SimpleDeepSearcher: Deep Information Seeking via Web-Powered Reasoning Trajectory Synthesis

TL;DR

SimpleDeepSearcher is a lightweight framework that improves deep search capabilities of large language models by synthesizing high-quality training data through realistic web search simulations, avoiding complex training and high computational costs.

Contribution

It introduces a data-centric approach using realistic web search simulations and multi-criteria data curation to enhance deep search performance without complex training paradigms.

Findings

Significant improvements over RL-based baselines on five benchmarks.

High-quality training data synthesized from only 871 samples.

Effective deep search achieved through strategic data engineering.

Abstract

Retrieval-augmented generation (RAG) systems have advanced large language models (LLMs) in complex deep search scenarios requiring multi-step reasoning and iterative information retrieval. However, existing approaches face critical limitations that lack high-quality training trajectories or suffer from the distributional mismatches in simulated environments and prohibitive computational costs for real-world deployment. This paper introduces SimpleDeepSearcher, a lightweight yet effective framework that bridges this gap through strategic data engineering rather than complex training paradigms. Our approach synthesizes high-quality training data by simulating realistic user interactions in live web search environments, coupled with a multi-criteria curation strategy that optimizes the diversity and quality of input and output side. Experiments on five benchmarks across diverse domains demonstrate that SFT on only 871 curated samples yields significant improvements over RL-based baselines. Our work establishes SFT as a viable pathway by systematically addressing the data-scarce bottleneck, offering practical insights for efficient deep search systems. Our code is available at https://github.com/RUCAIBox/SimpleDeepSearcher.