Task-Aware Retrieval Augmentation for Dynamic Recommendation

TL;DR

TarDGR addresses generalization gaps in dynamic graph recommendations caused by temporal shifts by introducing a task-aware retrieval augmentation framework. It combines a Task-Aware Evaluation Mechanism to automatically identify beneficial historical subgraphs with a Graph Transformer-based Task-Aware Model that fuses semantic and structural signals, further enhanced by BiSCL pretraining. During inference, TarDGR retrieves task-relevant subgraphs and fuses them with the query to improve representation and robustness across time, evidenced by superior results on Taobao, Koubei, and Amazon datasets. The work demonstrates that task-aligned retrieval increases information relevance for dynamic recommendations, offering a practical path to more robust, time-adaptive systems.

Abstract

Dynamic recommendation systems aim to provide personalized suggestions by modeling temporal user-item interactions across time-series behavioral data. Recent studies have leveraged pre-trained dynamic graph neural networks (GNNs) to learn user-item representations over temporal snapshot graphs. However, fine-tuning GNNs on these graphs often results in generalization issues due to temporal discrepancies between pre-training and fine-tuning stages, limiting the model's ability to capture evolving user preferences. To address this, we propose TarDGR, a task-aware retrieval-augmented framework designed to enhance generalization capability by incorporating task-aware model and retrieval-augmentation. Specifically, TarDGR introduces a Task-Aware Evaluation Mechanism to identify semantically relevant historical subgraphs, enabling the construction of task-specific datasets without manual labeling. It also presents a Graph Transformer-based Task-Aware Model that integrates semantic and structural encodings to assess subgraph relevance. During inference, TarDGR retrieves and fuses task-aware subgraphs with the query subgraph, enriching its representation and mitigating temporal generalization issues. Experiments on multiple large-scale dynamic graph datasets demonstrate that TarDGR consistently outperforms state-of-the-art methods, with extensive empirical evidence underscoring its superior accuracy and generalization capabilities.

Figures (7)

• Figure 1: Comparison of current methods with TarDGR in dynamic graph recommendation.
• Figure 2: Task-Aware Retrieval Recommendation.
• Figure 3: Overview of the TarDGR framework.
• Figure 4: Performance comparison of TarDGR and other RAG methods.
• Figure 5: Training resource experiments for the Graph Transformer-based Aware Model applied to TarDGR.