多任务学习论文趋势 2023–2026

主要结论

• 多任务学习论文从 2023 年的 288 篇增长到 2026 年的 402 篇，三年间扩大到约 1.4 倍。
• NeurIPS 是发多任务学习论文最多的会议，2023–2026 年共 399 篇，占四个会议总量的 34%。
• 交叉方向里涨得最快的是数据集与基准：同时属于两个主题的论文从 2023 年的 6 篇增至 2026 年的 16 篇。
• 在「元学习与迁移」的 5 个主题里，它的 2026 年论文数排第 2；同组增长最快的基础模型适配是 3.8 倍，多任务学习是 1.4 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 2026A Mahalanobis Margin \texorpdfstring{$\gamma_{\min}$}{gamma\_min} Bound on Task Confusion in Pretrained Class-Incremental Learning: From Infeasibility to Exponential Attenuation
• ICML 2026A Systematic Study of Behavioral Cloning for Scientific Data Annotation
• ICLR 2026Automatic Stage Lighting Control: Is it a Rule-Driven Process or Generative Task?
• CVPR 20263D-Aware Multi-Task Learning with Cross-View Correlations for Dense Scene Understanding
• NeurIPS 2025A High-Dimensional Statistical Method for Optimizing Transfer Quantities in Multi-Source Transfer Learning
• ICML 2025ATA: Adaptive Task Allocation for Efficient Resource Management in Distributed Machine Learning
• ICLR 2025CodeMMLU: A Multi-Task Benchmark for Assessing Code Understanding & Reasoning Capabilities of CodeLLMs
• CVPR 2025ACAttack: Adaptive Cross Attacking RGB-T Tracker via Multi-Modal Response Decoupling
• NeurIPS 2024CoMix: A Comprehensive Benchmark for Multi-Task Comic Understanding
• ICML 2024Bayesian Uncertainty for Gradient Aggregation in Multi-Task Learning
• ICLR 2024Class Incremental Learning via Likelihood Ratio Based Task Prediction
• CVPR 2024Differentiable Information Bottleneck for Deterministic Multi-view Clustering
• NeurIPS 2023A Spectral Algorithm for List-Decodable Covariance Estimation in Relative Frobenius Norm
• ICML 2023A Conditional Normalizing Flow for Accelerated Multi-Coil MR Imaging
• ICLR 2023An Extensible Multi-modal Multi-task Object Dataset with Materials
• CVPR 2023AMT: All-Pairs Multi-Field Transforms for Efficient Frame Interpolation
• NeurIPS 2026A Scalable Multi-Task Model for Virtual Sensors
• ICML 2026Alleviating Observation Bias via Causal-Invariant Meta-Learning for Unbalanced Incomplete Multi-view Clustering
• ICLR 2026Beyond Hearing: Learning Task-agnostic ExG Representations from Earphones via Physiology-informed Tokenization
• CVPR 2026Any Resolution Any Geometry From Multi-View To Multi-Patch

全部多任务学习论文

共 91 篇，第 2 / 2 页

51. ICML 2026LEMUR: Learned Multi-Vector Retrieval
52. ICML 2026LiME: Lightweight Mixture of Experts for Efficient Multimodal Multi-task Learning
53. ICML 2026MAGIC: Multi-Granularity Language-Informed Image Clustering
54. ICML 2026Masked Multi-path Contrast with Confidence-Gated Semantic Imputation for Incomplete Multi-view Clustering
55. ICML 2026MIMO-LP: A Multi-Input Multi-Output Framework for Subgraph-based Link Prediction
56. ICML 2026MM-Spectrum: Multimodal Multi-spectral Molecular Structural Elucidation with a Stable MoE Framework
57. ICML 2026Multi-Head Attention as a Source of Catastrophic Forgetting in MoE Transformers
58. ICML 2026Multi-Task Bayesian In-Context Learning
59. ICML 2026Multi-Task GRPO: Reliable LLM Reasoning Across Tasks
60. ICML 2026Multi-task Linear Regression without Eigenvalue Lower Bounds: Adaptivity, Robustness, and Safety
61. ICML 2026Multi-Way Representation Alignment
62. ICML 2026Multimodal Fusion via Self-Consistent Task-Gradient Fields
63. ICML 2026NBCG: Nash-Bargained Causal Game for Long-Tailed Multi-Label NLP
64. ICML 2026Near-Optimal and Efficient First-Order Algorithm for Multi-Task Learning with Shared Linear Representation
65. ICML 2026OPTION: Optimal Transport–Guided Flow Matching for Incomplete and Unaligned Multi-View Clustering
66. ICML 2026Orthogonal Model Merging
67. ICML 2026Personalized Additive Modeling for Multi-level Federated Learning
68. ICML 2026Post-Hoc Merging is Not Enough: Many-Shot Model Merging with Loss-Gap Balancing
69. ICML 2026Predicting evolutionary rate as a pretraining task improves genome language model representations
70. ICML 2026Probabilistic Performance Guarantees for Multi-Task Reinforcement Learning
71. ICML 2026Problem Distributions as Tasks: Repurposing Meta Learning for Generative Combinatorial Optimization towards Multi-task Pretraining and Adaptation
72. ICML 2026Rethinking Human Intent-to-CAD: Parametric CAD Model Generation via Cooperative Multi-Task Alignment and Spatial-Aware Reinforcement Learning
73. ICML 2026Saliency-Aware Model Merging
74. ICML 2026Shared Lexical Task Representations Explain Behavioral Variability In LLMs
75. ICML 2026Sketch-Based Low-Rank Model Merging with Shared Circulant Transforms
76. ICML 2026Structured Expert Routing with Multi-View Task Priors for Offline Meta-Reinforcement Learning
77. ICML 2026Structured Multi-modal Graph Disentanglement for Psychiatric Diagnosis
78. ICML 2026SyMerge: From Non-Interference to Synergistic Merging via Single-Layer Adaptation
79. ICML 2026T-GINEE: A Tensor-Based Multi-Graph Representation Learning
80. ICML 2026TACTIC: Task-Aware Sparse Coordination Graphs for Multi-Task Multi-agent Reinforcement Learning
81. ICML 2026Task-Aware Exploration via a Predictive Bisimulation Metric
82. ICML 2026Task-Aware Structured Memory for Dynamic Multi-modal In-Context Learning
83. ICML 2026The Geometry of Updates: Fisher Alignment at Vocabulary Scale
84. ICML 2026TimeMRA: LLM-Empowered Time Series Forecasting via Multi-Scale Retrieval-Augmented Representations
85. ICML 2026Transformers learn factored representations
86. ICML 2026Transporting Task Vectors across Different Architectures without Training
87. ICML 2026Unifying Heterogeneous Multi-Modal Remote Sensing Detection Via Language-Pivoted Pretraining
88. ICML 2026WarmServe: Enabling One-for-Many GPU Prewarming for Multi-LLM Serving
89. ICML 2026Weaving in the Clouds: Achieving Synergistic Collaboration among LLM Agents via Federated Learning
90. ICML 2026What Linear Probes Miss: Multi-View Probing for Weight-Space Learning
91. ICML 2026When Shared Knowledge Hurts: Spectral Over-Accumulation in Model Merging

常见问题

2023–2026 年四大会议一共有多少篇多任务学习论文？

共 1,181 篇，按年份分别是 2023 年 288 篇、2024 年 227 篇、2025 年 264 篇、2026 年 402 篇。

NeurIPS 2026 有多少篇多任务学习论文？

NeurIPS 2026 有 126 篇多任务学习论文，上一届是 101 篇。

ICLR 2026 有多少篇多任务学习论文？

ICLR 2026 有 57 篇多任务学习论文，上一届是 40 篇。

ICML 2026 有多少篇多任务学习论文？

ICML 2026 有 91 篇多任务学习论文，上一届是 47 篇。

CVPR 2026 有多少篇多任务学习论文？

CVPR 2026 有 128 篇多任务学习论文，上一届是 76 篇。

哪个会议的多任务学习论文最多？

NeurIPS，2023–2026 年共 399 篇。