The functional realization of proteins and other biological macromolecules depends on conformational dynamics and allosteric regulation, and elucidating their molecular mechanisms is an important foundation for understanding life processes. Molecular dynamics simulations are a powerful tool for investigating conformational evolution at the atomic level. However, traditional methods are limited by simulation timescales and high free-energy barriers, making it difficult to effectively capture rare conformations and their transition pathways. As a result, the development of enhanced sampling techniques has become key to overcoming this bottleneck. As a classical enhanced sampling technique, metadynamics suffers from several shortcomings, including strong dependence on collective variables and significant errors caused by bias potential accumulation. This article reviews three major improvement strategies. The first combines stochastic resetting with metadynamics, using trajectory-resetting mechanisms to improve sampling efficiency while avoiding the difficulty of optimizing collective variables. The second, SinkMeta, employs a “sinking” bias effect to enable efficient exploration of specific regions and paths. The third, OPES-based hybrid methods, improve the stability of free-energy estimation by optimizing the target distribution or the way the bias is constructed. These methods provide new ideas for characterizing free-energy landscapes and studying conformational transitions in complex biological systems, while also promoting the continued development of enhanced sampling methodologies.