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 3 / 6 页

101. ICML 2025Provably Cost-Sensitive Adversarial Defense via Randomized Smoothing
102. ICML 2025Ranked from Within: Ranking Large Multimodal Models Without Labels
103. ICML 2025Relative Error Fair Clustering in the Weak-Strong Oracle Model
104. ICML 2025Representative Ranking for Deliberation in the Public Sphere
105. ICML 2025Rethinking the Temperature for Federated Heterogeneous Distillation
106. ICML 2025SafeArena: Evaluating the Safety of Autonomous Web Agents
107. ICML 2025Safety Alignment Can Be Not Superficial With Explicit Safety Signals
108. ICML 2025Safety Reasoning with Guidelines
109. ICML 2025SafetyAnalyst: Interpretable, Transparent, and Steerable Safety Moderation for AI Behavior
110. ICML 2025Scaling Trends in Language Model Robustness
111. ICML 2025Selective Preference Aggregation
112. ICML 2025SEMU: Singular Value Decomposition for Efficient Machine Unlearning
113. ICML 2025ShieldAgent: Shielding Agents via Verifiable Safety Policy Reasoning
114. ICML 2025Speak Easy: Eliciting Harmful Jailbreaks from LLMs with Simple Interactions
115. ICML 2025SPRI: Aligning Large Language Models with Context-Situated Principles
116. ICML 2025STAIR: Improving Safety Alignment with Introspective Reasoning
117. ICML 2025Statistical Collusion by Collectives on Learning Platforms
118. ICML 2025STD-FD: Spatio-Temporal Distribution Fitting Deviation for AIGC Forgery Identification
119. ICML 2025Stealix: Model Stealing via Prompt Evolution
120. ICML 2025Suitability Filter: A Statistical Framework for Classifier Evaluation in Real-World Deployment Settings
121. ICML 2025Sum-of-Parts: Self-Attributing Neural Networks with End-to-End Learning of Feature Groups
122. ICML 2025Tensor Product Neural Networks for Functional ANOVA Model
123. ICML 2025Test-Time Multimodal Backdoor Detection by Contrastive Prompting
124. ICML 2025The Elicitation Game: Evaluating Capability Elicitation Techniques
125. ICML 2025The Illusion of Role Separation: Hidden Shortcuts in LLM Role Learning (and How to Fix Them)
126. ICML 2025The Jailbreak Tax: How Useful are Your Jailbreak Outputs?
127. ICML 2025The Limits of Predicting Agents from Behaviour
128. ICML 2025The Perils of Optimizing Learned Reward Functions: Low Training Error Does Not Guarantee Low Regret
129. ICML 2025TOPLOC: A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference
130. ICML 2025Towards Attributions of Input Variables in a Coalition
131. ICML 2025Towards Black-Box Membership Inference Attack for Diffusion Models
132. ICML 2025Towards flexible perception with visual memory
133. ICML 2025Towards Global-level Mechanistic Interpretability: A Perspective of Modular Circuits of Large Language Models
134. ICML 2025TRUST-VLM: Thorough Red-Teaming for Uncovering Safety Threats in Vision-Language Models
135. ICML 2025Trustworthy Machine Learning through Data-Specific Indistinguishability
136. ICML 2025Uncertainty-Based Extensible Codebook for Discrete Federated Learning in Heterogeneous Data Silos
137. ICML 2025Validating Mechanistic Interpretations: An Axiomatic Approach
138. ICML 2025Weak-to-Strong Jailbreaking on Large Language Models
139. ICML 2025What makes an Ensemble (Un) Interpretable?
140. ICML 2025Which Attention Heads Matter for In-Context Learning?
141. ICML 2025WMarkGPT: Watermarked Image Understanding via Multimodal Large Language Models
142. ICML 2025X-Hacking: The Threat of Misguided AutoML
143. NeurIPS 2025AdaptDel: Adaptable Deletion Rate Randomized Smoothing for Certified Robustness
144. NeurIPS 2025Adaptive Defense against Harmful Fine-Tuning for Large Language Models via Bayesian Data Scheduler
145. NeurIPS 2025Additive Models Explained: A Computational Complexity Approach
146. NeurIPS 2025Adjacent Words, Divergent Intents: Jailbreaking Large Language Models via Task Concurrency
147. NeurIPS 2025Alignment of Large Language Models with Constrained Learning
148. NeurIPS 2025Analogy-based Multi-Turn Jailbreak against Large Language Models
149. NeurIPS 2025Backdoor Cleaning without External Guidance in MLLM Fine-tuning
150. NeurIPS 2025Boosting the Uniqueness of Neural Networks Fingerprints with Informative Triggers

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。