cs papers | Gist.Science

Progressive Context Enrichment for LLM-Based MILP Formulation in Open-Pit Mine Production Scheduling

This paper proposes a progressive context-enrichment framework that enhances the reliability of Large Language Models in generating Mixed Integer Linear Programming formulations for open-pit mine production scheduling by systematically augmenting prompts with simulator data and code annotations, revealing that while simulator code improves objective accuracy, annotated code is superior for ensuring constraint feasibility.

Mustavi Ibne Masum, Thiago Eustaquio Alves de Oliveira, Mahzabeen Emu2026-07-23

💻 computer science

AI-Driven Spline-Based Segmentation of Handwritten Physico-Mathematical Documents

This paper presents an AI-driven segmentation pipeline for handwritten physico-mathematical documents that utilizes shape-preserving quadratic splines to model oscillatory baselines and generate curvature-adaptive cutting masks, thereby enhancing detection accuracy and structural reconstruction for downstream OCR tasks.

Vasyl Zalizko2026-07-23

💻 computer science

Metric-Based Few-Shot Learning Framework for User-Defined Vocabulary Registration in Electromyography-Based Silent Speech Interfaces

This study proposes a metric-based few-shot learning framework, specifically utilizing Prototypical Networks, to enable efficient and computationally lightweight registration of user-defined vocabularies in EMG-based silent speech interfaces without requiring repeated model retraining.

Hyeseong Jeon, Jangjay Sohn, Chang-Hwan Im2026-07-23

💻 computer science

Validated cross-source synthesis of structural links in a compiled medical knowledge base: an independent literature check with a discrimination control

This paper reports an independently validated demonstration that a proprietary engine can correctly synthesize structural medical links across multiple sources by surfacing nine verified connections from external literature and proving selectivity through a discrimination control, establishing a reliable foundation for future computational discovery without hallucination.

Youssef Ghaly2026-07-23

💻 computer science

Gradient Alignment Dynamics in Mixture-of Experts: The Gaussian Attractor, Laminar Flow Regime, and Output-Space Isolation

This paper establishes a unified theoretical framework for Mixture-of-Experts (MoE) that identifies cross-expert gradient alignment as the root cause of expert collapse, introduces the Gradient Pathway Hypothesis and Reynolds number diagnostics to characterize these dynamics, and demonstrates that architectural output-space isolation via Mosaic Tile hard routing effectively eliminates catastrophic forgetting and expert collapse across vision and language tasks.

庆君张2026-07-23

💻 computer science

Achieving 100% Accuracy in Steel Surface Defect Classification Using a Custom High-Performance CNN

This paper proposes a custom high-performance CNN architecture that achieves perfect 100% accuracy and state-of-the-art performance on the NEU-CLS dataset for classifying six types of steel surface defects, demonstrating the superiority of task-specific designs over general-purpose models in industrial quality control.

Mohamma S2026-07-23

💻 computer science

Introducing HPC-Enabled Computational Research Training in Africa: A Pilot Study

This paper presents a pilot study of the Center for Computational Research and Education (CCRE), a training initiative in Africa that successfully guided a small cohort of early-career researchers through a structured, mentored pathway to develop high-performance computing skills despite significant infrastructure and mentorship barriers.

James Afful2026-07-23

💻 computer science

InVAER: An AI Framework for Real-Time Accident Detection via Adaptive Hyperparameter Optimization and Multi-Cue Validation

The paper introduces InVAER, a novel AI framework that combines adaptive hyperparameter optimization and multi-cue spatio-temporal validation to achieve real-time, high-precision accident detection with automated geospatial emergency coordination, significantly reducing false positives and response delays on edge devices.

Rahul Panja, MD Ajij, Prasenjit Paul2026-07-23

💻 computer science

AI–IoT–Digital Twin Framework for Predictive Maintenance in Smart Manufacturing

This study proposes and validates a novel five-layer AI–IoT–Digital Twin framework that integrates real-time data, predictive analytics, and augmented reality to achieve significant improvements in manufacturing efficiency, sustainability, and human-machine collaboration while aligning with Industry 5.0 values.

P.R. Sekhar Reddy, S. Saravanan2026-07-23

💻 computer science

Blockchain-focused Cyber Security for the Detection and Protection of Denial of Service (DoS) Attacks

This study proposes a blockchain-integrated cybersecurity framework that utilizes a Convolutional Neural Network (CNN) model to achieve 98.3% detection accuracy and 100% prevention capability against Denial of Service (DoS) attacks in AI-based network systems.

Eshetie Teka Mitiku, Tibebe Beshah, Habtamu Kassaw2026-07-23

← Previous Next →