视频理解论文趋势 2023–2026

主要结论

• 视频理解论文从 2023 年的 319 篇增长到 2026 年的 832 篇，三年间扩大到约 2.6 倍。
• CVPR 是发视频理解论文最多的会议，2023–2026 年共 1,037 篇，占四个会议总量的 46%。
• 交叉方向里涨得最快的是扩散模型：同时属于两个主题的论文从 2023 年的 8 篇增至 2026 年的 116 篇。
• 在「计算机视觉」的 5 个主题里，它的 2026 年论文数排第 3；同组增长最快的图像生成是 2.2 倍，视频理解是 2.6 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 2026Articulation in Prime: Primitive-Based Articulated Object Understanding from a Single Casual Video
• ICML 2026A Single Layer to Explain Them All: Understanding Massive Values in Large Language Models
• ICLR 2026Action-Guided Attention for Video Action Anticipation
• CVPR 20264D-RGPT Toward Region-level 4D Understanding via Perceptual Distillation
• NeurIPS 20254DGCPro: Efficient Hierarchical 4D Gaussian Compression for Progressive Volumetric Video Streaming
• ICML 2025$\infty$-Video: A Training-Free Approach to Long Video Understanding via Continuous-Time Memory Consolidation
• ICLR 2025Answer, Assemble, Ace: Understanding How LMs Answer Multiple Choice Questions
• CVPR 20254D-Fly: Fast 4D Reconstruction from a Single Monocular Video
• NeurIPS 2024$E^3$: Exploring Embodied Emotion Through A Large-Scale Egocentric Video Dataset
• ICML 2024Enhancing Sufficient Dimension Reduction via Hellinger Correlation
• ICLR 2024Don't Judge by the Look: Towards Motion Coherent Video Representation
• CVPR 2024Dancing with Still Images: Video Distillation via Static-Dynamic Disentanglement
• NeurIPS 2023A Multi-modal Global Instance Tracking Benchmark (MGIT): Better Locating Target in Complex Spatio-temporal and Causal Relationship
• ICML 2023Dissecting the Effects of SGD Noise in Distinct Regimes of Deep Learning
• ICLR 2023A Theoretical Understanding of Shallow Vision Transformers: Learning, Generalization, and Sample Complexity
• CVPR 20231000 FPS HDR Video With a Spike-RGB Hybrid Camera
• NeurIPS 2026AsymHP: Load-Balanced Sparse Attention for Video Diffusion Transformers
• ICML 2026AnyMod-LLVE: Low-Light Video Enhancement with Modality-Agnostic Inference
• ICLR 2026Emotions Where Art Thou: Understanding and Characterizing the Emotional Latent Space of Large Language Models
• CVPR 2026Accelerating Streaming Video Large Language Models via Hierarchical Token Compression

全部视频理解论文

共 114 篇，第 2 / 3 页

51. ICML 2026Precision-Induced Miscalibration: Understanding and Correcting Confidence Distortion in Quantized Neural Networks
52. ICML 2026Privacy-Aware Video Anomaly Detection through Orthogonal Subspace Projection
53. ICML 2026Proact-VL: A Proactive VideoLLM for Real-Time AI Companions
54. ICML 2026ProtoKV: Streaming Video Understanding under Delayed Query with Summary-State Memory
55. ICML 2026Real Data Lies: Unveiling and Closing the Quality Shortcut in Generalizable AI-Generated Video Detection
56. ICML 2026Reasoning LLM Improves Speaker Recognition in Long-form TV Dramas
57. ICML 2026Rethinking Visual Intelligence: Insights from Video Pretraining
58. ICML 2026Return of Frustratingly Easy Unsupervised Video Domain Adaptation
59. ICML 2026Rh-3DGS: Robust Open-Vocabulary Scene Understanding via Riemannian Huber Distillation and Manifold-Aware Sampling
60. ICML 2026Robust-U1: Can MLLMs Self-Recover Corrupted Visual Content for Robust Understanding?
61. ICML 2026RVAS: Referring Video Active Exploration and Segmentation
62. ICML 2026Seeing is Understanding: Unlocking Causal Attention into Modality-Mutual Attention for Multimodal LLMs
63. ICML 2026Seeing Symbols, Missing Structure: A Real-World Handwritten Mathematical Expression Recognition Benchmark for Large Models
64. ICML 2026SegPVSG: Panoptic Video Scene Graph Generation via Temporal Focusing and Generative Augmentation
65. ICML 2026SemanticNVS: Improving Semantic Scene Understanding in Generative Novel View Synthesis
66. ICML 2026SGMD: Score Gradient Matching Distillation for Few-Step Video Diffusion Distillation
67. ICML 2026Smoothing Slot Attention Iterations and Recurrences
68. ICML 2026SpaceVista: All-Scale Visual Spatial Reasoning from mm to km
69. ICML 2026Spectral Reach: Understanding Neural Scaling as Progress into the Spectral Tail
70. ICML 2026TapSampling: Inference-Time Sampling with a Task-Progress-Understanding Verifier for Robotic Manipulation
71. ICML 2026Think in Cloud, Look at Edges: Semantic-Driven Query Decomposition for Efficient Video Reasoning
72. ICML 2026TimeChat-Captioner: Scripting Multi-Scene Videos with Time-Aware and Structural Audio-Visual Captions
73. ICML 2026Toward Safe Quantization-Aware Fine-tuning: Understanding and Mitigating Safety Alignment Degradation
74. ICML 2026Toward Understanding Adversarial Distillation: Why Robust Teachers Fail
75. ICML 2026Towards Theoretical Understanding of Transformer Test-Time Computing: Investigation on In-Context Linear Regression
76. ICML 2026Towards Trustworthy Video Anomaly Understanding: A Class-Guided Chain-of-Evaluation Metric and An Anomaly-focused Meta-Benchmark
77. ICML 2026Towards Understanding Adam Convergence on Highly Degenerate Polynomials
78. ICML 2026Towards Understanding Massive Activations in Attention Sink Mechanism
79. ICML 2026Towards Understanding Modality Interaction in Multimodal Language Models via Partial Information Decomposition
80. ICML 2026Towards Understanding Steering Strength
81. ICML 2026Trajectory-Stabilized Inference for Diffusion-Based Video Inpainting
82. ICML 2026Two-Layer Linear Auto-Regressive Models Estimate Latent States
83. ICML 2026Understanding Behavior Cloning with Action Quantization
84. ICML 2026Understanding Catastrophic Forgetting In LoRA via Mean-Field Attention Dynamics
85. ICML 2026Understanding Data Temporality Impact on Large Language Models Pre-training
86. ICML 2026Understanding Dynamic Compute Allocation in Recurrent Transformers
87. ICML 2026Understanding Dynamics of Adam in Zero-Sum Games: An ODE Approach
88. ICML 2026Understanding Generalization and Forgetting in In-Context Continual Learning
89. ICML 2026Understanding Generalization from Embedding Dimension and Distributional Convergence
90. ICML 2026Understanding MARS: When Scaling Momentum Provably Helps
91. ICML 2026Understanding Performance Collapse in Layer-Pruned Large Language Models via Decision Representation Transitions
92. ICML 2026Understanding Private Learning From Feature Perspective
93. ICML 2026Understanding SAM through Minimax Perspective
94. ICML 2026Understanding Self-Supervised Learning via Latent Distribution Matching
95. ICML 2026Understanding the Ability of LLMs to Handle Character-Level Perturbation
96. ICML 2026Understanding the Gaps in Satisficing Bandits
97. ICML 2026Understanding the Parameter Space Geometry of Transformers Encoding Boolean Functions
98. ICML 2026Understanding Truncated Positional Encodings for Graph Neural Networks
99. ICML 2026Unison: Benchmarking Unified Multimodal Models via Synergistic Understanding and Generation
100. ICML 2026Universal Skeleton Understanding via Differentiable Rendering and MLLMs

常见问题

2023–2026 年四大会议一共有多少篇视频理解论文？

共 2,261 篇，按年份分别是 2023 年 319 篇、2024 年 431 篇、2025 年 679 篇、2026 年 832 篇。

NeurIPS 2026 有多少篇视频理解论文？

NeurIPS 2026 有 229 篇视频理解论文，上一届是 222 篇。

ICLR 2026 有多少篇视频理解论文？

ICLR 2026 有 148 篇视频理解论文，上一届是 95 篇。

ICML 2026 有多少篇视频理解论文？

ICML 2026 有 114 篇视频理解论文，上一届是 64 篇。

CVPR 2026 有多少篇视频理解论文？

CVPR 2026 有 341 篇视频理解论文，上一届是 298 篇。

哪个会议的视频理解论文最多？

CVPR，2023–2026 年共 1,037 篇。