The dynamic characteristics of species composition, community structure, important value (IV), diversity, mortalities, and recruitments in the primary forest communities were analyzed based on the vegetation survey of a 2 hm² dynamic monitoring plot in the Mulun National Nature Reserve in karst peak-cluster depression in 2007, 2012, and 2017. The results showed that there were 123 species of woody plants belonging to 95 genera and 46 families in 2017, which increased by 2 families and reduced by 5 genera and 7 species compared with those of 2007, and increased by 2 families and reduced by 4 genera and 4 species compared with those of 2012. The dominant species in the plot were not obviously changed in past 10 years. The number of species with IV ≥ 1 decreased by 3 in 2017 in comparison with those of 2007. However, the top five species remained the same but with different IV orders of the individual species. In general, the average diameter at breast height (DBH), tree height, and biomass of the communities increased over time. Specifically, there were no significant differences in mean diameter and tree height between 2007 and 2012, while there were significant differences between 2007 and 2017. The Shannon-wiener index of the forest community gradually decreased while the Simpson index and Pielou index gradually increased. On the whole, the DBH size-class structure of all species in the forest community exhibited a reverse J-shaped pattern, which indicated that the forest community was in a good regeneration status. The average annual mortality of species with DBH ≥ 1.0 cm was 8.23% and the average annual recruitment rate of species with DBH ≥ 1.0 cm was 5.73% between 2007 and 2017. The number of Mallotus japonicas and Liquidambar formosana in the forest community showed an increasing trend and their dominant position was gradually increasing. In addition, there was about 35% of 20 species in the status of insufficient regeneration, which led to a recession with large variation in species composition and structure of the community. Understanding of regeneration mechanism of the dominant species and vegetation structure dynamics in a forest community could help predict the forest future trend.