Saved in:
Bibliographic Details
Main Authors: Chu, Zheng, Wang, Xiao, Hong, Jack, Fan, Huiming, Huang, Yuqi, Yang, Yue, Xu, Guohai, Zhao, Chenxiao, Xiang, Cheng, Hu, Shengchao, Kuang, Dongdong, Liu, Ming, Qin, Bing, Yu, Xing
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.14234
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908835443113984
author Chu, Zheng
Wang, Xiao
Hong, Jack
Fan, Huiming
Huang, Yuqi
Yang, Yue
Xu, Guohai
Zhao, Chenxiao
Xiang, Cheng
Hu, Shengchao
Kuang, Dongdong
Liu, Ming
Qin, Bing
Yu, Xing
author_facet Chu, Zheng
Wang, Xiao
Hong, Jack
Fan, Huiming
Huang, Yuqi
Yang, Yue
Xu, Guohai
Zhao, Chenxiao
Xiang, Cheng
Hu, Shengchao
Kuang, Dongdong
Liu, Ming
Qin, Bing
Yu, Xing
contents Large language models are transitioning from generalpurpose knowledge engines to realworld problem solvers, yet optimizing them for deep search tasks remains challenging. The central bottleneck lies in the extreme sparsity of highquality search trajectories and reward signals, arising from the difficulty of scalable longhorizon task construction and the high cost of interactionheavy rollouts involving external tool calls. To address these challenges, we propose REDSearcher, a unified framework that codesigns complex task synthesis, midtraining, and posttraining for scalable searchagent optimization. Specifically, REDSearcher introduces the following improvements: (1) We frame task synthesis as a dualconstrained optimization, where task difficulty is precisely governed by graph topology and evidence dispersion, allowing scalable generation of complex, highquality tasks. (2) We introduce toolaugmented queries to encourage proactive tool use rather than passive recall.(3) During midtraining, we strengthen core atomic capabilities knowledge, planning, and function calling substantially reducing the cost of collecting highquality trajectories for downstream training. (4) We build a local simulated environment that enables rapid, lowcost algorithmic iteration for reinforcement learning experiments. Across both textonly and multimodal searchagent benchmarks, our approach achieves stateoftheart performance. To facilitate future research on longhorizon search agents, we will release 10K highquality complex text search trajectories, 5K multimodal trajectories and 1K text RL query set, and together with code and model checkpoints.
format Preprint
id arxiv_https___arxiv_org_abs_2602_14234
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle REDSearcher: A Scalable and Cost-Efficient Framework for Long-Horizon Search Agents
Chu, Zheng
Wang, Xiao
Hong, Jack
Fan, Huiming
Huang, Yuqi
Yang, Yue
Xu, Guohai
Zhao, Chenxiao
Xiang, Cheng
Hu, Shengchao
Kuang, Dongdong
Liu, Ming
Qin, Bing
Yu, Xing
Artificial Intelligence
Computation and Language
Large language models are transitioning from generalpurpose knowledge engines to realworld problem solvers, yet optimizing them for deep search tasks remains challenging. The central bottleneck lies in the extreme sparsity of highquality search trajectories and reward signals, arising from the difficulty of scalable longhorizon task construction and the high cost of interactionheavy rollouts involving external tool calls. To address these challenges, we propose REDSearcher, a unified framework that codesigns complex task synthesis, midtraining, and posttraining for scalable searchagent optimization. Specifically, REDSearcher introduces the following improvements: (1) We frame task synthesis as a dualconstrained optimization, where task difficulty is precisely governed by graph topology and evidence dispersion, allowing scalable generation of complex, highquality tasks. (2) We introduce toolaugmented queries to encourage proactive tool use rather than passive recall.(3) During midtraining, we strengthen core atomic capabilities knowledge, planning, and function calling substantially reducing the cost of collecting highquality trajectories for downstream training. (4) We build a local simulated environment that enables rapid, lowcost algorithmic iteration for reinforcement learning experiments. Across both textonly and multimodal searchagent benchmarks, our approach achieves stateoftheart performance. To facilitate future research on longhorizon search agents, we will release 10K highquality complex text search trajectories, 5K multimodal trajectories and 1K text RL query set, and together with code and model checkpoints.
title REDSearcher: A Scalable and Cost-Efficient Framework for Long-Horizon Search Agents
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2602.14234