Undergraduate · Research Intern · Incoming PhD Student

Hongcan Guo郭泓灿

Beijing University of Posts and Telecommunications / ByteDance Seed Vision / HKU CDS

“Begin with addition, end with subtraction. Do not multiply entities beyond necessity.”
“先做加法，再做减法。如无必需，无增实体。”

I am an undergraduate at the School of Artificial Intelligence, BUPT, and a research intern at ByteDance Seed Vision. I will join the School of Computing and Data Science, The University of Hong Kong, as an AI PhD student in September 2026, advised by Prof. Hengshuang Zhao. I study the mathematical foundations and learning dynamics of modern generative and reasoning models — with a soft spot for elegant formulations.

This site collects my research updates, publications, and occasional notes. Please feel free to reach out.

Interests Natural Language Processing Reinforcement Learning Generative Models Unsupervised Learning Mathematics

Portrait of Hongcan Guo

I.

News

2025/11/08 One paper accepted to AAAI 2026 Oral.
2025/09/19 One paper accepted to NeurIPS 2025 Oral.
2025/06/26 One paper accepted to ICCV 2025.
2025/06/09 One paper accepted to IEEE JSAC.
2025/03/22 One paper accepted to IEEE JSAC.

II.

Education

The University of Hong Kong

Incoming AI PhD Student, School of Computing and Data Science

Advisor: Prof. Hengshuang Zhao

2026.9～2030.4 (expected)

Beijing University of Posts and Telecommunications

B.Eng. in Artificial Intelligence

2022 — Present

III.

Experience

ByteDance Seed

Research Intern

Diffusion models for unified generation tasks.

2025.06 — Present

Institute of Artificial Intelligence, China Telecom (TeleAI)

Research Intern

Reinforcement learning for large language models.

2024.10 — 2025.06

IV.

Selected Publications

^* Equal contribution ^✉ Corresponding author

Cola DLM main figure

arXiv Continuous Latent Diffusion Language Model

Hongcan Guo, Qinyu Zhao, Yian Zhao, Shen Nie, Rui Zhu, Qiushan Guo, Feng Wang, Tao Yang, Hengshuang Zhao, Guoqiang Wei, Yan Zeng

Preprint, arXiv:2605.06548, 2026

Project Paper Code Model Blog Zhihu

NeurIPS 2025 MoE teaser

NeurIPS 2025 Advancing Expert Specialization for Better Mixture-of-Experts

Hongcan Guo^*, Haolang Lu^*, Guoshun Nan^✉, Bolun Chu, Jialin Zhuang, Yuan Yang, Wenhao Che, Sicong Leng, Qimei Cui, Xudong Jiang

Annual Conference on Neural Information Processing Systems (NeurIPS), 2025 Oral

PDF Code

AAAI 2026 Rolling Stone teaser

AAAI 2026 A Rolling Stone Gathers No Moss: Adaptive Policy Optimization for Stable Self-Evaluation in Large Multimodal Models

Wenkai Wang, Hongcan Guo, Zheqi Lv, Shengyu Zhang^✉

Annual AAAI Conference on Artificial Intelligence (AAAI), 2026 Oral

PDF Code

ICCV 2025 VideoMiner teaser

ICCV 2025 VideoMiner: Iteratively Grounding Key Frames of Hour-Long Videos via Tree-based Group Relative Policy Optimization

Xinye Cao^*, Hongcan Guo^*, Jiawen Qian^*, Guoshun Nan^✉, Chao Wang, Yuqi Pan, Tianhao Hou, Xiaojuan Wang, Yutong Gao

International Conference on Computer Vision (ICCV), 2025

PDF Code

All publications →

V.

Recent Notes

2026/05

Rethinking Text Modeling: What Cola DLM Actually Is Doing

All notes →

VI.

Visitors