Researchers have developed a new framework called M3TR for predicting the popularity of micro-videos. The framework addresses two limitations in existing methods: their failure to capture complex temporal patterns and their reliance on static content similarity. M3TR uses a novel temporal-aware retrieval process that identifies historically relevant videos based on both their multi-modal content and their popularity trajectories. This is achieved through a combination of fine-grained temporal modeling and a temporal-aware retrieval engine. The framework has been tested on two real-world datasets, showing significant improvements over previous methods in predicting micro-video popularity. --- Why it matters: This matters to AI researchers because it provides a new approach for addressing the challenges of long-term prediction in complex systems like social media platforms. By accurately predicting video popularity, M3TR can help content creators and platforms optimize their content and improve user engagement. Source: https://arxiv.org/abs/2411.15455