生成流论文趋势 2023–2026

主要结论

• 生成流论文从 2023 年的 73 篇增长到 2026 年的 139 篇，三年间扩大到约 1.9 倍。
• NeurIPS 是发生成流论文最多的会议，2023–2026 年共 130 篇，占四个会议总量的 30%。
• 交叉方向里涨得最快的是应用 · chemistry physics and earth sciences：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 5 篇。
• 在「概率与贝叶斯」的 5 个主题里，它的 2026 年论文数排第 3；同组增长最快的高斯过程是 2.2 倍，生成流是 1.9 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 2026Exact Recovery of Lipschitz Orthogonal Coordinate Transformations via Constrained Normalizing Flows
• ICML 2026Learning-to-Optimize via Deep Unfolded Flows
• ICLR 2026Frozen Priors, Fluid Forecasts: Prequential Uncertainty for Low-Data Deployment with Pretrained Generative Models
• CVPR 2026AnchorFlow Training-Free 3D Editing via Latent Anchor-Aligned Flows
• NeurIPS 2025Amortized Sampling with Transferable Normalizing Flows
• ICML 2025Annealing Flow Generative Models Towards Sampling High-Dimensional and Multi-Modal Distributions
• ICLR 2025Beyond Squared Error: Exploring Loss Design for Enhanced Training of Generative Flow Networks
• CVPR 2025DiverseFlow: Sample-Efficient Diverse Mode Coverage in Flows
• NeurIPS 2024Learning Distributions on Manifolds with Free-Form Flows
• ICML 2024A Computational Framework for Solving Wasserstein Lagrangian Flows
• ICLR 2024Fast and unified path gradient estimators for normalizing flows
• CVPR 2024Distributionally Generative Augmentation for Fair Facial Attribute Classification
• NeurIPS 2023Canonical normalizing flows for manifold learning
• ICML 2023MonoFlow: Rethinking Divergence GANs via the Perspective of Wasserstein Gradient Flows
• ICLR 2023AE-FLOW: Autoencoders with Normalizing Flows for Medical Images Anomaly Detection
• CVPR 2023Delving Into Discrete Normalizing Flows on SO(3) Manifold for Probabilistic Rotation Modeling
• NeurIPS 2026From Baselines to Transport Geodesics: Axiomatic Attribution via Optimal Generative Flows
• ICML 2026Hyperbolic RQ-VAE enhanced Generative Recommendation with Differential-Length Codebook Strategy
• ICLR 2026JAPAN: Joint Adaptive Prediction Areas with Normalising Flow
• CVPR 2026BUSSARD Normalizing Flows for Bijective Universal Scene-Specific Anomalous Relationship Detection

全部生成流论文

共 22 篇，第 1 / 1 页

1. CVPR 2025Auto Cherry-Picker: Learning from High-quality Generative Data Driven by Language
2. CVPR 2025Can Generative Video Models Help Pose Estimation?
3. CVPR 2025Decompositional Neural Scene Reconstruction with Generative Diffusion Prior
4. CVPR 2025DiverseFlow: Sample-Efficient Diverse Mode Coverage in Flows
5. CVPR 2025Exploration-Driven Generative Interactive Environments
6. CVPR 2025GCC: Generative Color Constancy via Diffusing a Color Checker
7. CVPR 2025Generalizable Object Keypoint Localization from Generative Priors
8. CVPR 2025Generative Densification: Learning to Densify Gaussians for High-Fidelity Generalizable 3D Reconstruction
9. CVPR 2025Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens
10. CVPR 2025Generative Photomontage
11. CVPR 2025GIF: Generative Inspiration for Face Recognition at Scale
12. CVPR 2025IDEA-Bench: How Far are Generative Models from Professional Designing?
13. CVPR 2025LookingGlass: Generative Anamorphoses via Laplacian Pyramid Warping
14. CVPR 2025Omni-ID: Holistic Identity Representation Designed for Generative Tasks
15. CVPR 2025PERSE: Personalized 3D Generative Avatars from A Single Portrait
16. CVPR 2025PhysicsGen: Can Generative Models Learn from Images to Predict Complex Physical Relations?
17. CVPR 2025RoboTwin: Dual-Arm Robot Benchmark with Generative Digital Twins
18. CVPR 2025Segment Any-Quality Images with Generative Latent Space Enhancement
19. CVPR 2025Symbolic Representation for Any-to-Any Generative Tasks
20. CVPR 2025Training Data Provenance Verification: Did Your Model Use Synthetic Data from My Generative Model for Training?
21. CVPR 2025Traversing Distortion-Perception Tradeoff using a Single Score-Based Generative Model
22. CVPR 2025Unveiling Differences in Generative Models: A Scalable Differential Clustering Approach

常见问题

2023–2026 年四大会议一共有多少篇生成流论文？

共 429 篇，按年份分别是 2023 年 73 篇、2024 年 98 篇、2025 年 119 篇、2026 年 139 篇。

NeurIPS 2026 有多少篇生成流论文？

NeurIPS 2026 有 46 篇生成流论文，上一届是 37 篇。

ICLR 2026 有多少篇生成流论文？

ICLR 2026 有 37 篇生成流论文，上一届是 35 篇。

ICML 2026 有多少篇生成流论文？

ICML 2026 有 32 篇生成流论文，上一届是 25 篇。

CVPR 2026 有多少篇生成流论文？

CVPR 2026 有 24 篇生成流论文，上一届是 22 篇。

哪个会议的生成流论文最多？

NeurIPS，2023–2026 年共 130 篇。