@inproceedings{song-etal-2025-smart,
title = "Smart-Searcher: Incentivizing the Dynamic Knowledge Acquisition of {LLM}s via Reinforcement Learning",
author = "Song, Huatong and
Jiang, Jinhao and
Tian, Wenqing and
Chen, Zhipeng and
Wu, Yuhuan and
Zhao, Jiahao and
Min, Yingqian and
Zhao, Xin and
Fang, Lei and
Wen, Ji-Rong",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-emnlp.731/",
doi = "10.18653/v1/2025.findings-emnlp.731",
pages = "13572--13586",
ISBN = "979-8-89176-335-7",
abstract = "Large Language Models (LLMs) are powerful but prone to hallucinations due to static knowledge. Retrieval-Augmented Generation (RAG) helps by injecting external information, but current methods often are costly, generalize poorly, or ignore the model{'}s internal knowledge.In this paper, we introduce Smart-Searcher, a novel framework designed to train LLMs to adaptively leverage both internal and external knowledge sources. Smart-Searcher employs a two-stage training strategy: an initial SFT Cold-start phase for preliminary format learning, followed by RL for Dynamic Knowledge Acquisition. The RL stage uses outcome-supervision to encourage exploration, incorporates a reward mechanism for internal knowledge utilization, and integrates a memorization mechanism to continuously assimilate retrieved information, thereby enriching the model{'}s internal knowledge. By leveraging internal knowledge and external search engine, the model continuously improves its capabilities, enabling efficient retrieval-augmented reasoning.Our experiments demonstrate that Smart-Searcher outperforms previous RAG and reasoning methods and achieves efficient retrieval.The code is available at \url{https://github.com/RUCAIBox/R1-Searcher-plus}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="song-etal-2025-smart">
<titleInfo>
<title>Smart-Searcher: Incentivizing the Dynamic Knowledge Acquisition of LLMs via Reinforcement Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Huatong</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinhao</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenqing</namePart>
<namePart type="family">Tian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhipeng</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuhuan</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiahao</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yingqian</namePart>
<namePart type="family">Min</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lei</namePart>
<namePart type="family">Fang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ji-Rong</namePart>
<namePart type="family">Wen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christos</namePart>
<namePart type="family">Christodoulopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolyn</namePart>
<namePart type="family">Rose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Violet</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-335-7</identifier>
</relatedItem>
<abstract>Large Language Models (LLMs) are powerful but prone to hallucinations due to static knowledge. Retrieval-Augmented Generation (RAG) helps by injecting external information, but current methods often are costly, generalize poorly, or ignore the model’s internal knowledge.In this paper, we introduce Smart-Searcher, a novel framework designed to train LLMs to adaptively leverage both internal and external knowledge sources. Smart-Searcher employs a two-stage training strategy: an initial SFT Cold-start phase for preliminary format learning, followed by RL for Dynamic Knowledge Acquisition. The RL stage uses outcome-supervision to encourage exploration, incorporates a reward mechanism for internal knowledge utilization, and integrates a memorization mechanism to continuously assimilate retrieved information, thereby enriching the model’s internal knowledge. By leveraging internal knowledge and external search engine, the model continuously improves its capabilities, enabling efficient retrieval-augmented reasoning.Our experiments demonstrate that Smart-Searcher outperforms previous RAG and reasoning methods and achieves efficient retrieval.The code is available at https://github.com/RUCAIBox/R1-Searcher-plus.</abstract>
<identifier type="citekey">song-etal-2025-smart</identifier>
<identifier type="doi">10.18653/v1/2025.findings-emnlp.731</identifier>
<location>
<url>https://aclanthology.org/2025.findings-emnlp.731/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>13572</start>
<end>13586</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Smart-Searcher: Incentivizing the Dynamic Knowledge Acquisition of LLMs via Reinforcement Learning
%A Song, Huatong
%A Jiang, Jinhao
%A Tian, Wenqing
%A Chen, Zhipeng
%A Wu, Yuhuan
%A Zhao, Jiahao
%A Min, Yingqian
%A Zhao, Xin
%A Fang, Lei
%A Wen, Ji-Rong
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Findings of the Association for Computational Linguistics: EMNLP 2025
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-335-7
%F song-etal-2025-smart
%X Large Language Models (LLMs) are powerful but prone to hallucinations due to static knowledge. Retrieval-Augmented Generation (RAG) helps by injecting external information, but current methods often are costly, generalize poorly, or ignore the model’s internal knowledge.In this paper, we introduce Smart-Searcher, a novel framework designed to train LLMs to adaptively leverage both internal and external knowledge sources. Smart-Searcher employs a two-stage training strategy: an initial SFT Cold-start phase for preliminary format learning, followed by RL for Dynamic Knowledge Acquisition. The RL stage uses outcome-supervision to encourage exploration, incorporates a reward mechanism for internal knowledge utilization, and integrates a memorization mechanism to continuously assimilate retrieved information, thereby enriching the model’s internal knowledge. By leveraging internal knowledge and external search engine, the model continuously improves its capabilities, enabling efficient retrieval-augmented reasoning.Our experiments demonstrate that Smart-Searcher outperforms previous RAG and reasoning methods and achieves efficient retrieval.The code is available at https://github.com/RUCAIBox/R1-Searcher-plus.
%R 10.18653/v1/2025.findings-emnlp.731
%U https://aclanthology.org/2025.findings-emnlp.731/
%U https://doi.org/10.18653/v1/2025.findings-emnlp.731
%P 13572-13586
Markdown (Informal)
[Smart-Searcher: Incentivizing the Dynamic Knowledge Acquisition of LLMs via Reinforcement Learning](https://aclanthology.org/2025.findings-emnlp.731/) (Song et al., Findings 2025)
ACL
- Huatong Song, Jinhao Jiang, Wenqing Tian, Zhipeng Chen, Yuhuan Wu, Jiahao Zhao, Yingqian Min, Xin Zhao, Lei Fang, and Ji-Rong Wen. 2025. Smart-Searcher: Incentivizing the Dynamic Knowledge Acquisition of LLMs via Reinforcement Learning. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 13572–13586, Suzhou, China. Association for Computational Linguistics.