This book includes the revised selected papers presented at the CCF 19th International Conference on Service Science, ICSS 2026, held in Macau, China, during May 15-17, 2025.

The 33 full papers presented here were carefully reviewed and selected from 91 submissions. ICSS features a unique mix of academic, industrial, and crossdisciplinary topics, and provides a platform for the presentation and exchange of research results and practical experiences as well as education development on serviceology.

Inhalt

.- Paper Session 1
.- Diffusion-Enhanced Multi-Modal Anomaly Diagnosis with Adaptive Feature Fusion in Microservice Systems.
.- Automous End-to-End SOH Prediction Services for Battery Systems via Temporal-Contrastive Representation Learning.
.- MsFormer: Enabling Robust Predictive Maintenance Services for Industrial Devices.
.- Deep Reinforcement Learning-Based in Satellite Edge Computing.
.- Governance-as-a-Service for Asset-Intensive Education Systems: Unified Asset Records and Executable Audit Predicates.
.- Dependency-Constrained Task Offloading Method in Vehicular Edge Computing.
.- TDSE-Net: A Lightweight Causal Dilated Convolutional Method for Edge-Oriented Battery SOH Estimation.
.- Paper Session 2
.- Future-Aware Service Composition Verification in Cloud-Network Integrated MEC Environments.
.- Delay-Aware GPU Scheduling for Deep Learning in Kubernetes: Toward Better Fairness and Efficiency.
.- FedGout: A Multi-modal Federated Learning Framework for Gout Stage Prediction.
.- Paper Session 3
.- A Privacy-Preserving Secure Data Sharing Scheme in Federated Learning with Robustness Guarantee.
.- CoAL-RAG: A Complexity-Aware Legal Retrieval-Augmented Generation Method.
.- Research on Retrieval-Augmented Large Language Models for Digital Textbooks.
.- LDFA: A Novel LLM-Driven Hyper-Heuristic Method for Task Offloading in Vehicular Edge Computing.
.- HA-EDD: Heterogeneity-Aware Edge Data Distribution with Bandwidth and Deadline Constraints.
.- A Siamese Mamba Network for Nonferrous Metal Ore Classification Using Low- and High-energy X-ray Spectral Images.
.- Data Extraction Attacks Against Diffusion-Based Language Model Services.
.- TACO: Cost-Efficient LLM Reasoning Services via Dynamic Convergence Detection.
.- Paper Session 4
.- Trading the Future: A Dynamic Deep Reinforcement Learning Framework for Adaptive Portfolio Management.
.- DCSRec: Disentangled Contrastive Collaborative Filtering-based Web Service Recommendation.
.- Commit Message Generation Service for Web Repositories via Code Analysis and Adversarial Learning.
.- A Railway Track Extraction Method from UAV Aerial Imagery Based on MFC-UNet.
.- Dynamic Uncertainty-Aware Cascading of Large and Small Models for Toxic Text Detection.
.- Beyond Text Similarity: Evaluating Vision-Language Models for Image-to-Code Services.
.- Collaborative Knowledge Distillation for Meta-path Correlations-aware Service Classification.
.- Improving the Fairness of News Recommendation Services for Cold Users.
.- Paper Session 5
.- MoECLIP: Object-Agnostic Mixture of Experts Prompt Learning for Zero-shot Anomaly Detection.
.- Enabling Intelligent Services with Heterogeneous and Multimodal Knowledge by RAG.
.- Attributing Success: Measuring Individual Module Contribution in LLM Workflows.
.- Deep Reinforcement Learning-Based Microservice Deployment for Cloud-Edge Collaborative Systems.
.- Semantic Stories for Next Activity Prediction with Retrieval-Augmented Generation and Fine Tuning.
.- TabColor: A System of Table Question Answering with Visual Annotations for Distinguishing Header Types.
.- MAMDOC: Multi-Agent Multimodal Document Visual Question Answering.