Decoupled and Divergence-Conditioned Prompt for Multi-domain Dynamic Graph Foundation Models

TL;DR

This paper introduces DyGFM, a novel multi-domain dynamic graph foundation model that employs decoupled prompting and divergence-aware mechanisms to improve generalization and reduce negative transfer across diverse dynamic graph domains.

Contribution

The work proposes a new multi-domain dynamic graph foundation model with semantic-temporal decoupling, divergence-conditioned prompting, and a cross-domain routing mechanism, addressing challenges in multi-domain dynamic graph learning.

Findings

DyGFM outperforms 12 state-of-the-art baselines on node classification.

DyGFM achieves superior effectiveness and efficiency in link prediction tasks.

Extensive experiments validate DyGFM's ability to generalize across multiple dynamic graph domains.

Abstract

Dynamic graphs are ubiquitous in real-world systems, and building generalizable dynamic Graph Foundation Models has become a frontier in graph learning. However, dynamic graphs from different domains pose fundamental challenges to unified modeling, as their semantic and temporal patterns are inherently inconsistent, making the multi-domain pre-training difficult. Consequently, the widely used "pretrain-then-finetune" paradigm often suffers from severe negative knowledge transfer. To the best of our knowledge, there exists no multi-domain dynamic GFM. In this work, we propose DyGFM, a Dynamic Graph Foundation Model over multiple domains based on decoupled and divergence-conditioned prompting. To disentangle transferable semantics from the domain-specific dynamics, we introduce a dual-branch pre-training strategy with semantic-temporal decoupling. To alleviate negative transfer during domain adaptation, we further develop a cross-domain routing mechanism with divergence-aware expert selection. To enable efficient downstream fine-tuning, we design a divergence-conditioned prompt generator that injects lightweight, learnable graph prompts tailored to semantic and temporal traits. Extensive experiments on continuous dynamic graph benchmarks demonstrate that DyGFM consistently outperforms 12 state-of-the-art baselines on both node classification and link prediction tasks, achieving superior effectiveness and efficiency.