检索增强生成论文趋势 2023–2026
RAG 与检索增强
统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 1,200 篇被归入该主题的论文。
主要结论
- 检索增强生成论文从 2023 年的 102 篇增长到 2026 年的 549 篇，三年间扩大到约 5.4 倍。
- NeurIPS 是发检索增强生成论文最多的会议，2023–2026 年共 391 篇，占四个会议总量的 33%。
- 交叉方向里涨得最快的是AI 智能体：同时属于两个主题的论文从 2023 年的 0 篇增至 2026 年的 75 篇。
- 在「自然语言」的 5 个主题里，它的 2026 年论文数排第 2；同组增长最快的代码生成是 4.1 倍，检索增强生成是 5.4 倍。
各会议论文数
按年份和会议统计
| 年份 | NeurIPS | ICLR | ICML | CVPR | 合计 |
|---|---|---|---|---|---|
| 2023 | 19 | 29 | 17 | 37 | 102 |
| 2024 | 50 | 34 | 50 | 49 | 183 |
| 2025 | 136 | 111 | 57 | 62 | 366 |
| 2026 | 186 | 129 | 126 | 108 | 549 |
代表论文
按年份和会议均匀挑选，方便快速看到每一届的样貌
- NeurIPS 2026A Theory of Time-Sensitive Language Generation: Sparse Hallucination Beats Mode Collapse
- ICML 2026``Someone Hid It!'': Query-Agnostic Black-Box Attacks on LLM-Based Retrieval
- ICLR 2026AdaCache: Adaptive Caching and Context Augmentation for Efficient LLM Serving
- CVPR 2026ADSeeker A Knowledge-Grounded Reasoning Framework for Industry Anomaly Detection and
- NeurIPS 2025AceSearcher: Bootstrapping Reasoning and Search for LLMs via Reinforced Self-Play
- ICML 2025Are LLMs Prescient? A Continuous Evaluation using Daily News as the Oracle
- ICLR 2025A Theory for Token-Level Harmonization in Retrieval-Augmented Generation
- CVPR 2025AirRoom: Objects Matter in Room Reidentification
- NeurIPS 2024$\texttt{ConflictBank}$: A Benchmark for Evaluating the Influence of Knowledge Conflicts in LLMs
- ICML 2024A Statistical Framework for Data-dependent Retrieval-Augmented Models
- ICLR 2024BTR: Binary Token Representations for Efficient Retrieval Augmented Language Models
- CVPR 2024Evaluating Transferability in Retrieval Tasks: An Approach Using MMD and Kernel Methods
- NeurIPS 2023A Riemannian Exponential Augmented Lagrangian Method for Computing the Projection Robust Wasserstein Distance
- ICML 2023Everyone's Preference Changes Differently: A Weighted Multi-Interest Model For Retrieval
- ICLR 2023Decompositional Generation Process for Instance-Dependent Partial Label Learning
- CVPR 2023Are Binary Annotations Sufficient? Video Moment Retrieval via Hierarchical Uncertainty-Based Active Learning
- NeurIPS 2026Accelerating Neural Network Training with Augmented Koopman Dynamics
- ICML 2026$\tau$-Knowledge: Evaluating Conversational Agents over Unstructured Knowledge
- ICLR 2026Aria: an Agent for Retrieval and Iterative Auto-Formalization via Dependency Graph
- CVPR 2026Animator-Centric Skeleton Generation on Objects with Fine-Grained Details
全部检索增强生成论文
共 186 篇，第 4 / 4 页
- NeurIPS 2026SAGE: Semantic Ambiguity Guided Capacity Expansion for Retrieval-Augmented Generation
- NeurIPS 2026SciHazard: A Benchmark for Measuring Scientific Safety Risks with Decomposed Harm Scoring
- NeurIPS 2026Search-Augmented Masked Diffusion Models for Constrained Generation
- NeurIPS 2026SEISMOS: A Statistical Signal Detection Framework for Semantic Chunking
- NeurIPS 2026ShiftRAG: Bypassing the Textual Bottleneck via Decoupled Learning and Continuous Soft Tokens
- NeurIPS 2026SiliciclasticReservoirs: A Million-Reservoir Dataset and Flow-Matching Foundation Model for 3D Siliciclastic Reservoir Generation
- NeurIPS 2026SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation
- NeurIPS 2026Sinkhorn Based Associative Memory Retrieval Using Spherical Hellinger Kantorovich Dynamics
- NeurIPS 2026SkillCIR: Intent-Guided Skill Composition for Training-Free Composed Image Retrieval
- NeurIPS 2026Source-Causal Control of Historical Context in Longitudinal Radiology Report Generation
- NeurIPS 2026SpecHop: Continuous Speculation for Accelerating Multi-Hop Retrieval Agents
- NeurIPS 2026SPEXT: A Decoupled Multi-Spectral Foundation Model for Earth Observation and Vision-Language Grounding
- NeurIPS 2026Spherical Boltzmann machines: a solvable theory of learning and generation in energy-based models
- NeurIPS 2026SteerCast: Retrieval-Based Latent Steering for Decoder-Only Time Series Forecasting
- NeurIPS 2026The Future of Facts: Tracing the Factual Generation-Verification Gap
- NeurIPS 2026The Key to Going Linear: Analysis-Driven Transformer Linearization
- NeurIPS 2026The Override Gap: A Magnitude Account of Knowledge Conflict Failure in Hypernetwork-Based Instant LLM Adaptation
- NeurIPS 2026The Sharp Directions Are Against You: Curvature Analysis of Activation Steering
- NeurIPS 2026The Text Uncanny Valley: Non-Monotonic Performance Degradation in LLM Information Retrieval
- NeurIPS 2026TIGER-FG: Text-Guided Implicit Fine-Grained Grounding for E-commerce Retrieval
- NeurIPS 2026TKCAM: Text and Keyframe to Camera Trajectory Generation
- NeurIPS 2026TopoGraphRAG-Bench: Evaluating Multimodal GraphRAG on Layout-Grounded Evidence Reasoning
- NeurIPS 2026Towards Explainable Industrial Anomaly Detection via Knowledge-Guided Latent Reasoning
- NeurIPS 2026Towards Generalizable Partially Relevant Video Retrieval
- NeurIPS 2026Train-free Data Poisoning Attack against Retrieval-augmented Diffusion Models
- NeurIPS 2026TRIDENT: Post-Selection Evidence Accountability for Token-Efficient RAG
- NeurIPS 2026TRIM: A Theory of Retrieval with Incremental Memory -- Defect, Benefit, and Critical Horizon
- NeurIPS 2026Understanding Private Evolution as Learning-Augmented Clustering
- NeurIPS 2026VecDBLens: A Modular Framework for Diagnosing Vector Databases Retrieval Pipelines
- NeurIPS 2026VeriContest: A Competitive-Programming Benchmark for Verifiable Code Generation
- NeurIPS 2026VideoMDM: Towards 3D Human Motion Generation From 2D Supervision
- NeurIPS 2026WebNavigator: Global Web Navigation via Interaction Graph Retrieval
- NeurIPS 2026When Does Graph Retrieval Become Answer-Supporting Evidence? A Diagnostic Audit of GraphRAG
- NeurIPS 2026When Poison Meets Structure: Topology-based Defense against Poisoning Attack on Graph-based Retrieval-Augmented Generation
- NeurIPS 2026Where Root Cause Analysis Fails: A Retrieval-Reranking Decomposition
- NeurIPS 2026WorldMemArena: Evaluating Multimodal Agent Memory Through Action–World Interaction
在 Claude、Cursor 里直接问这个趋势
接上 Paperlist MCP 后，写综述、挑方向或准备投稿时，不用离开对话框就能拿到这页的数字：每年各会议的论文数、增长倍数、涨得最快的子方向和代表论文，AI 会按你的问题整理。
可以直接复制这句话问 AI
用 paperlist 查检索增强生成在 NeurIPS 2023–2026 的论文趋势，哪一年涨得最快？
AI 会调用的工具：get_topic_stats {"topic": "retrieval-augmented-generation"}
常见问题
- 2023–2026 年四大会议一共有多少篇检索增强生成论文？
- 共 1,200 篇，按年份分别是 2023 年 102 篇、2024 年 183 篇、2025 年 366 篇、2026 年 549 篇。
- NeurIPS 2026 有多少篇检索增强生成论文？
- NeurIPS 2026 有 186 篇检索增强生成论文，上一届是 136 篇。
- ICLR 2026 有多少篇检索增强生成论文？
- ICLR 2026 有 129 篇检索增强生成论文，上一届是 111 篇。
- ICML 2026 有多少篇检索增强生成论文？
- ICML 2026 有 126 篇检索增强生成论文，上一届是 57 篇。
- CVPR 2026 有多少篇检索增强生成论文？
- CVPR 2026 有 108 篇检索增强生成论文，上一届是 62 篇。
- 哪个会议的检索增强生成论文最多？
- NeurIPS，2023–2026 年共 391 篇。