MM-DyGraph Benchmark for Multimodal Dynamic Graph Learning
November 9, 2026
MM-DyGraph is a new benchmark for learning on dynamic graphs where nodes and edges contain text and image signals. The dataset includes fine-grained temporal evolution and demonstrates that simple feature concatenation often leads to prediction inconsistencies.
HOW THIS AFFECTS YOU
●
builderYou can use this benchmark to test how well your models handle graphs that change over time with mixed text and image data.
●
researcherThis provides a standardized way to evaluate how multimodal signals interact in evolving graph structures.