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 190 篇，第 3 / 4 页

101. NeurIPS 2025Measuring the Faithfulness of Thinking Drafts in Large Reasoning Models
102. NeurIPS 2025Mind the Gap: Bridging Thought Leap for Improved Chain-of-Thought Tuning
103. NeurIPS 2025miniF2F-Lean Revisited: Reviewing Limitations and Charting a Path Forward
104. NeurIPS 2025MINT-CoT: Enabling Interleaved Visual Tokens in Mathematical Chain-of-Thought Reasoning
105. NeurIPS 2025Mitigating Overthinking in Large Reasoning Models via Manifold Steering
106. NeurIPS 2025MM-Agent: LLM as Agents for Real-world Mathematical Modeling Problem
107. NeurIPS 2025More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Models
108. NeurIPS 2025Multipole Attention for Efficient Long Context Reasoning
109. NeurIPS 2025NaturalReasoning: Reasoning in the Wild with 2.8M Challenging Questions
110. NeurIPS 2025No Experts, No Problem: Avoidance Learning from Bad Demonstrations
111. NeurIPS 2025OMEGA: Can LLMs Reason Outside the Box in Math? Evaluating Exploratory, Compositional, and Transformative Generalization
112. NeurIPS 2025Omni-R1: Reinforcement Learning for Omnimodal Reasoning via Two-System Collaboration
113. NeurIPS 2025On Learning Verifiers and Implications to Chain-of-Thought Reasoning
114. NeurIPS 2025On the Bias of Next-Token Predictors Toward Systematically Inefficient Reasoning: A Shortest-Path Case Study
115. NeurIPS 2025Once Upon an Input: Reasoning via Per-Instance Program Synthesis
116. NeurIPS 2025One Token Embedding Is Enough to Deadlock Your Large Reasoning Model
117. NeurIPS 2025Optimizing Anytime Reasoning via Budget Relative Policy Optimization
118. NeurIPS 2025Optimizing Chain-of-Thought Reasoners via Gradient Variance Minimization in Rejection Sampling and RL
119. NeurIPS 2025PolyMath: Evaluating Mathematical Reasoning in Multilingual Contexts
120. NeurIPS 2025Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning
121. NeurIPS 2025ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models
122. NeurIPS 2025QFFT, Question-Free Fine-Tuning for Adaptive Reasoning
123. NeurIPS 2025QiMeng-CodeV-R1: Reasoning-Enhanced Verilog Generation
124. NeurIPS 2025QuestBench: Can LLMs ask the right question to acquire information in reasoning tasks?
125. NeurIPS 2025R-KV: Redundancy-aware KV Cache Compression for Reasoning Models
126. NeurIPS 2025R$^2$ec: Towards Large Recommender Models with Reasoning
127. NeurIPS 2025R2R: Efficiently Navigating Divergent Reasoning Paths with Small-Large Model Token Routing
128. NeurIPS 2025RealMath: A Continuous Benchmark for Evaluating Language Models on Research-Level Mathematics
129. NeurIPS 2025ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs
130. NeurIPS 2025Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought
131. NeurIPS 2025Reasoning Is Not a Race: When Stopping Early Beats Going Deeper
132. NeurIPS 2025Reasoning Models Better Express Their Confidence
133. NeurIPS 2025Reasoning Models Hallucinate More: Factuality-Aware Reinforcement Learning for Large Reasoning Models
134. NeurIPS 2025Reasoning Path Compression: Compressing Generation Trajectories for Efficient LLM Reasoning
135. NeurIPS 2025Reasoning Planning for Language Models
136. NeurIPS 2025Reinforcement Learning for Out-of-Distribution Reasoning in LLMs: An Empirical Study on Diagnosis-Related Group Coding
137. NeurIPS 2025Reinforcement Learning for Reasoning in Large Language Models with One Training Example
138. NeurIPS 2025Reinforcement Learning Teachers of Test Time Scaling
139. NeurIPS 2025Rethinking Fine-Tuning when Scaling Test-Time Compute: Limiting Confidence Improves Mathematical Reasoning
140. NeurIPS 2025S-GRPO: Early Exit via Reinforcement Learning in Reasoning Models
141. NeurIPS 2025SATURN: SAT-based Reinforcement Learning to Unleash LLMs Reasoning
142. NeurIPS 2025Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model Reasoning
143. NeurIPS 2025Scaling Speculative Decoding with Lookahead Reasoning
144. NeurIPS 2025Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
145. NeurIPS 2025SCOUT: Teaching Pre-trained Language Models to Enhance Reasoning via Flow Chain-of-Thought
146. NeurIPS 2025Self-Verifying Reflection Helps Transformers with CoT Reasoning
147. NeurIPS 2025SemCoT: Accelerating Chain-of-Thought Reasoning through Semantically-Aligned Implicit Tokens
148. NeurIPS 2025ShorterBetter: Guiding Reasoning Models to Find Optimal Inference Length for Efficient Reasoning
149. NeurIPS 2025SIGMA: Refining Large Language Model Reasoning via Sibling-Guided Monte Carlo Augmentation
150. NeurIPS 2025Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space

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。