检索记忆论文趋势 2023–2026

主要结论

• 检索记忆论文从 2023 年的 23 篇增长到 2026 年的 151 篇，三年间扩大到约 6.6 倍。
• NeurIPS 是发检索记忆论文最多的会议，2023–2026 年共 120 篇，占四个会议总量的 41%。
• 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 25 篇。
• 在「长上下文与记忆」的 5 个主题里，它的 2026 年论文数排第 3；同组增长最快的长上下文是 5.2 倍，检索记忆是 6.6 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 2026ADMIT: Support-Gated Memory-Write Admission for Document QA Agents
• ICML 2026Attention with Routed-Memory for Learnable Sparse Control
• ICLR 2026AssoMem: Scalable Memory QA with Multi-Signal Associative Retrieval
• CVPR 2026DREAM Document Recognition with Explicit Adaptive Memory
• NeurIPS 2025EfficientNav: Towards On-Device Object-Goal Navigation with Navigation Map Caching and Retrieval
• ICML 2025In-Context Denoising with One-Layer Transformers: Connections between Attention and Associative Memory Retrieval
• ICLR 2025BitStack: Any-Size Compression of Large Language Models in Variable Memory Environments
• CVPR 2025MAD: Memory-Augmented Detection of 3D Objects
• NeurIPS 2024A Gradient Accumulation Method for Dense Retriever under Memory Constraint
• ICML 2024Think Before You Act: Decision Transformers with Working Memory
• ICLR 2024STanHop: Sparse Tandem Hopfield Model for Memory-Enhanced Time Series Prediction
• CVPR 2024Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval
• NeurIPS 2023Augmented Memory Replay-based Continual Learning Approaches for Network Intrusion Detection
• ICML 2023Memory-Based Meta-Learning on Non-Stationary Distributions
• ICLR 2023Associative Memory Augmented Asynchronous Spatiotemporal Representation Learning for Event-based Perception
• CVPR 2023Memory-Friendly Scalable Super-Resolution via Rewinding Lottery Ticket Hypothesis
• NeurIPS 2026Automated Reformulation of Robust Optimization via Memory-Augmented Large Language Models
• ICML 2026Dense associative memory for Gaussian distributions
• ICLR 2026FreeKV: Boosting KV Cache Retrieval for Efficient LLM Inference
• CVPR 2026Interactive Tracking A Human-in-the-Loop Paradigm with Memory-Augmented Adaptation

全部检索记忆论文

共 38 篇，第 1 / 1 页

1. ICML 2026AMA-Bench: Evaluating Long-Horizon Memory for Agentic Applications
2. ICML 2026Attention with Routed-Memory for Learnable Sparse Control
3. ICML 2026CiteGuard: Conformal False-Discovery Control for Faithful Retrieval-Augmented Generation
4. ICML 2026CrossQ: Task-Aligned Cross-Token Conditional Quantization for Late Interaction Retrieval
5. ICML 2026Dense associative memory for Gaussian distributions
6. ICML 2026E2Former-V2: On-the-Fly Equivariant Attention with Linear Activation Memory
7. ICML 2026From Retrieval to Translation: Translating Query into Graph-level Clues for Retrieval-Augmented Generation
8. ICML 2026From Volume to Value: Preference-Aligned Memory Construction for On-Device RAG
9. ICML 2026GAM-RAG: Gain-Adaptive Memory for Evolving Retrieval in Retrieval-Augmented Generation
10. ICML 2026Generalized Linear Bandits with Memory
11. ICML 2026Geometric Entropy and Retrieval Phase Transitions in Continuous Thermal Dense Associative Memory
12. ICML 2026Global Merger-Arbitrage Forecasting with Language Models
13. ICML 2026GradMem: Learning to Write Context into Memory with Test-Time Gradient Descent
14. ICML 2026HGMem: Hypergraph-based Working Memory to Improve Multi-step RAG for Long-Context Complex Relational Modeling
15. ICML 2026Hyperbolic Associative Memory Networks
16. ICML 2026HypRAG: Hyperbolic Dense Retrieval for Retrieval Augmented Generation
17. ICML 2026LazyAttention: Efficient Retrieval-Augmented Generation with Deferred Positional Encoding
18. ICML 2026Linguistic Nepotism: Trading-off Quality for Language Preference in Multilingual RAG
19. ICML 2026MemIncept: Steering LLM Agents via Cooperative Stealthy Memory Injections
20. ICML 2026Memory Caching: RNNs with Growing Memory
21. ICML 2026Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents
22. ICML 2026Mitigating Error Accumulation in Continuous Navigation via Memory-Augmented Kalman Filtering
23. ICML 2026Navigating Massive Visual Context in Retrieval-Augmented Generation via Multimodal Memory Graph
24. ICML 2026No More K-means: Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval
25. ICML 2026Panini: Continual Learning in Token Space via Structured Memory
26. ICML 2026Predictive Prefetching for Retrieval-Augmented Generation
27. ICML 2026RAD: Retrieval High-quality Demonstrations to Enhance Decision-making
28. ICML 2026RADAR: Defending RAG Dynamically against Retrieval Corruption
29. ICML 2026RAG without Forgetting: Continual Query-Infused Key Memory
30. ICML 2026Reranker Helps, but Not Enough: Towards Strong Poisoning Attacks Against Retrieval-Augmented Generation
31. ICML 2026Retrieval-Aware Distillation for Transformer-SSM Hybrids
32. ICML 2026Retriever Portfolios: A Principled Approach to Adaptive RAG
33. ICML 2026RGMem: Renormalization Group–inspired Memory Evolution for Language Agents
34. ICML 2026SE-GA: Memory-Augmented Self-Evolution for GUI Agents
35. ICML 2026SimpleMem: Efficient Lifelong Memory for LLM Agents
36. ICML 2026The Crowded Embedding Space: A Mean-Field Mechanism for Emergent Marginalization in Retrieval-Augmented Agents
37. ICML 2026Token-Free Hierarchical Indexing for RAG beyond LLM-based Summarization
38. ICML 2026Understand and Accelerate Memory Processing Pipeline for Large Language Model Inference

常见问题

2023–2026 年四大会议一共有多少篇检索记忆论文？

共 290 篇，按年份分别是 2023 年 23 篇、2024 年 44 篇、2025 年 72 篇、2026 年 151 篇。

NeurIPS 2026 有多少篇检索记忆论文？

NeurIPS 2026 有 64 篇检索记忆论文，上一届是 34 篇。

ICLR 2026 有多少篇检索记忆论文？

ICLR 2026 有 31 篇检索记忆论文，上一届是 19 篇。

ICML 2026 有多少篇检索记忆论文？

ICML 2026 有 38 篇检索记忆论文，上一届是 14 篇。

CVPR 2026 有多少篇检索记忆论文？

CVPR 2026 有 18 篇检索记忆论文，上一届是 5 篇。

哪个会议的检索记忆论文最多？

NeurIPS，2023–2026 年共 120 篇。