Title: Neural Sequential Survival Modeling for Dynamic Conversion Likelihood in Delayed-Response Advertising Abstract: We introduce a neural framework that models delayed conversions by jointly learning user intent from pre-click sequences and dynamically updating conversion likelihood using post-click behavioral traces. Leveraging GRU-based attention and survival analysis, our approach adapts hazard functions in real time based on post-click user activity—transforming static delay assumptions into dynamic, behavior-aware predictions. Evaluated on real e-commerce data, the model improves CVR ranking accuracy by up to 8.02% AUC and significantly enhances calibration under sparse conversion signals. This work bridges sequential user modeling and survival analysis to capture evolving purchase intent, offering a behavior-sensitive solution for delayed-response ad targeting.