Tuan-Anh Bui

Me and two boys

I am Research Fellow at the Department of Data Science and AI, Monash University. My research interest lies in the intersection between Generative AI and Trustworthy Machine Learning. For example, my research focuses on how to ensure that models like ChatGPT do not respond to harmful queries asking to create a bomb, or that models like Stable Diffusion do not generate sexual images. I got my Ph.D. from Monash University in November 2023, under the supervision of Prof. Dinh Phung and Dr. Trung Le. My thesis can be found here.

Before that, I had one year working as Research Engineer at Credit AI Lab, Trusting Social, and two years working as Research Assistant at Singapore University of Technology and Design with Prof. Ngai Man Cheung and Dr. Trung Tran.

news

Mar 21, 2025	It was a great pleasure for me to present our works about Unlearning Concepts to the Machine Learning team at Canva. The slides are available here. I’m glad to see many interesting and practical/industry-related questions from the audience and see how our research can be applied to their real-world problems.
Feb 28, 2025	I’m excited to share that I am officially a Chief Investigator of the Trustworthy Generative AI: Towards Safe and Aligned Foundation Models project, funded by the Department of Defence, Australia with an $800K AUD grant. The project focuses on four key areas of modern foundation models: Certification - Alignment - Multimodality - Personalization, where I am leading the Personalization stream. Our goal is to push the boundaries of safe and aligned generative AI, ensuring its responsible deployment in real-world applications. The project is led by Professor Dinh Phung and co-led by a team of experts from the Faculty of IT, Monash University, where I am honored to be part of.
Feb 27, 2025	I’m excited to share that our paper “Preserving Clusters in Prompt Learning for Unsupervised Domain Adaptation” (led by Long Vuong) has been accepted to CVPR 2025! While CLIP-based methods for Unsupervised Domain Adaptation (UDA) have shown promise, they face limitations in target domain generalization due to embedding distribution shifts. In this paper, we propose a novel approach that exploits the geometric relationships between visual and text embeddings through optimal transport theory. By leveraging clustering behavior in multi-modal embeddings and reference predictions from source prompts, our method achieves superior performance in target-prompt learning and representation quality.
Jan 23, 2025	Hooray! I’m thrilled to finally share that our work has been accepted to ICLR 2025! This is more than just an acceptance—I’m truly proud that all reviewers recognized and appreciated the originality and creativity of our approach to concept unlearning, with a clear motivation and comprehensive experiments. The paper can be found here
Oct 4, 2024	Excited to share another paper that I am very proud of. This paper is an extension of our NeurIPS 2024 paper, where we dive deeper into the impact of erasing one concept to the others, but this time, we focus on the choice of target concepts. The paper can be found here. Our paper’s name was inspired by the movie “Fantastic Beasts and Where to Find Them”. Hopefully, the reviewers enjoy it as much as the movie .

latest posts

Mar 15, 2025	MS-Diffusion - Multi-subject Zero-shot Image Personalization with Layout Guidance (ICLR 2025)
Mar 14, 2025	Unlearning LLMs
Mar 8, 2025	Foundation of Diffusion Models

selected publications

2025

Fantastic Targets for Concept Erasure in Diffusion Models and Where to Find Them

Tuan-Anh Bui, Vu Trang, Vuong Long, and 4 more authors

International Conference on Learning Representations (ICLR), 2025

Bib PDF Code Slides

@article{bui2025adaptive,
  title = {Fantastic Targets for Concept Erasure in Diffusion Models and Where to Find Them},
  author = {Bui, Tuan-Anh and Trang, Vu and Long, Vuong and Le, Trung and Montague, Paul and Abraham, Tamas and Phung, Dinh},
  journal = {International Conference on Learning Representations (ICLR)},
  year = {2025},
  url = {https://arxiv.org/abs/2501.18950},
}

Hiding and Recovering Knowledge in Text-to-Image Diffusion Models via Learnable Prompts

Tuan-Anh Bui, Khanh Doan, Trung Le, and 3 more authors

ICLR 2025 DeLTa Workshop, 2025

Bib PDF Code Slides

@article{bui2024removing,
  title = {Hiding and Recovering Knowledge in Text-to-Image Diffusion Models via Learnable Prompts},
  author = {Bui, Tuan-Anh and Doan, Khanh and Le, Trung and Montague, Paul and Abraham, Tamas and Phung, Dinh},
  journal = {ICLR 2025 DeLTa Workshop},
  year = {2025},
  url = {https://arxiv.org/abs/2403.12326},
}

2024

Erasing Undesirable Concepts in Diffusion Models with Adversarial Preservation

Tuan-Anh Bui, Vuong Long, Khanh Doan, and 4 more authors

Advances in Neural Information Processing Systems (NeurIPS), 2024

Bib PDF Code Slides

@article{bui2024adversarial,
  title = {Erasing Undesirable Concepts in Diffusion Models with Adversarial Preservation},
  author = {Bui, Tuan-Anh and Long, Vuong and Doan, Khanh and Le, Trung and Montague, Paul and Abraham, Tamas and Phung, Dinh},
  journal = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2024},
  url = {https://tuananhbui89.github.io/projects/adversarial-preservation},
}

Diversity-Aware Agnostic Ensemble of Sharpness Minimizers

Tuan-Anh Bui*, Vy Vo*, Tung Pham, and 2 more authors

Preprint, 2024

Bib PDF Code

@article{bui2024diversity,
  title = {Diversity-Aware Agnostic Ensemble of Sharpness Minimizers},
  author = {Bui${*}$, Tuan-Anh and Vo${*}$, Vy and Pham, Tung and Phung, Dinh and Le, Trung},
  journal = {Preprint},
  year = {2024},
  url = {https://arxiv.org/abs/2403.13204},
}

2023

Optimal transport model distributional robustness

Van-Anh Nguyen, Trung Le, Tuan-Anh Bui, and 2 more authors

Advances in Neural Information Processing Systems (NeurIPS), 2023

Bib PDF Code

@article{nguyen2024optimal,
  title = {Optimal transport model distributional robustness},
  author = {Nguyen, Van-Anh and Le, Trung and Bui, Tuan-Anh and Do, Thanh-Toan and Phung, Dinh},
  journal = {Advances in Neural Information Processing Systems (NeurIPS)},
  volume = {36},
  year = {2023},
  url = {https://proceedings.neurips.cc/paper_files/paper/2023/hash/4b91825aec2ed35150f1d3e8fb195556-Abstract-Conference.html},
}

Generating Adversarial Examples with Task Oriented Multi-Objective Optimization

Tuan-Anh Bui, Trung Le, He Zhao, and 3 more authors

Transactions on Machine Learning Research (TMLR), 2023

Bib PDF Code

@article{bui2023generating,
  title = {Generating Adversarial Examples with Task Oriented Multi-Objective Optimization},
  author = {Bui, Tuan-Anh and Le, Trung and Zhao, He and Tran, Quan and Montague, Paul and Phung, Dinh},
  journal = {Transactions on Machine Learning Research (TMLR)},
  issn = {2835-8856},
  year = {2023},
  url = {https://openreview.net/forum?id=2f81Q622ww},
}

2022

A Unified Wasserstein Distributional Robustness Framework for Adversarial Training

Tuan-Anh Bui, Trung Le, Quan Tran, and 2 more authors

In International Conference on Learning Representations (ICLR), 2022

Bib PDF Code

@inproceedings{bui2021unified,
  title = {A Unified Wasserstein Distributional Robustness Framework for Adversarial Training},
  author = {Bui, Tuan-Anh and Le, Trung and Tran, Quan and Zhao, He and Phung, Dinh},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2022},
}

2021

Understanding and achieving efficient robustness with adversarial supervised contrastive learning

Tuan-Anh Bui, Trung Le, He Zhao, and 3 more authors

arXiv preprint arXiv:2101.10027, 2021

Bib PDF Code

@article{bui2021understanding,
  title = {Understanding and achieving efficient robustness with adversarial supervised contrastive learning},
  author = {Bui, Tuan-Anh and Le, Trung and Zhao, He and Montague, Paul and Camtepe, Seyit and Phung, Dinh},
  journal = {arXiv preprint arXiv:2101.10027},
  year = {2021},
}

2020

Improving adversarial robustness by enforcing local and global compactness

Tuan-Anh Bui, Trung Le, He Zhao, and 4 more authors

In European Conference on Computer Vision (ECCV), 2020

Bib PDF Code Poster Slides

@inproceedings{bui2020improving,
  title = {Improving adversarial robustness by enforcing local and global compactness},
  author = {Bui, Tuan-Anh and Le, Trung and Zhao, He and Montague, Paul and deVel, Olivier and Abraham, Tamas and Phung, Dinh},
  booktitle = {European Conference on Computer Vision (ECCV)},
  pages = {209--223},
  year = {2020},
  organization = {Springer},
}

2019

Improving GAN with neighbors embedding and gradient matching

Ngoc-Trung Tran*, Tuan-Anh Bui*, and Ngai-Man Cheung

In Proceedings of the AAAI conference on artificial intelligence (AAAI), 2019

Bib PDF Code

@inproceedings{tran2019improving,
  title = {Improving GAN with neighbors embedding and gradient matching},
  author = {Tran${*}$, Ngoc-Trung and Bui${*}$, Tuan-Anh and Cheung, Ngai-Man},
  booktitle = {Proceedings of the AAAI conference on artificial intelligence (AAAI)},
  volume = {33},
  number = {01},
  pages = {5191--5198},
  year = {2019},
}

2018

Dist-gan: An improved gan using distance constraints

Ngoc-Trung Tran, Tuan-Anh Bui, and Ngai-Man Cheung

In Proceedings of the European conference on computer vision (ECCV), 2018

Bib PDF Code

@inproceedings{tran2018dist,
  title = {Dist-gan: An improved gan using distance constraints},
  author = {Tran, Ngoc-Trung and Bui, Tuan-Anh and Cheung, Ngai-Man},
  booktitle = {Proceedings of the European conference on computer vision (ECCV)},
  pages = {370--385},
  year = {2018},
}