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 1,882 篇，第 25 / 38 页

  1. ICML 2026Per-example Gradients: a New Frontier for Understanding and Improving Optimizers
  2. ICML 2026Perceptrons and Localization of Attention’s Mean-Field Landscape
  3. ICML 2026Personalized Image Generation via Human-in-the-loop Bayesian Optimization
  4. ICML 2026PICACO: Pluralistic In-Context Value Alignment via Total Correlation Optimization
  5. ICML 2026PLaID++: A Preference Aligned Language Model for Targeted Inorganic Materials Design
  6. ICML 2026PLANTAIN: Plan-Answer Interleaved Reasoning
  7. ICML 2026Plug-and-Play Spiking Operators: Breaking the Nonlinearity Bottleneck in Spiking Transformers
  8. ICML 2026PoMtVRS: Preference-Optimized Multi-Task Vehicle Routing Solver with Preference Gating
  9. ICML 2026Positive–Unlabeled Reinforcement Learning Distillation for On-Premise Small Models
  10. ICML 2026Post-Training Language Models for Crosslingual Consistency
  11. ICML 2026Post-Training with Policy Gradients: Optimality and the Base Model Barrier
  12. ICML 2026Powerful and Theoretically Guaranteed Independence Testing on Heterogeneous Federated Clients
  13. ICML 2026Preference Goal Tuning: Post-Training as Latent Control for Frozen Policies
  14. ICML 2026Preference-based Antibody Expression Ranking: Scaling with Large-scale Weak Supervision
  15. ICML 2026Preference-Calibrated Optimization with Score-Level Distribution Alignment for Text-to-Image Diffusion Model Unlearning
  16. ICML 2026Preference-Enhanced Reinforcement Learning for Pluralistic Image Inpainting
  17. ICML 2026Preference-Modulated Structural Attention for Multi-Objective Combinatorial Optimization
  18. ICML 2026Pressure Reveals Character: Behavioural Alignment Evaluation at Depth
  19. ICML 2026Probabilistic Modeling of Latent Agentic Substructures in Deep Neural Networks
  20. ICML 2026ProMeCD: Unifying Long-Tailed and Noisy Label Learning via White-Box Control
  21. ICML 2026Prototype-guided Bilateral Alignment Multimodal Federated Learning
  22. ICML 2026Pruning at Initialisation through the lens of Graphon Limit: Convergence, Expressivity, and Generalisation
  23. ICML 2026PS-PPO : Prefix-Sampling PPO for Critic-Free RLHF
  24. ICML 2026QiMeng-PerceptOS: Semantic-Aware Kernel Optimization for OS-Intensive Workloads via Hardware-Software Alignment
  25. ICML 2026Quadratically Regularized Optimal Transport: Localization Bounds and Affine Case Analysis
  26. ICML 2026Quantifying the Salience of Geo-Cultural Values for Pluralistic Safety Alignment
  27. ICML 2026Rare Event Analysis of Large Language Models
  28. ICML 2026RBCBF: Decoding Time Safety Alignment via Risk Guided Rollback and Barrier Control
  29. ICML 2026Real-Time Aligned Reward Model beyond Semantics
  30. ICML 2026REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
  31. ICML 2026REAR: Test-time Preference Realignment through Reward Decomposition
  32. ICML 2026Reasoning Models Struggle to Control their Chains of Thought
  33. ICML 2026Reasoning over Boundaries: Enhancing Specification Alignment via Test-time Deliberation
  34. ICML 2026Recontextualization Mitigates Specification Gaming Without Modifying the Specification
  35. ICML 2026RED-HDP-HMM: Observation-Dependent Durations for Bayesian Nonparametric Sequential Models
  36. ICML 2026RedDebate: Safer Responses Through Multi-Agent Red Teaming Debates
  37. ICML 2026Regularized Discriminative Alignment for Deep Representations under Label Shift
  38. ICML 2026Reinforcement Learning with Pairwise Preferences in Long-Term Decision Problems
  39. ICML 2026Reliability-Aware LLM Alignment from Inconsistent Human Feedback
  40. ICML 2026Reliable Confidence Alignment for Generalized Category Discovery
  41. ICML 2026Reliable Neighborhood-Aware Multi-View Outlier Detection
  42. ICML 2026Removing Sandbagging in LLMs by Training with Weak Supervision
  43. ICML 2026Resilient Coresets and Clustering
  44. ICML 2026Resolution as a Direction: Vector-Panning Feature Alignment for Cross-Resolution Re-Identification
  45. ICML 2026Rethinking Feature Alignment in Generalist Graph Anomaly Detection: A Relational Fingerprint-based Approach
  46. ICML 2026Return-to-Go Is More Than a Number: Q-Guided Alignment for Return-Conditioned Supervised Learning
  47. ICML 2026Revisiting Padded Transformer Expressivity: Which Architectural Choices Matter and Which Don't
  48. ICML 2026Revisiting Robustness for LLM Safety Alignment via Selective Geometry Control
  49. ICML 2026Reward Auditor: Inference on Reward Modeling Suitability in Real-World Perturbed Scenarios
  50. ICML 2026Reward Shaping for (Inference-Time) Alignment: A Stackelberg Game Perspective

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