The paper discusses challenges in Temporal Interaction Graphs (TIGs) used in systems like e-commerce and social networks. Existing models face issues with timely predictions and lack versatility for diverse tasks. To address this, the authors introduce TIGPrompt, a framework that integrates with existing TIG models and bridges the temporal and semantic gaps. The framework uses a temporal prompt generator to provide temporally-aware prompts for various tasks. The effectiveness and efficiency of TIGPrompt are demonstrated through multiple experiments.
Publication date: 12 Feb 2024
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2402.06326
