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 577 篇，第 3 / 12 页

101. ICLR 2026CMT-Benchmark: A Benchmark for Condensed Matter Theory Built by Expert Researchers
102. ICLR 2026CogFlow: Bridging Perception and Reasoning through Knowledge Internalization for Visual Mathematical Problem Solving
103. ICLR 2026Compositional-ARC: Assessing Systematic Generalization in Abstract Spatial Reasoning
104. ICLR 2026Conditional Advantage Estimation for Reinforcement Learning in Large Reasoning Models
105. ICLR 2026Continuous Chain of Thought Enables Parallel Exploration and Reasoning
106. ICLR 2026CORE: Concept-Oriented Reinforcement for Bridging the Definition–Application Gap in Mathematical Reasoning
107. ICLR 2026CoT-Evo: Evolutionary Distillation of Chain-of-Thought for Scientific Reasoning
108. ICLR 2026Curriculum Reinforcement Learning from Easy to Hard Tasks Improves LLM Reasoning
109. ICLR 2026CyclicReflex: Improving Reasoning Models via Cyclical Reflection Token Scheduling
110. ICLR 2026DAG-Math: Graph-Guided Mathematical Reasoning in LLMs
111. ICLR 2026Deep Think with Confidence
112. ICLR 2026DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning
113. ICLR 2026DIVA-GRPO: Enhancing Multimodal Reasoning through Difficulty-Adaptive Variant Advantage
114. ICLR 2026Diversity-Enhanced Reasoning for Subjective Questions
115. ICLR 2026Divide and Abstract: Autoformalization via Decomposition and Abstraction Learning
116. ICLR 2026DRAGON: Guard LLM Unlearning in Context via Negative Detection and Reasoning
117. ICLR 2026DRIFT: Decompose, Retrieve, Illustrate, then Formalize Theorems
118. ICLR 2026DRPO: Efficient Reasoning via Decoupled Reward Policy Optimization
119. ICLR 2026Dynamic Early Exit in Reasoning Models
120. ICLR 2026Echoes as Anchors: Probabilistic Costs and Attention Refocusing in LLM Reasoning
121. ICLR 2026Efficient Reasoning with Balanced Thinking
122. ICLR 2026Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning
123. ICLR 2026Energy-Based Transformers are Scalable Learners and Thinkers
124. ICLR 2026Equilibrium Language Models
125. ICLR 2026Evaluating Cross-Modal Reasoning Ability and Problem Charactaristics with Multimodal Item Response Theory
126. ICLR 2026Evaluating Language Models' Evaluations of Games
127. ICLR 2026EvolProver: Advancing Automated theorem proving by Evolving Formalized Problems via Symmetry and Difficulty
128. ICLR 2026ExGRPO: Learning to Reason from Prior Successes
129. ICLR 2026Expanding Reasoning Potential in Foundation Model by Learning Diverse Chains of Thought Patterns
130. ICLR 2026Explain in Your Own Words: Improving Reasoning via Token-Selective Dual Knowledge Distillation
131. ICLR 2026FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning
132. ICLR 2026FATE: A Formal Benchmark Series for Frontier Algebra of Multiple Difficulty Levels
133. ICLR 2026From Abstract to Contextual: What LLMs Still Cannot Do in Mathematics
134. ICLR 2026Front-Loading Reasoning: The Synergy between Pretraining and Post-Training Data
135. ICLR 2026FROST: Filtering Reasoning Outliers with Attention for Efficient Reasoning
136. ICLR 2026Generative Adversarial Reasoner: Enhancing LLM Reasoning with Adversarial Reinforcement Learning
137. ICLR 2026GEPO: Group Expectation Policy Optimization for Stable Heterogeneous Reinforcement Learning
138. ICLR 2026Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation
139. ICLR 2026Helix: Evolutionary Reinforcement Learning for Open-Ended Scientific Problem Solving
140. ICLR 2026Heterogeneous Front-Door Effects: Debiased Estimation with Quasi-Oracle Guarantees
141. ICLR 2026Hidden Patterns in Chain-of-Thought Reasoning
142. ICLR 2026Hilbert: Recursively Building Formal Proofs with Informal Reasoning
143. ICLR 2026HiPO: Self-Hint Policy Optimization for RLVR
144. ICLR 2026Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization
145. ICLR 2026Inductive Reasoning for Temporal Knowledge Graphs with Emerging Entities
146. ICLR 2026Is it Thinking or Cheating? Detecting Implicit Reward Hacking by Measuring Reasoning Effort
147. ICLR 2026Is the Reversal Curse a Binding Problem? Uncovering Limitations of Transformers from a Basic Generalization Failure
148. ICLR 2026KaVa: Latent Reasoning via Compressed KV-Cache Distillation
149. ICLR 2026Landscape of Thoughts: Visualizing the Reasoning Process of Large Language Models
150. ICLR 2026Latent Thinking Optimization: Your Latent Reasoning Language Model Secretly Encodes Reward Signals in its Latent Thoughts

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。