arXiv:2604.04426v1 Announce Type: new Abstract: Existing research on LLM agent security mainly focuses on prompt injection and unsafe input/output behaviors. However, as agents increasingly rely on third-party tools and MCP servers, a new class of supply-chain threats has emerged, where malicious behaviors are embedded in seemingly benign tools, silently hijacking agent execution, leaking sensitive data, or triggering unauthorized actions. Despite their growing impact, there is currently no comprehensive benchmark for evaluating such threats. To bridge this gap, we introduce SC-Inject-Bench, a large-scale benchmark comprising over 10,000 malicious MCP tools grounded in a taxonomy of 25+ attack types derived from MITRE ATT&CK targeting supply-chain threats. We observe that existing MCP scanners and semantic guardrails perform poorly on this benchmark. Motivated by this finding, we propose ShieldNet, a network-level guardrail framework that detects supply-chain poisoning by observing real network interactions rather than surface-level tool traces. ShieldNet integrates a man-in-the-middle (MITM) proxy and an event extractor to identify critical network behaviors, which are then processed by a lightweight classifier for attack detection. Extensive experiments show that ShieldNet achieves strong detection performance (up to 0.995 F-1 with only 0.8% false positives) while introducing little runtime overhead, substantially outperforming existing MCP scanners and LLM-based guardrails.

arXiv:2511.16145v2 Announce Type: replace-cross Abstract: Time series anomaly detection (TSAD) is a critical data mining task often constrained by label scarcity. Consequently, current research predominantly focuses on Unsupervised Time-series Anomaly Detection (UTAD), relying on increasingly complex architectures to model normal data distributions. However, this algorithm-centric trend often overlooks the significant performance gains achievable from limited anomaly labels available in practical scenarios. This paper challenges the premise that algorithmic complexity is the optimal path for TSAD. Instead of proposing another intricate unsupervised model, we present a comprehensive benchmark and empirical study to rigorously compare supervised and unsupervised paradigms. To isolate the value of labels, we introduce \stand, a deliberately minimalist supervised baseline. Extensive experiments on five public datasets demonstrate that: (1) Labels matter more than models: under a limited labeling budget, simple supervised models significantly outperform complex state-of-the-art unsupervised methods; (2) Supervision yields higher returns: the performance gain from minimal supervision far exceeds the incremental gains from architectural innovations; and (3) Practicality: \stand~exhibits superior prediction consistency and anomaly localization compared to unsupervised counterparts. These findings advocate for a paradigm shift in TSAD research, urging the community to prioritize data-centric label utilization over purely algorithmic complexity. The code and benchmark are publicly available at https://github.com/EmorZz1G/STAND.

Background: Early detection of Alzheimer disease (AD) is essential for timely intervention; yet, diagnostic performance varies widely across modalities and datasets. Recent multimodal artificial intelligence (AI) models have made significant progress, but the evidence base remains fragmented due to heterogeneous datasets, modeling frameworks, and reporting quality. Objective: This systematic review aimed to analyze studies on multimodal AI models for AD diagnosis, prognosis, and risk prediction over 5 years. We evaluated dataset characteristics, modality combinations, modeling strategies, performance metrics, and methodological limitations. We further discuss real-world implications and translational pathways. Methods: Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines, we systematically searched PubMed, IEEE Xplore, Scopus, ACM Digital Library, Cochrane, and arXiv, with the final datasets last searched on November 15, 2025. Studies applying multimodal machine learning or deep learning to AD, mild cognitive impairment, and dementia outcomes were included, whereas studies using a single modality or lacking sufficient methodological detail were excluded. QUADAS-2 (Revised Quality Assessment of Diagnostic Accuracy Studies tool) assessed risk of bias. Extracted performance results were synthesized across 4 major multimodal dataset families. Results: A total of 66 studies met the inclusion criteria. Across datasets, multimodal models consistently outperformed single-modal baselines. Alzheimer’s Disease Neuroimaging Initiative–based diagnosis achieved an average accuracy of 92.5% (SD 3.8%), while mild cognitive impairment–conversion models achieved an average area under the curve (AUC) of 0.922 (SD 0.045), and several fusion architectures reported AUCs above 0.95. In contrast, UK Biobank risk-prediction studies reported an average AUC of 0.84 (SD 0.056), and this reflects performance in large, population-based datasets. DementiaBank speech-language studies achieved an average AUC of 0.813 (SD 0.042), and cross-lingual AD detection achieved an accuracy of 77% (SD 6.5%). Self-collected multimodal datasets demonstrated average accuracies around 96% (SD 2.4%), but their generalizability is limited due to small sample sizes and single-center designs. Conclusions: This systematic review demonstrates that multimodal AI models consistently outperform single-modal models for AD diagnosis, prognosis, and risk prediction by integrating complementary biological, clinical, and behavioral information. Unlike prior reviews, this review provides a unified synthesis across heterogeneous clinical, imaging, genetic, and linguistic datasets, enabling cross-domain comparison of modeling strategies and performance. However, the generalizability of reported performance was limited due to substantial heterogeneity in dataset composition, outcome definitions, and validation, and prevalent risks of bias. By evaluating these factors, this review clarifies where current evidence is robust and where caution is warranted. The findings highlight the need for standardized multimodal benchmarks, transparent evaluation protocols, and clinically grounded model design to enable reliable real-world deployment. Overall, this work advances the field by framing multimodal AI not only as a performance-driven tool but also as a translational framework for equitable, interpretable, and scalable AD diagnosis. Trial Registration: PROSPERO CRD420251241895;

arXiv:2603.08332v1 Announce Type: cross Abstract: The proliferation of fake reviews, often produced by organized groups, undermines consumer trust and fair competition on online platforms. These groups employ sophisticated strategies that evade traditional detection methods, particularly in cold-start scenarios involving newly launched products with sparse data. To address this, we propose the \underline{D}iversity- and \underline{S}imilarity-aware \underline{D}ynamic \underline{G}raph \underline{A}ttention-enhanced \underline{G}raph \underline{C}onvolutional \underline{N}etwork (DS-DGA-GCN), a new graph learning model for detecting fake reviewer groups. DS-DGA-GCN achieves robust detection since it focuses on the joint relationships among products, reviews, and reviewers by modeling product-review-reviewer networks. DS-DGA-GCN also achieves adaptive detection by integrating a Network Feature Scoring (NFS) system and a new dynamic graph attention mechanism. The NFS system quantifies network attributes, including neighbor diversity, network self-similarity, as a unified feature score. The dynamic graph attention mechanism improves the adaptability and computational efficiency by captures features related to temporal information, node importance, and global network structure. Extensive experiments conducted on two real-world datasets derived from Amazon and Xiaohongshu demonstrate that DS-DGA-GCN significantly outperforms state-of-the-art baselines, achieving accuracies of up to \textbf{89.8\% and 88.3\%}, respectively.

arXiv:2603.08398v1 Announce Type: cross Abstract: In this work, we reveal that Large Language Models (LLMs) possess intrinsic behavioral plasticity-akin to chameleons adapting their coloration to environmental cues-that can be exposed through token-conditional generation and stabilized via reinforcement learning. Specifically, by conditioning generation on carefully selected token prefixes sampled from responses exhibiting desired behaviors, LLMs seamlessly adapt their behavioral modes at inference time (e.g., switching from step-by-step reasoning to direct answering) without retraining. Based on this insight, we propose Token-Conditioned Reinforcement Learning (ToCoRL), a principled framework that leverages RL to internalize this chameleon-like plasticity, transforming transient inference-time adaptations into stable and learnable behavioral patterns. ToCoRL guides exploration with token-conditional generation and keep enhancing exploitation, enabling emergence of appropriate behaviors. Extensive experiments show that ToCoRL enables precise behavioral control without capability degradation. Notably, we show that large reasoning models, while performing strongly on complex mathematics, can be effectively adapted to excel at factual question answering, which was a capability previously hindered by their step-by-step reasoning patterns.

arXiv:2602.17686v2 Announce Type: replace-cross Abstract: Distilling Chain-of-Thought (CoT) reasoning from large language models into compact student models presents a fundamental challenge: teacher rationales are often too verbose for smaller models to faithfully reproduce. Existing approaches either compress reasoning into single-step, losing the interpretability that makes CoT valuable. We present a three-stage curriculum learning framework that addresses this capacity mismatch through progressive skill acquisition. First, we establish structural understanding via masked shuffled reconstruction. Second, we apply Group Relative Policy Optimization (GRPO) on masked completion tasks, enabling the model to discover its own balance between accuracy and brevity. Third, we identify persistent failure cases and guide the student to internalize teacher knowledge through targeted rewriting, again optimized with GRPO. Experiments on GSM8K demonstrate that our approach enables Qwen2.5-3B-Base to achieve an 11.29 percent accuracy improvement while reducing output length by 27.4 percent, surpassing both instruction-tuned variants and prior distillation methods.