Forest fires play a pivotal role in influencing ecosystem evolution, exerting a profound impact on plant diversity and community stability. Understanding post-fire recovery strategies holds significant scientific importance for the ecological succession and restoration of forest ecosystems. This study utilized Partial Least Squares Path Modeling (PLS-PM) to investigate dynamic relationships among plant species diversity, phylogenetic diversity, soil properties, and community stability during various recovery stages (5-year, 15-year, and 23-year) following wildfires on the northeastern margin of the Qinghai-Tibet Plateau. The findings revealed: (1) Over time, species richness significantly decreased (p< 0.05 or p< 0.01), while species diversity and dominance increased, resulting in uniform species distribution. Community stability progressively improved, with increased species compositional similarity. (2) Throughout succession, phylogenetic diversity (PD) significantly decreased (p< 0.01), accompanied by rising Mean Pairwise Distance (MPD) and Mean Nearest Taxon Distance (MNTD). Net Relatedness Index (NRI) shifted from positive to negative, indicating an increasing aggregation and dominance of plants with similar evolutionary traits in burned areas. Early succession witnessed simultaneous environmental filtering and competitive exclusion, shifting predominantly to competitive exclusion in later stages. (3) PLS-PM revealed that in the early recovery stage, soil properties mainly affected community stability, while species diversity metamorphosed into the primary factor in the mid-to-late stages. In summary, this study showed that plant diversity and phylogenetic variation were successful in revealing changes in community structure during the succession process. Soil characteristics functioned as selective barriers for plant communities during succession, and community stability underwent a multi-faceted and dynamic process. The soil-plant dynamic feedback continuously enhanced soil conditions and community vegetation structure thereby augmenting stability. Post-fire vegetation gradually transitioned towards the original native state, demonstrating inherent ecological self-recovery capabilities in the absence of secondary disturbances.