活动预告 | CodeWisdom软件智能化开发与运维学术系列报告 第15期：利用大语言模型进行可靠的代码转换和软件规范生成...

Danning Xie

PhD candidate in the Computer Science Department at Purdue University‍

Summary

Title

Harnessing Large Language Models for Reliable Code Transformation and Software Specification Generation

Abstract

Large Language Models (LLMs) have shown great promise in software engineering (SE) tasks, leveraging their ability to understand and generate code. However, their inherent limitations—such as hallucinations, token constraints, and reduced performance with long inputs—present challenges for tasks requiring a comprehensive understanding of the codebase or document. In this presentation, I will introduce recent work that harnesses LLMs for reliable code transformation and software specification generation. Specifically, we leverage a hybrid approach combining LLMs with program analysis to address the critical challenge of recovering symbol information in decompiled code, including both names and types for local variables and user-defined data structures. In addition, I will discuss our broader efforts to integrate LLMs with traditional techniques to enhance reliability and soundness in software specification generation, showcasing how this hybrid approach bridges the gap between LLM capabilities and the demands of complex software engineering challenges.

Speaker

Danning Xie is a PhD candidate in the Computer Science Department at Purdue University, where she is advised by Professor Lin Tan and Professor Xiangyu Zhang. Her research interests lie in the fields of software engineering, software testing, and software analysis. Recently, her work has centered on harnessing the power of Large Language Models (LLMs) for code-related tasks.

Schedule

❖

时间：

2024年12月27日 周五上午 10:30  

❖

腾讯会议：

会议号：978 883 396

密码：123456

❖

地点：

复旦大学江湾校区 2号交叉学科楼A2003