联邦边缘推理论文趋势 2023–2026

主要结论

• 联邦边缘推理论文从 2023 年的 43 篇增长到 2026 年的 45 篇，三年间扩大到约 1.0 倍。
• ICML 是发联邦边缘推理论文最多的会议，2023–2026 年共 60 篇，占四个会议总量的 33%。
• 交叉方向里涨得最快的是端侧推理：同时属于两个主题的论文从 2023 年的 3 篇增至 2026 年的 15 篇。
• 在「边缘与物联网」的 5 个主题里，它的 2026 年论文数排第 2；同组增长最快的传感融合是 3.8 倍，联邦边缘推理是 1.0 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 2026Does Sparse Connectivity Improve Generalization? Convolutional Networks Below the Edge of Stability
• ICML 2026Breaking the Capacity Bottleneck in Model-Heterogeneous Federated Learning via Gradual Model Restoration
• ICLR 2026Beyond Aggregation: Guiding Clients in Heterogeneous Federated Learning
• CVPR 2026E2-SCI Elastic Edge-Cloud Speculative Decoding via Credit Inertia
• NeurIPS 2025Class-wise Balancing Data Replay for Federated Class-Incremental Learning
• ICML 2025Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability
• ICLR 2025Connecting Federated ADMM to Bayes
• CVPR 2025Beyond Local Sharpness: Communication-Efficient Global Sharpness-aware Minimization for Federated Learning
• NeurIPS 2024DapperFL: Domain Adaptive Federated Learning with Model Fusion Pruning for Edge Devices
• ICML 2024Achieving Lossless Gradient Sparsification via Mapping to Alternative Space in Federated Learning
• ICLR 2024FedTrans: Client-Transparent Utility Estimation for Robust Federated Learning
• CVPR 2024FedAS: Bridging Inconsistency in Personalized Federated Learning
• NeurIPS 2023Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
• ICML 2023Achieving Linear Speedup in Non-IID Federated Bilevel Learning
• ICLR 2023Combating Exacerbated Heterogeneity for Robust Models in Federated Learning
• CVPR 2023How To Prevent the Poor Performance Clients for Personalized Federated Learning?
• NeurIPS 2026Edge of Stability Selectively Shapes Learning Across the Data Distribution
• ICML 2026Influence-Disentangled Federated Training: Learning Models That Are Easy to Unlearn
• ICLR 2026DeepAFL: Deep Analytic Federated Learning
• CVPR 2026Edge-Focused Super-Resolution for Omnidirectional Images with Spherical Geometric Augmentation

全部联邦边缘推理论文

共 18 篇，第 1 / 1 页

1. ICML 2026Benchmarking at the Edge of Comprehension
2. ICML 2026Breaking the Capacity Bottleneck in Model-Heterogeneous Federated Learning via Gradual Model Restoration
3. ICML 2026Conflicting Biases at the Edge of Stability: Norm versus Sharpness Regularization
4. ICML 2026Credibility-Aware Weighting Federated Causal Discovery for Time Series
5. ICML 2026Edge-colored Clustering in Hypergraphs: A MaxECC Approximation
6. ICML 2026FedEMoE: Improving Personalization on Heterogeneous Federated Learning via Elastic Mixture of Experts Architecture
7. ICML 2026FedFit: Federated Dynamic Sparse Training via Fisher Information scoring
8. ICML 2026FedGain: Toward Negative-Gain-Free Client Collaboration in Federated Learning
9. ICML 2026FedReLa: Imbalanced Federated Learning via Re-Labeling
10. ICML 2026Influence-Disentangled Federated Training: Learning Models That Are Easy to Unlearn
11. ICML 2026Lightweight Federated Incremental Learning via Decoupled Replay
12. ICML 2026Momentum Further Constrains Sharpness at the Edge of Stochastic Stability
13. ICML 2026No Retraining at Edge: Efficient Resource-Aware Mixed-Precision Quantization via Federated Supernet Learning
14. ICML 2026Non-Euclidean Gradient Descent Operates at the Edge of Stability
15. ICML 2026PipeSD: An Efficient Cloud-Edge Collaborative Pipeline Inference Framework with Speculative Decoding
16. ICML 2026RADE: Random Add-Drop Edge as a Regularizer
17. ICML 2026ShapCCS: Shapley-Driven Client Coreset Selection in Federated Learning
18. ICML 2026Trustworthy Federated Label Distribution Learning under Annotation Quality Disparity

常见问题

2023–2026 年四大会议一共有多少篇联邦边缘推理论文？

共 180 篇，按年份分别是 2023 年 43 篇、2024 年 47 篇、2025 年 45 篇、2026 年 45 篇。

NeurIPS 2026 有多少篇联邦边缘推理论文？

NeurIPS 2026 有 15 篇联邦边缘推理论文，上一届是 20 篇。

ICLR 2026 有多少篇联邦边缘推理论文？

ICLR 2026 有 7 篇联邦边缘推理论文，上一届是 10 篇。

ICML 2026 有多少篇联邦边缘推理论文？

ICML 2026 有 18 篇联邦边缘推理论文，上一届是 10 篇。

CVPR 2026 有多少篇联邦边缘推理论文？

CVPR 2026 有 5 篇联邦边缘推理论文，上一届是 5 篇。

哪个会议的联邦边缘推理论文最多？

ICML，2023–2026 年共 60 篇。