Branch-and-Browse: Efficient and Controllable Web Exploration with Tree-Structured Reasoning and Action Memory

TL;DR

Branch-and-Browse is a novel web agent framework that combines structured reasoning, memory, and efficient execution to improve multi-step web exploration tasks with higher success rates and reduced time.

Contribution

It introduces a tree-structured reasoning and action memory framework for controllable and efficient web exploration by LLM-based agents.

Findings

Achieves 35.8% success rate on WebArena benchmark.

Reduces execution time by up to 40.4% compared to state-of-the-art.

Demonstrates improved reasoning depth and efficiency in web tasks.

Abstract

Autonomous web agents powered by large language models (LLMs) show strong potential for performing goal-oriented tasks such as information retrieval, report generation, and online transactions. These agents mark a key step toward practical embodied reasoning in open web environments. However, existing approaches remain limited in reasoning depth and efficiency: vanilla linear methods fail at multi-step reasoning and lack effective backtracking, while other search strategies are coarse-grained and computationally costly. We introduce Branch-and-Browse, a fine-grained web agent framework that unifies structured reasoning-acting, contextual memory, and efficient execution. It (i) employs explicit subtask management with tree-structured exploration for controllable multi-branch reasoning, (ii) bootstraps exploration through efficient web state replay with background reasoning, and (iii) leverages a page action memory to share explored actions within and across sessions. On the WebArena benchmark, Branch-and-Browse achieves a task success rate of 35.8\% and reduces execution time by up to 40.4\% relative to state-of-the-art methods. These results demonstrate that Branch-and-Browse is a reliable and efficient framework for LLM-based web agents.