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 1,363 篇，第 11 / 28 页

  1. ICLR 2025Robust System Identification: Finite-sample Guarantees and Connection to Regularization
  2. ICLR 2025Robustness Inspired Graph Backdoor Defense
  3. ICLR 2025Round and Round We Go! What makes Rotary Positional Encodings useful?
  4. ICLR 2025RouteLLM: Learning to Route LLMs from Preference Data
  5. ICLR 2025RRM: Robust Reward Model Training Mitigates Reward Hacking
  6. ICLR 2025SafeDiffuser: Safe Planning with Diffusion Probabilistic Models
  7. ICLR 2025Safety Alignment Should be Made More Than Just a Few Tokens Deep
  8. ICLR 2025Safety Layers in Aligned Large Language Models: The Key to LLM Security
  9. ICLR 2025SafeWatch: An Efficient Safety-Policy Following Video Guardrail Model with Transparent Explanations
  10. ICLR 2025Sail into the Headwind: Alignment via Robust Rewards and Dynamic Labels against Reward Hacking
  11. ICLR 2025SaLoRA: Safety-Alignment Preserved Low-Rank Adaptation
  12. ICLR 2025Scalable Extraction of Training Data from Aligned, Production Language Models
  13. ICLR 2025Scaling up the Banded Matrix Factorization Mechanism for Large Scale Differentially Private ML
  14. ICLR 2025SEAL: Safety-enhanced Aligned LLM Fine-tuning via Bilevel Data Selection
  15. ICLR 2025SEBRA : Debiasing through Self-Guided Bias Ranking
  16. ICLR 2025SecureGS: Boosting the Security and Fidelity of 3D Gaussian Splatting Steganography
  17. ICLR 2025Seeing Eye to AI: Human Alignment via Gaze-Based Response Rewards for Large Language Models
  18. ICLR 2025Selective Unlearning via Representation Erasure Using Domain Adversarial Training
  19. ICLR 2025Self-Boosting Large Language Models with Synthetic Preference Data
  20. ICLR 2025SELF-EVOLVED REWARD LEARNING FOR LLMS
  21. ICLR 2025Self-Improvement in Language Models: The Sharpening Mechanism
  22. ICLR 2025Self-Improving Robust Preference Optimization
  23. ICLR 2025Self-Play Preference Optimization for Language Model Alignment
  24. ICLR 2025Self-play with Execution Feedback: Improving Instruction-following Capabilities of Large Language Models
  25. ICLR 2025Sensitivity Verification for Additive Decision Tree Ensembles
  26. ICLR 2025SeRA: Self-Reviewing and Alignment of LLMs using Implicit Reward Margins
  27. ICLR 2025Severing Spurious Correlations with Data Pruning
  28. ICLR 2025Shape as Line Segments: Accurate and Flexible Implicit Surface Representation
  29. ICLR 2025Shh, don't say that! Domain Certification in LLMs
  30. ICLR 2025SimPER: A Minimalist Approach to Preference Alignment without Hyperparameters
  31. ICLR 2025Simple yet Effective Incomplete Multi-view Clustering: Similarity-level Imputation and Intra-view Hybrid-group Prototype Construction
  32. ICLR 2025SimulPL: Aligning Human Preferences in Simultaneous Machine Translation
  33. ICLR 2025SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal
  34. ICLR 2025Sparse components distinguish visual pathways & their alignment to neural networks
  35. ICLR 2025Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment
  36. ICLR 2025Spurious Forgetting in Continual Learning of Language Models
  37. ICLR 2025Stealthy Shield Defense: A Conditional Mutual Information-Based Approach against Black-Box Model Inversion Attacks
  38. ICLR 2025Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction
  39. ICLR 2025Strategic Classification With Externalities
  40. ICLR 2025Strong Preferences Affect the Robustness of Preference Models and Value Alignment
  41. ICLR 2025Style Outweighs Substance: Failure Modes of LLM Judges in Alignment Benchmarking
  42. ICLR 2025Super(ficial)-alignment: Strong Models May Deceive Weak Models in Weak-to-Strong Generalization
  43. ICLR 2025SuperCorrect: Advancing Small LLM Reasoning with Thought Template Distillation and Self-Correction
  44. ICLR 2025Support is All You Need for Certified VAE Training
  45. ICLR 2025Surgical, Cheap, and Flexible: Mitigating False Refusal in Language Models via Single Vector Ablation
  46. ICLR 2025Tailoring Mixup to Data for Calibration
  47. ICLR 2025Taming Overconfidence in LLMs: Reward Calibration in RLHF
  48. ICLR 2025Tamper-Resistant Safeguards for Open-Weight LLMs
  49. ICLR 2025TASAR: Transfer-based Attack on Skeletal Action Recognition
  50. ICLR 2025Test-time Adaptation for Regression by Subspace Alignment

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