Machine learning · Cognition · Human–AI interaction

Shaoyang Zhang.

张劭扬Also known as Onion Green

I study how AI systems work, how they fail, and how they can better understand the people they serve.

I’m a computer science undergraduate at the University of Michigan, with a background in mechanical engineering at Shanghai Jiao Tong University. My research connects model mechanisms, world-model safety, and human–AI interaction.

Shaoyang Zhang

Ann Arbor, Michigan

Exploring PhD opportunities
Fall 2027

01 / SELECTED WORK

Research & publications

All work ↗
Normal and attacked pipelines: input image, imagined video, and downstream IDM action
Conceptual problem overview
Ongoing· Project lead · World-model safety

Targeted attacks on world models

I lead research on world-model hijacking: targeted adversarial attacks that aim to steer model behavior toward an attacker-chosen objective.

Methods and results will accompany a public release.

Standard AttnResidual compared with OASIS
OASIS, Figure 1 · Enlarge ↗
NeurIPS 2026· COLM 2026 Efficient Reasoning Workshop · Spotlight

Attention Sinks and Outliers in Attention Residuals

Haozheng Luo*, Haoran Dai*, Jingyuan Huang*, Shaoyang Zhang, Xi Chen, Eric Hanchen Jiang, Yijiang Li, Chenghao Qiu, Chenwei Xu, Zhenyu Pan, Haotian Zhang, Binghui Wang, and Yan Chen

Understanding how attention sinks and activation outliers interact with cross-layer routing, and using null-aware attention to improve model robustness.

02 / RESEARCH

From mechanisms to people

Research & projects ↗
01 / MECHANISMS

What happens
inside the model?

I study attention, activation structure, and internal representations to understand how model behavior arises.

02 / RELIABILITY

Can we trust
what it predicts?

I study world-model hijacking and targeted adversarial threats to model behavior.

03 / PEOPLE

Does it understand
the person?

I build and study user simulation systems that account for evolving knowledge and multi-turn interaction.

03 / NOTEBOOK

Questions I’m thinking about

All notes ↗
04 / A LITTLE MORE

Engineering, with
a human question.

My path has moved from mechanical engineering and robotics to language models and human–AI interaction. I’m drawn to problems where understanding a system’s internal mechanisms changes how we evaluate and use it.

Beyond research, I follow Vocaloid music, especially Hatsune Miku. The connection between technology, expression, and human meaning is a lasting source of curiosity.

More about me ↗