Knowledge Graph Retrieval-Augmented Generation (KG-RAG) extends the RAG paradigm by incorporating structured knowledge from knowledge graphs, enabling Large Language Models (LLMs) to perform more precise and explainable reasoning. While KG-RAG improves factual accuracy in complex tasks, existing KG-RAG models are often severely overconfident, producing high-confidence predictions even when retrieved sub-graphs are incomplete or unreliable, which raises concerns for deployment in high-stakes domains. To address this issue, we propose Ca2KG, a Causality-aware Calibration framework for KG-RAG. Ca2KG integrates counterfactual prompting, which exposes retrieval-dependent uncertainties in knowledge quality and reasoning reliability, with a panel-based re-scoring mechanism that stabilises predictions across interventions. Extensive experiments on two complex QA datasets demonstrate that Ca2KG consistently improves calibration while maintaining or even enhancing predictive accuracy. The source code can be found at~ https://aisuko.github.io/ca2kg/.

Observation-based trajectory prediction for systems with unknown dynamics is essential in fields such as physics and biology. Most existing approaches are limited to learning within a single system with fixed dynamics patterns. However, many real-world applications require learning across systems with evolving dynamics patterns, a challenge that has been largely overlooked. To address this, we systematically investigate the problem of Continual Dynamics Learning (CDL), examining task configurations and evaluating the applicability of existing techniques, while identifying key challenges. In response, we propose the Mode-switching Graph ODE (MS-GODE) model, which integrates the strengths LG-ODE and sub-network learning with a mode-switching module, enabling efficient learning over varying dynamics. Moreover, we construct a novel benchmark of biological dynamic systems for CDL, Bio-CDL, featuring diverse systems with disparate dynamics and significantly enriching the research field of machine learning for dynamic systems. Our code available at \url{https://github.com/QueuQ/MS-GODE}.

Memory replay based techniques have shown great success for continual learning with incrementally accumulated Euclidean data. Directly applying them to continually expanding networks, however, leads to the potential memory explosion problem due to the need to buffer representative nodes and their associated topological neighborhood structures. To this end, we systematically analyze the key challenges in the memory explosion problem, and present a general framework,i.e., Parameter Decoupled Graph Neural Networks (PDGNNs) with Topology-aware Embedding Memory (TEM), to tackle this issue. The proposed framework not only reduces the memory space complexity from O (ndL) to O (n)1: memory budget, d: average node degree, L: the radius of the GNN receptive field, but also fully utilizes the topological information for memory replay. Specifically, PDGNNs decouple trainable parameters from the computation ego-subnetwork viaTopology-aware Embeddings (TEs), which compress ego-subnetworks into compact vectors (i.e., TEs) to reduce the memory consumption. Based on this framework, we discover a unique pseudo-training effect in continual learning on expanding networks and this effect motivates us to develop a novel coverage maximization sampling strategy that can enhance the performance with a tight memory budget. Thorough empirical studies demonstrate that, by tackling the memory explosion problem and incorporating topological information into memory replay, PDGNNs with TEM significantly outperform state-of-the-art techniques, especially in the challenging class-incremental setting.

Memory replay, which stores a subset of historical data from previous tasks to replay while learning new tasks, exhibits state-of-the-art performance for various continual learning applications on the Euclidean data. While topological information plays a critical role in characterizing graph data, existing memory replay-based graph learning techniques only store individual nodes for replay and do not consider their associated edge information. To this end, based on the message-passing mechanism in graph neural networks (GNNs), we present the Ricci curvature-based graph sparsification technique to perform continual graph representation learning. Specifically, we first develop the subgraph episodic memory (SEM) to store the topological information in the form of computation subgraphs. Next, we sparsify the subgraphs such that they only contain the most informative structures (nodes and edges). The informativeness is evaluated with the Ricci curvature, a theoretically justified metric to estimate the contribution of neighbors to represent a target node. In this way, we can reduce the memory consumption of a computation subgraph from to and enable GNNs to fully utilize the most informative topological information for memory replay. Besides, to ensure the applicability on large graphs, we also provide the theoretically justified surrogate for the Ricci curvature in the sparsification process, which can greatly facilitate the computation. Finally, our empirical studies show that SEM outperforms state-of-the-art approaches significantly on four different public datasets. Unlike existing methods, which mainly focus on task incremental learning (task-IL) setting, SEM also succeeds in the challenging class incremental learning (class-IL) setting in which the model is required to distinguish all learned classes without task indicators and even achieves comparable performance to joint training, which is the performance upper bound for continual learning.

Despite significant advances in graph representation learning, little attention has been paid to the more practical continual learning scenario in which new categories of nodes (e.g., new research areas in citation networks, or new types of products in co-purchasing networks) and their associated edges are continuously emerging, causing catastrophic forgetting on previous categories. Existing methods either ignore the rich topological information or sacrifice plasticity for stability. To this end, we present Hierarchical Prototype Networks (HPNs) which extract different levels of abstract knowledge in the form of prototypes to represent the continuously expanded graphs. Specifically, we first leverage a set of Atomic Feature Extractors (AFEs) to encode both the elemental attribute information and the topological structure of the target node. Next, we develop HPNs to adaptively select relevant AFEs and represent each node with three levels of prototypes. In this way, whenever a new category of nodes is given, only the relevant AFEs and prototypes at each level will be activated and refined, while others remain uninterrupted to maintain the performance over existing nodes. Theoretically, we first demonstrate that the memory consumption of HPNs is bounded regardless of how many tasks are encountered. Then, we prove that under mild constraints, learning new tasks will not alter the prototypes matched to previous data, thereby eliminating the forgetting problem. The theoretical results are supported by experiments on five datasets, showing that HPNs not only outperform state-of-the-art baseline techniques but also consume relatively less memory. Code and datasets are available at https://github.com/QueuQ/HPNs.

Memory replay, which stores a subset of representative historical data from previous tasks to replay while learning new tasks, exhibits state-of-the-art performance for various continual learning applications on Euclidean data. While topological information plays a critical role in characterizing graph data, existing memory replay based graph learning techniques only store individual nodes for replay and do not consider their associated edge information. To this end, we propose a sparsified subgraph memory (SSM), which sparsifies the selected computation graphs into fixed size before storing them into the memory. In this way, we can reduce the memory consumption of a computation subgraph from $\mathcal{O}(d^{L})$ to $\mathcal{O}(1)$, and for the first time enable GNNs to utilize the explicit topological information for memory replay. Finally, our empirical studies show that SSM outperforms state-of-the-art approaches by up to 27.8% on four different public datasets. Unlike existing methods which focus on task incremental learning (task-IL) setting, SSM succeeds in the challenging class incremental learning (class-IL) setting in which a model is required to distinguish all learned classes without task indicators, and even achieves comparable performance to joint training which is the performance upper bound for continual learning. Our code is available at https://github.com/QueuQ/SSM.

Continual learning on graph data, which aims to accommodate new tasks over newly emerged graph data while maintaining the model performance over existing tasks, is attracting increasing attention from the community. Unlike continual learning on Euclidean data ($\textit{e.g.}$, images, texts, etc.) that has established benchmarks and unified experimental settings, benchmark tasks are rare for Continual Graph Learning (CGL). Moreover, due to the variety of graph data and its complex topological structures, existing works adopt different protocols to configure datasets and experimental settings. This creates a great obstacle to compare different techniques and thus hinders the development of CGL. To this end, we systematically study the task configurations in different application scenarios and develop a comprehensive Continual Graph Learning Benchmark (CGLB) curated from different public datasets. Specifically, CGLB contains both node-level and graph-level continual graph learning tasks under task-incremental (currently widely adopted) and class-incremental (more practical, challenging, yet underexplored) settings, as well as a toolkit for training, evaluating, and visualizing different CGL methods. Within CGLB, we also systematically explain the difference among these task configurations by comparing them to classical continual learning settings. Finally, we comprehensively compare state-of-the-art baselines on CGLB to investigate their effectiveness. Given CGLB and the developed toolkit, the barrier to exploring CGL has been greatly lowered and researchers can focus more on the model development without worrying about tedious work on pre-processing of datasets or encountering unseen pitfalls. The benchmark and the toolkit are available through https://github.com/QueuQ/CGLB.

Graph convolutional models have gained impressive successes on skeleton based human action recognition task. As graph convolution is a local operation, it cannot fully investigate non-local joints that could be vital to recognizing the action. For example, actions like typing and clapping request the cooperation of two hands, which are distant from each other in a human skeleton graph. Multiple graph convolutional layers thus tend to be stacked together to increase receptive field, which brings in computational inefficiency and optimization difficulty. But there is still no guarantee that distant joints (e.g. two hands) can be well integrated. In this paper, we propose a context aware graph convolutional network (CA-GCN). Besides the computation of localized graph convolution, CA-GCN considers a context term for each vertex by integrating information of all other vertices. Long range dependencies among joints are thus naturally integrated in context information, which then eliminates the need of stacking multiple layers to enlarge receptive field and greatly simplifies the network. Moreover, we further propose an advanced CA-GCN, in which asymmetric relevance measurement and higher level representation are utilized to compute context information for more flexibility and better performance. Besides the joint features, our CA-GCN could also be extended to handle graphs with edge (limb) features. Extensive experiments on two real-world datasets demonstrate the importance of context information and the effectiveness of the proposed CA-GCN in skeleton based action recognition.

Body joints, directly obtained from a pose estimation model, have proven effective for action recognition. Existing works focus on analyzing the dynamics of human joints. However, except joints, humans also explore motions of limbs for understanding actions. Given this observation, we investigate the dynamics of human limbs for skeleton-based action recognition. Specifically, we represent an edge in a graph of a human skeleton by integrating its spatial neighboring edges (for encoding the cooperation between different limbs) and its temporal neighboring edges (for achieving the consistency of movements in an action). Based on this new edge representation, we devise a graph edge convolutional neural network (CNN). Considering the complementarity between graph node convolution and edge convolution, we further construct two hybrid networks by introducing different shared intermediate layers to integrate graph node and edge CNNs. Our contributions are twofold, graph edge convolution and hybrid networks for integrating the proposed edge convolution and the conventional node convolution. Experimental results on the Kinetics and NTU-RGB+D data sets demonstrate that our graph edge convolution is effective at capturing the characteristics of actions and that our graph edge CNN significantly outperforms the existing state-of-the-art skeleton-based action recognition methods.