WebThinker: Empowering Large Reasoning Models with Autonomous Search and Intelligent Report Generation Recent advancements in Large Reasoning Models (LRMs) have demonstrated remarkable capabilities in mathematical reasoning, code generation, and scientific problem-solving. However, these models face significant limitations when tackling real-world research tasks that require dynamic access to external knowledge. The WebThinker framework, developed by researchers from Renmin University, Beihang AI Research Institute, and Huawei Poisson Lab, bridges this gap by integrating autonomous web exploration with advanced reasoning capabilities. This article explores its technical innovations, performance benchmarks, and practical applications. Breaking the Limitations of Traditional LRMs The Challenge of Static Knowledge …