Fake news has caused significant harm and disruption across various sectors of society. With the rapid advancement of the Internet and social media platforms, both academic and industrial communities have shown growing interest in multimodal fake news detection. In this work, we propose MERF (Multimodal Evidence Retrieval and Forensics with LVLM), a unified framework for multimodal fake news detection that leverages the reasoning capabilities of Large Vision-Language Models (LVLMs). While LVLMs outperform traditional Large Language Models (LLMs) in processing multimodal content, our study reveals that their reasoning abilities remain limited in the absence of sufficient supporting evidence. MERF addresses this challenge by integrating web-based content retrieval, reverse image search, and image manipulation detection into a coherent pipeline, enabling the model to generate informed and explainable veracity judgments. Specifically, our approach performs cross-modal consistency checking, retrieves corroborative information for both textual and visual content, and applies forensic analysis to detect potential visual forgeries. The aggregated evidence is then fed into the LVLM, facilitating comprehensive reasoning and evidence-based decision-making. Experimental results on two public benchmark datasets—Weibo and Twitter—demonstrate that MERF consistently outperforms state-of-the-art baselines across all major evaluation metrics, achieving substantial improvements in accuracy, robustness, and interpretability.
The rapid proliferation of Internet-of-Things (IoT) systems has led to increasingly heterogeneous, resource-constrained, and security-sensitive software deployments. In this context, detecting vulnerabilities in embedded and system-level IoT code has become particularly critical, as even a single exploitable flaw may compromise entire networks or devices. Vulnerability detection in real-world software continues to pose significant challenges due to the intricate nature of program semantics, the diversity of vulnerability patterns, and the limited explainability offered by current machine learning models. Although Large Language Models (LLMs) have exhibited remarkable capabilities in comprehending and reasoning about source code, their effectiveness is frequently compromised by inadequate domain knowledge and instances of hallucinated outputs. To address these limitations, we propose VulKnow, a retrieval-augmented framework for vulnerability detection that harnesses structured vulnerability knowledge alongside multi-stage prompt-based reasoning. Specifically, VulKnow first constructs a structured knowledge base derived from authoritative sources such as CWE/CVE reports and standard library specifications. Sub-sequently, it conducts context-aware knowledge retrieval and integrates the retrieved items into an LLM-based initial filtering module. To ensure high-confidence and verifiable results, we design a multi-stage reasoning pipeline that progressively validates vulnerabilities through semantic prompts and consistency checks. Experiments conducted on multiple real-world datasets (Linux, Qemu, Big-Vul) demonstrate that VulKnow significantly enhances precision, recall, and explainability compared to existing static analysis tools as well as LLM-only baselines. This positions VulKnow as a reliable solution for practical vulnerability auditing scenarios.
To address the challenges associated with identifying causal relationships among event mentions in the event causality identification (ECI) task, ECI has emerged as a pivotal area of research for comprehending event structures. Recent studies have leveraged Transformer-based models, augmented by auxiliary components, to develop effective contextual representations for causality prediction. A critical step in ECI models involves transforming intricate event context representations into causal label representations, thereby facilitating the logical score calculations necessary for both training and inference. However, existing models frequently depend on simplistic feedforward networks for this transformation process, which often struggle to bridge the semantic gap between complex event contexts and target causal labels, particularly in linguistically nuanced scenarios. To address these limitations, we propose Contrastive Learning for Event Causality Identification (CLECI), an innovative ECI framework that enhances representation learning through the integration of contrastive learning techniques, a generator-discriminator mechanism with causal label embeddings. In contrast to traditional direct transformation methods, CLECI generates latent causal label embeddings that filter out irrelevant information while aligning with potential label representations. By incorporating contrastive learning principles, CLECI further augments the discriminative capability of event representations by constructing positive and negative pairs of events. Experimental evaluations conducted on the EventStoryLine (ESL), Causal-TimeBank (CTB), and MECI datasets demonstrate that CLECI achieves competitive performance, with F1-score improvements of 4.3%, 7.9%, and 2.5%, respectively, compared with the strongest baseline methods, while maintaining strong robustness in complex and noisy multilingual event contexts.
Event representation in text is basic task for natural language processing. In this paper, an enhanced event representation framework using contrastive learning based on Gaussian embedding (EventGE) is proposed. To make the representation robustly for downstream tasks, Gaussian distribution for each event density embedding is constructed through specific Multi-layer Perceptron (MLP). Also we feed the Gaussian distribution for positive and negative samples to contrastive learning network to gain the accuracy representation by setting the loss function. The experimental results show the accuracy rate is improved significantly compared to the current state-of-the-art methods with some public datasets.