Achieving nonhysteretic and reversible (nHR) oxygen redox in sodium-ion batteries(SIBs) is critical for enhancing their energy density and practical viability. Herein, we present a high-throughput computational framework that integrates machine learning interatomic potentials (MLIPs) with density functional theory (DFT) to design high-entropy layered Na-based cathodes with improved redox reversibility. Using MLIP-based screening of 30,240 configurations derived from Na[Li1/6Mn2/6A1/6B1/6C1/6]O2 (A, B, C: 3d transition metals), Na[Li1/6Mn2/6Ti1/6V1/6Zn1/6]O2 (HE-NLMO) is identified as the most thermodynamically stable composition. With understanding of thermodyanmic phase stabilities, DFT calculations reveal that HE-NLMO enables spontaneous and structurally continuous Li-ion migration into tetrahedral sites within the Na layer during desodiation. Electronic structure investigations further show that the selective oxidation behaviours trigger the asymmetric Li position in the Na layer for easily returning it to the original site. This combination of structural and electronic flexibility facilitates a smooth Li-ion migration pathway and sustainable nHR oxygen redox. Our results underscore the effectiveness of entropy engineering and ML-DFT hybrid approaches in accelerating the discovery of high-performance Na-based cathodes.