数学推理论文趋势 2023–2026

主要结论

• 数学推理论文从 2023 年的 57 篇增长到 2026 年的 577 篇，三年间扩大到约 10 倍。
• NeurIPS 是发数学推理论文最多的会议，2023–2026 年共 469 篇，占四个会议总量的 43%。
• 交叉方向里涨得最快的是强化学习：同时属于两个主题的论文从 2023 年的 3 篇增至 2026 年的 140 篇。
• 在「神经符号」的 5 个主题里，它的 2026 年论文数排第 1；同组增长最快的可微推理是 2.5 倍，数学推理是 10 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 mathematical theory of balancing relational generalization and memorization
• ICML 2026ASyMOB: Algebraic Symbolic Mathematical Operations Benchmark
• ICLR 2026AutoGPS: Automated Geometry Problem Solving via Multimodal Formalization and Deductive Reasoning
• CVPR 2026Affine Perspective-Three-Point Problem
• NeurIPS 2025Are Language Models Efficient Reasoners? A Perspective from Logic Programming
• ICML 2025Autoformulation of Mathematical Optimization Models Using LLMs
• ICLR 2025Advancing Mathematical Reasoning in Language Models: The Impact of Problem-Solving Data, Data Synthesis Methods, and Training Stages
• CVPR 2025Bridging Viewpoint Gaps: Geometric Reasoning Boosts Semantic Correspondence
• NeurIPS 2024A Careful Examination of Large Language Model Performance on Grade School Arithmetic
• ICML 2024Accelerating Legacy Numerical Solvers by Non-intrusive Gradient-based Meta-solving
• ICLR 2024Don't Trust: Verify -- Grounding LLM Quantitative Reasoning with Autoformalization
• CVPR 2024E-GPS: Explainable Geometry Problem Solving via Top-Down Solver and Bottom-Up Generator
• NeurIPS 2023Evaluating and Improving Tool-Augmented Computation-Intensive Math Reasoning
• ICML 2023Automated Search for Conjectures on Mathematical Constants using Analysis of Integer Sequences
• ICLR 2023Differentiable Mathematical Programming for Object-Centric Representation Learning
• CVPR 2023Abstract Visual Reasoning: An Algebraic Approach for Solving Raven's Progressive Matrices
• NeurIPS 2026A Mechanistic Analysis of Looped Reasoning Language Models
• ICML 2026Beyond Theorem Proving: Formulation, Framework and Benchmark for Formal Problem-Solving
• ICLR 2026CogFlow: Bridging Perception and Reasoning through Knowledge Internalization for Visual Mathematical Problem Solving
• CVPR 2026Hilbert-Geo Solving Solid Geometric Problems by Neural-Symbolic Reasoning

全部数学推理论文

共 287 篇，第 2 / 6 页

51. ICLR 2026FATE: A Formal Benchmark Series for Frontier Algebra of Multiple Difficulty Levels
52. ICLR 2026From Abstract to Contextual: What LLMs Still Cannot Do in Mathematics
53. ICLR 2026Front-Loading Reasoning: The Synergy between Pretraining and Post-Training Data
54. ICLR 2026FROST: Filtering Reasoning Outliers with Attention for Efficient Reasoning
55. ICLR 2026Generative Adversarial Reasoner: Enhancing LLM Reasoning with Adversarial Reinforcement Learning
56. ICLR 2026GEPO: Group Expectation Policy Optimization for Stable Heterogeneous Reinforcement Learning
57. ICLR 2026Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation
58. ICLR 2026Helix: Evolutionary Reinforcement Learning for Open-Ended Scientific Problem Solving
59. ICLR 2026Heterogeneous Front-Door Effects: Debiased Estimation with Quasi-Oracle Guarantees
60. ICLR 2026Hidden Patterns in Chain-of-Thought Reasoning
61. ICLR 2026Hilbert: Recursively Building Formal Proofs with Informal Reasoning
62. ICLR 2026HiPO: Self-Hint Policy Optimization for RLVR
63. ICLR 2026Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization
64. ICLR 2026Inductive Reasoning for Temporal Knowledge Graphs with Emerging Entities
65. ICLR 2026Is it Thinking or Cheating? Detecting Implicit Reward Hacking by Measuring Reasoning Effort
66. ICLR 2026Is the Reversal Curse a Binding Problem? Uncovering Limitations of Transformers from a Basic Generalization Failure
67. ICLR 2026KaVa: Latent Reasoning via Compressed KV-Cache Distillation
68. ICLR 2026Landscape of Thoughts: Visualizing the Reasoning Process of Large Language Models
69. ICLR 2026Latent Thinking Optimization: Your Latent Reasoning Language Model Secretly Encodes Reward Signals in its Latent Thoughts
70. ICLR 2026Latent Veracity Inference for Identifying Errors in Stepwise Reasoning
71. ICLR 2026Latent-Guided Reasoning: Empowering Small LLMs with Large-Model Thinking
72. ICLR 2026LeanForPhysics: Comprehensive Reasoning Framework for University-level Physics in Lean4
73. ICLR 2026Learn More with Less: Uncertainty Consistency Guided Query Selection for RLVR
74. ICLR 2026Learning from Synthetic Data Improves Multi-hop Reasoning
75. ICLR 2026Learning Global Hypothesis Space for Enhancing Synergistic Reasoning Chain
76. ICLR 2026Learning to Reason in Structured In-context Environments with Reinforcement Learning
77. ICLR 2026Learning to Reason over Continuous Tokens with Reinforcement Learning
78. ICLR 2026Learning to Reason without External Rewards
79. ICLR 2026Let's Explore Step by Step: Generating Provable Formal Statements with Deductive Exploration
80. ICLR 2026LINGOLY-TOO: Disentangling Reasoning from Knowledge with Templatised Orthographic Obfuscation
81. ICLR 2026LLMs are Single-threaded Reasoners: Demystifying the Working Mechanism of Soft Thinking
82. ICLR 2026LoC-Decomp: LLM Autoformalization via Logical Concept Decomposition and Iterative Feedback Correction
83. ICLR 2026LogicReward: Incentivizing LLM Reasoning via Step-Wise Logical Supervision
84. ICLR 2026Long Chain-of-Thought Reasoning Across Languages
85. ICLR 2026Math Blind: Failures in Diagram Understanding Undermine Reasoning in MLLMs
86. ICLR 2026MATH-Beyond: A Benchmark for RL to Expand Beyond the Base Model
87. ICLR 2026Mathesis: Towards Formal Theorem Proving from Natural Languages
88. ICLR 2026MathFimer: Enhancing Mathematical Reasoning by Expanding Reasoning Steps through Fill-in-the-Middle Task
89. ICLR 2026MATHMO: Automated Mathematical Modeling Through Adaptive Search
90. ICLR 2026MathNet: A Global Multimodal Benchmark for Mathematical Reasoning and Retrieval
91. ICLR 2026MSCR: Exploring the Vulnerability of LLMs’ Mathematical Reasoning Abilities Using Multi-Source Candidate Replacement
92. ICLR 2026Native Reasoning Models: Training Language Models to Reason on Unverifiable Data
93. ICLR 2026Nemotron-CC-Math: A 133 Billion-Token-Scale High Quality Math Pretraining Dataset
94. ICLR 2026Neural Force Field: Few-shot Learning of Generalized Physical Reasoning
95. ICLR 2026NFT: Bridging Supervised Learning and Reinforcement Learning in Math Reasoning
96. ICLR 2026Not All Bits Are Equal: How Model Scale Changes Memory-Optimal Reasoning
97. ICLR 2026Nudging the Boundaries of LLM Reasoning
98. ICLR 2026Off-Trajectory Reasoning: Can LLMs Collaborate on Reasoning Trajectory?
99. ICLR 2026On the Reasoning Abilities of Masked Diffusion Language Models
100. ICLR 2026OpenThoughts: Data Recipes for Reasoning Models

常见问题

2023–2026 年四大会议一共有多少篇数学推理论文？

共 1,085 篇，按年份分别是 2023 年 57 篇、2024 年 137 篇、2025 年 314 篇、2026 年 577 篇。

NeurIPS 2026 有多少篇数学推理论文？

NeurIPS 2026 有 208 篇数学推理论文，上一届是 190 篇。

ICLR 2026 有多少篇数学推理论文？

ICLR 2026 有 175 篇数学推理论文，上一届是 61 篇。

ICML 2026 有多少篇数学推理论文？

ICML 2026 有 113 篇数学推理论文，上一届是 45 篇。

CVPR 2026 有多少篇数学推理论文？

CVPR 2026 有 81 篇数学推理论文，上一届是 18 篇。

哪个会议的数学推理论文最多？

NeurIPS，2023–2026 年共 469 篇。