稠密检索论文趋势 2023–2026

主要结论

• 稠密检索论文从 2023 年的 60 篇增长到 2026 年的 162 篇，三年间扩大到约 2.7 倍。
• CVPR 是发稠密检索论文最多的会议，2023–2026 年共 147 篇，占四个会议总量的 37%。
• 交叉方向里涨得最快的是检索记忆：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 21 篇。
• 在「信息检索」的 5 个主题里，它的 2026 年论文数排第 1；同组增长最快的交叉编码重排是 3.0 倍，稠密检索是 2.7 倍。

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 2026Half-Truths Break Similarity-Based Retrieval
• ICML 2026Deep Discriminative Structure Proxy Hashing for Cross-modal Retrieval
• ICLR 2026A Dense Subset Index for Collective Query Coverage
• CVPR 2026Beyond Caption-Based Queries in Video Moment Retrieval
• NeurIPS 2025Dense Associative Memory with Epanechnikov Energy
• ICML 2025Nonparametric Modern Hopfield Models
• ICLR 2025Contextual Document Embeddings
• CVPR 2025Adapting Dense Matching for Homography Estimation with Grid-based Acceleration
• NeurIPS 2024CARE: a Benchmark Suite for the Classification and Retrieval of Enzymes
• ICML 2024Compute Better Spent: Replacing Dense Layers with Structured Matrices
• ICLR 2024Adaptive Retrieval and Scalable Indexing for k-NN Search with Cross-Encoders
• CVPR 2024Dense Optical Tracking: Connecting the Dots
• NeurIPS 2023Bypassing spike sorting: Density-based decoding using spike localization from dense multielectrode probes
• ICML 2023Everyone's Preference Changes Differently: A Weighted Multi-Interest Model For Retrieval
• ICLR 2023Automating Nearest Neighbor Search Configuration with Constrained Optimization
• CVPR 2023Are Binary Annotations Sufficient? Video Moment Retrieval via Hierarchical Uncertainty-Based Active Learning
• NeurIPS 2026Hyperparameter Transfer for Dense Associative Memories
• ICML 2026Dense associative memory for Gaussian distributions
• ICLR 2026Beyond Instance-Level Alignment: Dual-Level Optimal Transport for Audio-Text Retrieval
• CVPR 2026Beyond Tie Points Satellite Image Block Adjustment based on Dense

全部稠密检索论文

共 162 篇，第 2 / 4 页

51. ICLR 2026A Biologically Plausible Dense Associative Memory with Exponential Capacity
52. ICLR 2026A Dense Subset Index for Collective Query Coverage
53. ICLR 2026Avey-B
54. ICLR 2026Beyond Instance-Level Alignment: Dual-Level Optimal Transport for Audio-Text Retrieval
55. ICLR 2026Beyond Sequential Reranking: Reranker-Guided Search Improves Reasoning Intensive Retrieval
56. ICLR 2026Dynamical properties of dense associative memory
57. ICLR 2026Efficient Discriminative Joint Encoders for Large Scale Vision-Language Reranking
58. ICLR 2026Evoking User Memory: Personalizing LLM via Recollection-Familiarity Adaptive Retrieval
59. ICLR 2026Fast Escape, Slow Convergence: Learning Dynamics of Phase Retrieval under Power-Law Data
60. ICLR 2026Fast Proteome-Scale Protein Interaction Retrieval via Residue-Level Factorization
61. ICLR 2026Frustratingly Simple Retrieval Improves Challenging, Reasoning-Intensive Benchmarks
62. ICLR 2026Grasp Any Region: Prompting MLLM to Understand the Dense World
63. ICLR 2026Guided Query Refinement: Multimodal Hybrid Retrieval with Test-Time Optimization
64. ICLR 2026Hybrid Reinforcement: when reward is sparse, better to be dense
65. ICLR 2026Knowledge Externalization: Reversible Unlearning and Modular Retrieval in Multimodal Large Language Models
66. ICLR 2026LightRetriever: A LLM-based Text Retrieval Architecture with Extremely Faster Query Inference
67. ICLR 2026LouisKV: Efficient KV Cache Retrieval for Long Input-Output Sequences
68. ICLR 2026MILCO: Learned Sparse Retrieval Across Languages via a Multilingual Connector
69. ICLR 2026Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource
70. ICLR 2026On the Theoretical Limitations of Embedding-Based Retrieval
71. ICLR 2026Pisces: Cryptography-based Private Retrieval-Augmented Generation with Dual-Path Retrieval
72. ICLR 2026RAEE: A Robust Retrieval-Augmented Early Exit Framework for Efficient Inference
73. ICLR 2026Revela: Dense Retriever Learning via Language Modeling
74. ICLR 2026Seq vs Seq: An Open Suite of Paired Encoders and Decoders
75. ICLR 2026SPRINT: Sparse-Dense Residual Fusion for Efficient Diffusion Transformers
76. ICLR 2026Tools are under-documented: Simple Document Expansion Boosts Tool Retrieval
77. ICML 2026``Someone Hid It!'': Query-Agnostic Black-Box Attacks on LLM-Based Retrieval
78. ICML 2026$\mathbb{R}^{2k}$ is Theoretically Large Enough for Embedding-based Top-$k$ Retrieval
79. ICML 2026AgentSelect: Benchmark for Narrative Query-to-Agent Recommendation
80. ICML 2026Beyond Global Alignment: Fine-Grained Motion-Language Retrieval via Pyramidal Shapley-Taylor Learning
81. ICML 2026Beyond Hamming: Query-Aware Decoding of Binary Cosine Sketches
82. ICML 2026Coverage, Not Averages: Semantic Stratification for Trustworthy Retrieval Evaluation
83. ICML 2026Deep Discriminative Structure Proxy Hashing for Cross-modal Retrieval
84. ICML 2026Dense associative memory for Gaussian distributions
85. ICML 2026DenseMLLM: Standard Multimodal LLMs for Dense Prediction
86. ICML 2026Dual Quaternion SE(3) Synchronization with Recovery Guarantees
87. ICML 2026Efficient, Property-Aligned Fan-Out Retrieval via RL-Compiled Diffusion
88. ICML 2026ElicitR: Unlocking Latent Reasoning in Dense Retrievers via Generative Regularization
89. ICML 2026Full-Batch Gradient Descent Outperforms One-Pass SGD: Sample Complexity Separation in Single-Index Learning
90. ICML 2026Geometric Entropy and Retrieval Phase Transitions in Continuous Thermal Dense Associative Memory
91. ICML 2026GRAPE: Let GRPO Supervise Query Rewriting by Ranking for Retrieval
92. ICML 2026Hierarchical Retrieval at Scale: Bridging Interpretability and Efficiency
93. ICML 2026HOBIT: Hardness Optimized Batch Sampling for InfoNCE Training
94. ICML 2026HypRAG: Hyperbolic Dense Retrieval for Retrieval Augmented Generation
95. ICML 2026Improving Topic Modeling by Distilling Soft Labels from Language Models
96. ICML 2026Language Model Networks: Supervision-Efficient Learning through Dense Communication
97. ICML 2026Learning with Admissibility: Robust Fuzzy Hashing for Cross-Modal Retrieval with Noisy Labels
98. ICML 2026mmBERT: A Modern Multilingual Encoder with Annealed Language Learning
99. ICML 2026Neural Thickets: Diverse Task Experts Are Dense Around Pretrained Weights
100. ICML 2026No More K-means: Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval

常见问题

2023–2026 年四大会议一共有多少篇稠密检索论文？

共 401 篇，按年份分别是 2023 年 60 篇、2024 年 92 篇、2025 年 87 篇、2026 年 162 篇。

NeurIPS 2026 有多少篇稠密检索论文？

NeurIPS 2026 有 51 篇稠密检索论文，上一届是 35 篇。

ICLR 2026 有多少篇稠密检索论文？

ICLR 2026 有 26 篇稠密检索论文，上一届是 12 篇。

ICML 2026 有多少篇稠密检索论文？

ICML 2026 有 35 篇稠密检索论文，上一届是 10 篇。

CVPR 2026 有多少篇稠密检索论文？

CVPR 2026 有 50 篇稠密检索论文，上一届是 30 篇。

哪个会议的稠密检索论文最多？

CVPR，2023–2026 年共 147 篇。