Differential evolution remains among the most effective evolutionary algorithms for global optimization, yet its performance depends on balancing exploration and exploitation. This paper proposes Guided Differential Evolution (GDE), which alternates explicitly between an exploration mode using the classical DE/rand/1 scheme and a guided mode that incorporates historical population information via a sliding window of previous generations. The latter steers candidate solutions toward a population centroid, reducing fluctuant behavior and improving convergence stability without added complexity. Retaining the classical DE structure, GDE augments it with a guidance mechanism and deterministic phase scheduling. Experimental validation on the CEC'2017 benchmark suite confirms competitive performance across diverse function categories, demonstrating the effectiveness of the proposed phase alternation strategy. https://github.com/espartan0007/GDE