]; (iii) singing/speech: application of machine learning, e.g. multimodal learning [MM'2022 Oral, TOMM'2024], multi-distribution learning (domain adaptation [ISMIR'2022, TOMM'2024], fairness [MM'2023]), to singing/speech techniques.

Email  /  CV  /  Google Scholar  /  Openreview  /  Linkedin  /  Twitter  /  Github

profile photo

News
  • [2025.03]: I passed my Ph.D. thesis proposal review!
  • [2025.02]: I was invited to give a talk titled On the Interpretability and Safety of Generative Models by research week open house of NUS!
  • [2025.01]: Two papers with one spotlight and one poster got accepted to International Conference on Learning Representations (ICLR'2025), one paper got accepted to Transactions on Machine Learning Research (TMLR'2025)!
  • [2025.01]: I gave a poster presentation about Agent Smith during Global Young Scientists Summit 2025!
  • [2024.10]: I was selected as a participant for Global Young Scientists Summit 2025!
  • [2024.07]: I received Dean's Graduate Research Excellence Award from School of Computing, NUS!
  • [2024.05]: One paper got accepted to International Conference on Machine Learning (ICML'2024)!
  • [2024.02]: One paper got accepted to ACM Transactions on Multimedia Computing Communications and Applications (TOMM'2024)!
  • [2024.02]: We released Agent Smith, which got posted as "Here Come the AI Worms" on WIRED Magazine!

Selected Research
* denotes equal contribution, † denotes correspondence. Please see my CV or Google Scholar for full list.
Generative Models
When Attention Sink Emerges in Language Models: An Empirical View
Xiangming Gu, Tianyu Pang†, Chao Du, Qian Liu, Fengzhuo Zhang, Cunxiao Du, Ye Wang†, Min Lin
International Conference on Learning Representations (ICLR), Singapore, Singapore, 2025. (Spotlight)
Also in Annual Conference on Neural Information Processing Systems Workshop on Attributing Model Behavior at Scale (ATTRIB @ NeurIPS), Vancouver, Canada, 2024. (Oral)
pdf / code / video / slides / poster
On Memorization in Diffusion Models
Xiangming Gu, Chao Du†, Tianyu Pang†, Chongxuan Li, Min Lin, Ye Wang
Transactions on Machine Learning Research (TMLR), 2025.
pdf / code
SkyLadder: Better and Faster Pretraining via Context Window Scheduling
Tongyao Zhu, Qian Liu†, Haonan Wang, Shiqi Chen, Xiangming Gu, Tianyu Pang, Min-Yen Kan
International Conference on Learning Representations Workshop on Open Science for Foundation Models (SCI-FM @ ICLR), Singapore, Singapore, 2025.
pdf / code
AI Safety
Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast
Xiangming Gu*, Xiaosen Zheng*, Tianyu Pang*†, Chao Du, Qian Liu, Ye Wang†, Jing Jiang†, Min Lin
International Conference on Machine Learning (ICML), Vienna, Austria, 2024.
Also in International Conference on Learning Representations Workshop on Large Language Model Agents (LLMAgents @ ICLR), Vienna, Austria, 2024.
pdf / project page / code / video / slides / ICML poster / GYSS poster / press
On Calibration of LLM-based Guard Models for Reliable Content Moderation
Hongfu Liu†, Hengguan Huang, Xiangming Gu, Hao Wang, Ye Wang
International Conference on Learning Representations (ICLR), Singapore, Singapore, 2025.
Also in Annual Conference on Neural Information Processing Systems Safe Generative AI Workshop (SafeGenAI @ NeurIPS), Vancouver, Canada, 2024. (Oral)
pdf / code
Speech and Singing
Automatic Lyric Transcription and Automatic Music Transcription from Multimodal Singing
Xiangming Gu, Longshen Ou, Wei Zeng, Jianan Zhang, Nicholas Wong, Ye Wang
ACM Transactions on Multimedia Computing Communications and Applications (TOMM), 2024.
pdf / code / data
Elucidate Gender Fairness in Singing Voice Transcription
Xiangming Gu, Wei Zeng, Ye Wang
ACM International Conference on Multimedia (MM), Ottawa, Canada, 2023.
pdf / code / video / poster
MM-ALT: A Multimodal Automatic Lyric Transcription System
Xiangming Gu*, Longshen Ou*, Danielle Ong, Ye Wang
ACM International Conference on Multimedia (MM), Lisbon, Portugal, 2022. (Oral, Top Paper Award)
pdf / appendix / project page / code / data / video / press
Transfer Learning of wav2vec 2.0 for Automatic Lyric Transcription
Longshen Ou*, Xiangming Gu*, Ye Wang
International Society for Music Information Retrieval Conference (ISMIR), Bengaluru, India, 2022.
pdf / code
Bayesian Deep Learning
Extrapolative Continuous-time Bayesian Neural Network for Fast Training-free Test-time Adaptation
Hengguan Huang†, Xiangming Gu, Hao Wang, Chang Xiao, Hongfu Liu, Ye Wang
Annual Conference on Neural Information Processing Systems (NeurIPS), New Orleans, USA, 2022.
pdf / code / video
Unsupervised Mismatch Localization in Cross-Modal Sequential Data with Application to Mispronunciations Localization
Wei Wei*, Hengguan Huang*, Xiangming Gu, Hao Wang, Ye Wang
Transactions on Machine Learning Research (TMLR), 2022.
pdf / code

Honors and Awards
Dean's Graduate Research Excellence Award, National University of Singapore, 2024
Research Incentive Award, National University of Singapore, 2023
Research Achievement Award, National University of Singapore, 2022
MM'22 Top Paper Award, Association for Computing Machinery, 2022
MM'22 Student Travel Grant, Association for Computing Machinery, 2022
President's Graduate Fellowship, National University of Singapore, 2021
Visiting Undergraduate Student Scholarship, Tsinghua University, 2020
Tsinghua's Friend- Zheng Geru Scholarship, Tsinghua University, 2018

Invited Talks or Posters
NUS Research Week Open House, invited talk on On the Interpretability and Safety of Generative Models , 2025
Global Young Scientists Summit, poster presentation on Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast, 2025

Academic Services
Conference reviewer for NeurIPS 2025/2024, ICML 2025, ICLR 2025, CVPR 2025, ICCV 2025/2023, ECCV 2024, ACL ARR 2025/2024, MM 2025/2024, IJCAI 2024, AISTATS 2025/2021
Workshop reviewer for (i) ICLR 2025 Workshop on Deep Generative Model in Machine Learning: Theory, Principle and Efficacy, (ii) NeurIPS 2024 Workshop on Attributing Model Behavior at Scale, (iii) NeurIPS 2024 Safe Generative AI Workshop
Journal reviewer for TOMM, TASLP, RA-L

Teaching
Teaching Assistant, CS4347/CS5647, Sound and Music Computing, Fall 2024
Teaching Assistant, CS6212, Topics in Media, Spring 2024
Teaching Assistant, CS5242, Neural Networks and Deep Learning, Spring 2023
Teaching Assistant, CS3244, Machine Learning, Fall 2022
Teaching Assistant, CS4243, Computer Vision and Pattern Recognition, Spring 2022


You've probably seen this website template before, thanks to Jon Barron.