step out.
Your little guide
Welcome Explore the research trail.

Latest news

01 / RESEARCH VISION

Research, connected.

RESEARCH GOAL

Interactive World Models

AI4Game
01 / Environments

OmniDrive

Multi-view driving world modeling

 02 / Spoken dialogue

C3

Evaluating complex conversations

 03 / Multi-party interaction

Duplex-MPE

Full-duplex group dialogue

 04 / Digital embodiment

Motion-Omni

Joint speech & full-body motion

02 / SELECTED WORK

Selected Publications.

Full paper list on Google Scholar

03 / OPEN SOURCE

Projects.

Resources for the
research community.

A RESEARCH RESOURCE

Multimodal Model Zoo

A structured, searchable collection of 114+ multimodal large language models. Explore architectures, tasks, benchmarks, papers, and code across autoregressive, diffusion, and hybrid approaches.

From vision–language and audio–language models to autonomous driving and 3D scene understanding.

ImageTextVideoSpeech3D
Explore GitHub

04 / THE JOURNEY

Experience.

Research & education.

Peking University

PhD student

EDUCATION

Seed, ByteDance

Research Intern

Contributed to foundation models for spoken dialogue.

INTERN

LightSpeed, Tencent

Research Intern

Developed C3 and Motion-Omni during an industry research internship to support AI teammates and spoken-dialogue assistants in the game Peacekeeper Elite.

INTERN

Xiamen University

Bachelor's degree

Research Intern at MAC Lab.

EDUCATION

05 / RECOGNITION

Honors &
awards.

Academic honors

  • Outstanding Graduate

    Xiamen University

    2025

  • Merit Student × 4

    Xiamen University

    2021–25

  • Golden Kapok Medal

    Xiamen University · 1 out of 1000

    2024

Competitions

  • Huawei ICT Competition

    1st place, China Area Top 0.1%

    2024

  • COMAP's MCM & ICM

    Meritorious Winner Top 5%

    2023

  • China Collegiate Programming Contest

    CCPC · National Silver Medal

    2023
Flag Counter
Artwork credits

RobotExpressive by Quaternius · Tomás Laulhé, adapted by Don McCurdy, under CC0. Blue-and-ivory materials and animation direction adapted for this site. Nature models by Quaternius and Kenney, under CC0. Rendered with Three.js (MIT). Model sources and licenses. Garden architecture, terrain and camera routes are original to this site.