报告题目(Title)： Testing Large Language Models with Natural Language Requirements

时间(Date & Time)：2023.12.19   1-2pm

地点(Location)：理科一号楼1621（燕园校区）Room 1621, Science Building #1 (Yanyuan)

主讲人(Speaker)：Xueqing Liu

邀请人(Host)：谢涛

报告摘要(Abstract)：

In this talk, I will introduce our recent work on capability tests for language models. Capability tests allow model developers to test language models’ failures by specifying their requirements as natural language specifications. First, I will introduce TestAug, a tool we developed for automatically generating capability-based tests using the GPT models; second, I will introduce our recent work on designing capability tests for detecting the failures of hate speech detection models in conforming to content policies.

主讲人简介(Bio)：

Xueqing Liu is an Assistant Professor at the Department of Computer Science at Stevens Institute of Technology. Her research interests are text mining, natural language processing, and their applications in software engineering and security. Her work is published in EMNLP, SIGIR, KDD, WWW, TKDE, and software engineering conferences such as RE and VL/HCC.

欢迎关注计算机学院微信公众号，了解更多讲座信息！

北京大学计算机学院