Gao
Zeyu
2000–2025
yield per unit area
of the best model
on Chinese questions
under pressure
complete reversals
How is a conclusion actually formed?
Today, I am especially interested in a question that appears across very different fields: When a model gives us an answer, how do we know whether the evidence behind that answer is reliable?
From regression models predicting agricultural production to experiments testing whether an AI system changes its judgment under social pressure, I use data and experiments to examine not only what a model concludes , but also why it reaches that conclusion.
Two questions.
One underlying concern.
Whether examining agricultural production or artificial intelligence, my research focuses on the reliability of evidence, models and judgment.
What Drives
Grain Production?
A statistical study of Shandong Province using 26 years of data, examining how planting area, irrigation and natural disasters relate to grain output — and whether those relationships can reliably predict future production.
Will AI Change
Its Mind?
An experimental study of AI sycophancy in Chinese-language learning, testing whether an AI model abandons its original judgment when users apply answer, challenge, emotional or authority pressure.