Recently, a new family of conjugate gradient methods with a Grippo-Lucidi type step length rule is proposed by Shi and Guo [A new family of conjugate gradient methods. Journal of Computational and Applied Mathematics, 2009, 224:444-457]. In this paper, we improve Shi and Guo’s method by adopting an improved Grippo-Lucidi type step length rule, and the improvement is twofold: (1) We drop the local Lipschitz constant in the step length, which is beyond the problem data and has to be further estimated; (2) The search direction $d_k$ only needs to satisfy the descent property instead of the sufficient descent property. The global convergence result of the modified method is established under some mild conditions. Preliminary numerical results are also reported to show the efficiency of the improved method.