Memristor-based nonvolatile architectures have been a promising platform for advancing artificial intelligence and neuromorphic techniques. The development of polymorphic memristors for artificial synapses, which can respond to varying degrees of synaptic weight under different stimuli, has contributed significantly to the efficient processing of massive data on system-on-chips. However, it is critical and challenging to develop multimodal or reconfigurable resistive switching, enabling the precise regulation of multiple resistive states and suppressing misreading. In this paper, we proposed the reconfigurable resistive switching that can be programmed by manipulating isomeric phases in vanadium dioxide (VO 2 ) nanostructures. The electric-driven mutual transformation of multiphase states facilitated the emergence of multiple resistive switching modes. By adjustment of the switching energy, the threshold switching processes can be precisely tailored, offering a unique programmable threshold in VO 2 -based resistive switching. We believe that the incorporation of multimodal threshold switching in memristors helps for responding to varying synaptic weights and holds great potential for intelligent memories and artificial synapses.