Mochi: Aligning Pre-training and Inference for Efficient Graph Foundation Models via Meta-Learning
This work addresses the efficiency and task unification challenges in Graph Foundation Models for graph learning practitioners, offering a practical solution that significantly reduces training time without sacrificing performance.
Mochi introduces a meta-learning-based training framework for Graph Foundation Models that aligns pre-training objectives with downstream inference, achieving competitive or superior performance across 25 graph datasets while requiring 8-27 times less training time than the strongest baseline.
We propose Mochi, a Graph Foundation Model that addresses task unification and training efficiency by adopting a meta-learning based training framework. Prior models pre-train with reconstruction-based objectives such as link prediction, and assume that the resulting representations can be aligned with downstream tasks through a separate unification step such as class prototypes. We demonstrate through synthetic and real-world experiments that this procedure, while simple and intuitive, has limitations that directly affect downstream task performance. To address these limitations, Mochi pre-trains on few-shot episodes that mirror the downstream evaluation protocol, aligning the training objective with inference rather than relying on a post-hoc unification step. We show that Mochi, along with its more powerful variant Mochi++, achieves competitive or superior performance compared to existing Graph Foundation Models across 25 real-world graph datasets spanning node classification, link prediction, and graph classification, while requiring 8$\sim$27 times less training time than the strongest baseline.