机制设计论文趋势 2023–2026

主要结论

• 机制设计论文从 2023 年的 140 篇增长到 2026 年的 426 篇，三年间扩大到约 3.0 倍。
• NeurIPS 是发机制设计论文最多的会议，2023–2026 年共 394 篇，占四个会议总量的 38%。
• 交叉方向里涨得最快的是多模态大模型：同时属于两个主题的论文从 2023 年的 5 篇增至 2026 年的 42 篇。
• 在「博弈与规划」的 5 个主题里，它的 2026 年论文数排第 2；同组增长最快的经典规划是 3.3 倍，机制设计是 3.0 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 2026A Generalized Tikhonov Layer for Interpretable-by-design Graph Neural Networks
• ICML 2026Affine-Scaled Attention: Towards Flexible and Stable Transformer Attention
• ICLR 2026Bayesian Attention Mechanism: A Probabilistic Framework for Positional Encoding and Context Length Extrapolation
• CVPR 2026FEAT Fashion Editing and Try-On from Any Design
• NeurIPS 2025A Closer Look at Graph Transformers: Cross-Aggregation and Beyond
• ICML 2025Benign Overfitting in Token Selection of Attention Mechanism
• ICLR 2025AdaManip: Adaptive Articulated Object Manipulation Environments and Policy Learning
• CVPR 2025A Selective Re-learning Mechanism for Hyperspectral Fusion Imaging
• NeurIPS 2024AFBench: A Large-scale Benchmark for Airfoil Design
• ICML 2024CarbonNovo: Joint Design of Protein Structure and Sequence Using a Unified Energy-based Model
• ICLR 2024Adaptive Instrument Design for Indirect Experiments
• CVPR 2024InceptionNeXt: When Inception Meets ConvNeXt
• NeurIPS 2023An Optimal and Scalable Matrix Mechanism for Noisy Marginals under Convex Loss Functions
• ICML 2023AbODE: Ab initio antibody design using conjoined ODEs
• ICLR 2023Conditional Antibody Design as 3D Equivariant Graph Translation
• CVPR 2023Exploring the Relationship Between Architectural Design and Adversarially Robust Generalization
• NeurIPS 2026A Unified Framework for Uniform-Price Resource Allocation Mechanisms
• ICML 2026Anatomy of Massive Activations and Attention Sinks
• ICLR 2026How Muon’s Spectral Design Benefits Generalization: A Study on Imbalanced Data
• CVPR 2026Parallax to Align Them All An OmniParallax Attention Mechanism for

全部机制设计论文

共 175 篇，第 1 / 4 页

1. NeurIPS 2026$\text{PartConcepts}$: A Unified Mechanism for Fine-Grained Part Localization and Generation
2. NeurIPS 2026$R^2E$: A Role-driven Reward Evolutionary Framework for Automated Reward Function Design
3. NeurIPS 2026A Generalized Tikhonov Layer for Interpretable-by-design Graph Neural Networks
4. NeurIPS 2026A Unified Framework for Uniform-Price Resource Allocation Mechanisms
5. NeurIPS 2026Agentic Neural Architecture Search
6. NeurIPS 2026Agree to Disagree: Multimodal Autonomous Negotiation and Calibration for Entity Representation Learning
7. NeurIPS 2026Algorithmic Impact Reveals the Hidden Structure of Alignment
8. NeurIPS 2026Aligning AI Teams
9. NeurIPS 2026All-Addition Spiking Diffusion Models with Attention Enhancement
10. NeurIPS 2026AlloGen: Conformation-Selective Binder Design with Differential State Scoring
11. NeurIPS 2026An Equivariance Principle for Optimizer Design: Symmetry-Compatible Updates for Embeddings, LM Heads, and MoE Routers
12. NeurIPS 2026Analyzing Learning-Dynamics Across Loss and Prediction Choices for Diffusion Models
13. NeurIPS 2026Archimedean Copula Inference via Taylor-Mode AD
14. NeurIPS 2026Assign and Add: A Mechanistic Study of Compositional Arithmetic
15. NeurIPS 2026Attention Sinks and Outliers in Attention Residuals
16. NeurIPS 2026AutoManifold: Agentic Design of Data Visualisation Algorithms via Manifold Embedding
17. NeurIPS 2026BA-T: An Iterative Transformer for Two-View Bundle Adjustment
18. NeurIPS 2026Bayesian Causal Experimental Design for CATE Estimation under Noncompliance
19. NeurIPS 2026Benchmarks Design under Data Scarcity: From Coarse Labels to Diagnostic Evaluation of Biosynthetic Gene Cluster Models
20. NeurIPS 2026Benign Reinforcement Learning Can Amplify Latent Backdoors
21. NeurIPS 2026Best Arm Identification in Generalized Linear Bandits via Hybrid Feedback
22. NeurIPS 2026Beyond ICA: Identifiability by Symmetry Breaking
23. NeurIPS 2026Beyond MMSE: Enhancing PnP Restoration with ProxiMAP
24. NeurIPS 2026Block Sphere Vector Quantization
25. NeurIPS 2026Bonobo: Efficient Library-Scale Generation for De Novo Antibody Design
26. NeurIPS 2026Bootstrapped Bipartite Actor-Critic for Diffusion RL
27. NeurIPS 2026Bug or Feature$^2$: Weight Drift, Activation Sparsity, and Spikes
28. NeurIPS 2026Can Hybrid-Parallel Planning Support Alternating Model–Strategy Design? Dependency-Keyed Per-Layer Primitives for Re-Planning
29. NeurIPS 2026CANDO: Cooperative Agentic Network for Layout Design Optimization
30. NeurIPS 2026Cauchy Scientific Networks: Loss–Architecture Alignment and Its Limit
31. NeurIPS 2026CHARM+: Cross-Hardware Attention with Re-Merge Multistream Mechanism
32. NeurIPS 2026Cluster with Auctions for Vector Search
33. NeurIPS 2026Coherence Mechanisms for Provable Self-Improvement
34. NeurIPS 2026CompactAttention: Accelerating Chunked Prefill with Block-Union KV Selection
35. NeurIPS 2026Computational Dynamic Mechanism Design
36. NeurIPS 2026Continuous-depth Deep Gaussian Processes
37. NeurIPS 2026Convergence of Near-Linear Width ReLU Networks with Unbalanced Initialization
38. NeurIPS 2026Decoding the Critique Mechanism in Large Reasoning Models
39. NeurIPS 2026Deep Double Q-learning
40. NeurIPS 2026DeltaMomentum: A Key-Value based Anisotropic Momentum Update via Delta Rule
41. NeurIPS 2026Detect Anything in Graphic Design: Element-Level Rewards for Autoregressive Detection
42. NeurIPS 2026Differentiating Network Design Objectives for Balancing Cost and Distance
43. NeurIPS 2026Disentanglement as Identifiable Pushforward Factorisation
44. NeurIPS 2026Distribution Free Fourier Sparsity Testing
45. NeurIPS 2026Does Weight Decay Enhance Training Stability?
46. NeurIPS 2026EGCA: A Spectral Perspective on Forward Process Design in Diffusion Models
47. NeurIPS 2026Epigenomics-Guided Flow Matching for 3D Genome Super-Resolution
48. NeurIPS 2026Ergodic Trajectory Design by Learned Pushforward Maps: Provable Coverage via Conditional Flow Matching
49. NeurIPS 2026EvoDiagram: Agentic Editable Diagram Creation via Design Expertise Evolution
50. NeurIPS 2026Explaining and Preventing Alignment Collapse in Iterative RLHF

常见问题

2023–2026 年四大会议一共有多少篇机制设计论文？

共 1,037 篇，按年份分别是 2023 年 140 篇、2024 年 200 篇、2025 年 271 篇、2026 年 426 篇。

NeurIPS 2026 有多少篇机制设计论文？

NeurIPS 2026 有 175 篇机制设计论文，上一届是 108 篇。

ICLR 2026 有多少篇机制设计论文？

ICLR 2026 有 65 篇机制设计论文，上一届是 58 篇。

ICML 2026 有多少篇机制设计论文？

ICML 2026 有 128 篇机制设计论文，上一届是 65 篇。

CVPR 2026 有多少篇机制设计论文？

CVPR 2026 有 58 篇机制设计论文，上一届是 40 篇。

哪个会议的机制设计论文最多？

NeurIPS，2023–2026 年共 394 篇。