BrowseMaster: Towards Scalable Web Browsing via Tool-Augmented Programmatic Agent Pair

2025-08-12

AI Summary

Existing LLM-based agents struggle to balance search breadth and reasoning depth in large-scale web retrieval: sequential querying leads to insufficient coverage, while noisy raw inputs disrupt coherence in multi-step reasoning. This paper proposes a planning-execution dual-agent collaborative framework, enabled by programmatic tool augmentation to enforce role specialization—where the planner agent dynamically generates search strategies, and the executor agent precisely invokes tools and synthesizes structured evidence. We further introduce task-constrained strategy optimization and cross-lingual support. Evaluated on English and Chinese benchmarks, our approach achieves 30.0 and 46.5 points, respectively—significantly outperforming both open-source and closed-source baselines. It is the first method to achieve a substantive trade-off between scalability and reasoning quality, establishing a scalable, robust paradigm for complex, reasoning-intensive information retrieval.

Technology Category

• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Planning with Language Models
• Multiagent Systems: Multiagent Planning

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Effective information seeking in the vast and ever-growing digital landscape requires balancing expansive search with strategic reasoning. Current large language model (LLM)-based agents struggle to achieve this balance due to limitations in search breadth and reasoning depth, where slow, serial querying restricts coverage of relevant sources and noisy raw inputs disrupt the continuity of multi-step reasoning. To address these challenges, we propose BrowseMaster, a scalable framework built around a programmatically augmented planner-executor agent pair. The planner formulates and adapts search strategies based on task constraints, while the executor conducts efficient, targeted retrieval to supply the planner with concise, relevant evidence. This division of labor preserves coherent, long-horizon reasoning while sustaining broad and systematic exploration, overcoming the trade-off that limits existing agents. Extensive experiments on challenging English and Chinese benchmarks show that BrowseMaster consistently outperforms open-source and proprietary baselines, achieving scores of 30.0 on BrowseComp-en and 46.5 on BrowseComp-zh, which demonstrates its strong capability in complex, reasoning-heavy information-seeking tasks at scale.

Problem

• Balancing expansive search with strategic reasoning in digital landscapes
• Overcoming slow serial querying and noisy raw inputs in LLM agents
• Enhancing multi-step reasoning and broad exploration in web browsing

Innovation

• Tool-augmented planner-executor agent pair
• Planner formulates adaptive search strategies
• Executor conducts efficient targeted retrieval