cs papers | Gist.Science

Client-Side Ephemeral De-Identification of Protected Health Information (PHI) in Multi-Center Clinical Trial Analysis and Large Language Model Workflows: Validating Zero-Trust Data Sanitization Under HIPAA Safe Harbor Section 164.514(b)

This paper presents and validates Zero-Trust Data Sanitization (ZTDS), a client-side, on-device architectural framework that performs real-time, HIPAA Safe Harbor-compliant de-identification of Protected Health Information within volatile memory before transmission, thereby enabling secure, zero-trust utilization of public Large Language Models in clinical research without requiring Business Associate Agreements or risking data exfiltration.

Ilya Sibiryakov2026-09-22

💻 computer science

An Imbalance-Aware and Structure-Enriched Framework for SMS Spam Detection Using Ensemble Learning

This paper proposes a novel hybrid framework for SMS spam detection that integrates multi-criteria decision making for feature selection, SMOTE for addressing class imbalance, and a Stacking-based ensemble classifier to achieve superior performance in precision, F1-score, and AUC on benchmark datasets.

Mahdi Sarbazi, Mohammad Fathi, Keyhan Khamforoosh2026-09-22

💻 computer science

Federated PPO: Federated Proximal Policy Optimization for Multi-Robot Collision Avoidance

This paper proposes Federated Proximal Policy Optimization (FedPPO), a privacy-preserving federated learning approach that enables efficient and adaptable multi-robot collision avoidance in both homogeneous and heterogeneous systems by aggregating locally trained model parameters to overcome the limitations of independent and shared policy methods.

Xing An, Limeng Chao, Celimuge Wu2026-09-22

💻 computer science

OntoCacheRAG: Ontology-Driven Selective Cache Invalidation for Knowledge-Graph-Augmented Retrieval Systems

OntoCacheRAG is an ontology-driven framework that resolves the trade-off between correctness and efficiency in Knowledge Graph-augmented Retrieval-Augmented Generation systems by employing subsumption-aware reasoning to perform fine-grained, selective cache invalidation, thereby eliminating the need for costly full-cache flushing while ensuring semantic freshness.

Nimas Ayu Untariyati, Kusworo Adi, Aris Puji Widodo, M. Teduh Uliniansyah2026-09-22

💻 computer science

Testing Frontier LLMs on Indian Statutory Interpretation: The Indian Construction Canon Benchmarking (ICCB) -Pilot Benchmark

This paper introduces the ICCB-Pilot benchmark to evaluate frontier LLMs on Indian statutory interpretation, revealing that while models can often predict correct legal outcomes, they fail to correctly identify and apply the underlying canons of construction, suggesting their reasoning relies on surface pattern-matching rather than genuine jurisprudential understanding.

Yashu Bansal, Gaurav Dahiya, Aditi Tiwari, Akshobhya N Saralaya, Dana Susan Kurian, Devyani Singh, Nidhi Singh K V2026-09-22

💻 computer science

A Domain-Tuned Multimodal Agent for NWP Forecast Intelligence

This paper presents a domain-tuned multimodal agent built on a fine-tuned LLaMA4-Scout model and a specialized image analysis module, designed to enhance operational weather forecasting at the National Centre for Medium Range Weather Forecasting (NCMRWF) by integrating meteorological text, technical reports, and visual data into a comprehensive intelligent assistant.

Avinash Chalumuri, Pranav Singh, Anitha Gera, Ashish Routray, Preveen Kumar Devarajan, V S Prasad2026-09-21

💻 computer science

Detecting Evidence of LLM Contributions to Open Access Biomedical Literature: A Bibliometric Analysis

This bibliometric analysis of nearly 3 million open-access biomedical articles reveals a significant post-2022 increase in LLM-associated terminology, indicating rapid adoption of generative AI tools and highlighting the urgent need for transparency and detection methods to preserve scientific integrity.

Gabriel M. Peterson, Marilyn H. Oermann, Jacqueline K. Owens, Gary F. Templeton, Hannah Bailey, Heather Carter-Templeton2026-09-21

💻 computer science

Classification-Aware and DSIS-Targeted Path Editing Based on the Theory of Network Wave for Wireless Multi-Hop Networks

This paper proposes a classification-aware and DSIS-targeted path editing framework based on the Theory of Network Wave that optimizes wireless multi-hop routes by strategically substituting, inserting, or deleting relays to minimize interference-spacing and improve throughput or delay while adhering to strict resource and structural constraints.

Penghui Wang, Bo Li2026-09-21

💻 computer science

An Explainable Multimodal Framework for Adolescent Short Video Addiction Risk Prediction

This study proposes an explainable multimodal framework that integrates diverse behavioral and psychological data with a dual-attention mechanism and XGBoost to dynamically predict adolescent short-video addiction risks and their negative impacts on sleep, academics, emotions, and social adaptation.

Qing Bian2026-09-21

💻 computer science

Generative AI Usage Patterns and User Profiles Among Researchers: Evidence from a Large-Scale Survey in China

This study utilizes large-scale survey data from Chinese researchers to extend the Unified Theory of Acceptance and Use of Technology with machine learning, identifying three distinct GenAI user profiles—"Pioneering Explorers," "Pragmatic Coping Users," and "Alienated Users"—that reveal how academic standing, institutional pressures, and ethical concerns shape divergent adoption patterns in the context of AI for Science.

weitong yuan, jiahao zheng2026-09-21

← Previous Next →