The unchecked proliferation of online information underscores the need for transparent and efficient fact-checking systems. While some methods leverage structured representations like knowledge graphs, significant gaps persist in efficiently and accurately extracting relevant information. To address this, we introduce EZ-Check, a novel framework that combines explainability with a fast, reliable, and effective querying mechanism for knowledge graphs, leveraging resources such as ConceptNet. At its core, EZ-Check integrates structured knowledge graph relationships with advanced language models to assess textual authenticity and extract relevant information precisely. EZ-Check achieves more than a 10% relative F1 improvement on FEVER and UKP Snopes compared to representative claim-only and evidence-based baselines, including BERT-as-KB, GPT-3.5 Turbo, KGAT, and TARSA, while reducing inference time by approximately 50%. On fact-checking datasets such as FEVER and UKP Snopes, it achieved state-of-the-art results, surpassing baselines without relying on external evidence sentences. Moreover, EZ-Check delivers clear, interpretable justifications, setting a new standard for both performance and explainability in fact verification based on our testing against baselines.

The spread of infectious diseases in networked populations poses significant challenges for public health intervention strategies. Traditional centrality-based and heuristic network dismantling approaches prioritize highly connected nodes but often fail to account for individual mortality risk, limiting their effectiveness in minimizing overall fatalities. While recent advances in machine learning have improved intervention strategies, existing models largely focus on reducing disease transmission rather than directly targeting mortality outcomes. To address this gap, we propose a reinforcement learning-based framework that integrates graph representation learning to identify and remove high-risk nodes, thereby maximizing network fragmentation while minimizing overall deaths. The framework is trained using synthetic networks and evaluated on five synthetic and four real-world datasets, benchmarking its performance against state-of-the-art network dismantling methods [graph dismantling with machine learning (GDM), generalized network dismantling (GND), and graph enhanced reinforcement learning (GERL)]. Experimental results demonstrate that the proposed method consistently outperforms baseline approaches, particularly in scale-free and community-structured networks, where targeted node removal significantly weakens network connectivity and suppresses epidemic spread. Moreover, in real-world networks, the method achieves lower cumulative death rates and higher epidemic thresholds, highlighting its robustness in controlling disease propagation. By incorporating mortality risk into network representation learning, the proposed framework offers a scalable, adaptive, and socially responsible approach to epidemic mitigation, misinformation control, and network resilience enhancement.

Knowledge Graphs (KGs), such as DBpedia and ConceptNet, enhance Natural Language Processing (NLP) applications by providing structured information. However, extracting accurate data from KGs is challenging due to issues in entity detection, disambiguation, and relation classification, which often lead to errors and inefficiencies. We introduce Attention2Query (A2Q) , an attention-driven approach that directly ranks and selects the most relevant facts, thus minimizing error propagation. A2Q centres on three key contributions: (1) Focused Node Selection , which streamlines graph traversal; (2) Global Attention Alignment , improving retrieval by comparing facts against the query text; and (3) Contextual Re-ranking , enabling on-the-fly adjustments of fact importance based on evolving query context. Experimental results across multiple tasks and datasets show that A2Q substantially outperforms baseline methods, including those in zero-shot settings, achieving higher retrieval accuracy with reduced computational overhead. • Novel on-the-go fact retrieval framework for KGs using attention mechanisms. • The reranking strategy improves retrieval accuracy with minimal context. • A versatile approach reduces retrieval of non-relevant data from KGs. • A2Q outperforms baselines in data quality and efficiency in zero-shot settings.

Network coding is an effective idea to boost the capacity of wireless networks, and a variety of studies have explored its advantages in different scenarios. However, there is not much analytical study on throughput and end-to-end delay of network coding in multi-hop wireless networks considering the specifications of IEEE 802.11 Distributed Coordination Function. In this paper, we utilize queuing theory to propose an analytical framework for bidirectional unicast flows in multi-hop wireless mesh networks. We study the throughput and end-to-end delay of inter-flow network coding under the IEEE 802.11 standard with CSMA/CA random access and exponential back-off time considering clock freezing and virtual carrier sensing, and formulate several parameters such as the probability of successful transmission in terms of bit error rate and collision probability, waiting time of packets at nodes, and retransmission mechanism. Our model uses a multi-class queuing network with stable queues, where coded packets have a nonpreemptive higher priority over native packets, and forwarding of native packets is not delayed if no coding opportunities are available. Finally, we use computer simulations to verify the accuracy of our analytical model.

In recent years, a significant amount of research has been conducted to explore the benefits of network coding in different scenarios, from both theoretical and simulation perspectives. In this paper, we utilize queueing theory to propose an analytical framework for bidirectional unicast flows in multi-hop wireless mesh networks, and study throughput of inter-flow network coding. We analytically determine performance metrics such as the probability of successful transmission in terms of collision probability, and feedback mechanism and retransmission. Regarding the coding process, our model uses a multi-class queueing network where coded packets are separated from native packets and have a non-preemptive higher priority over native packets, and both queues are in a stable state. Finally, we use simulations to verify the accuracy of our analytical model.

In recent years, network coding has emerged as an innovative method that helps wireless network approaches its maximum capacity, by combining multiple unicasts in one broadcast. However, the majority of research conducted in this area is yet to fully utilize the broadcasting nature of wireless networks, and still assumes fixed route between the source and destination that every packet should travel through. This assumption not only limits coding opportunities, but can also cause buffer overflow in some specific intermediate nodes. Although some studies considered scattering of the flows dynamically in the network, they still face some limitations. This paper explains pros and cons of some prominent research in network coding and proposes FlexONC (Flexible and Opportunistic Network Coding) as a solution to such issues. The performance results show that FlexONC outperforms previous methods especially in worse quality networks, by better utilizing redundant packets spread in the network.

There is increased demand to run applications on mobile devices in recent years. Although many of these alluring applications are resource-intensive, they expect to get the same performance on mobile devices as on powerful non-mobile computers. On the other hand, considerations such as weight, size, and mobility impose constraints on mobile devices and restrict their processor speed, memory size and battery lifetime. Cyber foraging ameliorates this performance disparity by utilizing nearby non-mobile computers called surrogates to run the whole or parts of applications, which are offloaded from mobile devices. However, cyber foraging is not suitable for all circumstances. In this paper, we propose a mechanism to determine the best location, either a mobile device or best surrogate(s) around, to run an application by calculating the cost of offloading the task of running the application to each location according to the context's metrics such as specifications of mobile device, surrogates, network, and application. Experimental results show that our proposed mechanism almost always selects the best one between local execution of the application on the mobile device and task offloading to surrogate(s) according to the current input size of the application, with respect to latency and energy consumption.