Integrating Technological Innovation with Industrial Innovation (the“Tech-Industry Integration”） represents the core pathway for cultivating new quality productive forces, and the cross-sectoral transfer of key core technologies is the most direct manifestation of this integration. Using China’s patent assignment data from 2015 to 2024, this paper employs the LDA topic model to precisely identify key core technologies and innovatively constructs a two-mode (“academic-research institutions”and “enterprises”) network. On this basis, we apply the Temporal Exponential Random Graph Model (TERGM) to investigate the intrinsic mechanisms of network evolution and further conduct simulations to examine network resilience and bottom-line security. The findings are fourfold. First, the network evolution exhibits significant path dependence and a “Matthew effect”, whereby prior collaboration ties and supply-chain relationships provide a solid trust foundation for cross-sectoral technology flows. Second, the driving factors of network evolution are asymmetrical across agents: the “hard power” of academic-research institutions and the “soft power” of enterprises are the two primary drivers. Third, the evolutionary mechanisms display marked industrial heterogeneity—the “core-periphery” structure is most dominant in traditional industries, whereas structured cooperative networks in emerging and future industries remain underdeveloped. Fourth, while agglomerative evolution entails a high “vulnerability” risk due to over-reliance on a few hubs, in terms of resilience against extreme shocks, enterprise-led diversified cooperative structures are increasingly replacing single academic-research institutions as the key pillar for safeguarding the bottom-line security of cross-sectoral technology coordination and industrial security.