逻辑与神经论文趋势 2023–2026

主要结论

• 逻辑与神经论文从 2023 年的 162 篇降到 2026 年的 86 篇，热度在回落。
• NeurIPS 是发逻辑与神经论文最多的会议，2023–2026 年共 173 篇，占四个会议总量的 47%。
• 交叉方向里涨得最快的是社会层面 · accountability transparency and interpretability：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 8 篇。
• 在「神经符号」的 5 个主题里，它的 2026 年论文数排第 2；同组增长最快的数学推理是 10 倍，逻辑与神经是 0.5 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 2026B$^3$-PWL: GPU-Batched Branch-and-Bound for Piecewise-Linear Optimization with SOS2 Constraints
• ICML 2026Length Generalization Bounds for Transformers
• ICLR 2026From Neural Networks to Logical Theories: The Correspondence between Fibring Modal Logics and Fibring Neural Networks
• CVPR 2026Med-CMR A Fine-Grained Benchmark Integrating Visual Evidence and Clinical Logic
• NeurIPS 2025Characterizing the Expressivity of Fixed-Precision Transformer Language Models
• ICML 2025Counterfactual Graphical Models: Constraints and Inference
• ICLR 2025Beyond the convexity assumption: Realistic tabular data generation under quantifier-free real linear constraints
• CVPR 2025Learning Affine Correspondences by Integrating Geometric Constraints
• NeurIPS 2024Beyond Primal-Dual Methods in Bandits with Stochastic and Adversarial Constraints
• ICML 2024A Field Guide for Pacing Budget and ROS Constraints
• ICLR 2024LogicMP: A Neuro-symbolic Approach for Encoding First-order Logic Constraints
• CVPR 2024MS-MANO: Enabling Hand Pose Tracking with Biomechanical Constraints
• NeurIPS 2023Aligning Gradient and Hessian for Neural Signed Distance Function
• ICML 2023Eventual Discounting Temporal Logic Counterfactual Experience Replay
• ICLR 2023Global Explainability of GNNs via Logic Combination of Learned Concepts
• CVPR 2023Handy: Towards a High Fidelity 3D Hand Shape and Appearance Model
• NeurIPS 2026Batched Stochastic Linear Bandits with 1-Bit Communication Constraints
• ICML 2026KANFIS: A Neuro-Symbolic Framework for Interpretable and Uncertainty-Aware Learning
• ICLR 2026The Logical Expressiveness of Topological Neural Networks
• CVPR 2026NeuroRule Bridging Vision and Logic with Differentiable Rule Induction

全部逻辑与神经论文

共 366 篇，第 2 / 8 页

51. ICML 2026The Theory and Practice of MAP Inference over Non-Convex Constraints
52. ICML 2026Towards Achieving Optimal Strong Regret and Constraint Violation via Computationally Efficient Model-free RL
53. ICML 2026Two-Stage Unit Tying for Simplifying Differentiable Logic Gate Networks
54. ICML 2026Unlearning in Diffusion Models: A Unified Framework with KL Divergence and Likelihood Constraints
55. NeurIPS 2026A Unified Neural Architecture for Variable-Wise Shape Constraints
56. NeurIPS 2026B$^3$-PWL: GPU-Batched Branch-and-Bound for Piecewise-Linear Optimization with SOS2 Constraints
57. NeurIPS 2026Batched Stochastic Linear Bandits with 1-Bit Communication Constraints
58. NeurIPS 2026Beyond the Half Approximation: Fair and Efficient Online Class Matching
59. NeurIPS 2026COMPOSE: Composing Future Theorems from Citations and Formal Structure
60. NeurIPS 2026Constrained MDPs with Trajectory Constraints
61. NeurIPS 2026Expressive Power of Deep Homomorphism Networks over Relational Databases
62. NeurIPS 2026Federated Logic Gate Networks via Boolean Feature Selection
63. NeurIPS 2026Ferrogen: Generative Pipeline for Guided Search of Novel Ferroelectric Material for Logic and Memory
64. NeurIPS 2026Finite-Sample and Communication-Efficient Networked Information Aggregation
65. NeurIPS 2026GeoX: Mastering Geospatial Reasoning Through Self-Play and Verifiable Rewards
66. NeurIPS 2026Graph Learning Should Move Beyond Restrictive Views of Spectral and Message-Passing GNNs
67. NeurIPS 2026Harnessing Accurate and Automatic Trend Detection in Data Streams via Tbps-Level Inference
68. NeurIPS 2026Interpretable Multiway-Split Trees
69. NeurIPS 2026Learning Unbiased Permutations via Flow Matching
70. NeurIPS 2026Mining Logic under Uncertainty: Probabilistic Soft Logic with Energy-Based Inference for Chain-of-Thought Verification
71. NeurIPS 2026Nearly Optimal Fixed-Confidence Best-Arm Identification with 1-Bit Feedback
72. NeurIPS 2026Neural Networks With Dense Weights Are Not Universal Approximators
73. NeurIPS 2026Normative Networks for Source Separation via Local Plasticity and Dendritic Computation
74. NeurIPS 2026Overcoming Kernel Redundancy for Scaling Logic Gate Networks
75. NeurIPS 2026ProjKAN: Model Compression via KAN Projections to Bridge the Hypothesis and Capacity Gaps
76. NeurIPS 2026Reconciling Safety and Performance via Dual-Expert Offline Imitation Learning
77. NeurIPS 2026Reliable Clustering and Quantization via Distortion-Constrained Optimal Transport
78. NeurIPS 2026Richer or More Gates? Fan-In Trade-offs in Learnable Logic Circuits
79. NeurIPS 2026Runtime Monitoring of Perception-Based Autonomous Systems via Embedding Temporal Logic
80. NeurIPS 2026Safe Offline Reinforcement Learning using Behavior Regularisation and Latent Feasibility-Guidance
81. NeurIPS 2026Safe-PG-LQR: Provably Safe, Convergent and Optimal Model-Free Linear Quadratic Regulator with Hard Constraints
82. NeurIPS 2026SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
83. NeurIPS 2026Small Model Portfolios for Many Deployment Profiles: Submodular Coverage under Bundled Constraints
84. NeurIPS 2026SynGeo: Synergizing Seeing and Proving through Revisable Geometric States
85. NeurIPS 2026Temporal Smoothness Constraints on Efficient Neurobiological Codes Imply Temporal Specialization
86. NeurIPS 2026TGPO: Temporal Grounded Policy Optimization for Signal Temporal Logic Tasks
87. CVPR 2025Estimating Body and Hand Motion in an Ego-sensed World
88. CVPR 2025Learning Affine Correspondences by Integrating Geometric Constraints
89. CVPR 2025LOGICZSL: Exploring Logic-induced Representation for Compositional Zero-shot Learning
90. CVPR 2025MotionPRO: Exploring the Role of Pressure in Human MoCap and Beyond
91. CVPR 2025PLeaS - Merging Models with Permutations and Least Squares
92. CVPR 2025SCFlow2: Plug-and-Play Object Pose Refiner with Shape-Constraint Scene Flow
93. CVPR 2025Towards Cost-Effective Learning: A Synergy of Semi-Supervised and Active Learning
94. CVPR 2025UMotion: Uncertainty-driven Human Motion Estimation from Inertial and Ultra-wideband Units
95. ICLR 2025Beyond the convexity assumption: Realistic tabular data generation under quantifier-free real linear constraints
96. ICLR 2025Circuit Transformer: A Transformer That Preserves Logical Equivalence
97. ICLR 2025Enhance Multi-View Classification Through Multi-Scale Alignment and Expanded Boundary
98. ICLR 2025Faster Algorithms for Structured Linear and Kernel Support Vector Machines
99. ICLR 2025GeoILP: A Synthetic Dataset to Guide Large-Scale Rule Induction
100. ICLR 2025Interference Among First-Price Pacing Equilibria: A Bias and Variance Analysis

常见问题

2023–2026 年四大会议一共有多少篇逻辑与神经论文？

共 366 篇，按年份分别是 2023 年 162 篇、2024 年 64 篇、2025 年 54 篇、2026 年 86 篇。

NeurIPS 2026 有多少篇逻辑与神经论文？

NeurIPS 2026 有 32 篇逻辑与神经论文，上一届是 23 篇。

ICLR 2026 有多少篇逻辑与神经论文？

ICLR 2026 有 20 篇逻辑与神经论文，上一届是 11 篇。

ICML 2026 有多少篇逻辑与神经论文？

ICML 2026 有 25 篇逻辑与神经论文，上一届是 12 篇。

CVPR 2026 有多少篇逻辑与神经论文？

CVPR 2026 有 9 篇逻辑与神经论文，上一届是 8 篇。

哪个会议的逻辑与神经论文最多？

NeurIPS，2023–2026 年共 173 篇。