Scholars have long employed a divisive lens to view science and humanities, with the former
rapidly developing to replace human labor while the latter ultimately oriented around human
interpretations.
Such division in perspective also manifests itself in the division of scholarly communities
that perceive themselves as having little to do with each other.
However, my experience researching at the intersection of the two revealed that this made us miss out on an enormous potential for the two domains to co-evolute and recursively augment each other. This motivates me to pursue an integrated approach to research that combines both.
I believe that it is the questions we ask, not the methods we employ, that can positively impact the world.
Publications
(* = Equal Contribution)
Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviors
Jing Huang*, Junyi Tao*, Thomas Icard, Diyi Yang, Christopher Potts
In International Conference on Machine Learning (ICML), 2025.
Inference and
Verbalization Functions During In-Context Learning
Junyi Tao*, Xiaoyin Chen*, Nelson F. Liu
In Findings of the ACL: EMNLP, 2024.
Personal
Besides reading and writing, I enjoy designing and playing the ancient Chinese guqin (古琴) and zither (古筝) in the Stanford Baipu Chinese Music Ensemble.