推理模型论文趋势 2023–2026

链式推理、测试时计算与显式推理能力

统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 3,512 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

推理模型论文数，按年份和会议
年份NeurIPSICLRICMLCVPR合计
20232715710933470
2024121846343311
202550020315970932
20263616525822041,799

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

全部推理模型论文

年份全部2023202420252026
会议全部NeurIPSICLRICMLCVPR

共 3,512 篇，第 54 / 71 页

  1. NeurIPS 2025TCM-Ladder: A Benchmark for Multimodal Question Answering on Traditional Chinese Medicine
  2. NeurIPS 2025Teaching Language Models to Reason with Tools
  3. NeurIPS 2025TempSamp-R1: Effective Temporal Sampling with Reinforcement Fine-Tuning for Video LLMs
  4. NeurIPS 2025Test-Time Adaptation by Causal Trimming
  5. NeurIPS 2025The Dual Nature of Plasticity Loss in Deep Continual Learning: Dissection and Mitigation
  6. NeurIPS 2025The Emergence of Abstract Thought in Large Language Models Beyond Any Language
  7. NeurIPS 2025The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models
  8. NeurIPS 2025The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awareness
  9. NeurIPS 2025The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
  10. NeurIPS 2025The Overthinker's DIET: Cutting Token Calories with DIfficulty-AwarE Training
  11. NeurIPS 2025The Promise of RL for Autoregressive Image Editing
  12. NeurIPS 2025The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning
  13. NeurIPS 2025Think Only When You Need with Large Hybrid-Reasoning Models
  14. NeurIPS 2025Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens
  15. NeurIPS 2025Think Silently, Think Fast: Dynamic Latent Compression of LLM Reasoning Chains
  16. NeurIPS 2025Think-RM: Enabling Long-Horizon Reasoning in Generative Reward Models
  17. NeurIPS 2025ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planning
  18. NeurIPS 2025ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning
  19. NeurIPS 2025Thinker: Learning to Think Fast and Slow
  20. NeurIPS 2025Thinking in Character: Advancing Role-Playing Agents with Role-Aware Reasoning
  21. NeurIPS 2025Thinking vs. Doing: Improving Agent Reasoning by Scaling Test-Time Interaction
  22. NeurIPS 2025Thinkless: LLM Learns When to Think
  23. NeurIPS 2025ThinkSound: Chain-of-Thought Reasoning in Multimodal LLMs for Audio Generation and Editing
  24. NeurIPS 2025Thoughts Are All Over the Place: On the Underthinking of Long Reasoning Models
  25. NeurIPS 2025TimE: A Multi-level Benchmark for Temporal Reasoning of LLMs in Real-World Scenarios
  26. NeurIPS 2025Titans: Learning to Memorize at Test Time
  27. NeurIPS 2025TokenSqueeze: Performance-Preserving Compression for Reasoning LLMs
  28. NeurIPS 2025TOMCAT: Test-time Comprehensive Knowledge Accumulation for Compositional Zero-Shot Learning
  29. NeurIPS 2025Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrieval
  30. NeurIPS 2025Tool-Augmented Spatiotemporal Reasoning for Streamlining Video Question Answering Task
  31. NeurIPS 2025Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties
  32. NeurIPS 2025TopoPoint: Enhance Topology Reasoning via Endpoint Detection in Autonomous Driving
  33. NeurIPS 2025Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning
  34. NeurIPS 2025Tracking and Understanding Object Transformations
  35. NeurIPS 2025Training a Scientific Reasoning Model for Chemistry
  36. NeurIPS 2025Training Language Models to Reason Efficiently
  37. NeurIPS 2025Trajectory Balance with Asynchrony: Decoupling Exploration and Learning for Fast, Scalable LLM Post-Training
  38. NeurIPS 2025Transformers Provably Learn Chain-of-Thought Reasoning with Length Generalization
  39. NeurIPS 2025Tropical Attention: Neural Algorithmic Reasoning for Combinatorial Algorithms
  40. NeurIPS 2025Trust, But Verify: A Self-Verification Approach to Reinforcement Learning with Verifiable Rewards
  41. NeurIPS 2025TRUST: Test-Time Refinement using Uncertainty-Guided SSM Traverses
  42. NeurIPS 2025Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training
  43. NeurIPS 2025UFT: Unifying Supervised and Reinforcement Fine-Tuning
  44. NeurIPS 2025Understanding and Mitigating Numerical Sources of Nondeterminism in LLM Inference
  45. NeurIPS 2025Uni-MuMER: Unified Multi-Task Fine-Tuning of Vision-Language Model for Handwritten Mathematical Expression Recognition
  46. NeurIPS 2025Unified Multimodal Chain-of-Thought Reward Model through Reinforcement Fine-Tuning
  47. NeurIPS 2025UniGen: Enhanced Training & Test-Time Strategies for Unified Multimodal Understanding and Generation
  48. NeurIPS 2025UniPixel: Unified Object Referring and Segmentation for Pixel-Level Visual Reasoning
  49. NeurIPS 2025UniTransfer: Video Concept Transfer via Progressive Spatio-Temporal Decomposition
  50. NeurIPS 2025Unlabeled Data Can Provably Enhance In-Context Learning of Transformers

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 查推理模型在 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": "reasoning-models"}

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推理模型论文？
共 3,512 篇，按年份分别是 2023 年 470 篇、2024 年 311 篇、2025 年 932 篇、2026 年 1,799 篇。
NeurIPS 2026 有多少篇推理模型论文？
NeurIPS 2026 有 361 篇推理模型论文，上一届是 500 篇。
ICLR 2026 有多少篇推理模型论文？
ICLR 2026 有 652 篇推理模型论文，上一届是 203 篇。
ICML 2026 有多少篇推理模型论文？
ICML 2026 有 582 篇推理模型论文，上一届是 159 篇。
CVPR 2026 有多少篇推理模型论文？
CVPR 2026 有 204 篇推理模型论文，上一届是 70 篇。
哪个会议的推理模型论文最多？
NeurIPS，2023–2026 年共 1,253 篇。

「前沿系统」下的其他主题