CAISA Lab

CAISA group welcomes Tianyi Zhang

Tianyi Zhang from University of Pennsylvania is visiting CAISA group in the next three months. A brief introduction of Tianyi: “I am passionate about building intelligent agents that emulate human understanding and reasoning of world events. In contrast to human learning, which assimilates and accommodates information into brain schemas, a significant challenge with current Language Models (LMs), including the SOTA GPT-4, is their inability to automatically acquire and anchor structured knowledge in the network. This deficiency leads to unreliable reasoning and hallucinations. To alleviate it, my research directs LMs to construct and reason with structured and symbolic representations. These efforts include event extraction, schema induction, entity-state tracking, natural to symbolic language translation and reasoning.” More about Tianyi.