2786 papers
This paper proposes Conservative Proxy Prompting (CPP), a parameter-efficient framework that enhances the reliability of frozen vision-language models under missing modalities by constructing feature-space proxy representations and conservatively regulating their contribution to avoid misleading signals.
This study demonstrates that while visual-based page retrieval significantly improves evidence selection and downstream answer quality in Document Visual Question Answering compared to lexical methods, substantial performance gaps persist due to answer generation limitations even when the correct page is provided, and hybrid fusion offers no additional benefit over strong visual retrievers alone.
AgentCare-Guard is a capability-constrained reference monitor that secures agentic FHIR transactions by verifying workload identity, binding approvals to specific payloads and contexts, and ensuring atomic execution, thereby achieving 100% defense against authorization-boundary attacks in synthetic evaluations where existing models failed.
This paper presents a multi-agent clinical decision support system that leverages MIMIC-IV data and large language models to perform real-time patient similarity retrieval and generate temporally-aware clinical analyses through a six-agent orchestration workflow.
SynNat-BERT is an interpretable, dual-modal self-supervised framework that leverages 1.7 million unlabeled molecules from synthetic and natural-product spaces to achieve robust blood-brain barrier permeability prediction and successfully identify novel permeable scaffolds, outperforming existing baselines in both accuracy and generalization.
This paper proposes and evaluates an interpretable Digital Twin-safe scheduling framework for deadline-aware task allocation in dynamic fog-edge computing, demonstrating that a lightweight analytical Digital Twin achieves competitive performance with a hand-crafted heuristic and significantly outperforms value-based deep reinforcement learning baselines while providing robust safety guarantees and predictive diagnostics.
This paper proposes a bilateral game-theoretic framework for heterogeneous data centers that jointly optimizes task migration and acceptance decisions by modeling load-dependent computing efficiency and rejection penalties, thereby achieving superior system utility and energy efficiency compared to existing static or unilateral approaches.
This paper proposes SOC2I, an AI-driven framework that adapts Security Operation Center capabilities to Unmanned Aerial Vehicles by integrating real-time anomaly detection, risk quantification, and autonomous mitigation to bridge the gap between conventional network defense and aerial cyber-physical security.
Execution Governance AI (EGA) V9 is a deterministic runtime framework that enables trustworthy autonomous workflows by combining replay consistency, provenance verification, and fail-closed containment to achieve 100% detection of execution divergence with zero false positives or negatives and no additional language-model overhead, while transparently acknowledging specific limitations in post-incident restrictions and side-effect guarantees.
This paper introduces Freeze-to-Confirm (F2C), a recovery strategy for stateful streaming open-vocabulary segmentation that temporarily halts adaptation to confirm persistent risk before restoring model states, thereby significantly improving performance across multiple benchmarks while maintaining low latency.