Xinyi Chen

Human intelligence relies on collaboration, where individuals combine diverse expertise to achieve outcomes that exceed individual capabilities. Inspired by this, my research focuses on improving the efficiency of large language models (LLMs) through multi-agent collaboration. The goal is to integrate the complementary knowledge of specialized LLMs and optimize their interactions to address complex tasks. This work targets two key challenges: (1) effectively integrating diverse, distributed knowledge across agents and (2) designing resource-efficient strategies to enhance cooperative performance. By developing frameworks modeled on human-inspired collaboration, this research aims to create scalable, data-efficient methods to advance LLM capabilities in solving complex problems.