Deep Research Pretraining via Predictive Navigation

2026-08-01

AI Summary

This work addresses the inefficiency of conventional deep research agents, which rely on costly environment interaction trajectories. The authors propose Deep Research Pretraining (DRP), a novel framework that, for the first time, automatically extracts predictive navigation signals from the citation or hyperlink structures of static text to construct search–read–write trajectories, enabling offline pretraining without real-time retrieval. Leveraging the Qwen3-14B-Base model, they perform one billion token pretraining on DRP-Paper and DRP-Web, followed by fine-tuning with minimal trajectory data and reinforcement learning. Experiments demonstrate that DRP surpasses full-data baselines using only one-quarter of the supervised data and consistently improves performance across multiple benchmarks—including DeepResearch Bench, ResearchQA, WebWalkerQA, and SimpleQA—validating the effectiveness and generalization capability of evidence-guided navigation learning.

Abstract

Deep research agents are often trained on expensive, environment-grounded tool-use trajectories that require repeated retrieval, document inspection, and report evaluation. We introduce Deep Research Pretraining (DRP), an offline framework that derives predictive navigation supervision from naturally occurring evidence structures. Given a citation-bearing or hyperlinked passage, DRP constructs a proxy research objective, recovers linked evidence and graph-related alternatives, and converts them into search-open-write trajectories. This teaches models what to search for, which documents to inspect, and how to synthesize evidence, without a live retrieval environment or executed policy rollout. We instantiate DRP on scholarly citation graphs (DRP-Paper) and Wikipedia hyperlinks (DRP-Web), continually pretrain separate Qwen3-14B-Base models on 1B tokens, and fine-tune them on controlled fractions of 13K agent trajectories. Across five independently sampled subsets at each low-data budget, both variants consistently outperform matched no-DRP models on DeepResearch Bench. With one quarter of the SFT data, DRP-Web even surpasses a fixed no-DRP full-data checkpoint, with gains transferring to ResearchQA, WebWalkerQA, and SimpleQA. Starting from matched low-data SFT checkpoints, the DRP-Web advantage also persists through subsequent agentic RL. Source-matched and evidence-mismatch controls indicate that these improvements arise from evidence-conditioned navigation rather than domain exposure or agent-format imitation. DRP thus provides a promising complementary approach to trajectory-based agent training.

Problem

• deep research agents
• tool-use trajectories
• evidence navigation
• offline pretraining
• retrieval-based reasoning

Innovation

• Deep Research Pretraining
• Predictive Navigation
• Evidence-conditioned Learning
• Offline Agent Training
• Search-Open-Write Trajectories