对齐与安全论文趋势 2023–2026

RLHF、偏好优化、红队与安全对齐

统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 4,558 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

对齐与安全论文数，按年份和会议
年份NeurIPSICLRICMLCVPR合计
202313887101112438
2024348156246125875
20254554733001351,363
20264806235512281,882

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

全部对齐与安全论文

年份全部2023202420252026
会议全部NeurIPSICLRICMLCVPR

共 455 篇，第 5 / 10 页

  1. NeurIPS 2025Inference-time Alignment in Continuous Space
  2. NeurIPS 2025Inference-Time Personalized Alignment with a Few User Preference Queries
  3. NeurIPS 2025Inference-Time Reward Hacking in Large Language Models
  4. NeurIPS 2025Inference-Time Scaling for Flow Models via Stochastic Generation and Rollover Budget Forcing
  5. NeurIPS 2025Inference-Time Text-to-Video Alignment with Diffusion Latent Beam Search
  6. NeurIPS 2025InfiFPO: Implicit Model Fusion via Preference Optimization in Large Language Models
  7. NeurIPS 2025Information Retrieval Induced Safety Degradation in AI Agents
  8. NeurIPS 2025Information-Theoretic Reward Decomposition for Generalizable RLHF
  9. NeurIPS 2025Infrequent Exploration in Linear Bandits
  10. NeurIPS 2025Integrating Drug Substructures and Longitudinal Electronic Health Records for Personalized Drug Recommendation
  11. NeurIPS 2025Intermediate Domain Alignment and Morphology Analogy for Patent-Product Image Retrieval
  12. NeurIPS 2025InterMT: Multi-Turn Interleaved Preference Alignment with Human Feedback
  13. NeurIPS 2025IPAD: Inverse Prompt for AI Detection - A Robust and Interpretable LLM-Generated Text Detector
  14. NeurIPS 2025JADE: Joint Alignment and Deep Embedding for Multi-Slice Spatial Transcriptomics
  15. NeurIPS 2025JailBound: Jailbreaking Internal Safety Boundaries of Vision-Language Models
  16. NeurIPS 2025Keeping an Eye on LLM Unlearning: The Hidden Risk and Remedy
  17. NeurIPS 2025KL Penalty Control via Perturbation for Direct Preference Optimization
  18. NeurIPS 2025KL-Regularized RLHF with Multiple Reference Models: Exact Solutions and Sample Complexity
  19. NeurIPS 2025LABridge: Text–Image Latent Alignment Framework via Mean-Conditioned OU Process
  20. NeurIPS 2025Language Model Behavioral Phases are Consistent Across Architecture, Training Data, and Scale
  21. NeurIPS 2025Language Models Are Capable of Metacognitive Monitoring and Control of Their Internal Activations
  22. NeurIPS 2025Large language models can learn and generalize steganographic chain-of-thought under process supervision
  23. NeurIPS 2025LARGO: Latent Adversarial Reflection through Gradient Optimization for Jailbreaking LLMs
  24. NeurIPS 2025LASeR: Learning to Adaptively Select Reward Models with Multi-Arm Bandits
  25. NeurIPS 2025Latent Harmony: Synergistic Unified UHD Image Restoration via Latent Space Regularization and Controllable Refinement
  26. NeurIPS 2025LayerNavigator: Finding Promising Intervention Layers for Efficient Activation Steering in Large Language Models
  27. NeurIPS 2025Learning “Partner-Aware” Collaborators in Multi-Party Collaboration
  28. NeurIPS 2025Learning a Cross-Modal Schrödinger Bridge for Visual Domain Generalization
  29. NeurIPS 2025Learning single index models via harmonic decomposition
  30. NeurIPS 2025Learning to Add, Multiply, and Execute Algorithmic Instructions Exactly with Neural Networks
  31. NeurIPS 2025Learning to Steer: Input-dependent Steering for Multimodal LLMs
  32. NeurIPS 2025Less is More: Improving LLM Alignment via Preference Data Selection
  33. NeurIPS 2025Let a Neural Network be Your Invariant
  34. NeurIPS 2025Leveraging Importance Sampling to Detach Alignment Modules from Large Language Models
  35. NeurIPS 2025Leveraging robust optimization for llm alignment under distribution shifts
  36. NeurIPS 2025LeVo: High-Quality Song Generation with Multi-Preference Alignment
  37. NeurIPS 2025Lifelong Safety Alignment for Language Models
  38. NeurIPS 2025LLM Safety Alignment is Divergence Estimation in Disguise
  39. NeurIPS 2025LLMs Encode Harmfulness and Refusal Separately
  40. NeurIPS 2025LongVPO: From Anchored Cues to Self-Reasoning for Long-Form Video Preference Optimization
  41. NeurIPS 2025Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy and Research
  42. NeurIPS 2025macOSWorld: A Multilingual Interactive Benchmark for GUI Agents
  43. NeurIPS 2025Many LLMs Are More Utilitarian Than One
  44. NeurIPS 2025Matchings Under Biased and Correlated Evaluations
  45. NeurIPS 2025Measuring and Guiding Monosemanticity
  46. NeurIPS 2025Mechanism Design for LLM Fine-tuning with Multiple Reward Models
  47. NeurIPS 2025Median Selection with Noisy and Structural Information
  48. NeurIPS 2025Mesh-RFT: Enhancing Mesh Generation via Fine-grained Reinforcement Fine-Tuning
  49. NeurIPS 2025Meta-Learning Objectives for Preference Optimization
  50. NeurIPS 2025MGUP: A Momentum-Gradient Alignment Update Policy for Stochastic Optimization

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 查对齐与安全在 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": "alignment-and-safety"}

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对齐与安全论文？
共 4,558 篇，按年份分别是 2023 年 438 篇、2024 年 875 篇、2025 年 1,363 篇、2026 年 1,882 篇。
NeurIPS 2026 有多少篇对齐与安全论文？
NeurIPS 2026 有 480 篇对齐与安全论文，上一届是 455 篇。
ICLR 2026 有多少篇对齐与安全论文？
ICLR 2026 有 623 篇对齐与安全论文，上一届是 473 篇。
ICML 2026 有多少篇对齐与安全论文？
ICML 2026 有 551 篇对齐与安全论文，上一届是 300 篇。
CVPR 2026 有多少篇对齐与安全论文？
CVPR 2026 有 228 篇对齐与安全论文，上一届是 135 篇。
哪个会议的对齐与安全论文最多？
NeurIPS，2023–2026 年共 1,421 篇。

「对齐与负责任 AI」下的其他主题