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 1,085 篇，第 16 / 22 页

751. NeurIPS 2025Deliberation on Priors: Trustworthy Reasoning of Large Language Models on Knowledge Graphs
752. NeurIPS 2025Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning
753. NeurIPS 2025DisCO: Reinforcing Large Reasoning Models with Discriminative Constrained Optimization
754. NeurIPS 2025Diversity-Aware Policy Optimization for Large Language Model Reasoning
755. NeurIPS 2025DLoFT: Gradient-Decoupled Fine-Tuning for Generalizable Long Chain-of-Thought Reasoning
756. NeurIPS 2025Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
757. NeurIPS 2025Don’t Think Longer, Think Wisely: Optimizing Thinking Dynamics for Large Reasoning Models
758. NeurIPS 2025DreamPRM: Domain-reweighted Process Reward Model for Multimodal Reasoning
759. NeurIPS 2025DSRF: A Dynamic and Scalable Reasoning Framework for Solving RPMs
760. NeurIPS 2025DuetGraph: Coarse-to-Fine Knowledge Graph Reasoning with Dual-Pathway Global-Local Fusion
761. NeurIPS 2025DyFlow: Dynamic Workflow Framework for Agentic Reasoning
762. NeurIPS 2025DynaAct: Large Language Model Reasoning with Dynamic Action Spaces
763. NeurIPS 2025Efficiently Scaling LLM Reasoning Programs with Certaindex
764. NeurIPS 2025Enigmata: Scaling Logical Reasoning in Large Language Models with Synthetic Verifiable Puzzles
765. NeurIPS 2025EvaLearn: Quantifying the Learning Capability and Efficiency of LLMs via Sequential Problem Solving
766. NeurIPS 2025Evaluating the Inductive Abilities of Large Language Models: Why Chain-of-Thought Reasoning Sometimes Hurts More Than Helps
767. NeurIPS 2025Evolutionary Reasoning Does Not Arise in Standard Usage of Protein Language Models
768. NeurIPS 2025Execution Guided Line-by-Line Code Generation
769. NeurIPS 2025ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning
770. NeurIPS 2025Far from the Shallow: Brain-Predictive Reasoning Embedding through Residual Disentanglement
771. NeurIPS 2025First SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Training
772. NeurIPS 2025Flexible Realignment of Language Models
773. NeurIPS 2025From Euler to AI: Unifying Formulas for Mathematical Constants
774. NeurIPS 2025Generalizable Reasoning through Compositional Energy Minimization
775. NeurIPS 2025GPO: Learning from Critical Steps to Improve LLM Reasoning
776. NeurIPS 2025GRIP: A Graph-Based Reasoning Instruction Producer
777. NeurIPS 2025HARDMath2: A Benchmark for Applied Mathematics Built by Students as Part of a Graduate Class
778. NeurIPS 2025HELM: Hyperbolic Large Language Models via Mixture-of-Curvature Experts
779. NeurIPS 2025Heterogeneous Adversarial Play in Interactive Environments
780. NeurIPS 2025How Far Are We from Optimal Reasoning Efficiency?
781. NeurIPS 2025Hybrid Latent Reasoning via Reinforcement Learning
782. NeurIPS 2025Hybrid-Balance GFlowNet for Solving Vehicle Routing Problems
783. NeurIPS 2025Improve Temporal Reasoning in Multimodal Large Language Models via Video Contrastive Decoding
784. NeurIPS 2025Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning
785. NeurIPS 2025Incentivizing LLMs to Self-Verify Their Answers
786. NeurIPS 2025Incentivizing Reasoning for Advanced Instruction-Following of Large Language Models
787. NeurIPS 2025Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving of Inequalities
788. NeurIPS 2025IneqSearch: Hybrid Reasoning for Olympiad Inequality Proofs
789. NeurIPS 2025Is PRM Necessary? Problem-Solving RL Implicitly Induces PRM Capability in LLMs
790. NeurIPS 2025Knot So Simple: A Minimalistic Environment for Spatial Reasoning
791. NeurIPS 2025KTAE: A Model-Free Algorithm to Key-Tokens Advantage Estimation in Mathematical Reasoning
792. NeurIPS 2025Large Language Models Think Too Fast To Explore Effectively
793. NeurIPS 2025Latent Refinement via Flow Matching for Training-free Linear Inverse Problem Solving
794. NeurIPS 2025Learning Interestingness in Automated Mathematical Theory Formation
795. NeurIPS 2025Learning to Solve Complex Problems via Dataset Decomposition
796. NeurIPS 2025Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RL
797. NeurIPS 2025Let LRMs Break Free from Overthinking via Self-Braking Tuning
798. NeurIPS 2025LIMOPro: Reasoning Refinement for Efficient and Effective Test-time Scaling
799. NeurIPS 2025Look-Ahead Reasoning on Learning Platforms
800. NeurIPS 2025Lost in Transmission: When and Why LLMs Fail to Reason Globally

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。