社会层面 · safety论文趋势 2023–2026

主要结论

• 社会层面 · safety论文从 2023 年的 4 篇增长到 2026 年的 370 篇，三年间扩大到约 93 倍。
• ICML 是发社会层面 · safety论文最多的会议，2023–2026 年共 364 篇，占四个会议总量的 53%。
• 交叉方向里涨得最快的是对齐与安全：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 3；同组增长最快的社会层面 · privacy是 6.6 倍，社会层面 · safety是 93 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 2026BarrierSteer: LLM Safety via Learning Barrier Steering
• ICML 2026Anti-Backdoor Coreset Selection via Cumulative Entropy
• ICLR 2026Bi-Criteria Metric Distortion
• CVPR 2026Omni-MMSI Toward Identity-attributed Social Interaction Understanding
• NeurIPS 2025Safety Pretraining: Toward the Next Generation of Safe AI
• ICML 2025$S^2$FGL: Spatial Spectral Federated Graph Learning
• ICLR 2025The Hidden Cost of Waiting for Accurate Predictions
• CVPR 2025Do We Really Need Curated Malicious Data for Safety Alignment in Multi-modal Large Language Models?
• NeurIPS 2024Quantifying the Bitter Lesson: How Safety Benchmarks Measure Capabilities Instead of Safety
• ICML 2024Optimally Improving Cooperative Learning in a Social Setting
• ICLR 2024The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Language Models
• CVPR 2024Digital Life Project: Autonomous 3D Characters with Social Intelligence
• NeurIPS 2023Strategyproof Voting under Correlated Beliefs
• ICML 2023Adaptive Coordination in Social Embodied Rearrangement
• NeurIPS 2026Beyond Feature Disruption: Boundary-Diverting Unlearnable Examples against Linear Probing
• ICML 2026AutoControl Arena: Synthesizing Executable Test Environments for Frontier AI Risk Evaluation
• ICLR 2026PRISON: Unmasking the Criminal Potential of Large Language Models
• CVPR 2026Spot The Ball A Benchmark for Visual Social Inference
• NeurIPS 2025SAFEx: Analyzing Vulnerabilities of MoE-Based LLMs via Stable Safety-critical Expert Identification
• ICML 2025Adversaries Can Misuse Combinations of Safe Models

全部社会层面 · safety论文

共 273 篇，第 2 / 6 页

51. ICML 2025Fair Clustering via Alignment
52. ICML 2025FairICP: Encouraging Equalized Odds via Inverse Conditional Permutation
53. ICML 2025Fast Estimation of Partial Dependence Functions using Trees
54. ICML 2025Flow Matching for Denoised Social Recommendation
55. ICML 2025From Pixels to Perception: Interpretable Predictions via Instance-wise Grouped Feature Selection
56. ICML 2025Gandalf the Red: Adaptive Security for LLMs
57. ICML 2025GaussMarker: Robust Dual-Domain Watermark for Diffusion Models
58. ICML 2025Graph Inverse Style Transfer for Counterfactual Explainability
59. ICML 2025How Do Transformers Learn Variable Binding in Symbolic Programs?
60. ICML 2025Improving LLM Safety Alignment with Dual-Objective Optimization
61. ICML 2025Improving Out-of-Distribution Detection via Dynamic Covariance Calibration
62. ICML 2025Improving Out-of-Distribution Detection with Markov Logic Networks
63. ICML 2025Inducing, Detecting and Characterising Neural Modules: A Pipeline for Functional Interpretability in Reinforcement Learning
64. ICML 2025InfoCons: Identifying Interpretable Critical Concepts in Point Clouds via Information Theory
65. ICML 2025Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviors
66. ICML 2025Jacobian Sparse Autoencoders: Sparsify Computations, Not Just Activations
67. ICML 2025Just Enough Shifts: Mitigating Over-Refusal in Aligned Language Models with Targeted Representation Fine-Tuning
68. ICML 2025Learning Classifiers That Induce Markets
69. ICML 2025Learning Safety Constraints for Large Language Models
70. ICML 2025Learning to Route LLMs with Confidence Tokens
71. ICML 2025Leveraging Predictive Equivalence in Decision Trees
72. ICML 2025LIMEFLDL: A Local Interpretable Model-Agnostic Explanations Approach for Label Distribution Learning
73. ICML 2025LlavaGuard: An Open VLM-based Framework for Safeguarding Vision Datasets and Models
74. ICML 2025MIB: A Mechanistic Interpretability Benchmark
75. ICML 2025Mind the Gap: A Practical Attack on GGUF Quantization
76. ICML 2025Minimalist Concept Erasure in Generative Models
77. ICML 2025Mixture of Experts Made Intrinsically Interpretable
78. ICML 2025MP-Nav: Enhancing Data Poisoning Attacks against Multimodal Learning
79. ICML 2025MTL-UE: Learning to Learn Nothing for Multi-Task Learning
80. ICML 2025Near-Optimal Decision Trees in a SPLIT Second
81. ICML 2025Not All Wrong is Bad: Using Adversarial Examples for Unlearning
82. ICML 2025On Explaining Equivariant Graph Networks via Improved Relevance Propagation
83. ICML 2025On the Impact of Performative Risk Minimization for Binary Random Variables
84. ICML 2025On the Vulnerability of Applying Retrieval-Augmented Generation within Knowledge-Intensive Application Domains
85. ICML 2025Optimizing Adaptive Attacks against Watermarks for Language Models
86. ICML 2025OR-Bench: An Over-Refusal Benchmark for Large Language Models
87. ICML 2025PASS: Private Attributes Protection with Stochastic Data Substitution
88. ICML 2025Pointwise Information Measures as Confidence Estimators in Deep Neural Networks: A Comparative Study
89. ICML 2025POROver: Improving Safety and Reducing Overrefusal in Large Language Models with Overgeneration and Preference Optimization
90. ICML 2025Position: AI Safety Must Embrace an Antifragile Perspective
91. ICML 2025Position: AI Safety should prioritize the Future of Work
92. ICML 2025Position: Editing Large Language Models Poses Serious Safety Risks
93. ICML 2025Position: Generative AI Regulation Can Learn from Social Media Regulation
94. ICML 2025Position: Humanity Faces Existential Risk from Gradual Disempowerment
95. ICML 2025Position: Political Neutrality in AI Is Impossible — But Here Is How to Approximate It
96. ICML 2025Position: Strong Consumer Protection is an Inalienable Defense for AI Safety in the United States
97. ICML 2025Position: The Right to AI
98. ICML 2025Position: We Need Responsible, Application-Driven (RAD) AI Research
99. ICML 2025Prediction via Shapley Value Regression
100. ICML 2025Preference learning made easy: Everything should be understood through win rate

常见问题

2023–2026 年四大会议一共有多少篇社会层面 · safety论文？

共 688 篇，按年份分别是 2023 年 4 篇、2024 年 41 篇、2025 年 273 篇、2026 年 370 篇。

NeurIPS 2026 有多少篇社会层面 · safety论文？

NeurIPS 2026 有 121 篇社会层面 · safety论文，上一届是 131 篇。

ICLR 2026 有多少篇社会层面 · safety论文？

ICLR 2026 有 20 篇社会层面 · safety论文，上一届是 11 篇。

ICML 2026 有多少篇社会层面 · safety论文？

ICML 2026 有 226 篇社会层面 · safety论文，上一届是 130 篇。

CVPR 2026 有多少篇社会层面 · safety论文？

CVPR 2026 有 3 篇社会层面 · safety论文，上一届是 1 篇。

哪个会议的社会层面 · safety论文最多？

ICML，2023–2026 年共 364 篇。