This book constitutes the proceedings of the 10th China Conference on Knowledge Graph and Semantic Computing, CCKS 2025, held in Fuzhou, China, during September 1921, 2025.

The 22 full papers presented in this book were carefully reviewed and selected from 112 submissions. They were organized into the following topical sections: Knowledge Graph Construction and Integration; Large Models Enhanced by Knowledge Graphs; Applications of Knowledge Graphs and Large Models/Agents; Open Resources for Knowledge Graphs and Large Models; and Evaluations.

Inhalt

.- Knowledge Graph Construction and Integration.

.- A Cross-Subgraph Attention Fusion and Comparison Method for Contrastive Learning Based Knowledge Graph Completion.

.- A Preliminary Attempt to Generate a Sichuan Dialect Handbook by LLMs.

.- EDREL: Document-level Relation Extraction with Evidence and Logical Rules.

.- Knowledge Retrieval-Augmented Interest learning for Recommendation.

.- Large Models Enhanced by Knowledge Graphs.

.- Multi-granularity Hierarchical RAG for Welding Parameter Recommendation.

.- Lag-Relative Sparse Attention In Long Context Training.

.- VIKA: Vectorized Indispensable Knowledge-subgraph Augmentation for Large Language Models.

.- Traff-LLM: A Spatio-Temporal Knowledge-Guided Large Language Model for Traffic Flow Prediction.

.- Applications of Knowledge Graphs and Large Models/Agents.

.- MAEPS: Multi-Agent Event Prediction System Based on Human Expert Team Collaboration Simulation.

.- Zero-shot Instruction Generation via Dual-Alignment Instruction Wrappers with Summary-Text fused instruction wrappers.

.- FEFT: A Feedback-enhanced Evaluation Fine-tuning Framework for Financial Report Summarization.

.- Iterative Generation Method for Factual QA in Large Language Models Based on Semantic Entropy Verification.

.- Open Resources for Knowledge Graphs and Large Models.

.- Autism Children Education Knowledge Graph: Construction and Validation.

.- C-Voice: Culturally-grounded Multi-dimensional Alignment of LLMs with Chinese Social Values.

.- Evaluations.

.- HiParse-RAG: A High-Fidelity Document Parsing and Hybrid Retrieval Multi-Model Fusion Framework for Complex Academic Question Answering.

.- HybriDoc: An Adaptive Multi-Path Framework for End-to-End Document Structure Extraction.

.- Pre-training for Document Structure Extraction with Lightweight Model Architecture.

.- Robust Detection of AI-Generated Text: Insights on Evolving LLMs and Adversarial Data.

.- A Fact-Aware Cascaded Framework for Dynamic-granularity Timeline Summarization.

.- Multi-Agent for Dynamic-Granularity Timeline Summarization.

.- Advancing Grounded Multimodal NER via Self-Reflective Prompt Refinement and Visual Noise Mitigation.

.- ReFineG: Synergizing Small Supervised Models and LLMs for Low-Resource Grounded Multimodal NER.