Branch-and-Browse: Efficient and Controllable Web Exploration with Tree-Structured Reasoning and Action Memory
Researchers have introduced Branch-and-Browse, a new framework designed to enhance the capabilities of large language model (LLM)-powered web agents. This framework addresses limitations in reasoning depth and efficiency found in current approaches by employing a tree-structured exploration method for multi-branch reasoning and incorporating contextual memory. Branch-and-Browse also features efficient web state replay and a page action memory to share explored actions across sessions, leading to improved performance on benchmarks. AI
IMPACT Enhances LLM web agent efficiency and controllability for complex tasks.