深度学习 · self supervised learning论文趋势 2023–2026
ICML/NeurIPS 深度学习方向：self_supervised_learning
统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 1,958 篇被归入该主题的论文。
主要结论
- 深度学习 · self supervised learning论文从 2023 年的 455 篇增长到 2026 年的 534 篇，三年间扩大到约 1.2 倍。
- ICLR 是发深度学习 · self supervised learning论文最多的会议，2023–2026 年共 693 篇，占四个会议总量的 35%。
- 交叉方向里涨得最快的是通用机器学习 · unsupervised and semisupervised learning：同时属于两个主题的论文从 2023 年的 23 篇增至 2026 年的 86 篇。
- 在「深度学习」的 11 个主题里，它的 2026 年论文数排第 4；同组增长最快的扩散模型是 4.1 倍，深度学习 · self supervised learning是 1.2 倍。
各会议论文数
按年份和会议统计
| 年份 | NeurIPS | ICLR | ICML | CVPR | 合计 |
|---|---|---|---|---|---|
| 2023 | 160 | 108 | 41 | 146 | 455 |
| 2024 | 110 | 195 | 62 | 105 | 472 |
| 2025 | 147 | 195 | 72 | 83 | 497 |
| 2026 | 110 | 195 | 135 | 94 | 534 |
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 2026Attending on Attention ($A^2$): Smaller Self-Supervised ViTs Localize Better Than Larger Ones
- ICML 20263D-DLP: Self-supervised 3D Object-centric Scene Representation Learning
- ICLR 2026A Single Architecture for Representing Invariance Under Any Space Group
- CVPR 2026A Stitch in Time Learning Procedural Workflow via Self-Supervised Plackett-Luce
- NeurIPS 2025$\textit{HiMaCon:}$ Discovering Hierarchical Manipulation Concepts from Unlabeled Multi-Modal Data
- ICML 2025An Augmentation-Aware Theory for Self-Supervised Contrastive Learning
- ICLR 2025Approaching Rate-Distortion Limits in Neural Compression with Lattice Transform Coding
- CVPR 2025A Unified Framework for Heterogeneous Semi-supervised Learning
- NeurIPS 2024Intrinsic Self-Supervision for Data Quality Audits
- ICML 2024Binning as a Pretext Task: Improving Self-Supervised Learning in Tabular Domains
- ICLR 2024A Graph is Worth 1-bit Spikes: When Graph Contrastive Learning Meets Spiking Neural Networks
- CVPR 2024Exploring Efficient Asymmetric Blind-Spots for Self-Supervised Denoising in Real-World Scenarios
- NeurIPS 2023A Graph-Theoretic Framework for Understanding Open-World Semi-Supervised Learning
- ICML 2023A Closer Look at Self-Supervised Lightweight Vision Transformers
- ICLR 2023A Multi-Grained Self-Interpretable Symbolic-Neural Model For Single/Multi-Labeled Text Classification
- CVPR 2023Are Data-Driven Explanations Robust Against Out-of-Distribution Data?
- NeurIPS 2026AudioSphere: Towards Self-Supervised Spatial Audio Representation Models
- ICML 20263DPoV: Improving 3D understanding via Patch Ordering on Videos
- ICLR 2026Angle K-Means
- CVPR 2026Action Motifs Self-Supervised Hierarchical Representation of Human Body Movements
全部深度学习 · self supervised learning论文
共 1,956 篇，第 25 / 40 页
- ICLR 2024InfoCon: Concept Discovery with Generative and Discriminative Informativeness
- ICLR 2024Information Retention via Learning Supplemental Features
- ICLR 2024Initializing Models with Larger Ones
- ICLR 2024Instructive Decoding: Instruction-Tuned Large Language Models are Self-Refiner from Noisy Instructions
- ICLR 2024Investigating the Benefits of Projection Head for Representation Learning
- ICLR 2024Is ImageNet worth 1 video? Learning strong image encoders from 1 long unlabelled video
- ICLR 2024Knowledge Distillation Based on Transformed Teacher Matching
- ICLR 2024Label-Agnostic Forgetting: A Supervision-Free Unlearning in Deep Models
- ICLR 2024Language Model Self-improvement by Reinforcement Learning Contemplation
- ICLR 2024Large-scale Training of Foundation Models for Wearable Biosignals
- ICLR 2024Layer-wise linear mode connectivity
- ICLR 2024LDReg: Local Dimensionality Regularized Self-Supervised Learning
- ICLR 2024Learning 3D Particle-based Simulators from RGB-D Videos
- ICLR 2024Learning from Label Proportions: Bootstrapping Supervised Learners via Belief Propagation
- ICLR 2024Learning Multi-Faceted Prototypical User Interests
- ICLR 2024Learning to Compose: Improving Object Centric Learning by Injecting Compositionality
- ICLR 2024Learning to design protein-protein interactions with enhanced generalization
- ICLR 2024Learning to Embed Time Series Patches Independently
- ICLR 2024LEMON: Lossless model expansion
- ICLR 2024LiDAR: Sensing Linear Probing Performance in Joint Embedding SSL Architectures
- ICLR 2024LightHGNN: Distilling Hypergraph Neural Networks into MLPs for 100x Faster Inference
- ICLR 2024Linear Log-Normal Attention with Unbiased Concentration
- ICLR 2024Looped Transformers are Better at Learning Learning Algorithms
- ICLR 2024Masked Autoencoders with Multi-Window Local-Global Attention Are Better Audio Learners
- ICLR 2024Masked Completion via Structured Diffusion with White-Box Transformers
- ICLR 2024Masked Distillation Advances Self-Supervised Transformer Architecture Search
- ICLR 2024MCM: Masked Cell Modeling for Anomaly Detection in Tabular Data
- ICLR 2024Mean Field Theory in Deep Metric Learning
- ICLR 2024Memorization in Self-Supervised Learning Improves Downstream Generalization
- ICLR 2024MERT: Acoustic Music Understanding Model with Large-Scale Self-supervised Training
- ICLR 2024Mind Your Augmentation: The Key to Decoupling Dense Self-Supervised Learning
- ICLR 2024Modulate Your Spectrum in Self-Supervised Learning
- ICLR 2024MogaNet: Multi-order Gated Aggregation Network
- ICLR 2024MOTOR: A Time-to-Event Foundation Model For Structured Medical Records
- ICLR 2024Multi-resolution HuBERT: Multi-resolution Speech Self-Supervised Learning with Masked Unit Prediction
- ICLR 2024Multilinear Operator Networks
- ICLR 2024Multiscale Positive-Unlabeled Detection of AI-Generated Texts
- ICLR 2024Neural Fourier Transform: A General Approach to Equivariant Representation Learning
- ICLR 2024Neural Language of Thought Models
- ICLR 2024Neural Spectral Methods: Self-supervised learning in the spectral domain
- ICLR 2024Never Train from Scratch: Fair Comparison of Long-Sequence Models Requires Data-Driven Priors
- ICLR 2024Non-negative Contrastive Learning
- ICLR 2024Object centric architectures enable efficient causal representation learning
- ICLR 2024On Harmonizing Implicit Subpopulations
- ICLR 2024On the Role of Discrete Tokenization in Visual Representation Learning
- ICLR 2024On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes
- ICLR 2024One Step of Gradient Descent is Provably the Optimal In-Context Learner with One Layer of Linear Self-Attention
- ICLR 2024Only Pay for What Is Uncertain: Variance-Adaptive Thompson Sampling
- ICLR 2024Out-of-Distribution Detection with Negative Prompts
- ICLR 2024Out-Of-Domain Unlabeled Data Improves Generalization
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 查深度学习 · self supervised learning在 ICLR 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": "dl-self-supervised-learning"}
常见问题
- 2023–2026 年四大会议一共有多少篇深度学习 · self supervised learning论文？
- 共 1,958 篇，按年份分别是 2023 年 455 篇、2024 年 472 篇、2025 年 497 篇、2026 年 534 篇。
- NeurIPS 2026 有多少篇深度学习 · self supervised learning论文？
- NeurIPS 2026 有 110 篇深度学习 · self supervised learning论文，上一届是 147 篇。
- ICLR 2026 有多少篇深度学习 · self supervised learning论文？
- ICLR 2026 有 195 篇深度学习 · self supervised learning论文，上一届是 195 篇。
- ICML 2026 有多少篇深度学习 · self supervised learning论文？
- ICML 2026 有 135 篇深度学习 · self supervised learning论文，上一届是 72 篇。
- CVPR 2026 有多少篇深度学习 · self supervised learning论文？
- CVPR 2026 有 94 篇深度学习 · self supervised learning论文，上一届是 83 篇。
- 哪个会议的深度学习 · self supervised learning论文最多？
- ICLR，2023–2026 年共 693 篇。