端上学习论文趋势 2023–2026

主要结论

• 端上学习论文从 2023 年的 122 篇降到 2026 年的 6 篇，热度在回落。
• NeurIPS 是发端上学习论文最多的会议，2023–2026 年共 91 篇，占四个会议总量的 68%。
• 在「边缘与物联网」的 5 个主题里，它的 2026 年论文数排第 4；同组增长最快的传感融合是 3.8 倍，端上学习是 0.0 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 2026MobileMoE: Scaling On-Device Mixture of Experts
• ICML 2026Lower Bounds for Frank-Wolfe on Strongly Convex Sets
• CVPR 2026DeDelayed Deleting Remote Inference Delay via On-Device Correction
• ICLR 2025Robust Function-Calling for On-Device Language Model via Function Masking
• CVPR 2025EdgeTAM: On-Device Track Anything Model
• ICML 2024MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases
• CVPR 2024Device-Wise Federated Network Pruning
• NeurIPS 2023[Re] Numerical influence of ReLU'(0) on backpropagation
• ICML 2023Adversarial Collaborative Learning on Non-IID Features
• CVPR 2023Efficient On-Device Training via Gradient Filtering
• ICML 2026On the Fragility of Data Attribution When Learning Is Distributed
• CVPR 2026Mobile-VTON High-Fidelity On-Device Virtual Try-On
• ICML 2024On Online Experimentation without Device Identifiers
• CVPR 2024Cloud-Device Collaborative Learning for Multimodal Large Language Models
• NeurIPS 2023[Re] On the Reproducibility of CartoonX
• ICML 2023FlexRound: Learnable Rounding based on Element-wise Division for Post-Training Quantization
• CVPR 2023TMO: Textured Mesh Acquisition of Objects With a Mobile Device by Using Differentiable Rendering
• ICML 2026Efficient Preference Poisoning Attack on Offline RLHF
• NeurIPS 2023A U-turn on Double Descent: Rethinking Parameter Counting in Statistical Learning
• ICML 2023Towards Robust Graph Incremental Learning on Evolving Graphs

全部端上学习论文

共 90 篇，第 2 / 2 页

51. NeurIPS 2023On student-teacher deviations in distillation: does it pay to disobey?
52. NeurIPS 2023On the Asymptotic Learning Curves of Kernel Ridge Regression under Power-law Decay
53. NeurIPS 2023On the Convergence of Black-Box Variational Inference
54. NeurIPS 2023On the Convergence of CART under Sufficient Impurity Decrease Condition
55. NeurIPS 2023On the Convergence of Encoder-only Shallow Transformers
56. NeurIPS 2023On the Convergence of No-Regret Learning Dynamics in Time-Varying Games
57. NeurIPS 2023On the Convergence to a Global Solution of Shuffling-Type Gradient Algorithms
58. NeurIPS 2023On the Exploitability of Instruction Tuning
59. NeurIPS 2023On the Gini-impurity Preservation For Privacy Random Forests
60. NeurIPS 2023On the Importance of Feature Separability in Predicting Out-Of-Distribution Error
61. NeurIPS 2023On the Learnability of Multilabel Ranking
62. NeurIPS 2023On the Minimax Regret for Online Learning with Feedback Graphs
63. NeurIPS 2023On the Overlooked Structure of Stochastic Gradients
64. NeurIPS 2023On the Pareto Front of Multilingual Neural Machine Translation
65. NeurIPS 2023On the Robustness of Removal-Based Feature Attributions
66. NeurIPS 2023On the Role of Entanglement and Statistics in Learning
67. NeurIPS 2023On the Sublinear Regret of GP-UCB
68. NeurIPS 2023On the Trade-off of Intra-/Inter-class Diversity for Supervised Pre-training
69. NeurIPS 2023On Transfer of Adversarial Robustness from Pretraining to Downstream Tasks
70. NeurIPS 2023Out-of-distribution Detection Learning with Unreliable Out-of-distribution Sources
71. NeurIPS 2023Posterior Contraction Rates for Matérn Gaussian Processes on Riemannian Manifolds
72. NeurIPS 2023Prioritizing Samples in Reinforcement Learning with Reducible Loss
73. NeurIPS 2023Provable Advantage of Curriculum Learning on Parity Targets with Mixed Inputs
74. NeurIPS 2023Provable Training for Graph Contrastive Learning
75. NeurIPS 2023RECESS Vaccine for Federated Learning: Proactive Defense Against Model Poisoning Attacks
76. NeurIPS 2023ReMaX: Relaxing for Better Training on Efficient Panoptic Segmentation
77. NeurIPS 2023Saving 100x Storage: Prototype Replay for Reconstructing Training Sample Distribution in Class-Incremental Semantic Segmentation
78. NeurIPS 2023Self-Adaptive Motion Tracking against On-body Displacement of Flexible Sensors
79. NeurIPS 2023SoTTA: Robust Test-Time Adaptation on Noisy Data Streams
80. NeurIPS 2023State Regularized Policy Optimization on Data with Dynamics Shift
81. NeurIPS 2023Structure Learning with Adaptive Random Neighborhood Informed MCMC
82. NeurIPS 2023Students Parrot Their Teachers: Membership Inference on Model Distillation
83. NeurIPS 2023Tempo Adaptation in Non-stationary Reinforcement Learning
84. NeurIPS 2023Tight Risk Bounds for Gradient Descent on Separable Data
85. NeurIPS 2023Time Series Kernels based on Nonlinear Vector AutoRegressive Delay Embeddings
86. NeurIPS 2023To Stay or Not to Stay in the Pre-train Basin: Insights on Ensembling in Transfer Learning
87. NeurIPS 2023Towards Data-Algorithm Dependent Generalization: a Case Study on Overparameterized Linear Regression
88. NeurIPS 2023Towards Semi-Structured Automatic ICD Coding via Tree-based Contrastive Learning
89. NeurIPS 2023Training on Foveated Images Improves Robustness to Adversarial Attacks
90. NeurIPS 2023Unlocking Deterministic Robustness Certification on ImageNet

常见问题

2023–2026 年四大会议一共有多少篇端上学习论文？

共 134 篇，按年份分别是 2023 年 122 篇、2024 年 4 篇、2025 年 2 篇、2026 年 6 篇。

NeurIPS 2026 有多少篇端上学习论文？

NeurIPS 2026 有 1 篇端上学习论文。

ICLR 2025 有多少篇端上学习论文？

ICLR 2025 有 1 篇端上学习论文。

ICML 2026 有多少篇端上学习论文？

ICML 2026 有 3 篇端上学习论文。

CVPR 2026 有多少篇端上学习论文？

CVPR 2026 有 2 篇端上学习论文，上一届是 1 篇。

哪个会议的端上学习论文最多？

NeurIPS，2023–2026 年共 91 篇。