Jin, Yiqiao, et al. "Predicting Information Pathways Across Online Communities." arXiv preprint arXiv:2306.02259 (2023).

This research tackles the complex problem of predicting how information travels across online communities, a challenge known as community-level information pathway prediction (CLIPP). By focusing on the diffusion of YouTube videos on Reddit, the study gathers extensive multi-modal datasets to understand how content spreads and to construct community influence graphs (CIGs). The researchers introduce a cutting-edge dynamic graph framework named INPAC, which utilizes CIGs to account for the changing nature and multi-modal aspects of content propagation. INPAC demonstrates superior performance in predicting content transmission trajectories in both familiar (warm-start) and unfamiliar (cold-start) scenarios, outshining seven other benchmark models. This work has profound implications for amplifying beneficial information and curtailing misinformation online.