信息检索：2025–2026 年度综述

An annual survey of the cs.IR arXiv category, covering research published from September 2025 – August 2026. Built on a close-reading set of 132 papers and organised into 37 chapters, it maps the year by the field's research sub-directions.

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

• Information Retrieval and Dense Multimodal Benchmarking (20.9%) — Develops dense retrieval models and multimodal benchmarks for measuring retrieval quality on heterogeneous content.
• LLM-Based Recommendation and Agentic Ranking (11.1%) — Uses large language models to drive recommendation and ranking, including agentic interfaces over user-item interactions.
• Agentic AI Search and Semantic Vector Retrieval (10.9%) — Explores agentic search agents and semantic vector retrieval for open-ended and task-oriented information access.
• Multi-Agent Systems and Multi-Hop Reasoning Frameworks (10.4%) — Investigates multi-agent architectures and multi-hop reasoning for decomposing complex queries across several retrieval steps.
• Knowledge Graphs and Retrieval Augmented Question Answering (9.7%) — Combines knowledge graphs with retrieval-augmented question answering to support fact-grounded responses.
• Advances in Multimodal and Cross-Domain Sequential Recommendation (8.1%) — Models user behaviour across multimodal and cross-domain settings to improve sequential recommendation.
• Large Language Models and Their Applications (7.8%) — Surveys large language model capabilities and their application to information-seeking and text-processing tasks.
• Generative Recommendation and Retrieval (5.7%) — Explores generative and diffusion-based approaches to recommendation and retrieval using semantic identifiers.
• Reinforcement and Contrastive Learning for Multimodal Representation (5.6%) — Applies reinforcement and contrastive learning to learn multimodal representations for ranking and retrieval.
• Retrieval-Augmented Generation and Multimodal Frameworks (5.3%) — Studies retrieval-augmented generation pipelines that ground language models in external knowledge, spanning multimodal and cross-modal retrieval.
• Advances in Conversational and Agentic Recommender Systems (4.6%) — Investigates conversational and agentic recommender systems that interact with users to refine recommendations.

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.