LLM-Enhanced Recommendation, Remote Cardiac Injury Evidence, and Flood Disaster Modeling

Abstract

LLM-enhanced sequential recommendation systems model user behavior through dynamic mechanisms that can adapt to changing preferences and historical interaction patterns. Similar dynamic modeling challenges appear in flood disaster relocation, where households respond to evolving risk, policy incentives, and social constraints. Biomedical evidence on hepatic ischemia-reperfusion and remote cardiac injury introduces another form of dynamic system behavior, where local injury can trigger downstream cardiac effects through macrophage inflammasome activation. These three domains share an emphasis on temporal progression, indirect effects, and heterogeneous responses. Sequential recommendation focuses on preference evolution, biomedical research focuses on mechanistic injury pathways, and flood relocation modeling focuses on household movement before disaster. This literature cluster supports cross-domain thinking about adaptive intelligent systems that model sequences, causal pathways, and risk-sensitive decisions.