Jiepeng Jin [金杰鹏]

Jiepeng Jin [金杰鹏]

Incoming Direct Ph.D. student @ ZJU CS

đź‘‹ Hi there! I am Jiepeng Jin

I am a fourth-year undergraduate student at Zhejiang University (ZJU), majoring in Computer Science & Technology and minoring in ACEE (Advanced Class of Engineering Education) at the Chu Kochen Honors College.

Starting from Fall 2026, I will continue my journey as a Direct Ph.D. student at ZJU CS.

Currently, I am taking my first steps into the world of First-Principles AI. While I am still refining my specific research direction, I am particularly interested in building autonomous and grounded intelligence from the ground up.

The “Restart”

For a long time, I was following a predefined “correct track,” easily securing high scores through imitation. However, a deep disillusionment gradually grew within me: I began to feel that much of contemporary AI research can resemble “playing with LEGO blocks”—iteratively patching models and chasing benchmarks without sufficiently understanding the underlying mechanisms.

Recently, I decided to press the “Restart” button. I am no longer content with being an imitator or a passive navigator. To put this shift into practice, I am currently working through Stanford University’s CS336 (Language Modeling from Scratch) to better understand how complex systems are built from first principles. Inspired by the research philosophy of Yang Xu, I aim not only to interpret existing systems, but to construct new ones.

Research Taste: What I Avoid

I have developed a clearer “compass” for the type of research I tend to avoid:

  • Unscalable Toy-Model Analysis: I am cautious about post-hoc analyses on small models that lack the capacity to scale up. In the era of scaling laws, conclusions drawn from toy models may not reliably generalize to the frontier of large-scale systems.
  • Agentic “Harness” Wrapping: I am less interested in approaches that primarily wrap black-box models with tool-calling loops. While sometimes useful, this style of work can risk creating an illusion of progress without advancing core understanding. I am more interested in building the underlying systems themselves—the engine rather than the wrapper.

Mission & Philosophy

  • Evolution through Correction: “The earlier you correct a mistake, the better. No matter how high the cost seems today, it is the lowest price you will ever pay.”
  • Beyond Work, Towards Value: I aim to pursue long-term, intrinsically meaningful directions and create value through original and sustained contributions.
  • Win-Win Leadership: I value growing together with others—building meaningful collaborations and achieving shared progress.
  • Stay Hungry: “When your dreams are big enough, you remain humble and hungry.”

Let’s Connect

I’m always open to interesting conversations, potential collaborations, or simply connecting with new people.

News

  • [2026.04] Initiated my “Restart” and launched this academic homepage
  • [2025.12] Admitted to the Direct Ph.D. program at ZJU CS

Education

  • Ph.D. in Computer Science, Zhejiang University (ZJU CS), 2026 - 2031 (Expected)
  • B.E. in Computer Science & Technology, Zhejiang University (ZJU CS), 2022 - 2026 (Expected)
    • Minor in ACEE, Chu Kochen Honors College