STRUCTUREDAGENT: Planning with AND/OR Trees for Long-Horizon Web Tasks

TL;DR

STRUCTUREDAGENT introduces a hierarchical planning framework with AND/OR trees and structured memory, significantly enhancing long-horizon web task performance by improving planning, memory, and interpretability.

Contribution

It presents a novel hierarchical planning approach with dynamic AND/OR trees and structured memory, addressing memory and planning limitations of existing web agents.

Findings

Improves performance on long-horizon web tasks

Enables interpretable hierarchical plans

Outperforms standard LLM-based agents on benchmarks

Abstract

Recent advances in large language models (LLMs) have enabled agentic systems for sequential decision-making. Such agents must perceive their environment, reason across multiple time steps, and take actions that optimize long-term objectives. However, existing web agents struggle on complex, long-horizon tasks due to limited in-context memory for tracking history, weak planning abilities, and greedy behaviors that lead to premature termination. To address these challenges, we propose STRUCTUREDAGENT, a hierarchical planning framework with two core components: (1) an online hierarchical planner that uses dynamic AND/OR trees for efficient search and (2) a structured memory module that tracks and maintains candidate solutions to improve constraint satisfaction in information-seeking tasks. The framework also produces interpretable hierarchical plans, enabling easier debugging and facilitating human intervention when needed. Our results on WebVoyager, WebArena, and custom shopping benchmarks show that STRUCTUREDAGENT improves performance on long-horizon web-browsing tasks compared to standard LLM-based agents.