ABOUT / YIDAN HUANG关于 / 黄一丹

The person behind the questions.那些问题背后,是我。

Researcher, engineer, and a believer that a model's behavior should be traceable to the data and decisions that shaped it.研究者、工程师。我相信模型的行为应该能追溯到塑造它的数据与决策。

BEIJING, CHINACASIA / UCASJD EXPLORE
Yidan Huang outdoorsPORTRAIT / 2026

HELLO, I'M YIDAN你好,我是黄一丹

I care about the parts of AI systems that are easy to overlook.我关心 AI 系统里最容易被忽略的部分。

I'm pursuing a master's degree in Computer Application Technology at the Institute of Automation, Chinese Academy of Sciences, through UCAS. I also work as an LLM algorithm intern at JD Explore Academy.我在中国科学院大学攻读计算机应用技术硕士,培养单位为中国科学院自动化研究所;同时在京东探索研究院担任大模型算法实习生。

My research spans efficient reasoning, on-policy distillation, generative and multimodal models, and long-horizon agents. I study when reasoning can be compressed, how verification should shape a teacher target, whether models truly read external features, and how agents retain evidence and act at the right moment. The common thread is information flow: what a model receives, what it preserves, and what finally shapes its behavior.我的研究涵盖高效推理、在线蒸馏、生成与多模态模型,以及长程智能体。我研究推理何时可以压缩、验证结果如何约束教师目标、模型能否读懂外部特征,以及智能体如何保留证据并在恰当时刻行动。贯穿其中的是信息流:模型接收什么、保留什么,最终又是什么塑造了它的行为。

Before that, I studied Information and Computing Science at Huazhong Agricultural University, graduating first in my cohort. Applied work in computer vision and scientific forecasting taught me to start with the real operating conditions, not only the model architecture.本科就读于华中农业大学信息与计算科学专业,以专业第一毕业。早期在计算机视觉与科学预测上的应用工作,让我习惯先看真实使用条件,再谈模型结构。

See my CV查看简历 ↗Email me给我写信 ↗

THE PATH SO FAR走过的路

An evolving research direction.逐渐清晰的研究方向。

2021—2025

A mathematical foundation打好数学与计算基础

B.S. in Information and Computing Science, Huazhong Agricultural University. GPA 3.92/4.00, rank 1/102, twice awarded the National Scholarship.华中农业大学信息与计算科学本科。GPA 3.92/4.00,专业排名 1/102,两次获得国家奖学金。

2025—

From perception to agents从感知走向智能体

M.S. at CASIA / UCAS, studying foundation models, multimodal systems, and agents that operate over long horizons.在中国科学院自动化研究所 / 中国科学院大学攻读硕士,研究基座模型、多模态系统与长程智能体。

2026—

Research at model scale走进大模型训练现场

LLM algorithm intern at JD Explore Academy, working on post-training systems and search-agent research.在京东探索研究院担任大模型算法实习生,从事后训练系统与搜索智能体研究。