AutoWebGLM: A Large Language Model-based Web Navigating Agent

Hanyu Lai , Xiao Liu , Iat Long Iong , Shuntian Yao , Yuxuan Chen , Pengbo Shen , Hao Yu , Hanchen Zhang , Xiaohan Zhang , Yuxiao Dong , Jie Tang

🏛 Institutions: Tsinghua , Zhipu , Beijing University of Posts and Telecommunications , University of Chinese Academy of Sciences
📅 Date: April 4, 2024
📑 Publisher: KDD 2024
💻 Env: Web
🔑 Keywords: model benchmark AutoWebBench HTML simplification curriculum training reinforcement learning

TLDR

AutoWebGLM is a web-navigation agent built on ChatGLM3-6B that combines HTML simplification, hybrid human-AI trajectory construction, and reinforcement learning with rejection sampling. The paper also introduces the bilingual AutoWebBench benchmark for real-world web navigation and uses it together with other benchmarks to evaluate the system.

Open paper Code Report issue