English · 中文
I study how machines build internal models of the world —
and what it means for a model to understand rather than merely predict.
An undergraduate at USTC, currently doing research in world models, based in Beijing.
Interests revolve around representation learning, embodied intelligence, and AI agent systems.
Tinkering with personal AI infrastructure: Assist Agent, self-hosted toolchains, multi-platform development environments, and public web surfaces under luonto.me.
Music — ヨルシカ, あたらよ, and Midwest Emo have accompanied most of the thinking here; also plenty of instrumental.
Sports — badminton (haven't played in a while); hiking (a 3-day, 2-night field expedition to Mount Qiyun with the university's field-research club).
Photography — landscape, street.
Anime — 少女終末旅行, のんのんびより; 響け!ユーフォニアム, けいおん!, ぼっち・ざ・ろっく!; うさぎドロップ, スーパーカブ, ゆるキャン△; 罗小黑战记, 葬送のフリーレン.
|
World Models How to build representations of Action in the image domain / construct World Model architectures where Action and Video mutually reinforce each other — aiming to raise success rates on embodied-intelligence tasks. |
AI Agent AI Agent Benchmark construction: probing the capability frontier of agents. Computer / Mobile Using Tools: building agents' tool-use capabilities. |
Robotics RoboMaster 2025/2026 Robowalker team member. |
Beijing · yurunjin@mail.ustc.edu.cn