<p>Differential Evolution (DE) is a robust algorithm for global optimization; yet, state-of-the-art variants still face challenges such as premature convergence, parameter sensitivity, and stagnation risks, particularly in complex, multimodal landscapes. To address these issues, this paper proposes the Adaptive Stagnation Guided Differential Evolution (ASGDE) algorithm. ASGDE introduces three core mechanisms: (1) an adaptive stagnation detection strategy that identifies stagnated individuals and dynamically adjusts the number of guiding neighbors to enhance search diversity; (2) an improved guide individual selection method that allows stagnated individuals to learn from a broader range of top-performing peers, facilitating escape from local optima; and (3) a nonlinear population size adjustment scheme using a Sigmoid function, which maintains a larger population for exploration in the early stages and reduces it to accelerate convergence later. Extensive experimental results on the CEC2014 and CEC2011 benchmarks demonstrate that ASGDE exhibits strong competitiveness when compared with 25 other state-of-the-art algorithms. The proposed approach significantly improves the balance between exploration and exploitation, showing particular strength in solving high-dimensional and multimodal problems.</p>