长上下文论文趋势 2023–2026

主要结论

• 长上下文论文从 2023 年的 66 篇增长到 2026 年的 341 篇，三年间扩大到约 5.2 倍。
• NeurIPS 是发长上下文论文最多的会议，2023–2026 年共 298 篇，占四个会议总量的 36%。
• 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 45 篇。
• 在「长上下文与记忆」的 5 个主题里，它的 2026 年论文数排第 1；同组增长最快的检索记忆是 6.6 倍，长上下文是 5.2 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 2026Accelerating Long-Context LLM Prefill via Layer-wise Progressive Token Pruning in Local Deployment
• ICML 2026AdaSplash-2: Faster Differentiable Sparse Attention
• ICLR 2026An evolutionary perspective on modes of learning in Transformers
• CVPR 2026AR2-4FV Anchored Referring and Re-identification for Long-Term Grounding in Fixed-View
• NeurIPS 2025A Controllable Examination for Long-Context Language Models
• ICML 2025CateKV: On Sequential Consistency for Long-Context LLM Inference Acceleration
• ICLR 2025$\text{D}_{2}\text{O}$: Dynamic Discriminative Operations for Efficient Long-Context Inference of Large Language Models
• CVPR 2025BIMBA: Selective-Scan Compression for Long-Range Video Question Answering
• NeurIPS 2024[Re] CUDA: Curriculum of Data Augmentation for Long‐tailed Recognition
• ICML 2024Benchmarking and Building Long-Context Retrieval Models with LoCo and M2-BERT
• ICLR 2024CLEX: Continuous Length Extrapolation for Large Language Models
• CVPR 2024GigaTraj: Predicting Long-term Trajectories of Hundreds of Pedestrians in Gigapixel Complex Scenes
• NeurIPS 2023A Single 2D Pose with Context is Worth Hundreds for 3D Human Pose Estimation
• ICML 2023CAB: Comprehensive Attention Benchmarking on Long Sequence Modeling
• ICLR 2023$k$NN Prompting: Beyond-Context Learning with Calibration-Free Nearest Neighbor Inference
• CVPR 2023AutoAD: Movie Description in Context
• NeurIPS 2026AdaCal: Adaptive Calibration for Robust Sparse Attention in Long-Context LLMs
• ICML 2026Artificial Hippocampus Networks for Efficient Long-Context Modeling
• ICLR 2026AutoSP: Unlocking Long-Context LLM Training Via Compiler-Based Sequence Parallelism
• CVPR 2026AstraNav-Memory Contexts Compression for Long Memory

全部长上下文论文

共 98 篇，第 2 / 2 页

51. ICML 2026MoRGen: Mixture-of-Resolutions Generative Forecasting for Irregularly Sampled Medical Time-Series Data
52. ICML 2026MuonSSM: Orthogonalizing State Space Models for Sequence Modeling
53. ICML 2026NanoSpec: Accelerating Speculative Decoding using Minimalist In-Context Vocabularies
54. ICML 2026OBCache: Optimal Brain KV Cache Pruning for Efficient Long-Context LLM Inference
55. ICML 2026OPIC: Enhancing Language Model Merging via Optimizing In-Context Capability
56. ICML 2026ParaTool: Shifting Tool Representations from Context to Parameters
57. ICML 2026ParisKV: Fast and Drift-Robust KV-Cache Retrieval for Long-Context LLMs
58. ICML 2026PersistBench: When Should Long-Term Memories Be Forgotten by LLMs?
59. ICML 2026PLASH: Provably Linear-Time Attention with Selective Higher-Order Feature Sketching
60. ICML 2026Predicting Future Utility: Global Combinatorial Optimization for Task-Agnostic KV Cache Eviction
61. ICML 2026Prism: Spectral-Aware Block-Sparse Attention
62. ICML 2026Probing How Scalable Table Data Enhances General Long-Context Reasoning
63. ICML 2026ProphetKV: User-Query-Driven Selective Recomputation for Efficient KV Cache Reuse in Retrieval-Augmented Generation
64. ICML 2026Push, Pop, Parallelize: Stack-Augmented Linear Attention via the Delta Rule
65. ICML 2026RaBitQCache: Rotated Binary Quantization for KVCache in Long Context LLM Inference
66. ICML 2026Recursive Models for Long-Horizon Reasoning
67. ICML 2026RePo: Language Models with Context Re-Positioning
68. ICML 2026Residual Context Diffusion Language Models
69. ICML 2026Resolving the Timestep Scaling Paradox in Spiking Neural Networks with a Timestep-Scalable Neuron Model
70. ICML 2026Rethinking Genomic Modeling Through Optical Character Recognition
71. ICML 2026Rethinking Loss Reweighting for Imbalance Learning as an Inverse Problem: A Neural Collapse Point of View
72. ICML 2026Revealing Long-context Potential of Attention Heads via Frequency Kernels
73. ICML 2026Robust Filter Attention: Self-Attention as Precision-Weighted State Estimation
74. ICML 2026Safe In-Context Reinforcement Learning
75. ICML 2026SALE : Low-bit Estimation for Efficient Sparse Attention in Long-context LLM Prefilling
76. ICML 2026Scalable Option Learning in High-Throughput Environments
77. ICML 2026Scaling Long-Horizon Agent via Context Folding
78. ICML 2026Scout Before You Attend: Sketch-and-Walk Sparse Attention for Efficient LLM Inference
79. ICML 2026Shortcut-Resistant CAM Distillation for Long-Tailed Recognition
80. ICML 2026Shuffle the Context: RoPE-Perturbed Self-Distillation for Long-Context Adaptation
81. ICML 2026SKETCH: Semantic Key-Point Conditioning for Long-Horizon Vessel Trajectory Prediction
82. ICML 2026Softmax as Linear Attention in the Large-Prompt Regime: a Measure-based Perspective
83. ICML 2026Softplus Attention with Re-weighting Boosts Length Extrapolation in Large Language Models
84. ICML 2026SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?
85. ICML 2026Symmetry Reveals Layerwise Dynamics: How Transformers Perform In-Context Classification
86. ICML 2026Temporal Context Reinstatement Drives Episodic-Like Order Memory in Long-Context Language Models
87. ICML 2026The Devil is in the Spectrum: Mitigating Representation Collapse in LLMs via Topologically Regularized Side-Path
88. ICML 2026The First Drop of Ink: Nonlinear Impact of Misleading Information in Long-Context Reasoning
89. ICML 2026The Geometric Reasoner: Manifold-Informed Latent Foresight Search for Long-Context Reasoning
90. ICML 2026Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection
91. ICML 2026Training–Inference Consistent Segmented Execution for Long-Context LLMs
92. ICML 2026Transformers Can Learn Posterior Predictive Distributions In-Context
93. ICML 2026Transformers Efficiently Perform In-Context Logistic Regression via Normalized Gradient Descent
94. ICML 2026Understanding Generalization and Forgetting in In-Context Continual Learning
95. ICML 2026Universal Approximation with Softmax Attention
96. ICML 2026Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking
97. ICML 2026When to Memorize and When to Stop: Gated Recurrent Memory for Long-Context Reasoning
98. ICML 2026You Need Better Attention Priors

常见问题

2023–2026 年四大会议一共有多少篇长上下文论文？

共 838 篇，按年份分别是 2023 年 66 篇、2024 年 176 篇、2025 年 255 篇、2026 年 341 篇。

NeurIPS 2026 有多少篇长上下文论文？

NeurIPS 2026 有 130 篇长上下文论文，上一届是 98 篇。

ICLR 2026 有多少篇长上下文论文？

ICLR 2026 有 82 篇长上下文论文，上一届是 87 篇。

ICML 2026 有多少篇长上下文论文？

ICML 2026 有 98 篇长上下文论文，上一届是 53 篇。

CVPR 2026 有多少篇长上下文论文？

CVPR 2026 有 31 篇长上下文论文，上一届是 17 篇。

哪个会议的长上下文论文最多？

NeurIPS，2023–2026 年共 298 篇。