Mengzhou Xia

A study examines the fragile safety mechanisms in language models and proposes improvements.

Language models aim to improve science learning by providing personalized assistance.

Exploring how benign data can unintentionally produce harmful outputs in language models.