Abstract: The rapid development of artificial intelligence technology and the continuous upgrading of intelligent industries have greatly driven market demand for various memory types. This paper focuses on hafnium-based ferroelectric memory, which features high density and low power consumption and has attracted considerable attention due to its excellent process compatibility, strong scalability, high reliability, and high integration density. Nevertheless, its data retention capability remains a key reliability challenge for large-scale applications. This paper summarizes recent research progress on the retention characteristics of hafnium-based ferroelectric memories. It first introduces the working principles of hafnium-based ferroelectric and antiferroelectric memories and the characterization methods for data retention. Subsequently, it analyzes key mechanisms affecting retention, including imprint and temperature effects. Comparative analysis reveals that the notable differences in retention characteristics between hafnium-based ferroelectric and antiferroelectric memories stem primarily from their free energy models. Finally, the paper summarizes technical strategies to improve data retention, including element doping, annealing process optimization, and interface engineering.