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 370 篇，第 3 / 8 页

101. ICML 2026FormalJudge: A Neuro-Symbolic Paradigm for Agentic Oversight
102. ICML 2026From Parameter Dynamics to Risk Scoring: Quantifying Sample-Level Safety Degradation in LLM Fine-tuning
103. ICML 2026From Rashomon Theory to PRAXIS: Efficient Decision Tree Rashomon Sets
104. ICML 2026From Talking to Singing: A New Challenge for Audio-Visual Deepfake Detection
105. ICML 2026Frontier Models Can Take Actions at Low Probabilities
106. ICML 2026Furina: Fragmented Uncertainty-Driven Refusal Instability Attack
107. ICML 2026FUSE: Full‑spectrum Unlearnable Examples via Spectral Equalization
108. ICML 2026FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing
109. ICML 2026Gaming Consensus: Coordinated Manipulation in Crowdsourced Fact-Checking
110. ICML 2026GKD-Recruiter: Jointly Modeling Social and Task Heterogeneity for Spatial Crowdsourcing via Graph Knowledge Distillation
111. ICML 2026Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance
112. ICML 2026Guaranteed Optimal Compositional Explanations for Neurons
113. ICML 2026Hidden in Plain Sight -- Class Competition Focuses Attribution Maps
114. ICML 2026Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio
115. ICML 2026How does information access affect LLM monitors' ability to detect sabotage?
116. ICML 2026How to Avoid Debate: Scalable AI Safety via Doubly-Efficient Interactive Proofs
117. ICML 2026Hugging Carbon: Quantifying the Training Carbon Emissions of AI Models at Scale
118. ICML 2026Implicit Safety Alignment from Crowd Preferences
119. ICML 2026In-Training Defenses Against Emergent Misalignment in Language Models
120. ICML 2026Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models
121. ICML 2026Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit
122. ICML 2026Introspection Adapters: Training LLMs to Report Their Learned Behaviors
123. ICML 2026Is Data Shapley Not Better than Random in Data Selection? Ask NASH
124. ICML 2026It's a TRAP! Task-Redirecting Agent Persuasion Benchmark for Web Agents
125. ICML 2026Jailbreak Foundry: From Papers to Runnable Attacks for Reproducible Benchmarking
126. ICML 2026Jailbreak to Protect: Buffering and Reinforcing via Temporary Jailbreaking for Safe Fine-Tuning in Large Language Models
127. ICML 2026Knowing Bias, Doing Better: Mitigating Social Bias in LLMs via Know-Bias Neuron Enhancement
128. ICML 2026Language Model Circuits Are Sparse in the Neuron Basis
129. ICML 2026Learning Efficient Guardrails for Compliance
130. ICML 2026Learning to Watermark in the Latent Space of Generative Models
131. ICML 2026Linear Ensembles Wash Away Watermarks: On the Fragility of Distributional Perturbations in LLMs
132. ICML 2026Locate then Correct: Debiasing Attention Heads in CLIP
133. ICML 2026Low-Compute Watermark Removal via Dual-Domain Natural Projection
134. ICML 2026Making Models Unmergeable via Scaling-Sensitive Loss Landscape
135. ICML 2026Mechanisms of Introspective Awareness
136. ICML 2026Mechanistic Anomaly Detection via Functional Attribution
137. ICML 2026Mechanistic Interpretability as Statistical Estimation: A Variance Analysis
138. ICML 2026Meerkat-VL: Implicit Risk Safety Alignment in Multimodal LLMs via Perceptual Reasoning and Self-Verification
139. ICML 2026Membership Inference Attacks for Unseen Classes
140. ICML 2026MESA: Improving MoE Safety Alignment via Decentralized Expertise
141. ICML 2026Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization
142. ICML 2026Mitigating the Safety–Utility Trade-off in LLM Alignment via Adaptive Safe Context Learning
143. ICML 2026Monitoring Monitorability
144. ICML 2026More Capable, Less Cooperative? When LLMs Fail at Zero-Cost Collaboration
145. ICML 2026More Sail than Ballast: Addressing Harmful Knowledge Leakage in the Expansive Reasoning Space of LRMs
146. ICML 2026MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety
147. ICML 2026Multilingual Safety Alignment via Representation-Space Separability
148. ICML 2026Neural Vector Lyapunov–Razumikhin Certificates for Delayed Interconnected Systems
149. ICML 2026Old Habits Die Hard: How Conversational History Geometrically Traps LLMs
150. ICML 2026OmniVL-Guard: Towards Unified Vision-Language Forgery Detection and Grounding via Balanced RL

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。