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 551 篇，第 5 / 12 页

  1. ICML 2026FUSE: Frequency-domain Unification and Spectral Energy Alignment for Multi-modal Object Re-Identification
  2. ICML 2026GEM: Geometric Erasure by Contrastive Velocity Matching in Rectified Flows
  3. ICML 2026Generalizable and Composable Multi-Model Embedding Translation
  4. ICML 2026GeoAlign: Geometric Rollout Curation for Robust LLM Reinforcement Learning
  5. ICML 2026GeoDM: Geometry-aware Distribution Matching for Dataset Distillation
  6. ICML 2026Geometric Conformal Prediction with Spatial Ranks and Multivariate Quantiles
  7. ICML 2026Geometry of Reason: Spectral Signatures of Valid Mathematical Reasoning
  8. ICML 2026Geometry-Preserving Unsupervised Alignment for Heterogeneous Foundation Models
  9. ICML 2026GIST: Targeted Data Selection for Instruction Tuning via Coupled Optimization Geometry
  10. ICML 2026GLAD: Bidirectional Structure-Attribute Alignment via Latent Graph Diffusion Models
  11. ICML 2026Gradient Flow Dynamics and Implicit Bias of Diagonal Linear Networks under Infinitesimal Initialization
  12. ICML 2026Gradient Regularization Mitigates Reward Hacking in Reinforcement Learning from Human Feedback and Verifiable Rewards
  13. ICML 2026Gradient Testing and Estimation by Comparisons
  14. ICML 2026Graph Alignment for Benchmarking Graph Neural Networks and Learning Positional Encodings
  15. ICML 2026Graph Alignment via Dual-Pass Spectral Encoding and Latent Space Communication
  16. ICML 2026Graph-Preference Learning: Debiasing Network-Sampled Human Feedback for Target Welfare Estimation
  17. ICML 2026Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance
  18. ICML 2026Gromov-Wasserstein at Scale, Beyond Squared Norms
  19. ICML 2026Guaranteed Optimal Compositional Explanations for Neurons
  20. ICML 2026Hair-Trigger Alignment: Black-Box Evaluation Cannot Guarantee Post-Update Alignment
  21. ICML 2026Helpful to a Fault: Measuring Illicit Assistance in Multi-Turn, Multilingual LLM Agents
  22. ICML 2026Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio
  23. ICML 2026Hierarchical Image Tokenization for Multi-Scale Image Super Resolution
  24. ICML 2026High-Fidelity ANN-to-SNN Conversion via Closed-Loop CKA Distillation
  25. ICML 2026HO-SFL: Hybrid-Order Split Federated Learning with Backprop-Free Clients and Dimension-Free Aggregation
  26. ICML 2026How High is ‘High’? Rethinking the Roles of Dimensionality in Topological Data Analysis and Manifold Learning
  27. ICML 2026How Out-of-Distribution Detection Learning Theory Enhances Transformer: Learnability and Reliability
  28. ICML 2026How RLHF Amplifies Sycophancy
  29. ICML 2026How to Avoid Debate: Scalable AI Safety via Doubly-Efficient Interactive Proofs
  30. ICML 2026HumanLM: Simulating Users with State Alignment Beats Response Imitation
  31. ICML 2026IBMA: Information Bottleneck-Based Multimodal Alignment
  32. ICML 2026Identifying and Mitigating Errors in Gradient Aggregation of Distributed Data Parallel Training
  33. ICML 2026Implicit Preference Alignment for Human Image Animation
  34. ICML 2026Implicit Safety Alignment from Crowd Preferences
  35. ICML 2026Improved Bounds for Private and Robust Alignment
  36. ICML 2026In-Context Universal Approximation, Compositional Generalization, and Algorithm Emulation
  37. ICML 2026In-Training Defenses Against Emergent Misalignment in Language Models
  38. ICML 2026Incorporating Importance Weighting in Optimal Transport Based Domain Alignment
  39. ICML 2026Inference-Aware Meta-Alignment of LLMs via Non-Linear GRPO
  40. ICML 2026Inference-time Alignment with Rewards in Besov Spaces: Provable Advantages of Feature Learning and Multi-Step Policy Updates
  41. ICML 2026Instance-Specific Approximation Ratios for Correlation Clustering and Max-Cut
  42. ICML 2026Introspection Adapters: Training LLMs to Report Their Learned Behaviors
  43. ICML 2026IRIS: Implicit Reward-Guided Internal Sifting for Mitigating Multimodal Hallucination
  44. ICML 2026IRPM: Intergroup Relative Preference Modeling for Pointwise Generative Reward Models
  45. ICML 2026It's a TRAP! Task-Redirecting Agent Persuasion Benchmark for Web Agents
  46. ICML 2026ITSPACE: Monotone Gaussian Optimal Transport Updates
  47. ICML 2026IVQ: Structured and Lightweight Vector Quantization via Binary Hierarchical Composition Inspired by $\textit{IChing}$
  48. ICML 2026Jailbreaking Vision-Language Models Through the Visual Modality
  49. ICML 2026Language Bias in LVLMs: From In-Depth Analysis to Simple and Effective Mitigation
  50. ICML 2026Large-Scale Notification Dispatch with Bundle Treatments and Multi-Outcome Uplift 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