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 41 篇，第 1 / 1 页

1. CVPR 2024Digital Life Project: Autonomous 3D Characters with Social Intelligence
2. ICLR 2024Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!
3. ICLR 2024Gaining Wisdom from Setbacks: Aligning Large Language Models via Mistake Analysis
4. ICLR 2024GPT-4 Is Too Smart To Be Safe: Stealthy Chat with LLMs via Cipher
5. ICLR 2024Multilingual Jailbreak Challenges in Large Language Models
6. ICLR 2024Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions
7. ICLR 2024Social-Transmotion: Promptable Human Trajectory Prediction
8. ICLR 2024Some Fundamental Aspects about Lipschitz Continuity of Neural Networks
9. ICLR 2024The Alignment Problem from a Deep Learning Perspective
10. ICLR 2024The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Language Models
11. ICLR 2024Training Socially Aligned Language Models on Simulated Social Interactions
12. ICLR 2024Unprocessing Seven Years of Algorithmic Fairness
13. ICML 2024AI Control: Improving Safety Despite Intentional Subversion
14. ICML 2024Optimally Improving Cooperative Learning in a Social Setting
15. ICML 2024Position: Social Choice Should Guide AI Alignment in Dealing with Diverse Human Feedback
16. ICML 2024Run-Time Task Composition with Safety Semantics
17. ICML 2024Scalable AI Safety via Doubly-Efficient Debate
18. ICML 2024Truly No-Regret Learning in Constrained MDPs
19. NeurIPS 2024Aggregating Quantitative Relative Judgments: From Social Choice to Ranking Prediction
20. NeurIPS 2024Can an AI Agent Safely Run a Government? Existence of Probably Approximately Aligned Policies
21. NeurIPS 2024Can Simple Averaging Defeat Modern Watermarks?
22. NeurIPS 2024Computational Aspects of Bayesian Persuasion under Approximate Best Response
23. NeurIPS 2024Confidence Calibration of Classifiers with Many Classes
24. NeurIPS 2024Conformalized Multiple Testing after Data-dependent Selection
25. NeurIPS 2024Constrained Human-AI Cooperation: An Inclusive Embodied Social Intelligence Challenge
26. NeurIPS 2024DarkSAM: Fooling Segment Anything Model to Segment Nothing
27. NeurIPS 2024Jailbreaking Large Language Models Against Moderation Guardrails via Cipher Characters
28. NeurIPS 2024Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates
29. NeurIPS 2024Large Language Model Unlearning via Embedding-Corrupted Prompts
30. NeurIPS 2024Learning Social Welfare Functions
31. NeurIPS 2024Lisa: Lazy Safety Alignment for Large Language Models against Harmful Fine-tuning Attack
32. NeurIPS 2024Mission Impossible: A Statistical Perspective on Jailbreaking LLMs
33. NeurIPS 2024Motif-oriented influence maximization for viral marketing in large-scale social networks
34. NeurIPS 2024Navigating the Safety Landscape: Measuring Risks in Finetuning Large Language Models
35. NeurIPS 2024Quantifying the Bitter Lesson: How Safety Benchmarks Measure Capabilities Instead of Safety
36. NeurIPS 2024RedCode: Multi-dimensional Safety Benchmark for Code Agents
37. NeurIPS 2024Safe LoRA: The Silver Lining of Reducing Safety Risks when Finetuning Large Language Models
38. NeurIPS 2024Safety through feedback in Constrained RL
39. NeurIPS 2024Safetywashing: Do AI Safety Benchmarks Actually Measure Safety Progress?
40. NeurIPS 2024Uncovering Safety Risks of Large Language Models through Concept Activation Vector
41. NeurIPS 2024Using Noise to Infer Aspects of Simplicity Without Learning

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。