Marco DeepResearch: Unlocking Efficient Deep Research Agents via Verification-Centric Design

TL;DR

Marco DeepResearch introduces a verification-centric framework for deep research agents, enhancing their ability to conduct reliable, long-horizon investigations by integrating verification at data synthesis, trajectory construction, and inference stages.

Contribution

The paper presents a novel verification-centric design for deep research agents, improving accuracy and scalability across multiple stages of research tasks.

Findings

Outperforms 8B-scale agents on challenging benchmarks like BrowseComp.

Uses verification mechanisms to control question difficulty and ensure answer correctness.

Achieves near or better performance than 30B-scale agents with fewer resources.

Abstract

Deep research agents autonomously conduct open-ended investigations, integrating complex information retrieval with multi-step reasoning across diverse sources to solve real-world problems. To sustain this capability on long-horizon tasks, reliable verification is critical during both training and inference. A major bottleneck in existing paradigms stems from the lack of explicit verification mechanisms in QA data synthesis, trajectory construction, and test-time scaling. Errors introduced at each stage propagate downstream and degrade the overall agent performance. To address this, we present Marco DeepResearch, a deep research agent optimized with a verification-centric framework design at three levels: \textbf{(1)~QA Data Synthesis:} We introduce verification mechanisms to graph-based and agent-based QA synthesis to control question difficulty while ensuring answers are unique and correct; \textbf{(2)~Trajectory Construction:} We design a verification-driven trajectory synthesis method that injects explicit verification patterns into training trajectories; and \textbf{(3)~Test-time scaling:} We use Marco DeepResearch itself as a verifier at inference time and effectively improve performance on challenging questions. Extensive experimental results demonstrate that our proposed Marco DeepResearch agent significantly outperforms 8B-scale deep research agents on most challenging benchmarks, such as BrowseComp and BrowseComp-ZH. Crucially, under a maximum budget of 600 tool calls, Marco DeepResearch even surpasses or approaches several 30B-scale agents, like Tongyi DeepResearch-30B.