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

2026-05-13

AI Summary

This work addresses the challenge of modeling multi-domain dynamic graphs, which exhibit substantial heterogeneity in both semantic content and temporal patterns, rendering unified representation difficult and often leading to negative transfer under conventional pretraining-finetuning paradigms. To overcome these limitations, we propose DyGFM, the first foundation model tailored for multi-domain dynamic graphs. DyGFM introduces a novel dual-branch pretraining architecture that decouples semantic and temporal modeling, a cross-domain routing mechanism based on distributional discrepancy, and a lightweight difference-conditioned prompt generator. This design effectively mitigates negative transfer and enables efficient adaptation across domains. Extensive experiments demonstrate that DyGFM significantly outperforms twelve state-of-the-art baselines on node classification and link prediction tasks across multiple continuous dynamic graph benchmarks, achieving both superior performance and computational efficiency.

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.

Problem

• dynamic graph
• multi-domain
• foundation model
• negative transfer
• pre-training

Innovation

• Dynamic Graph Foundation Model
• Semantic-Temporal Decoupling
• Divergence-Conditioned Prompting
• Cross-Domain Routing
• Negative Transfer Mitigation