THGFM: Dual-Branch Temporal Heterogeneous Graph Fusion Model
Abstract
Temporal heterogeneous graphs offer a natural abstraction for dynamic relational systems in which diverse node and relation types co-exist and evolve over time. Learning on such graphs requires jointly modeling cross-type structural heterogeneity and the temporal dynamics of interactions, yet existing methods still struggle to reconcile parameter-efficient cross-type transfer with relation-aware specialization, and typically inject time only as additive features outside the attention kernel. We propose \textbf{THGFM}, a web-scale temporal heterogeneous graph fusion model that addresses both limitations within a unified dual-path architecture. THGFM couples a \textit{Shared-Space Temporal Attention} branch for parameter-efficient cross-type transfer with a \textit{Relational Type-Partitioned Temporal Attention} branch for relation-aware specialization, and integrates them through \textit{Dual-Path Relational--Shared Fusion}, instantiated with \textit{Type-Conditioned Non-Competitive Gated Sum Fusion}: a adaptive mechanism that assigns independent, type-conditioned feature-wise gates to the shared and specialized branches, allowing both to be amplified or suppressed without zero-sum competition. To directly incorporate relative time into the attention score, THGFM further introduces \textit{Rotary Temporal Attention}, which rotates queries and keys by half-phases of relative time before matching. THGFM consistently outperforms baseline graph transformer models on academic graphs benchmarks, delivering a +3.25% six-task mean gain, with peak relative gains of +12.37% on OAG-CS PV, +4.87% on PF-L2, and +1.18% on PF-L1, and +4.24%, +3.73%, and +4.61% on OGBN-MAG, HTAG-ArXiv, and HTAG-DBLP, respectively.
Related
- Sinkhorn-Optimized Capacity-Constrained Temporal Heterogeneous Graph Model for Startup–Investor Matching
- Benign Overfitting under Heterogeneous Input Fusion
- Global minimax risk and acquisition laws for heterogeneous information fusion
- Graph-Split Bayesian Causal Forest for Spatial Heterogeneous Treatment Effect Estimation
- Systematic modelling of a large-scale, historicised, well-detailed building land use, with the fusion of massive and heterogeneous datasets
BibTeX
@misc{peng2026thgfm,
title = {{THGFM: Dual-Branch Temporal Heterogeneous Graph Fusion Model}},
author = {Yixin Peng and Diego Collarana and Er Jin and Stefan Decker},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2607.27303},
eprint = {2607.27303},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2607.27303},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - THGFM: Dual-Branch Temporal Heterogeneous Graph Fusion Model
AU - Yixin Peng
AU - Diego Collarana
AU - Er Jin
AU - Stefan Decker
PY - 2026
DA - 2026/09/28
PB - arXiv
DO - 10.48550/arxiv.2607.27303
UR - https://arxiv.org/abs/2607.27303
AB - Temporal heterogeneous graphs offer a natural abstraction for dynamic relational systems in which diverse node and relation types co-exist and evolve over time. Learning on such graphs requires jointly modeling cross-type structural heterogeneity and the temporal dynamics of interactions, yet existing methods still struggle to reconcile parameter-efficient cross-type transfer with relation-aware specialization, and typically inject time only as additive features outside the attention kernel. We propose \textbf{THGFM}, a web-scale temporal heterogeneous graph fusion model that addresses both limitations within a unified dual-path architecture. THGFM couples a \textit{Shared-Space Temporal Attention} branch for parameter-efficient cross-type transfer with a \textit{Relational Type-Partitioned Temporal Attention} branch for relation-aware specialization, and integrates them through \textit{Dual-Path Relational--Shared Fusion}, instantiated with \textit{Type-Conditioned Non-Competitive Gated Sum Fusion}: a adaptive mechanism that assigns independent, type-conditioned feature-wise gates to the shared and specialized branches, allowing both to be amplified or suppressed without zero-sum competition. To directly incorporate relative time into the attention score, THGFM further introduces \textit{Rotary Temporal Attention}, which rotates queries and keys by half-phases of relative time before matching. THGFM consistently outperforms baseline graph transformer models on academic graphs benchmarks, delivering a +3.25% six-task mean gain, with peak relative gains of +12.37% on OAG-CS PV, +4.87% on PF-L_2, and +1.18% on PF-L_1, and +4.24%, +3.73%, and +4.61% on OGBN-MAG, HTAG-ArXiv, and HTAG-DBLP, respectively.
ER -