these natural modes of locomotion with engineered systems has been challenging. [11][12][13] In this respect, burgeoning effort has been devoted to developing new simulation tools, physical models, and experimental platforms. Among these engineering tools, soft robotic systems are promising in light of their biologically relevant mechanical compliance, deformability, and modes of locomotion. [4,[14][15][16][17] Untethered soft robots, which can freely move without requiring a physical connection to external hardware and power supplies, are good candidates for studying the locomotion modes of natural invertebrates. [18][19][20][21] Thanks to recent advances in fabrication methods, [22][23][24][25][26] functional materials, [27][28][29] and actuation strategies, [7,[30][31][32][33][34][35] roboticists have proposed several soft robots that are capable of mimicking some features of natural animal locomotion. For example, previous efforts with photo-responsive hydrogels-based robots demonstrated the ability to mimic peristaltic earthworm crawling. [36] Likewise, untethered soft robots powered with shape memory alloy have been shown to exhibit a variety of locomotion modes, including undulation, jumping, crawling through narrow space or walking over rough terrain. [37,38] The application of magnetic soft robots in biomimetic study has received growing attention due to their high controllability. Hu et al. developed a millimeter-scale film robot that achieves multiple locomotion modes,The efficient motility of invertebrates helps them survive under evolutionary pressures. Reconstructing the locomotion of invertebrates and decoupling the influence of individual basic motion are crucial for understanding their underlying mechanisms, which, however, generally remain a challenge due to the complexity of locomotion gaits. Herein, a magnetic soft robot to reproduce midge larva's key natural swimming gaits is developed, and the coupling effect between body curling and rotation on motility is investigated. Through the authors' systematically decoupling studies using programmed magnetic field inputs, the soft robot (named LarvaBot) experiences various coupled gaits, including biomimetic side-to-side flexures, and unveils that the optimal rotation amplitude and the synchronization of curling and rotation greatly enhance its motility. The LarvaBot achieves fast locomotion and upstream capability at the moderate Reynolds number regime. The soft robotics-based platform provides new insight to decouple complex biological locomotion, and design programmed swimming gaits for the fast locomotion of softbodied swimmers.