cs papers | Gist.Science

UniPrune: Unified Progressive Visual Token Pruning with Information-Aware Budget Allocation for Efficient LLaVA-Style Vision Language Models

UniPrune is a unified progressive visual token pruning framework that synergistically combines encoder-stage semantic importance-diversity pruning with LLM-stage pyramid-shaped dropping, enhanced by Information-Aware Budget Allocation and Cross-Stage Information Continuity mechanisms, to achieve superior efficiency and performance in LLaVA-style Vision Language Models even at extreme compression ratios.

Jianhua Cui, Meizhou Ding, Ziru Niu, Wei Wang, Lifeng Han, Peng Zhao2026-08-31

💻 computer science

Dual‑Domain Adaptive Fusion with Dynamic Sparse Attention for Low‑Light Image Enhancement

The paper proposes FDSA-Net, a frequency-guided dynamic sparse attention network that integrates an axis-based attention block and a multi-scale dual-domain adaptive fusion module to efficiently enhance low-light images while preserving fine details, achieving state-of-the-art performance on multiple benchmarks.

Huaping Zhou, Hong Chen, Kelei Sun2026-08-31

💻 computer science

AI-Powered Self-Healing Distributed Job Queue for Predictive Failure Detection and Root Cause Diagnosis in Kubernetes

This paper introduces \sys, an AI-powered self-healing framework for Kubernetes job queues that combines hybrid anomaly detection, root cause diagnosis, and reinforcement learning to autonomously predict failures and execute remediations, significantly reducing recovery time and improving failure prediction accuracy compared to existing reactive systems.

Nagaraja Hegde, Jasmine K S2026-08-30

💻 computer science

Conditional Generative Adversarial Network Simulation of Hypertrophic Scar Appearance Across Three Treatments

This study developed and internally evaluated a deep learning system using conditional generative adversarial networks to simulate the post-treatment appearance of hypertrophic scars following surgical excision, autologous fat injection, or micro-plasma radiofrequency, demonstrating preliminary feasibility for enhancing patient communication and expectation management.

Chengfei Li, Shiyi Li, Qiang Fu, Guiwen Zhou, Qian Wu, Fanting Meng, Xiao Xu, Minliang Chen2026-08-29

💻 computer science

Algorithmic Thermostat: LLMs Standardize Political Expression Without Shaping Semantic Stances

This paper proposes an "algorithmic thermostat" framework to demonstrate that while large language models exhibit cyclical fluctuations in their explicit political stances across versions due to alignment mechanisms targeting high-sensitivity topics, their underlying semantic positioning remains relatively stable, indicating that updates standardize expression rather than fundamentally reshape political ideologies.

Ting chen2026-08-28

💻 computer science

Red Panda Catboost Prediction Framework For Detecting Intrusion In Industrial Internet Of Things

This paper proposes a novel Red Panda Catboost Prediction System (RPCPS) that utilizes a red panda-inspired fitness algorithm for optimal feature selection to achieve 99.4% accuracy in detecting intrusion attacks within Industrial Internet of Things (IIoT) environments.

Shevale Rupali Ramdas, Monika Sharad Deshmukh2026-08-28

💻 computer science

An evidence-gated audit algorithm for low-altitude-economy readiness in mountainous Yunnan, China

This paper introduces EGSA, an evidence-gated spatial audit algorithm designed to screen county-level readiness for low-altitude economy development in mountainous Yunnan by prioritizing auditable administrative preparedness and field-investigation targeting over predictive demand forecasting, thereby addressing governance challenges in data-scarce, unevenly developed regions.

Runzhe Liu, Yun He, Dabin Wang2026-08-28

💻 computer science

Adaptive Lower Bound Evaluation for the Permutation Flowshop Scheduling Problem

This paper presents a systematic analysis and adaptive strategies for selecting machine pairs in the LB2 lower bound evaluation of the Permutation Flowshop Scheduling Problem, demonstrating that dynamically adjusting the number and selection of pairs can significantly improve branch-and-bound performance by balancing bound tightness and computational cost.

Alisa Vorokhta, Jan Gmys, Gwen Maudet, Mohand Mezmaz, Grégoire Danoy2026-08-28

💻 computer science

Category-Level Fine-Grained Sketch-Based 3D Shape Retrieval

This paper addresses the challenge of category-level fine-grained sketch-based 3D shape retrieval by introducing a new dataset with 54 sub-categories and proposing a method featuring specialized feature extractors and optimized cross-domain alignment to effectively retrieve 3D shapes matching the fine-grained details of input sketches.

Bingrui Wang, Yuan Zhou, Sun-Yuan Kung2026-08-28

💻 computer science

Action-aligned Representation Geometry in Vision–Language–Action Models

This paper reveals that Vision–Language–Action (VLA) models consistently develop structured, action-aligned latent geometry during training, which serves as a critical organizing principle that correlates with task performance, emerges hierarchically across layers, and is essential for effective action prediction and generalization.

Han Qi, Huangyuan Su, Liuyixin Shao, Na Li, Heng Yang2026-08-28

← Previous Next →