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 30 / 38 页

  1. NeurIPS 2026Attention-Based Soft Answer Sets
  2. NeurIPS 2026Auditing Sabotage Bench: A Benchmark for Detecting and Fixing Research Sabotage in ML Codebases
  3. NeurIPS 2026Auto-Rubric as Reward: From Implicit Preference to Explicit Generative Criteria
  4. NeurIPS 2026AutoHoney: Automating, Deploying, and Evaluating Scheming Honeypots Across Production Codebases
  5. NeurIPS 2026B$^3$-PWL: GPU-Batched Branch-and-Bound for Piecewise-Linear Optimization with SOS2 Constraints
  6. NeurIPS 2026BarrierSteer: LLM Safety via Learning Barrier Steering
  7. NeurIPS 2026Bayesian Optimization with Fisher Information Geometry: Gradient Bounds and Trust-Region Methods
  8. NeurIPS 2026BDC-Merge: Cross-Architecture Model Merging via Dependency Alignment
  9. NeurIPS 2026BECON: Belief-Conditioned Constrained Multi-Objective Reinforcement Learning under Drifting Preferences and Budgets
  10. NeurIPS 2026Benchmarking and Improving Monitors for Out-Of-Distribution Alignment Failure in LLMs
  11. NeurIPS 2026Bentkus-type asymptotic e-values
  12. NeurIPS 2026Beyond Isotropy in JEPAs: Hamiltonian Geometry and Symplectic Prediction
  13. NeurIPS 2026Beyond Pairwise Supervision: Spectral Characteristic Matching for Data-Efficient Multimodal Alignment
  14. NeurIPS 2026Beyond Row Alignment: Virtual-Camera-Aware Online Stereo Rectification
  15. NeurIPS 2026Beyond Semantic Alignment: Geometric Incomparability in Multi-Oracle Soft Fusion
  16. NeurIPS 2026Bias-Variance Optimized Preference Optimization for Large Reasoning Models
  17. NeurIPS 2026Bias, Measurement Error, and Double-Dipping: When Can GNN Convolutions Help Brain Connectome Prediction?
  18. NeurIPS 2026Bounding Global and Local Compression Error of Signal Parameterizations
  19. NeurIPS 2026Bridging Compute- and Data-Optimal Pretraining
  20. NeurIPS 2026Bridging Simulation and Reality: Geometry and Decision Alignment for Autonomous Driving
  21. NeurIPS 2026Bridging the Simulation-to-Experiment Gap with Adversarial Distribution Alignment
  22. NeurIPS 2026BSO: Safety Alignment Is Density Ratio Matching
  23. NeurIPS 2026Bug or Feature$^2$: Weight Drift, Activation Sparsity, and Spikes
  24. NeurIPS 2026ByteDistill: Cross-Tokenizer Distillation via Chunk-wise Byte-Level Distribution Alignment
  25. NeurIPS 2026CapTrack: Multifaceted Evaluation of Forgetting in LLM Post-Training
  26. NeurIPS 2026Cauchy Scientific Networks: Loss–Architecture Alignment and Its Limit
  27. NeurIPS 2026Causal pieces: analysing and improving spiking neural networks piece by piece
  28. NeurIPS 2026Certifiably Optimal Robust Angular Synchronization
  29. NeurIPS 2026Certification from Examples is Hard for Circuits and Transformers under Minimal Overparametrization
  30. NeurIPS 2026Chaining 2-FWL GNNs for Combinatorial Graph Alignment
  31. NeurIPS 2026CHoRD: Coordinating Scheduling and Data Placement for Efficient Deep Neural Network Inference on Chiplet-Based GPUs
  32. NeurIPS 2026Classification at the Edge of Stability: Unifying Self-Stabilization and Convergence Rates
  33. NeurIPS 2026Clearer Sight, Fewer Lies: Oriented Pickup Preference Optimization for Multimodal Hallucination Mitigation
  34. NeurIPS 2026ClinStab: Stability-Oriented Learning for Medical Time Series via Dual-Stream Alignment
  35. NeurIPS 2026Closed-Loop Alignment: Socially Coupled In-Context Learning and Relational Posterior Collapse
  36. NeurIPS 2026Coding with "Enemy": Can Human Developers Detect AI Agent Sabotage?
  37. NeurIPS 2026Community-Centered AI is Feasible and Beneficial for Impacted Communities
  38. NeurIPS 2026Commutator Memory: Sparse, Path-Local Reading and Steering in Language Models
  39. NeurIPS 2026Comparing Transformers and Hybrid Models at the Token Level
  40. NeurIPS 2026Complex Schrödinger Bridges
  41. NeurIPS 2026COMPOSE: Composing Future Theorems from Citations and Formal Structure
  42. NeurIPS 2026Computing Thiele Rules on Interval Elections and their Generalizations
  43. NeurIPS 2026Concept frustration: Aligning human concepts and machine representations
  44. NeurIPS 2026Conditional Counterfactual Mean Embeddings: Doubly Robust Estimation and Learning Rates
  45. NeurIPS 2026Conservative neural posterior estimation via distributionally robust training
  46. NeurIPS 2026Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves
  47. NeurIPS 2026Constraint-Aware Influence Estimation in Deep Constrained Learning
  48. NeurIPS 2026Correct, Route, Calibrate: Efficient Preference Optimization from Noisy, Heterogeneous Human Feedback
  49. NeurIPS 2026Counterfactual Estimation under Composite Treatments via Progressive Distribution Alignment
  50. NeurIPS 2026Counterfactual Maps: What They Are and How to Find Them

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