Retrieval-Augmented Generation enhances large language models by incorporating external knowledge, but deploying it in sensitive scenarios risks privacy leakage via malicious prompts. To address this, we propose a multi-agent framework that sanitizes retrieved content through semantic rewriting. By employing three specialized agents for privacy extraction, semantic analysis, and reconstruction, our approach collaboratively removes sensitive identifiers while preserving the semantic core. We evaluate the framework on the ChatDoctor and Wiki-PII datasets across six large language models. Experimental results demonstrate a significant reduction in privacy leakage under targeted attacks. For instance, we reduced targeted information exposure in LLaMA-3-8B from 144 instances in the baseline to just 1. Furthermore, we maintain strong contextual fidelity with a BLEU-1 score of 0.122, outperforming the existing SAGE method's 0.117. Finally, the framework operates as an asynchronous preprocessing module, introducing no additional latency to online inference, as all rewriting is executed as a one-time offline preprocessing step. To promote reproducibility, the source code of this work is publicly available at https://github.com/foursoils/Privacy-Preserving-RAG.
Bibliometric analysis is a powerful means of evaluating research. This article employs a sequential mixed-methods approach to enrich the understanding of bibliometric analysis and presents a structured guide to maximize the technique’s benefits. A thematically synthesized narrative commentary on bibliometric review papers, based on input from 53 senior editorial members of leading hospitality and tourism journals, constitutes the study’s foundation. Cutting-edge literature comprising 195 bibliometric investigations is also considered. Our evidence-based guidance provides scholars with a roadmap for conducting state-of-the-art bibliometric analyses. These suggestions are intended to improve the method’s quality and accessibility, affirming its utility. Ultimately, this paper seeks to enhance bibliometric studies’ clarity, rigor, and efficacy by presenting forward-looking, real-world insights. Its step-by-step guide will facilitate decision-making and advance the field’s theoretical and managerial understanding.
A comprehensive understanding of the slurry diffusion mechanism and grouting performance in complex fractured rocks is of great significance for guiding engineering practice and improving the scientific basis of grouting construction. In this study, a novel polyacrylate latex-modified cement (PLMC) slurry with high fluidity and injectability was developed, and its time-dependent viscosity characteristics and hydration behavior were systematically investigated. Based on a low-field nuclear magnetic resonance (LF-NMR) testing system, PLMC slurry diffusion experiments under various conditions were conducted on complex fractured sandstones, followed by seepage tests on the grout-consolidated specimens. The evolution patterns of key parameters, including effective grouting time, grout filling rate and total grout intake, were quantitatively analyzed, and the effects of fracture aperture and fracture development level on the anti-seepage performance of the consolidation body were examined. The results indicate that both fracture aperture and fracture development level show a linear positive correlation with the total grout intake, while the filling rate first increases and then decreases. Increasing grouting pressure significantly shortens the effective grouting time and improves the filling rate, whereas increasing the polymer-cement ratio prolongs the effective grouting time and reduces the filling rate, with negligible effect on the total grout intake. The permeability of the grout-consolidated body increases almost linearly with fracture aperture, but exhibits no clear trend with fracture development level. The findings of this study provide theoretical guidance and technical references for the selection and optimization of grouting parameters in complex fractured rock masses.
As artificial intelligence (AI) increasingly takes on the product designer role, an important yet understudied issue is how consumers evaluate the visual characteristics of AI design. This study examines how visual complexity shapes consumers’ willingness to pay (WTP). On the basis of four experiments involving diverse product contexts (food, household products, and apparel) and different design objects (products and packaging), we find that consumers are willing to pay more for visually simple AI designs than visually complex designs. Drawing on signaling theory and cue utilization theory, we demonstrate that visual simple (vs. complex) AI designs increase willingness to pay by enhancing perceived restraint, which subsequently increases perceived luxuriousness. Importantly, this effect does not emerge for luxury brands, suggesting that consumers rely more heavily on visual simplicity as a value signal when brand-based luxury cues are relatively weak. As a result, the findings offer directional guidance when firms consider the incorporation of AI-generated designs—particularly mainstream brands. Together, these findings contribute to the literature by shedding light on consumer aesthetic judgments in the domain of AI design.
Multimodal Large Language Models (MLLMs) excel at understanding generic visual content, such as landscapes, objects, and events, thanks to extensive datasets and advanced training regimes. However, their effectiveness in medical applications remains limited due to the inherent discrepancies between data and tasks in medical scenarios and those in the general domain. Existing medical MLLMs face the following critical deficiencies: 1) inadequate coverage of medical knowledge beyond imaging; 2) elevated propensity for hallucinations due to suboptimal data curation; and 3) limited reasoning capacity tailored to complex medical tasks. To address these challenges, we first propose a comprehensive data-curation procedure that 1) efficiently acquires rich medical knowledge data not only from medical imaging but also from extensive medical texts and general domain data; and 2) synthesizes high-quality medical captions, visual question answering, and reasoning samples. Leveraging the curated data, we build a multimodal dataset imbued with extensive medical knowledge and develop our medical-specialized MLLM, Lingshu-Med, which undergoes multi-stage training to embed the medical expertise and enhance task-solving capabilities progressively. We also investigate reinforcement learning with verifiable rewards to further refine Lingshu-Med's medical reasoning abilities. For rigorous assessment, we introduce MedEvalKit, a unified evaluation framework that consolidates the leading multimodal and textual medical benchmarks for standardized, fair, and efficient model assessment. On three core medical tasks-multimodal QA, textual QA, and radiology report generation, Lingshu-Med consistently outperforms existing multimodal baselines in most tasks. Moreover, we conduct five case studies drawn from real-world clinical scenarios that illustrate its practical utility in medical contexts.
