Abstract: Given the high complexity and uncertainty of university laboratory safety systems, and to elucidate the mechanisms underlying accident evolution and improve the accuracy of risk prediction, this study proposes a method for identifying critical paths in complex networks of laboratory accidents driven by the “2-4” model. Based on the “2-4” model, multidimensional risk identification is conducted to construct a network model of laboratory accident evolution, and the TOPSIS comprehensive evaluation method is employed to identify critical nodes in the network; By introducing a step-count penalty factor and probability correction, we break the idealized assumptions of purely topological networks to achieve quantitative prediction of evolutionary paths. Applying this method to analyze 186 laboratory accident records, we performed static evaluations of nodes using four centrality metrics and identified dynamic high-risk evolutionary paths through entropy weighting. This revealed the heterogeneous evolutionary mechanisms of different accident types, specifically: the behavior-driven and rapid-spread characteristics of fire accidents; the nonlinear coupling of phase transitions and behavioral deviations in explosion incidents; and the risk perception blind spots and active failure of protective barriers in poisoning incidents. By applying this method, targeted prevention, control, and interruption strategies for critical paths are proposed, thereby enhancing safety effectiveness under limited resource constraints.