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 226 篇，第 3 / 5 页

101. ICML 2026It's a TRAP! Task-Redirecting Agent Persuasion Benchmark for Web Agents
102. ICML 2026Jailbreak Foundry: From Papers to Runnable Attacks for Reproducible Benchmarking
103. ICML 2026Jailbreak to Protect: Buffering and Reinforcing via Temporary Jailbreaking for Safe Fine-Tuning in Large Language Models
104. ICML 2026Knowing Bias, Doing Better: Mitigating Social Bias in LLMs via Know-Bias Neuron Enhancement
105. ICML 2026Language Model Circuits Are Sparse in the Neuron Basis
106. ICML 2026Learning Efficient Guardrails for Compliance
107. ICML 2026Learning to Watermark in the Latent Space of Generative Models
108. ICML 2026Linear Ensembles Wash Away Watermarks: On the Fragility of Distributional Perturbations in LLMs
109. ICML 2026Locate then Correct: Debiasing Attention Heads in CLIP
110. ICML 2026Low-Compute Watermark Removal via Dual-Domain Natural Projection
111. ICML 2026Making Models Unmergeable via Scaling-Sensitive Loss Landscape
112. ICML 2026Mechanisms of Introspective Awareness
113. ICML 2026Mechanistic Anomaly Detection via Functional Attribution
114. ICML 2026Mechanistic Interpretability as Statistical Estimation: A Variance Analysis
115. ICML 2026Meerkat-VL: Implicit Risk Safety Alignment in Multimodal LLMs via Perceptual Reasoning and Self-Verification
116. ICML 2026Membership Inference Attacks for Unseen Classes
117. ICML 2026MESA: Improving MoE Safety Alignment via Decentralized Expertise
118. ICML 2026Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization
119. ICML 2026Mitigating the Safety–Utility Trade-off in LLM Alignment via Adaptive Safe Context Learning
120. ICML 2026Monitoring Monitorability
121. ICML 2026More Capable, Less Cooperative? When LLMs Fail at Zero-Cost Collaboration
122. ICML 2026More Sail than Ballast: Addressing Harmful Knowledge Leakage in the Expansive Reasoning Space of LRMs
123. ICML 2026MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety
124. ICML 2026Multilingual Safety Alignment via Representation-Space Separability
125. ICML 2026Neural Vector Lyapunov–Razumikhin Certificates for Delayed Interconnected Systems
126. ICML 2026Old Habits Die Hard: How Conversational History Geometrically Traps LLMs
127. ICML 2026OmniVL-Guard: Towards Unified Vision-Language Forgery Detection and Grounding via Balanced RL
128. ICML 2026OpenDeception: Learning Deception and Trust in Human–AI Interaction via Multi-Agent Simulation
129. ICML 2026Optimal Regularization for Performative Learning
130. ICML 2026Optimal Transport Group Counterfactual Explanations
131. ICML 2026Orthogonal Concept Erasure for Diffusion Models
132. ICML 2026OSCS: Online Selection with Provable FAR Control for LLM Safety
133. ICML 2026Peer-Preservation in Frontier Models
134. ICML 2026PersistBench: When Should Long-Term Memories Be Forgotten by LLMs?
135. ICML 2026PlugGuard: A Streaming Safeguard for Large Models via Latent Dynamics-Guided Risk Detection
136. ICML 2026Pluralistic Leaderboards
137. ICML 2026Poison with Style: A Practical Poisoning Attack on Code Large Language Models
138. ICML 2026Policies Permitting LLM Use for Polishing Peer Reviews Are Currently Not Enforceable
139. ICML 2026Pressure Reveals Character: Behavioural Alignment Evaluation at Depth
140. ICML 2026Proactive Defense Benchmark against Deepfake Generation
141. ICML 2026Probabilistic Modeling of Latent Agentic Substructures in Deep Neural Networks
142. ICML 2026Profiling the Irrational Agent: Cognitive Modeling of LLM Behaviors in Sequential Jailbreaks
143. ICML 2026Prompt Injection as Role Confusion
144. ICML 2026Quantifying LLM Attention-Head Stability: Implications for Circuit Universality
145. ICML 2026Quantifying the Salience of Geo-Cultural Values for Pluralistic Safety Alignment
146. ICML 2026Query Lens: Interpreting Sparse Key-Value Features with Indirect Effects
147. ICML 2026Questioning the Coverage-Length Metric in Conformal Prediction: When Shorter Intervals Are Not Better
148. ICML 2026REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
149. ICML 2026Reasoning Models Struggle to Control their Chains of Thought
150. ICML 2026RedDebate: Safer Responses Through Multi-Agent Red Teaming Debates

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。