Semantic IDs are crucial in generative recommendation, but they face a fundamental limitation: temporal information is not well incorporated. Time influences recommendations implicitly through session construction heuristics, preference alignment, or sequence order, while existing semantic ID learning remains entirely time-agnostic. This design conflates interactions under distinct temporal contexts into identical semantic representations, assuming that item semantics and user intent are temporally stationary. This assumption misaligns with real-world scenarios where evolving interaction rhythms play a central role. This work investigates how and when to explicitly incorporate time into semantic IDs for generative recommendation. First, we characterize the design space along three orthogonal dimensions of temporal signals and present a unified framework, ChronoID, for time-aware semantic ID learning. Then, by contributing a new time-explicit generation recommendation benchmark, ChronoID answers critical questions: what is the effective way to infuse time, how to design the architecture, and where the gains come from.
Blogger's Review: The introduction of ChronoID provides a fresh perspective on generative recommendation systems, highlighting the significance of temporal factors. By explicitly integrating time signals, ChronoID not only enhances recommendation accuracy but also points the way for future research, making it a worthy area for deeper exploration.