Dr. Tianyu Jiang

Dr. Jiang’s research in natural language processing and large language models touches on several areas relevant to CHaT. He works on computational semantics — including metaphor and multi-word expressions — and has also published in machine translation. A substantial thread of his research examines emotion and affect in language, from embodied emotion recognition to affective event understanding.

More recently, he has turned to AI trustworthiness and interpretability, particularly how LLMs internally align (or fail to align) with human values. His group uses causal and geometric methods to probe model internals and study behaviors like sycophancy, dishonesty, and moral reasoning.

Personal site: jiangtianyu.com