ZHIQIN YANG
opening the window to the bay…
Zhiqin (Brian) Yang
PhD @ HKUST · Collaborative Learning · LLMs
Click a marker to glide · WASD to walk · drag to look around
Zhiqin Yang

Zhiqin (Brian) Yang

PhD @ HKUST · Clear Water Bay

Toward AGI — what machines must learn, & how they should learn it

Portrait of Zhiqin Yang

Zhiqin (Brian) Yang

PhD student at HKUST, Clear Water Bay.

LLM ReasoningAgentic Models Collaborative LearningTrustworthy ML

I am a PhD student at HKUST, supervised by Prof. Yike Guo and Prof. Wei Xue. I also work closely with Prof. Bo Han and Prof. Yonggang Zhang at the TMLR group, HKBU.

Prior to this, I earned my M.S. from Beihang University (BUAA), advised by Prof. Hao Peng, and completed my B.S. at Nanjing University of Science and Technology (NJUST), where I spent four enriching years.

My research revolves around two intertwined questions on the path toward AGI: what a machine needs to learn to get there, and how a machine should learn it. Everything I work on — LLM reasoning, agentic models, and collaborative learning — is organized around these two questions.

Two questions drive everything I do, both on the path toward AGI: what a machine needs to learn to get there, and how it should learn it. The directions below are how I pursue them.

LLM Reasoning & Alignment

Preference alignment (rethinking DPO vs. RLHF), reasoning-data selection for RL, and linguistic principles for foundation models.

DPO / RLHFRL Post-trainingReasoning

Agentic Models & Memory

Human-symbiotic agent networks and on-the-fly agent memory optimization.

AgentsMemoryGovernance

Collaborative Learning

Robust learning under data heterogeneity, label deficiency, and noisy clients — with applications in privacy-preserving healthcare.

HeterogeneityNoisy ClientsPrivacy

Latest updates, newest first.

    Framed on the walls by topic — click any paper for its abstract & links. Full list on Scholar ↗

    2025 – present

    Ph.D. · The Hong Kong University of Science and Technology

    Advisors: Prof. Yike Guo & Prof. Wei Xue · Clear Water Bay, Hong Kong

    – 2024

    M.S. · Beihang University (BUAA)

    Advisor: Prof. Hao Peng · Outstanding Master Thesis Award

    B.S.

    Nanjing University of Science and Technology (NJUST)

    Four years of undergraduate study

    Nov 2025 – Feb 2026

    Algorithm Intern · Tencent LightSpeed

    Shenzhen · supervised by Dr. Dong Fang

    Research on preference alignment — rethinking the equivalence between DPO and RLHF.

    Aug 2025 – Oct 2025

    Algorithm Intern · HKGAI

    Hong Kong · supervised by Dr. Yonggang Zhang

    LLM SFT on a downstream domain: QA-pair dataset construction & cleaning; supervised fine-tuning of Qwen3-Next-80B-A3B-Instruct with Megatron + LoRA; and evaluation of the fine-tuned model.

    Jun 2024 – Sept 2024

    Algorithm Intern · GigaAI

    Beijing · supervised by Dr. Zheng Zhu

    Research on video generation — planning mechanical-arm manipulation in a simulation environment in a generative manner (e.g., video generation via DiT or SVD).

    • 🎖️
      Outstanding Master Thesis Award · Beihang University · 2024

    Away from research —

    • 🏀 Basketball & a devoted fan of Kobe Bryant.
    • 🎧 Live music — hip-hop & R&B, with Nous Underground (XAC) a favorite.
    • 📖 Chinese history, especially the Ming Dynasty.

    Happy to talk about any of these — or potential collaborations.

      Walked through the study? Leave a note by the sea — no account needed.

        Please feel free to reach out about collaborations or shared passions.