Dake Bu G X

卜大可・ウラナイタイカ

Phd Candidate, Department of Computer Science, City University of Hong Kong.

Hi! I’m currently a PhD candidate at Optima Group, Department of Computer Science, City University of Hong Kong, advised by Prof. Hau-San Wong and Prof. Qingfu Zhang since 2023 fall. I’m also fortunate to work closely with Prof. Wei Huang and Prof. Andi Han. Before that, I completed my B.S in 2023 in Mathematics at Xi’an Jiaotong University, advised by Prof. Hui Li and mentored by Prof. Jian Sun.

I’m currently an one-year intern at CFAR, A*STAR, where I’m supervised by Prof. Atsushi Nitanda. Prior to this, I spent one year (2024–2025) as a research intern at the University of Tokyo, working at the Deep Learning Theory Team at RIKEN AIP under the supervision of Prof. Taiji Suzuki. From May to June, 2026, I visited the School of Mathematics and Statistics at the University of Sydney as a visiting student, supervised by Prof. Andi Han. Earlier, from January to May 2023, I was a research assistant in the LOGO Lab at The Chinese University of Hong Kong, advised by Prof. Tianshu Yu.

My research broadly covers theoretical foundations of deep learning, distribution optimization, mean-field optimization, sampling, reinforcement learning and their applications.

News

2026-06 - Our github repo Auto-Lean-in-Sleep: Block Encoding for Quantum Computing is available, along with user-facing website with API (Under Constuction). The arXiv version is comming soon. Any feedback is welcome!
2026-06 - Our paper on A Plug-in Doob h transform-induced Token-Ordering Module for Diffusion Language Models ([code]) is selected as Oral, and theoretical explanation and solution to exploration dilemma in LLM post-training ([code]) is rewarded as spotlight in ICML 2026 Fogen.
2026-05 to 2026-06 - Visited the School of Mathematics and Statistics at the University of Sydney as a visiting student, supervised by Prof. Andi Han.
2026-05 - Gave a talk titled “Langevin Dynamics with Partial Structure: From Guided Generative Sampling to Mean-Field Feature Learning” at the University of Sydney.
2026-05 - Presented A Plug-in Doob h transform-induced Token-Ordering Module for Diffusion Language Models ([code]) at the AIVP-Joint workshop between A*STAR, RIKEN AIP, and NTU at Nanyang Technological University.
2026-05 - Honor to win the Gold Reviewer Award in ICML 2026.
2026-05 - One paper on Provable benefit of transformer curriculum post-training ([code]) accepted to ICML 2026 ([code]).
2026-04 - Two papers on Slowly Annealed Langevin Dynamics ([code]) and A Plug-in Doob h transform-induced Token-Ordering Module for Diffusion Language Models ([code]) are available online.
2025-11 - One paper on Provable benefit of transformer curriculum post-training is available on arXiv.
2025-01 — Two papers on theoretical foundation of task vector in In-Context Learning, and Multi-objective Reinforcement Learning with Lexicographic Rewards are accepted to ICML 2025.
2024-09 — One paper on In context learning with multi-concept word semantics is accepted to NeurIPS 2024.
2024-01 — One paper on theoretical foundation of neural active learning accepted to ICML 2025.

Education

Oct. 2023 – Present — PhD Candidate, Department of Computer Science, City University of Hong Kong (CityUHK).
Aug. 2019 – Jun. 2023 — School of Mathematics and Statistics, Xi’an Jiaotong University (XJTU).

Work Experience

Dec. 2025 - Present, - Research Intern, CFAR A*STAR
Dec. 2024 - Oct. 2023 - Research Intern, RIKEN AIP
Jan. 2023- May. 2023 - Research Assistant, The Chinese University of Hong Kong, Shenzhen

Service

Area Chair: ICML 2026 FOGEN; ICLR 2026 DeLTa.
Reviewer: ICML (2024; 2025; 2026); NeurIPS (2024; 2025; 2026); ICLR (2025; 2026).

Contact

Affiliation: Department of Computer Science, City University of Hong Kong
Location: Singapore; Hong Kong; Shenzhen
Email: <dakebu2-c[at]my.cityu[dot]edu[dot]hk>