报告名称：Consistency is All We Need

报告时间：5月9日9:00

报告地点：滴水湖国际软件学院中庭

报告摘要:

Defining "correctness" has long been a central concern in the application of formal methods. The emergence of large language models (LLMs) further complicates traditional notions of correctness, particularly in domains where formal definitions are inherently unattainable. In this talk, I will argue that in many contexts—especially those involving LLMs—consistency, rather than correctness, should serve as the primary standard. Through two case studies—one involving the application of LLMs to a software engineering task and another addressing an AI safety challenge—I will demonstrate how prioritizing consistency enables practical and effective solutions in settings where correctness is difficult or impossible to define.

报告人简介:

孙军(SUN Jun)是新加坡管理大学(SMU)的教授。他于2002年和2006年分别获得新加坡国立大学(NUS)计算机科学学士和博士学位，自2010年起担任教职。孙军的研究兴趣包括人工智能安全、软件工程、形式化方法等等。他喜欢设计算法，以解决具有实际应用的难题，也热爱生活。孙军2次获得李光耀学者称号。发表了100+篇CCF-A类期刊和会议论文，多次获得过ACM杰出论文奖。其参与研发的PAT模型检查器荣获“ICFEM 20年最具影响力系统奖”。更多信息请见: https://sunjun.site