<p>Grey Wolf Optimizer (GWO) is a prominent swarm intelligence algorithm that emulates the hunting behavior of grey wolves to solve optimization problems. While GWO performs well in various scenarios, its solution generation primarily relies on exploration, lacking effective exploitation around previously discovered optimal regions. In this work, we propose an improved version of GWO that incorporates best solutions directly into the generation of new solutions, enhancing both convergence speed and solution diversity. Additionally, we present a binary variant of the improved GWO, making it suitable for binary optimization problem, such as regression testing. The proposed binary GWO is applied to the test subset selection problem in regression testing, implemented on the Siemens test suite. Comparative analysis against other swarm intelligence algorithms on benchmark functions and regression testing demonstrates that our proposed variants outperform previous approaches in terms of speed and accuracy.</p>