Reversible shape-memory polymers (RSMPs) are highly desired for soft actuators due to the repeatability of deformation. Herein, a polyolefin elastomer vitrimer (POEV) was prepared by constructing a dynamic cross-linked network based on boronic ester bonds. POEV showed varied network relaxation in a wide temperature range due to hierarchical network relaxation, and then the entropy decreased and the relaxation of POEV chains was facilely controlled by temperature. The controllable relaxation of POEV by programming the temperature enabled the actuation domain with a reduction in entropy and the skeleton domain with a relatively high entropy can be built in POEV, greatly affecting the reversible shape memory effects (RSMEs). The topological rearrangement resulted from the activated exchange of dynamic covalent bonds, which enables POEV with good shape reconfigurability, and allows for complicated 3D shapes and shape-shifting on demand. More interestingly, combining the decreasing entropy of POEV chains and fully topological rearrangement tailored by temperature, hybrid aligned carbon nanotubes (CNTs) can be constructed in POEV via a two-stage training. Then, the aligned CNTs can enhance the elasticity and act as a hybrid skeleton for RSMEs, avoiding the negative impact of CNTs on the reversible actuation strain. The hierarchical network relaxation facilitates combining all these unusual properties in one shape memory network synergistically, paving new avenues for realizing smart materials with advanced RSME.