缩放定律论文趋势 2023–2026

主要结论

• 缩放定律论文从 2023 年的 60 篇增长到 2026 年的 248 篇，三年间扩大到约 4.1 倍。
• NeurIPS 是发缩放定律论文最多的会议，2023–2026 年共 221 篇，占四个会议总量的 37%。
• 交叉方向里涨得最快的是时间序列：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 24 篇。
• 在「优化与缩放」的 5 个主题里，它的 2026 年论文数排第 5；同组增长最快的低秩适配是 3.0 倍，缩放定律是 4.1 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 Unified Merge Calculus for Learning-Rate Scaling on Neural Computation Graphs
• ICML 2026A Capacity-Based Rationale for Multi-Head Attention
• ICLR 2026$\mathbf{Li_2}$: A Framework on Dynamics of Feature Emergence and Delayed Generalization
• CVPR 2026A Unified Framework for Knowledge Transfer in Bidirectional Model Scaling
• NeurIPS 2025Audits Under Resource, Data, and Access Constraints: Scaling Laws For Less Discriminatory Alternatives
• ICML 2025A Simple Model of Inference Scaling Laws
• ICLR 2025(Mis)Fitting Scaling Laws: A Survey of Scaling Law Fitting Techniques in Deep Learning
• CVPR 2025D^3: Scaling Up Deepfake Detection by Learning from Discrepancy
• NeurIPS 2024An exactly solvable model for emergence and scaling laws in the multitask sparse parity problem
• ICML 2024A Dynamical Model of Neural Scaling Laws
• ICLR 2024Principled Architecture-aware Scaling of Hyperparameters
• CVPR 2024From Activation to Initialization: Scaling Insights for Optimizing Neural Fields
• NeurIPS 2023Abide by the law and follow the flow: conservation laws for gradient flows
• ICML 2023Dissecting the Effects of SGD Noise in Distinct Regimes of Deep Learning
• ICLR 2023Distributionally Robust Post-hoc Classifiers under Prior Shifts
• CVPR 2023LargeKernel3D: Scaling Up Kernels in 3D Sparse CNNs
• NeurIPS 2026ADIS-Law: Unified Scaling Laws for Annealing-Phase Domain Injection in Large Language Models
• ICML 2026Balancing Learning Rates Across Layers: Exact Two-Step Dynamics and Optimal Scaling in Linear Neural Networks
• ICLR 2026A universal compression theory: Lottery ticket hypothesis and superpolynomial scaling laws
• CVPR 2026MatAnyone 2 Scaling Video Matting via a Learned Quality Evaluator

全部缩放定律论文

共 221 篇，第 2 / 5 页

51. NeurIPS 2026Rethinking Language Model Scaling under Transferable Hypersphere Optimization
52. NeurIPS 2026Scale Where It Matters: Training-Free Localized Scaling for Diffusion Models
53. NeurIPS 2026Scaling Arbitrary Architectures and Optimizers with Automatic Parameterization
54. NeurIPS 2026Scaling Genomic Language Modeling with Unified Corpus and Evaluation in Bacteria
55. NeurIPS 2026Scaling Laws and Tradeoffs in Recurrent Networks of Expressive Neurons
56. NeurIPS 2026Scaling Laws for SAE Training Data
57. NeurIPS 2026Scaling Linear Mode Connectivity and Merging to Billion Parameter Pretrained Transformers
58. NeurIPS 2026Scaling Neural Motor Decoding via Decoupled Behavioral Pretraining
59. NeurIPS 2026Scaling Optimization-Oriented Hypernetworks for Implicit Neural Representations
60. NeurIPS 2026Scaling Storm-Resolving Atmospheric AI Simulation to the Entire Planet
61. NeurIPS 2026SDS-LoRA: Overcoming Anisotropic Gradient Scaling in Low-Rank Adaptation
62. NeurIPS 2026Selected-Tail Reliability in Verifier-Guided Best-of-$N$ Inference
63. NeurIPS 2026SETA: Scaling Environments for Terminal Agents
64. NeurIPS 2026Sharp Capacity Scaling of Spectral Optimizers in Learning Associative Memory
65. NeurIPS 2026Sharp feature-learning transitions and Bayes-optimal neural scaling laws in extensive-width networks
66. NeurIPS 2026Sharpness, Stability, and Step-Size Scaling in Deep Polynomial Networks
67. NeurIPS 2026SWE-Crafter: Scaling Executable Multilingual Software Engineering Data with Meta-Skill Agents
68. NeurIPS 2026Systematic Scaling Analysis of Jailbreak Attacks in Large Language Models
69. NeurIPS 2026Tail Competition Explains Scaling in Best-of-N Sampling for Verifiable Problems
70. NeurIPS 2026Target-Budget Best-Function Identification for Scaling-Law-Guided Model Selection
71. NeurIPS 2026Test-time Scaling of Diffusions with Flow Maps
72. NeurIPS 2026Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching
73. NeurIPS 2026The Geometric Wall: Manifold Structure Predicts Layerwise Sparse Autoencoder Scaling Laws
74. NeurIPS 2026The Hidden Power of Scaling Factor in LoRA Optimization
75. NeurIPS 2026The Minimax Rate of Second-Order Calibration
76. NeurIPS 2026The Scaling Laws of Skills in LLM Agent Systems
77. NeurIPS 2026Theory of Optimal Learning Rate Schedules and Scaling Laws for a Random Feature Model
78. NeurIPS 2026Tokens-per-Parameter Coverage Is Critical for Robust LLM Scaling Law Extrapolation
79. NeurIPS 2026Training Optimal Large Diffusion Language Models
80. NeurIPS 2026What should post-training optimize? A test-time scaling law perspective
81. NeurIPS 2026When Less is More: The LLM Scaling Paradox in Context Compression
82. NeurIPS 2026When Reasoning Meets Its Laws
83. NeurIPS 2025$\mu$PC: Scaling Predictive Coding to 100+ Layer Networks
84. NeurIPS 2025Align Your Flow: Scaling Continuous-Time Flow Map Distillation
85. NeurIPS 2025AlphaZero Neural Scaling and Zipf's Law: a Tale of Board Games and Power Laws
86. NeurIPS 2025Audits Under Resource, Data, and Access Constraints: Scaling Laws For Less Discriminatory Alternatives
87. NeurIPS 2025Avoiding exp(R) scaling in RLHF through Preference-based Exploration
88. NeurIPS 2025BioCLIP 2: Emergent Properties from Scaling Hierarchical Contrastive Learning
89. NeurIPS 2025Can DPO Learn Diverse Human Values? A Theoretical Scaling Law
90. NeurIPS 2025Communication-Efficient Language Model Training Scales Reliably and Robustly: Scaling Laws for DiLoCo
91. NeurIPS 2025Complexity Scaling Laws for Neural Models using Combinatorial Optimization
92. NeurIPS 2025Compute-Optimal Scaling for Value-Based Deep RL
93. NeurIPS 2025Data Mixing Can Induce Phase Transitions in Knowledge Acquisition
94. NeurIPS 2025Data Mixture Optimization: A Multi-fidelity Multi-scale Bayesian Framework
95. NeurIPS 2025Datasets, Documents, and Repetitions: The Practicalities of Unequal Data Quality
96. NeurIPS 2025Dimension-adapted Momentum Outscales SGD
97. NeurIPS 2025DISC: Dynamic Decomposition Improves LLM Inference Scaling
98. NeurIPS 2025Does Thinking More Always Help? Mirage of Test-Time Scaling in Reasoning Models
99. NeurIPS 2025E2Former: An Efficient and Equivariant Transformer with Linear-Scaling Tensor Products
100. NeurIPS 2025Efficiently Scaling LLM Reasoning Programs with Certaindex

常见问题

2023–2026 年四大会议一共有多少篇缩放定律论文？

共 605 篇，按年份分别是 2023 年 60 篇、2024 年 96 篇、2025 年 201 篇、2026 年 248 篇。

NeurIPS 2026 有多少篇缩放定律论文？

NeurIPS 2026 有 82 篇缩放定律论文，上一届是 80 篇。

ICLR 2026 有多少篇缩放定律论文？

ICLR 2026 有 73 篇缩放定律论文，上一届是 48 篇。

ICML 2026 有多少篇缩放定律论文？

ICML 2026 有 78 篇缩放定律论文，上一届是 56 篇。

CVPR 2026 有多少篇缩放定律论文？

CVPR 2026 有 15 篇缩放定律论文，上一届是 17 篇。

哪个会议的缩放定律论文最多？

NeurIPS，2023–2026 年共 221 篇。