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 175 篇，第 3 / 4 页

101. ICLR 2026Parallel-R1: Towards Parallel Thinking via Reinforcement Learning
102. ICLR 2026Patching Gaps In LLM Reasoning With Interventional Training
103. ICLR 2026PEAR: Phase Entropy Aware Reward for Efficient Reasoning
104. ICLR 2026Plan and Budget: Effective and Efficient Test-Time Scaling on Reasoning Large Language Models
105. ICLR 2026Premise Selection for a Lean Hammer
106. ICLR 2026PRISM-Physics: Causal DAG-Based Process Evaluation for Physics Reasoning
107. ICLR 2026ProofBridge: Auto-Formalization of Natural Language Proofs in Lean via Joint Embeddings
108. ICLR 2026ProofOptimizer: Training Language Models to Simplify Proofs without Human Demonstrations
109. ICLR 2026PROS: Towards Compute-Efficient RLVR via Rollout Prefix Reuse
110. ICLR 2026Pruning Long Chain-of-Thought of Large Reasoning Models via Small-Scale Preference Optimization
111. ICLR 2026Pushing on Multilingual Reasoning Models with Language-Mixed Chain-of-Thought
112. ICLR 2026QuestA: Expanding Reasoning Capacity in LLMs via Question Augmentation
113. ICLR 2026R-Horizon: How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
114. ICLR 2026R-Zero: Self-Evolving Reasoning LLM from Zero Data
115. ICLR 2026RADAR: Reasoning–Ability and Difficulty-Aware Routing in Language Models
116. ICLR 2026RAIN-Merging: A Gradient-Free Method to Enhance Instruction Following in Large Reasoning Models with Preserved Thinking Format
117. ICLR 2026Random Policy Valuation is Enough for LLM Reasoning with Verifiable Rewards
118. ICLR 2026Real-Time Reasoning Agents in Evolving Environments
119. ICLR 2026Reasoning in Space via Grounding in the World
120. ICLR 2026Reasoning Language Model Inference Serving Unveiled: An Empirical Study
121. ICLR 2026Reasoning Models Can be Accurately Pruned Via Chain-of-Thought Reconstruction
122. ICLR 2026Reasoning Scaffolding: Distilling the Flow of Thought from LLMs
123. ICLR 2026ReForm: Reflective Autoformalization with Prospective Bounded Sequence Optimization
124. ICLR 2026Reinforced Latent Reasoning for LLM-based Recommendation
125. ICLR 2026Reinforcement Mid-Training
126. ICLR 2026Reliable Fine-Grained Evaluation of Natural Language Math Proofs
127. ICLR 2026Rethinking LLM Reasoning: From Explicit Trajectories to Latent Representations
128. ICLR 2026Rewriting Pre-Training Data Boosts LLM Performance in Math and Code
129. ICLR 2026RFEval: Benchmarking Reasoning Faithfulness under Counterfactual Reasoning Intervention in Large Reasoning Models
130. ICLR 2026RLAD: Training LLMs to Discover Abstractions for Solving Reasoning Problems
131. ICLR 2026RLP: Reinforcement as a Pretraining Objective
132. ICLR 2026RLVMR: Reinforcement Learning with Verifiable Meta-Reasoning Rewards for Robust Long-Horizon Agents
133. ICLR 2026RM-R1: Reward Modeling as Reasoning
134. ICLR 2026Sample Smart, Not Hard: Correctness-First Decoding for Better Reasoning in LLMs
135. ICLR 2026Scaf-GRPO: Scaffolded Group Relative Policy Optimization for Enhancing LLM Reasoning
136. ICLR 2026Scalable Chain of Thoughts via Elastic Reasoning
137. ICLR 2026Scheduling Your LLM Reinforcement Learning with Reasoning Trees
138. ICLR 2026Segment-Level Attribution for Selective Learning of Long Reasoning Traces
139. ICLR 2026Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens
140. ICLR 2026Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training
141. ICLR 2026Self-Refining Vision Language Model for Robotic Failure Detection and Reasoning
142. ICLR 2026SIM-CoT: Supervised Implicit Chain-of-Thought
143. ICLR 2026SLM-MUX: Orchestrating Small Language Models for Reasoning
144. ICLR 2026Slow-Fast Policy Optimization: Reposition-Before-Update for LLM Reasoning
145. ICLR 2026Smarter Not Harder: Generative Process Evaluation with Intrinsic-Signal Driving and Ability‑Adaptive Reward Shaping
146. ICLR 2026Soft Tokens, Hard Truths
147. ICLR 2026Solving the 2-norm k-hyperplane clustering problem via multi-norm formulations
148. ICLR 2026Stabilizing Policy Gradients for Sample-Efficient Reinforcement Learning in LLM Reasoning
149. ICLR 2026STITCH: Simultaneous Thinking and Talking with Chunked Reasoning for Spoken Language Models
150. ICLR 2026Strategic Obfuscation of Deceptive Reasoning in Language 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。