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 1,421 篇，第 4 / 29 页

  1. NeurIPS 2026Fast and slow gradient descent dynamics of logistic regression through weak alignment
  2. NeurIPS 2026Fast Approximate $\ell_p$ Chamfer Distance via Lopsided Embeddings and Structured JL
  3. NeurIPS 2026Fault Tolerance in Transformers Favors Output Alignment over Hidden-State Consistency
  4. NeurIPS 2026Fed-AGA: An Anchor Graph Alignment Framework for Federated Unaligned Multi-view Clustering
  5. NeurIPS 2026Federate the Router: Learning Language Model Routers with Sparse and Decentralized Evaluations
  6. NeurIPS 2026Feeling of Knowing in Large Language Models
  7. NeurIPS 2026Few-Shot Truly Benign DPO Attack for Jailbreaking LLMs
  8. NeurIPS 2026Fill the GAP: A Granular Alignment Paradigm for Visual Reasoning in Multimodal Large Language Models
  9. NeurIPS 2026Fine-tuning Does Not Reach All: Uneven Safety and Knowledge Dynamics in Language Models
  10. NeurIPS 2026FiTS: Interpretable Spiking Neurons via Frequency Selectivity and Temporal Shaping
  11. NeurIPS 2026FLIP: Fast and Accurate Global Lipschitz Estimation for Large Feedforward Networks
  12. NeurIPS 2026FOAM: Factored One-sided Adam-Moment for Practical and Scalable SOAP
  13. NeurIPS 2026FocusBranch: Combinatorial Branch-and-Bound for $\ell_0$ Neural Network Robustness Verification
  14. NeurIPS 2026Forest-Guided Semantic Transport for Label-Supervised Manifold Alignment
  15. NeurIPS 2026From Density Matrices to Phase Transitions in Deep Learning: Spectral Early Warnings and Interpretability
  16. NeurIPS 2026From Inexact Gradients to Byzantine Robustness: Acceleration and Optimization under Similarity
  17. NeurIPS 2026From Isolated feature to Orbits:\\Discovering Music Concepts via Multi-SAE Alignment
  18. NeurIPS 2026From Solver Trajectories to Teaching Trajectories: Cognition-Aligned Reasoning Distillation
  19. NeurIPS 2026From Squeezing to Grounding: Visual Guided DPO for Multimodal Hallucination Mitigation
  20. NeurIPS 2026Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention
  21. NeurIPS 2026GCD: Correcting Hidden-State Bias in Off-Policy Agentic RL
  22. NeurIPS 2026GEAR-Align: Grounding-Evidence-Aware Gradient Routing for Multimodal Alignment
  23. NeurIPS 2026GEM: Interpretable Language Models via Geometric Embedding Alignment
  24. NeurIPS 2026Generalized Adaptive Boosting and the Geometry of Mistakes
  25. NeurIPS 2026Geometric Alignment without Functional Equivalence: A Layer-wise Analysis of the Speech-Text Modality Gap
  26. NeurIPS 2026Geometry-Aware Directional Alignment for Coherent Model Merging
  27. NeurIPS 2026Geometry-Aware Optimal Transport: Fast Intrinsic Dimension and Wasserstein Distance Estimation
  28. NeurIPS 2026Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs
  29. NeurIPS 2026Gradient Descent on Two ReLU Neurons: Global Landscape and Bifurcation Dynamics
  30. NeurIPS 2026Gradient Descent’s Last Iterate is Often (slightly) Suboptimal
  31. NeurIPS 2026GradShield: Alignment Preserving Finetuning
  32. NeurIPS 2026GradTrack: Detecting Noisy Labels via Temporal Trajectories of Class-wise Gradient Misalignment
  33. NeurIPS 2026Graph Label Alignment: A Diagnostic Atlas for Graph Classification
  34. NeurIPS 2026Graph Learning Should Move Beyond Restrictive Views of Spectral and Message-Passing GNNs
  35. NeurIPS 2026Greedy Alignment Principle for Optimizer Selection
  36. NeurIPS 2026GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory
  37. NeurIPS 2026Guaranteed Jailbreaking Defense via Disrupt-and-Rectify Smoothing
  38. NeurIPS 2026Hallucination-Guided Unlearning: Using Hallucination Traces to Reveal Overfitted Memories
  39. NeurIPS 2026HARC: Coupling Harmfulness and Refusal Capabilities for Robust Safety Alignment
  40. NeurIPS 2026Harnessing Textual Refusal Directions for Multimodal Safety
  41. NeurIPS 2026HelpBench: Assessing the Ability of LLMs to Provide Privacy, Safety, and Security Advice
  42. NeurIPS 2026Hierarchical Cross-Class Part Alignment for Prototypical Part Networks via Hyperbolic Entailment
  43. NeurIPS 2026Hierarchical Graph Alignment for Cross-Modal 3D Scene Grounding
  44. NeurIPS 2026HoloGene: Learning to Lift Sliced Spatial Transcriptomics to Holistic 3D Gene Fields
  45. NeurIPS 2026Hyper Input Convex Neural Networks for Shape Constrained Learning and Optimal Transport
  46. NeurIPS 2026Hyperagents
  47. NeurIPS 2026Identification and Bounding of Joint Expectation over Potential Outcomes
  48. NeurIPS 2026Improved State Mixing in Higher-order and Block Diagonal Linear Recurrent Networks
  49. NeurIPS 2026Improving Audit Realism with Inference-Time Compute and Deployment Scaffolds
  50. NeurIPS 2026Indications of Belief-Guided Agency and Meta-Cognitive Monitoring in Large Language Models

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