数据库：2025–2026 年度综述

An annual survey of the cs.DB arXiv category, covering research published from September 2025 – August 2026. Built on a close-reading set of 100 papers and organised into 36 chapters, it maps the year by the field's research sub-directions.

Book framework (share of the year's output):

• Data Dependency Discovery and Table Mining (63.3%) — Focuses on efficient algorithms for discovering various data dependencies (functional, inclusion, conditional) and mining patterns in structured tables.
• LLM-Enhanced Text-to-SQL and GraphRAG (27.5%) — Addresses natural language to SQL translation, schema evolution robustness, and retrieval-augmented generation over structured and semi-structured data using Large Language Models.
• Approximate Nearest Neighbor and Vector Search (5.3%) — Optimizes high-dimensional vector indexing, filtering, and search efficiency for similarity retrieval in large-scale databases.
• Query Optimization and Execution Engine Design (2.4%) — Improves query processing performance through learned estimators, dynamic scheduling, and novel execution structures for analytical and temporal workloads.
• Relational Deep Learning and Graph Integration (0.9%) — Integrates deep learning models with relational database structures and property graphs to enhance querying and representation learning.
• Benchmarking and Evaluation Frameworks (0.3%) — Develops comprehensive benchmarks and evaluation methodologies for assessing database system performance, fairness, and query translation accuracy.
• Database Theory and Algorithmic Foundations (0.3%) — Explores fundamental algorithms for database repair, linearizability, and theoretical models like Yannakakis' algorithm.
• Theoretical Enumeration and Ranking Analysis (0.1%) — Investigates fundamental problems in listing projected trees, ranking stability, and explaining group biases in data models.

Each sub-direction is covered across three chapters — how it framed its questions, the strength of its key findings, and its open debates and frontiers — so the reader gets a traceable picture of what this part of the field actually produced over the year.