WebThinker: Empowering Large Reasoning Models with Deep Research Capability
By Xiaoxi Li, Jiajie Jin, Guanting Dong, Hongjin Qian, Yutao Zhu, Yongkang Wu, Ji-rong Wen, Zhicheng Dou. Neural Information Processing Systems, 2025
299 citations recorded by the source corpus.
Large reasoning models (LRMs), such as OpenAI-o1 and DeepSeek-R1, demonstrate impressive long-horizon reasoning capabilities. However, their reliance on static internal knowledge limits their performance on complex, knowledge-intensive tasks and hinders their ability to produce comprehensive research reports requiring synthesis of diverse web information. To address this, we propose WebThinker, a deep research agent that empowers LRMs to autonomously search the web, navigate among web pages, and draft reports during the reasoning process. WebThinker integrates a Deep Web Explorer module, enabling LRMs to dynamically search, navigate, and extract information from the web when encountering knowledge gaps. It also employs an Autonomous Think-Search-and-Draft strategy, allowing the model to seamlessly interleave reasoning, information gathering, and report writing in real time. To further enhance research tool utilization, we introduce an RL-based training strategy via iterative online Direct Preference Optimization (DPO). Extensive experiments on complex reasoning benchmarks (GPQA, GAIA, WebWalkerQA, HLE) and scientific report generation tasks (Glaive) demonstrate that WebThinker significantly outperforms existing methods and strong proprietary systems. Our approach enhances LRM reliability and applicability in complex scenarios, paving the way for more capable and versatile deep research systems. The code is available at https://github.com/RUC-NLPIR/WebThinker.
Subfields: Artificial Intelligence, Information Systems
View Paper: https://www.semanticscholar.org/paper/23f655a7a596ab5f431ab9e18d5c641de3f8afb9
Citations in the corpus
- Large Language Models for Information Retrieval: A Survey (2023, unknown): “Typical metrics include precision, recall, mean reciprocal rank (MRR) [221], mean average precision (MAP), and normalized discounted cumulative gain (nDCG) [222].”
- A Survey of Model Architectures in Information Retrieval (2025, unknown): “The commonly used smoothing methods (Zhai and Lafferty, 2004; Zhai et al., 2008) include Jalinek-Mercer smoothing, Dirichlet smoothing, etc.”
- Unlocking Multimodal Document Intelligence: From Current Triumphs to Future Frontiers of Visual Document Retrieval (2026, unknown): “For instance, while frameworks like WebThinker (Li et al., 2025f) and DeepResearcher (Zheng et al., 2025b) enable agents to interleave thought with web search actions, they often function as sophisticated tool-executors that reactively process information rather than as research strategists that…”