Ge Li
Boya Distinguished Professor at Peking University
Ge Li, Boya Distinguished Professor at Peking University, tenured professor at the School of Computer Science, and jointly appointed professor at the Institute for Artificial Intelligence. He is also a Changjiang Scholar under the Ministry of Education and serves as the deputy director of the Software Engineering Technical Committee of the China Computer Federation (CCF). He was an early pioneer in the international research on “deep learning–based program understanding and generation,” with many of his works recognized by scholars worldwide as “original contributions” and widely cited. His research team has remained active in the field of “large model–based software development technologies,” representing an internationally influential research force. The aiXcoder project, a successful transformation of his research, serves major aerospace projects, numerous large enterprises in the finance and IT sectors, and millions of developers.
Topic
Breakthroughs in Large Language Model Capabilities for Programmatic Logic Reasoning
Here’s a fluent and professional English translation of your abstract and outline: --- Large language models (LLMs) have already demonstrated significant value in assisting software development. However, their limitations in state-based reasoning and generalization are becoming increasingly apparent. These fundamental shortcomings are especially visible in tasks such as program generation, program understanding, and other areas of software engineering. This talk approaches the issue from the perspective of logical reasoning capabilities, exploring some of the core challenges and potential research advances for LLMs. The goal is to drive further improvements in LLM capabilities and advance intelligent software engineering powered by LLMs. **Outline:** 1. Discussion of fundamental issues in current large language model architectures 2. Key characteristics we expect from future large language models 3. Perspectives on the future development of large language models 4. Perspectives on intelligent software development
