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 208 篇，第 4 / 5 页

151. NeurIPS 2026Scaling Causal Reasoning with Increasingly Complex Causal Simulators
152. NeurIPS 2026Seeing or Rationalizing? Scene-Evidence-Guided Chain-of-Thought for Faithful 3D Multimodal Reasoning
153. NeurIPS 2026Seirênes: Adversarial Self-Play with Evolving Distractions for LLM Reasoning
154. NeurIPS 2026Self-Compacting Language Model Agents
155. NeurIPS 2026Semantic Search over 9 Million Mathematical Theorems
156. NeurIPS 2026SFT-then-RL Outperforms Mixed-Policy Methods for LLM Reasoning
157. NeurIPS 2026SH$^2$: A Mathematician-Curated benchmark for Assessing Research-level Math Capabilities of LLMs
158. NeurIPS 2026Simulating Students or Sycophantic Problem Solving? On Misconception Faithfulness of LLM Simulators
159. NeurIPS 2026Smart Picks in the Dark: Towards Efficient RLVR for Reasoning via Tracing Metacognitive Pivots
160. NeurIPS 2026SoDeArena: A Socially-Situated Reasoning Benchmark for Large Language Models
161. NeurIPS 2026Solving Every Step Is Not Enough: Milestone Oracles Reveal a Composition Gap in LLM Math Reasoning
162. NeurIPS 2026Sparse Attention as a Range Searching Problem: Towards an Inference-Efficient Index for KV Cache
163. NeurIPS 2026Sparse Expansion Utility: Identifying and Routing to Pivotal Steps in LLM Reasoning Chains
164. NeurIPS 2026SphMind: Towards Robust, Training-Free VLM-based Spatial Reasoning with a 360 Camera
165. NeurIPS 2026Stability in Multi-Step Reasoning via Jacobian-based Error Accumulation Analysis
166. NeurIPS 2026Stabilizing Knowledge, Promoting Reasoning: Dual-Token Constraints for RLVR
167. NeurIPS 2026STAR-Math: Multi-Agent Mathematical Reasoning under Persistent Meta-Strategic Supervision
168. NeurIPS 2026State of Thought Enables Endogenous Reasoning
169. NeurIPS 2026StateTree: Enhancing Long-term Dialogue Reasoning via Reinforcement Learning
170. NeurIPS 2026Step-by-Step Optimization-like Reasoning in LLMs over Expanding Search Spaces
171. NeurIPS 2026Step-wise Rubric Rewards for LLM Reasoning
172. NeurIPS 2026Stepwise Penalization for Length-Efficient Chain-of-Thought Reasoning
173. NeurIPS 2026Stop or Restart? Principled Inference Control for Large Reasoning Models via the Pandora's Box
174. NeurIPS 2026Strategist: Designing Agentic Reasoning at Scale
175. NeurIPS 2026STRIDE: Learnable Stepwise Language Feedback for LLM Reasoning
176. NeurIPS 2026Structural Rationale Distillation via Reasoning Space Compression
177. NeurIPS 2026TargetSage: Identifying Therapeutic Target Genes with Interpretable and Robust LLM Reasoning
178. NeurIPS 2026TARP: Trace-Anchored Regularization Prior for Retaining Stepwise Reasoning
179. NeurIPS 2026TaskGround: Structured Executable Task Inference for Full-Scene Household Reasoning
180. NeurIPS 2026Teach-to-Reason: Competition-Guided Reasoning with a Self-Improving Teacher
181. NeurIPS 2026TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning
182. NeurIPS 2026Test-Time Learning with an Evolving Library
183. NeurIPS 2026The Capability Frontier of GRPO
184. NeurIPS 2026The Era of Agentic Organization: Learning to Organize with Language Models
185. NeurIPS 2026The Reasoning Boundary Paradox: How Reinforcement Learning Constrains Language Models
186. NeurIPS 2026They Can See It but not Say It: Iterative Self Knowledge Re-expression in Visual Reasoning Relative Pose Identification
187. NeurIPS 2026TIGER: Bridging the Multimodal Reasoning-Access Gap via Modality Counterfactuals
188. NeurIPS 2026Time to Pay Attention! Understanding High Complexity Corpus Reasoning Tasks
189. NeurIPS 2026Train at the Moving Edge: Rollout-Efficient RL for Large Reasoning Models
190. NeurIPS 2026Training with Harnesses: On-Policy Harness Self-Distillation for Complex Reasoning
191. NeurIPS 2026Training-Free Dynamic Upcycling of Expert Language Models
192. NeurIPS 2026Treat Bias as Noise: Training Bias-Robust LLM Reasoning via Reinforcement Learning
193. NeurIPS 2026Triggering Generalist Reasoning via Predictive Uncertainty for Dual-System VLA
194. NeurIPS 2026TSQAgent: Rating Time Series Data Quality via Dedicated Agentic Reasoning
195. NeurIPS 2026Understanding the Out-of-Distribution Generalization of Chain-of-Thought Reasoning in LLMs
196. NeurIPS 2026Unifying Reasoning and Planning through Energy Minimization
197. NeurIPS 2026UniRAP: Towards Unified Part-level Physical Affordance Reasoning and Actionable Perception
198. NeurIPS 2026Value-Aware Stochastic KV Cache Eviction for Reasoning Models
199. NeurIPS 2026Verifiable Environments Are LEGO Bricks: Recursive Composition for Reasoning Generalization
200. NeurIPS 2026Verifier-Backed Hard Problem Generation for Mathematical Reasoning

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。