Egocentric visual query localization is vital for embodied AI and VR/AR, yet remains challenging due to camera motion, viewpoint changes, and appearance variations. We present EAGLE, a novel framework that leverages episodic appearance- and geometry-aware memory to achieve unified 2D-3D visual query localization in egocentric vision. Inspired by avian memory consolidation, EAGLE synergistically integrates segmentation guided by an appearance-aware meta-learning memory (AMM), with tracking driven by a geometry-aware localization memory (GLM). This memory consolidation mechanism, through structured appearance and geometry memory banks, stores high-confidence retrieval samples, effectively supporting both long- and short-term modeling of target appearance variations. This enables precise contour delineation with robust spatial discrimination, leading to significantly improved retrieval accuracy. Furthermore, by integrating the VQL-2D output with a visual geometry grounded Transformer (VGGT), we achieve a efficient unification of 2D and 3D tasks, enabling rapid and accurate back-projection into 3D space. Our method achieves state-of-the-art performance on the Ego4D-VQ benchmark.

Inspired by the success of large language models, there is a trend toward developing graph foundation models to conduct diverse downstream tasks in various domains. However, current models often require extra fine-tuning to apply their learned structural and semantic representations to new graphs, which limits their versatility. Recent breakthroughs in zero-shot inductive reasoning on knowledge graphs (KGs), offer us a new perspective on extending KG reasoning to general graph applications. In this paper, we introduce SCR, a unified graph reasoning framework designed to train on knowledge graphs and effectively generalize across a wide range of graph tasks and domains. We begin by designing the task-specific KG structures to establish a unified topology for different task formats. Then we propose semantic-conditioned message passing, a novel mechanism addressing the inherent semantic isolation in traditional KG reasoning, by jointly modeling structural and semantic invariance patterns in graph representations. Evaluated on 38 diverse datasets spanning node-, link-, and graph-level tasks, SCR achieves substantial performance gains over existing foundation models and supervised baselines, demonstrating its remarkable efficacy and adaptability.

We study Graph Neural Networks (GNNs)-based embedding techniques for knowledge graph (KG) reasoning. For the first time, we link the path redundancy issue in the state-of-the-art path encoding-based models to the transformation error in model training, which brings us new theoretical insights into KG reasoning, as well as high efficacy in practice. On the theoretical side, we analyze the entropy of transformation error in KG paths and point out query-specific redundant paths causing entropy increases. These findings guide us to maintain the shortest paths and remove redundant paths for minimized-entropy message passing. To achieve this goal, on the practical side, we propose an efficient Graph Percolation process motivated by the percolation phenomenon in Fluid Mechanics, and design a lightweight GNN-based KG reasoning framework called Graph Percolation Embeddings (GraPE)1. GraPE outperforms state-of-the-art methods in both transductive and inductive reasoning tasks, while requiring fewer training parameters and less inference time.

Knowledge graphs are more widely utilized to enhance recommendability and explainability. Reinforcement learning agents built to wander around the knowledge graph have been successfully applied in recommendation systems in a form of multi-hop relation reasoning. Some previous multi-hop methods relied on reinforcement learning of discrete actions, making agent space design challenging and a lack of clarity in the meaning of actions because of inconsistent action. To solve the aforementioned issues, we propose Continuous-action Walking-tendency Interest-oriented Path Reasoning (CWIPR), a novel and pioneering method that uses continuous actions provided by reinforcement learning agents to predict inference relations and the next entity. Meanwhile, to better interact with the knowledge graph through continuous actions, we firstly propose a graph search algorithm called the walking tendency algorithm. Moreover, we introduce an interest-oriented reward as the intrinsic reward that encourages the agent to balance the tendency between exploring the most similar entities and exploring the correct recommendation type to achieve more precise recommendations. We extensively evaluate our method on three real-world datasets from Amazon and obtain favorable performance compared with state-of-the-art methods.

Graph contrastive learning is a representative self-supervised graph learning that has demonstrated excellent performance in learning node representations. Despite the extensive studies on graph con- trastive learning models, most existing models are tailored to static graphs, hindering their application to real-world graphs which are often dynamically evolving. Directly applying these models to dynamic graphs brings in severe efficiency issues in repetitively updating the learned embeddings. To address this challenge, we propose IDOL, a novel contrastive learning framework for dynamic graph representation learning. IDOL conducts the graph propagation process based on a specially designed Personalized PageRank algorithm which can capture the topological changes incrementally. This effectively eliminates heavy recomputation while maintain- ing high learning quality. Our another main design is a topology-monitorable sampling strategy which lays the foundation of graph contrastive learning. We further show that the design in IDOL achieves a desired performance guarantee. Our experimental results on multiple dynamic graphs show that IDOL outperforms the strongest baselines on node classification tasks in various performance metrics.

Knowledge Graph (KG) Reasoning plays a vital role in various applications by predicting missing facts from existing knowledge. Inductive KG reasoning approaches based on Graph Neural Networks (GNNs) have shown impressive performance, particularly when reasoning with unseen entities and dynamic KGs. However, such state-of-the-art KG reasoning approaches encounter efficiency and scalability challenges on large-scale KGs due to the high computational costs associated with subgraph extraction - a key component in inductive KG reasoning. To address the computational challenge, we introduce TIGER, an inductive GNN training framework tailored for large-scale KG reasoning. TIGER employs a novel, efficient streaming procedure that facilitates rapid subgraph slicing and dynamic subgraph caching to minimize the cost of subgraph extraction. The fundamental challenge in TIGER lies in the optimal subgraph slicing problem, which we prove to be NP-hard. We propose a novel two-stage algorithm SiGMa to solve the problem practically. By decoupling the complicated problem into two classical ones, SiGMa achieves low computational complexity and high slice reuse. We also propose four new benchmarks for robust evaluation of large-scale inductive KG reasoning, the biggest of which performs on the Freebase KG (encompassing 86M entities, 285M edges). Through comprehensive experiments on state-of-the-art GNN-based KG reasoning models, we demonstrate that TIGER significantly reduces the running time of subgraph extraction, achieving an average 3.7× speedup relative to the basic training procedure.

The recent emergence of the promising paradigm of the Social Internet of Things (SIoT) is a result of an intelligent amalgamation of the social networking concepts with the Internet of Things (IoT) objects (also referred to as “things”) in an attempt to unravel the challenges of network discovery, navigability, and service composition. This is realized by facilitating the IoT objects to socialize with one another, i.e., similar to the social interactions amongst human beings. A fundamental issue that mandates careful attention is to thus establish, and over time, maintain trustworthy relationships amongst these IoT objects. Therefore, a trust framework for SIoT must include object-object interactions, the aspects of social relationships, credible recommendations, etc., however, the existing literature has only focused on some aspects of trust by primarily relying on the conventional approaches that govern linear relationships between input and output. In this paper, an artificial neural network-based trust framework,Trust–SIoT, has been envisaged for identifying the complex non-linear relationships between input and output in a bid to classify trustworthy objects. Moreover,Trust–SIoThas been designed for capturing a number of key trust metrics as input, i.e., direct trust by integrating both current and past interactions, reliability and benevolence of an object, credible recommendations, and the degree of relationship by employing knowledge graph embedding. Finally, we have performed extensive experiments to evaluate the performance ofTrust–SIoTvis-á-vis state-of-the-art heuristics on two real-world datasets. The results demonstrate thatTrust–SIoTachieves a higher F1-score and lower MAE and MSE scores.

Multi-label aerial image classification (MAIC) is a fundamental but challenging task for computer vision-based remote sensing applications. Existing MAIC models suffer from the insufficient semantic information of image and label representations. To this end, we integrate commonsense knowledge into the MAIC task and propose a novel Knowledge-augmented Concept Graph Learning (KCGL) framework. KCGL first collects relevant semantic concepts for each label from a commonsense knowledge graph ConceptNet. With the guidance of semantic concepts, an image decoupling module is employed to extract concept-specific image features from the input image. Then, KCGL constructs an individual concept graph for each image, in which nodes are corresponding to concept-specific image features and edges are their relations extracted from ConceptNet. Finally, the classification probability on each label is computed in the specific concept graph via a GCN-based encoder-decoder model. Experimental results prove that the proposed KCGL outperforms existing state-of-the-art MAIC models on two aerial image datasets.

Knowledge graph embedding (KGE) has shown great potential in automatic knowledge graph (KG) completion and knowledge-driven tasks. However, recent KGE models suffer from high training cost and large storage space, thus limiting their practicality in real-world applications. To address this challenge, based on the latest findings in the field of Contrastive Learning, we propose a novel KGE training framework called Hardness-aware Low-dimensional Embedding (HaLE). Instead of the traditional Negative Sampling, we design a new loss function based on query sampling that can balance two important training targets, Alignment and Uniformity. Furthermore, we analyze the hardness-aware ability of recent low-dimensional hyperbolic models and propose a lightweight hardness-aware activation mechanism, which can help the KGE models focus on hard instances and speed up convergence. The experimental results show that in the limited training time, HaLE can effectively improve the performance and training speed of KGE models on five commonly-used datasets. After training just a few minutes, the HaLE-trained models are competitive compared to the state-of-the-art models in both low- and high-dimensional conditions.

Link prediction based on knowledge graph embeddings (KGE) has recently drawn a considerable momentum. However, existing KGE models suffer from insufficient accuracy and hardly evaluate the confidence probability of each predicted triple. To fill this critical gap, we propose a novel confidence measurement method based on causal intervention, called Neighborhood Intervention Consistency (NIC). Unlike previous confidence measurement methods that focus on the optimal score in a prediction, NIC actively intervenes in the input entity vector to measure the robustness of the prediction result. The experimental results on ten popular KGE models show that our NIC method can effectively estimate the confidence score of each predicted triple. The top 10% triples with high NIC confidence can achieve 30% higher accuracy in the state-of-the-art KGE models.

Link prediction based on knowledge graph embeddings (KGE) aims to predict new triples to automatically construct knowledge graphs (KGs). However, recent KGE models achieve performance improvements by excessively increasing the embedding dimensions, which may cause enormous training costs and require more storage space. In this paper, instead of training high-dimensional models, we propose MulDE, a novel knowledge distillation framework, which includes multiple low-dimensional hyperbolic KGE models as teachers and two student components, namely Junior and Senior. Under a novel iterative distillation strategy, the Junior component, a low-dimensional KGE model, asks teachers actively based on its preliminary prediction results, and the Senior component integrates teachers’ knowledge adaptively to train the Junior component based on two mechanisms: relation-specific scaling and contrast attention. The experimental results show that MulDE can effectively improve the performance and training speed of low-dimensional KGE models. The distilled 32-dimensional model is competitive compared to the state-of-the-art high-dimensional methods on several widely-used datasets.

Distantly supervised relation extraction can label large amounts of unstructured text without human annotations for training. However, distant supervision inevitably accompanies with the wrong labeling problem, which can deteriorate the performance of relation extraction. What's more, the entity-pair information, which can enrich instance information, is still underutilized. In the light of these issues, we propose TMNN, a novel Neural Network framework with a Trade-off Mechanism, which combines the feature of text and entity pair on the sentence level to predict relations. Our proposed trade-off mechanism is a probability generation module to dynamically adjust the weights of text and corresponding entity pair for each sentence. Experimental results on a widely used dataset show that the proposed method reduces the noisy labels and achieves substantial improvement over the state-of-the-art methods.