In real-world multimodal web scenarios, graph-structured data often arrives in a streaming manner, making graph continual learning a crucial paradigm for continuously modeling such evolving structures. However, existing graph continual learning methods still face two fundamental challenges:
- Semantic-Structural Separation: Graph-based methods excel at modeling topological relationships but neglect deep semantics.
- Imbalanced Knowledge Transfer: Existing models fail to effectively leverage general knowledge gained from early tasks to benefit subsequent new tasks.
To address these issues, we propose a novel framework, UNleash Large Language Models PotentIal for Graph ConTinual Learning (UNIT). By fine-tuning a large language model only on the first task, we bridge the distributional gap between the pre-trained LLM corpus and the target task dataset to enhance the adaptability of LLMs for graph-structured tasks.
Meanwhile, we introduce an uncertainty-aware anchor generation mechanism to effectively preserve representative knowledge across tasks, avoiding neglect of universal knowledge learned from previous tasks. Additionally, we introduce structural confluence modeling to explicitly integrate graph topology information with semantic information, enhancing the collaborative capabilities between semantic understanding and structural modeling. Extensive experiments demonstrate that our proposed method achieves state-of-the-art performance in the graph continual learning task.
Blogger's Review: The UNIT framework proposed in this paper innovatively combines large language models with effective integration into graph structures, providing a novel solution for graph continual learning. By addressing the combination of semantics and structure, UNIT significantly enhances the efficiency of knowledge transfer, showcasing substantial practical application potential. Maintaining continuous knowledge updates in dynamic data environments will be a key research direction in the future.