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 17 / 92 页

801. ICLR 2026Towards Understanding Valuable Preference Data for Large Language Model Alignment
802. ICLR 2026Traceable Black-Box Watermarks For Federated Learning
803. ICLR 2026Transferable and Stealthy Adversarial Attacks on Large Vision-Language Models
804. ICLR 2026Translate Policy to Language: Flow Matching Generated Rewards for LLM Explanations
805. ICLR 2026Tree-based Dialogue Reinforced Policy Optimization for Red-Teaming Attacks
806. ICLR 2026TriQDef: Disrupting Semantic and Gradient Alignment to Prevent Adversarial Patch Transferability in Quantized Neural Networks
807. ICLR 2026Trust The Typical
808. ICLR 2026Truthful or Fabricated? Using Causal Attribution to Mitigate Reward Hacking in Explanations
809. ICLR 2026Truthfulness Despite Weak Supervision: Evaluating and Training LLMs Using Peer Prediction
810. ICLR 2026TurboBoA: Faster and Exact Attention-aware Quantization without Backpropagation
811. ICLR 2026Two-Way Is Better Than One: Bidirectional Alignment with Cycle Consistency for Exemplar-Free Class-Incremental Learning
812. ICLR 2026ULD-Net: Enabling Ultra-Low-Degree Fully Polynomial Networks for Homomorphically Encrypted Inference
813. ICLR 2026Understanding and Improving Continuous LLM Adversarial Training via In-context Learning Theory
814. ICLR 2026Understanding Sensitivity of Differential Attention through the Lens of Adversarial Robustness
815. ICLR 2026Uni-DPO: A Unified Paradigm for Dynamic Preference Optimization of LLMs
816. ICLR 2026UniCon: Unified Framework for Efficient Contrastive Alignment via Kernels
817. ICLR 2026Unified Privacy Guarantees for Decentralized Learning via Matrix Factorization
818. ICLR 2026Unifying Stable Optimization and Reference Regularization in RLHF
819. ICLR 2026Unlearning Isn't Invisible: Detecting Unlearning Traces in LLMs from Model Outputs
820. ICLR 2026Using cognitive models to reveal value trade-offs in language models
821. ICLR 2026Value Gradient Flow: Behavior-Regularized RL without Regularization
822. ICLR 2026VeriCoT: Neuro-symbolic Chain-of-Thought Validation via Logical Consistency Checks
823. ICLR 2026Verification and Co-Alignment via Heterogeneous Consistency for Preference-Aligned LLM Annotations
824. ICLR 2026VERIFY: A Novel Multi-Domain Dataset Grounding LTL in Contextual Natural Language via Provable Intermediate Logic
825. ICLR 2026ViPO: Visual Preference Optimization at Scale
826. ICLR 2026VisionTrim: Unified Vision Token Compression for Training-Free MLLM Acceleration
827. ICLR 2026VisJudge-Bench: Aesthetics and Quality Assessment of Visualizations
828. ICLR 2026VLSU: Mapping the Limits of Joint Multimodal Understanding for AI Safety
829. ICLR 2026WARP: Weight Teleportation for Attack-Resilient Unlearning Protocols
830. ICLR 2026Watch the Weights: Unsupervised monitoring and control of fine-tuned LLMs
831. ICLR 2026Watch your steps: Dormant Adversarial Behaviors that Activate upon LLM Finetuning
832. ICLR 2026WaterDrum: Watermark-based Data-centric Unlearning Metric
833. ICLR 2026Watermark-based Attribution of AI-Generated Content
834. ICLR 2026Weak-to-Strong Generalization with Failure Trajectories
835. ICLR 2026What Do Large Language Models Know About Opinions?
836. ICLR 2026What happens when generative AI models train recursively on each others' outputs?
837. ICLR 2026What matters for Representation Alignment: Global Information or Spatial Structure?
838. ICLR 2026What's In My Human Feedback? Learning Interpretable Descriptions of Preference Data
839. ICLR 2026What's the plan? Metrics for implicit planning in LLMs and their application to rhyme generation
840. ICLR 2026When Data is the Algorithm: A Systematic Study and Curation of Preference Optimization Datasets
841. ICLR 2026When Language Models Lose Their Mind: The Consequences of Brain Misalignment
842. ICLR 2026When Machine Learning Gets Personal: Evaluating Prediction and Explanation
843. ICLR 2026When Priors Backfire: On the Vulnerability of Unlearnable Examples to Pretraining
844. ICLR 2026When Style Breaks Safety: Defending LLMs Against Superficial Style Alignment
845. ICLR 2026When Weak LLMs Speak with Confidence, Preference Alignment Gets Stronger
846. ICLR 2026Where Did It Go Wrong? Attributing Undesirable LLM Behaviors via Representation Gradient Tracing
847. ICLR 2026Why DPO is a Misspecified Estimator and How to Fix It
848. ICLR 2026Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas
849. ICLR 2026Winter Soldier: Backdooring Language Models at Pre-Training with Indirect Data Poisoning
850. ICLR 2026Your Agent May Misevolve: Emergent Risks in Self-evolving LLM Agents

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。