对齐与安全论文趋势 2023–2026

主要结论

• 对齐与安全论文从 2023 年的 438 篇增长到 2026 年的 1,882 篇，三年间扩大到约 4.3 倍。
• NeurIPS 是发对齐与安全论文最多的会议，2023–2026 年共 1,421 篇，占四个会议总量的 31%。
• 交叉方向里涨得最快的是社会层面 · safety：同时属于两个主题的论文从 2023 年的 0 篇增至 2026 年的 137 篇。
• 在「对齐与负责任 AI」的 5 个主题里，它的 2026 年论文数排第 1；同组增长最快的社会层面 · safety是 93 倍，对齐与安全是 4.3 倍。

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

• NeurIPS 2026$\gamma$-weakly $\theta$-up-concavity: A Unified Framework for Non-Convex Optimization Beyond DR-Submodular and OSS Functions
• ICML 2026AICrypto: Evaluating Cryptography Capabilities of Large Language Models
• ICLR 2026A Fair Bayesian Inference through Matched Gibbs Posterior
• CVPR 2026Advancing Cancer Prognosis with Hierarchical Fusion of Genomic Proteomic and
• NeurIPS 2025$\Psi$-Sampler: Initial Particle Sampling for SMC-Based Inference-Time Reward Alignment in Score Models
• ICML 2025A Checks-and-Balances Framework for Context-Aware Ethical AI Alignment
• ICLR 2025$\sigma$-zero: Gradient-based Optimization of $\ell_0$-norm Adversarial Examples
• CVPR 2025AesthetiQ: Enhancing Graphic Layout Design via Aesthetic-Aware Preference Alignment of Multi-modal Large Language Models
• NeurIPS 2024A survey and benchmark of high-dimensional Bayesian optimization of discrete sequences
• ICML 2024A Global Geometric Analysis of Maximal Coding Rate Reduction
• ICLR 2024A unique M-pattern for micro-expression spotting in long videos
• CVPR 2024Decompose-and-Compose: A Compositional Approach to Mitigating Spurious Correlation
• NeurIPS 2023(S)GD over Diagonal Linear Networks: Implicit bias, Large Stepsizes and Edge of Stability
• ICML 2023A Robust Optimisation Perspective on Counterexample-Guided Repair of Neural Networks
• ICLR 2023Aligning Model and Macaque Inferior Temporal Cortex Representations Improves Model-to-Human Behavioral Alignment and Adversarial Robustness
• CVPR 2023Adjustment and Alignment for Unbiased Open Set Domain Adaptation
• NeurIPS 2026$\mathcal{P}$Torch: Narrowing the Gap Between Projection and Gradient-Based Learning
• ICML 2026Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment
• ICLR 2026A2D: Any-Order, Any-Step Safety Alignment for Diffusion Language Models
• CVPR 2026AlignPose Generalizable 6D Pose Estimation via Multi-view Feature-metric Alignment

全部对齐与安全论文

共 4,558 篇，第 51 / 92 页

2501. ICML 2025Active Reward Modeling: Adaptive Preference Labeling for Large Language Model Alignment
2502. ICML 2025Adapting While Learning: Grounding LLMs for Scientific Problems with Tool Usage Adaptation
2503. ICML 2025Adaptive Estimation and Learning under Temporal Distribution Shift
2504. ICML 2025Addressing Concept Mislabeling in Concept Bottleneck Models Through Preference Optimization
2505. ICML 2025ADHMR: Aligning Diffusion-based Human Mesh Recovery via Direct Preference Optimization
2506. ICML 2025AdvAgent: Controllable Blackbox Red-teaming on Web Agents
2507. ICML 2025Adversaries Can Misuse Combinations of Safe Models
2508. ICML 2025AdvPrompter: Fast Adaptive Adversarial Prompting for LLMs
2509. ICML 2025AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation
2510. ICML 2025Aligning Multimodal Representations through an Information Bottleneck
2511. ICML 2025Aligning Spoken Dialogue Models from User Interactions
2512. ICML 2025Aligning with Logic: Measuring, Evaluating and Improving Logical Preference Consistency in Large Language Models
2513. ICML 2025AlphaDPO: Adaptive Reward Margin for Direct Preference Optimization
2514. ICML 2025AlphaPO: Reward Shape Matters for LLM Alignment
2515. ICML 2025AMPO: Active Multi Preference Optimization for Self-play Preference Selection
2516. ICML 2025Antidote: Post-fine-tuning Safety Alignment for Large Language Models against Harmful Fine-tuning Attack
2517. ICML 2025Assessing Safety Risks and Quantization-aware Safety Patching for Quantized Large Language Models
2518. ICML 2025Automated Red Teaming with GOAT: the Generative Offensive Agent Tester
2519. ICML 2025Avoiding Catastrophe in Online Learning by Asking for Help
2520. ICML 2025Bayesian Optimization from Human Feedback: Near-Optimal Regret Bounds
2521. ICML 2025BDC-CLIP: Brownian Distance Covariance for Adapting CLIP to Action Recognition
2522. ICML 2025Benign Samples Matter! Fine-tuning On Outlier Benign Samples Severely Breaks Safety
2523. ICML 2025Beyond Bradley-Terry Models: A General Preference Model for Language Model Alignment
2524. ICML 2025BILBO: BILevel Bayesian Optimization
2525. ICML 2025Boosting Adversarial Robustness with CLAT: Criticality Leveraged Adversarial Training
2526. ICML 2025BOPO: Neural Combinatorial Optimization via Best-anchored and Objective-guided Preference Optimization
2527. ICML 2025Bounded Rationality for LLMs: Satisficing Alignment at Inference-Time
2528. ICML 2025BounDr.E: Predicting Drug-likeness via Biomedical Knowledge Alignment and EM-like One-Class Boundary Optimization
2529. ICML 2025Breaking Barriers: Combinatorial Algorithms for Non-Monotone Submodular Maximization with Sublinear Adaptivity and $1/e$ Approximation
2530. ICML 2025Calibrated Language Models and How to Find Them with Label Smoothing
2531. ICML 2025Calibrating Video Watch-time Predictions with Credible Prototype Alignment
2532. ICML 2025Can RLHF be More Efficient with Imperfect Reward Models? A Policy Coverage Perspective
2533. ICML 2025Cannot See the Forest for the Trees: Invoking Heuristics and Biases to Elicit Irrational Choices of LLMs
2534. ICML 2025CASE-Bench: Context-Aware SafEty Benchmark for Large Language Models
2535. ICML 2025Clipped SGD Algorithms for Performative Prediction: Tight Bounds for Stochastic Bias and Remedies
2536. ICML 2025Clipping Improves Adam-Norm and AdaGrad-Norm when the Noise Is Heavy-Tailed
2537. ICML 2025Clone-Robust AI Alignment
2538. ICML 2025CodeSteer: Symbolic-Augmented Language Models via Code/Text Guidance
2539. ICML 2025Comparing Comparisons: Informative and Easy Human Feedback with Distinguishability Queries
2540. ICML 2025Comparing Few to Rank Many: Active Human Preference Learning Using Randomized Frank-Wolfe Method
2541. ICML 2025Compelling ReLU Networks to Exhibit Exponentially Many Linear Regions at Initialization and During Training
2542. ICML 2025Compressing tree ensembles through Level-wise Optimization and Pruning
2543. ICML 2025Computing Optimal Transport Maps and Wasserstein Barycenters Using Conditional Normalizing Flows
2544. ICML 2025Conformal Tail Risk Control for Large Language Model Alignment
2545. ICML 2025ConfPO: Exploiting Policy Model Confidence for Critical Token Selection in Preference Optimization
2546. ICML 2025Constrain Alignment with Sparse Autoencoders
2547. ICML 2025Contextual Optimization Under Model Misspecification: A Tractable and Generalizable Approach
2548. ICML 2025Counting in Small Transformers: The Delicate Interplay between Attention and Feed-Forward Layers
2549. ICML 2025Cross-City Latent Space Alignment for Consistency Region Embedding
2550. ICML 2025CSV-Occ: Fusing Multi-frame Alignment for Occupancy Prediction with Temporal Cross State Space Model and Central Voting Mechanism

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。