Animals exhibit great diversity in their foraging behavior patterns in nature in order to survive. These behaviors vary depending on the environment and biological composition of the animal. These different ways of searching for food have inspired scientists to create mathematical models. Inspired by the nature of these animals and use them to solve complex problems or problems that traditional methods are unable to solve and since there is no comprehensive algorithm that solves all problems, researchers have resorted to hybridizing these algorithms with each other and with traditional methods to obtain a more efficient hybrid algorithm. In this paper, we derived a new conjugate coefficient for the conjugate gradient method, demonstrated its global convergence and descent, and incorporated it into the directed search phase of the artificial hummingbird algorithm When the algorithm is unable to find a point closest to the optimal solution. The results showed a significant improvement over the results of the original algorithm.