数学推理论文趋势 2023–2026

定理与解题

统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 1,085 篇被归入该主题的论文。

主要结论

每年论文数

各会议论文数

按年份和会议统计

数学推理论文数，按年份和会议
年份NeurIPSICLRICMLCVPR合计
20231917111057
202452343912137
2025190614518314
202620817511381577

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

全部数学推理论文

年份全部2023202420252026
会议全部NeurIPSICLRICMLCVPR

共 469 篇，第 7 / 10 页

  1. NeurIPS 2025Learning Interestingness in Automated Mathematical Theory Formation
  2. NeurIPS 2025Learning to Solve Complex Problems via Dataset Decomposition
  3. NeurIPS 2025Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RL
  4. NeurIPS 2025Let LRMs Break Free from Overthinking via Self-Braking Tuning
  5. NeurIPS 2025LIMOPro: Reasoning Refinement for Efficient and Effective Test-time Scaling
  6. NeurIPS 2025Look-Ahead Reasoning on Learning Platforms
  7. NeurIPS 2025Lost in Transmission: When and Why LLMs Fail to Reason Globally
  8. NeurIPS 2025MathArena: Evaluating LLMs on Uncontaminated Math Competitions
  9. NeurIPS 2025Measuring the Faithfulness of Thinking Drafts in Large Reasoning Models
  10. NeurIPS 2025Mind the Gap: Bridging Thought Leap for Improved Chain-of-Thought Tuning
  11. NeurIPS 2025miniF2F-Lean Revisited: Reviewing Limitations and Charting a Path Forward
  12. NeurIPS 2025MINT-CoT: Enabling Interleaved Visual Tokens in Mathematical Chain-of-Thought Reasoning
  13. NeurIPS 2025Mitigating Overthinking in Large Reasoning Models via Manifold Steering
  14. NeurIPS 2025MM-Agent: LLM as Agents for Real-world Mathematical Modeling Problem
  15. NeurIPS 2025More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Models
  16. NeurIPS 2025Multipole Attention for Efficient Long Context Reasoning
  17. NeurIPS 2025NaturalReasoning: Reasoning in the Wild with 2.8M Challenging Questions
  18. NeurIPS 2025No Experts, No Problem: Avoidance Learning from Bad Demonstrations
  19. NeurIPS 2025OMEGA: Can LLMs Reason Outside the Box in Math? Evaluating Exploratory, Compositional, and Transformative Generalization
  20. NeurIPS 2025Omni-R1: Reinforcement Learning for Omnimodal Reasoning via Two-System Collaboration
  21. NeurIPS 2025On Learning Verifiers and Implications to Chain-of-Thought Reasoning
  22. NeurIPS 2025On the Bias of Next-Token Predictors Toward Systematically Inefficient Reasoning: A Shortest-Path Case Study
  23. NeurIPS 2025Once Upon an Input: Reasoning via Per-Instance Program Synthesis
  24. NeurIPS 2025One Token Embedding Is Enough to Deadlock Your Large Reasoning Model
  25. NeurIPS 2025Optimizing Anytime Reasoning via Budget Relative Policy Optimization
  26. NeurIPS 2025Optimizing Chain-of-Thought Reasoners via Gradient Variance Minimization in Rejection Sampling and RL
  27. NeurIPS 2025PolyMath: Evaluating Mathematical Reasoning in Multilingual Contexts
  28. NeurIPS 2025Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning
  29. NeurIPS 2025ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models
  30. NeurIPS 2025QFFT, Question-Free Fine-Tuning for Adaptive Reasoning
  31. NeurIPS 2025QiMeng-CodeV-R1: Reasoning-Enhanced Verilog Generation
  32. NeurIPS 2025QuestBench: Can LLMs ask the right question to acquire information in reasoning tasks?
  33. NeurIPS 2025R-KV: Redundancy-aware KV Cache Compression for Reasoning Models
  34. NeurIPS 2025R$^2$ec: Towards Large Recommender Models with Reasoning
  35. NeurIPS 2025R2R: Efficiently Navigating Divergent Reasoning Paths with Small-Large Model Token Routing
  36. NeurIPS 2025RealMath: A Continuous Benchmark for Evaluating Language Models on Research-Level Mathematics
  37. NeurIPS 2025ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs
  38. NeurIPS 2025Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought
  39. NeurIPS 2025Reasoning Is Not a Race: When Stopping Early Beats Going Deeper
  40. NeurIPS 2025Reasoning Models Better Express Their Confidence
  41. NeurIPS 2025Reasoning Models Hallucinate More: Factuality-Aware Reinforcement Learning for Large Reasoning Models
  42. NeurIPS 2025Reasoning Path Compression: Compressing Generation Trajectories for Efficient LLM Reasoning
  43. NeurIPS 2025Reasoning Planning for Language Models
  44. NeurIPS 2025Reinforcement Learning for Out-of-Distribution Reasoning in LLMs: An Empirical Study on Diagnosis-Related Group Coding
  45. NeurIPS 2025Reinforcement Learning for Reasoning in Large Language Models with One Training Example
  46. NeurIPS 2025Reinforcement Learning Teachers of Test Time Scaling
  47. NeurIPS 2025Rethinking Fine-Tuning when Scaling Test-Time Compute: Limiting Confidence Improves Mathematical Reasoning
  48. NeurIPS 2025S-GRPO: Early Exit via Reinforcement Learning in Reasoning Models
  49. NeurIPS 2025SATURN: SAT-based Reinforcement Learning to Unleash LLMs Reasoning
  50. NeurIPS 2025Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model Reasoning

在 Claude、Cursor 里直接问这个趋势

接上 Paperlist MCP 后，写综述、挑方向或准备投稿时，不用离开对话框就能拿到这页的数字：每年各会议的论文数、增长倍数、涨得最快的子方向和代表论文，AI 会按你的问题整理。

可以直接复制这句话问 AI

用 paperlist 查数学推理在 NeurIPS 2023–2026 的论文趋势，哪一年涨得最快？

AI 会调用的工具：get_topic_stats {"topic": "math-reasoning"}

两分钟接入 Paperlist MCP（Claude Code、Cursor、Codex、ChatGPT）

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。

「神经符号」下的其他主题