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 2 / 5 页

51. ICML 2026Dissect and Prune: Enhancing Robustness in AI-Generated Image Detection
52. ICML 2026Dissecting the Safety Circuit: Neuronal Intervention for Transferable Adversarial Attacks on VLMs
53. ICML 2026Divergence Decoding: Inference-Time Unlearning via Auxiliary Models
54. ICML 2026Do Language Models Track Entities Across State Changes?
55. ICML 2026Don't Walk the Line: Boundary Guidance for Filtered Generation
56. ICML 2026Dual-branch Robust Unlearnable Examples
57. ICML 2026Efficient LLM Moderation with Multi-Layer Latent Prototypes
58. ICML 2026EMBGuard: Constructing Hazard-Aware Guardrails for Safe Planning in Embodied Agents
59. ICML 2026Emergence of Biased Consensus in Multi-Agent LLM Debates
60. ICML 2026Emergence of Hierarchical Emotion Organization in Large Language Models
61. ICML 2026Emergent Communication Under Misinformation
62. ICML 2026Enhancing Conformal Prediction via Class Similarity
63. ICML 2026Ensembling Sparse Autoencoders
64. ICML 2026Estimating Tail Risks in Language Model Output Distributions
65. ICML 2026Exactly Computing do-Shapley Values
66. ICML 2026ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model Behavior
67. ICML 2026Exploration Hacking: Can LLMs Learn to Resist RL Training?
68. ICML 2026Exploring Nonlinear Pathway in Parameter Space for Machine Unlearning
69. ICML 2026Eyes-on-Me: Scalable RAG Poisoning through Transferable Attention-Steering Attractors
70. ICML 2026Fair Decisions from Calibrated Scores: Achieving Optimal Classification While Satisfying Sufficiency
71. ICML 2026Faithful Mobile GUI Agents with Guided Advantage Estimator
72. ICML 2026Falling Trees: A Model Class for Interpretable Risk Prioritization
73. ICML 2026Finding Most Influential Sets
74. ICML 2026FLARE-AI: Flaw Reporting for AI
75. ICML 2026Fleet: Few Shots Lead Effective AI-generated Image Detection
76. ICML 2026FoeGlass: Simple In-Context Learning Is Enough for Red Teaming Audio Deepfake Detectors
77. ICML 2026ForceForget: Reinforcement Concept Removal for Enhancing Safety in Text-to-Image Models
78. ICML 2026FormalJudge: A Neuro-Symbolic Paradigm for Agentic Oversight
79. ICML 2026From Parameter Dynamics to Risk Scoring: Quantifying Sample-Level Safety Degradation in LLM Fine-tuning
80. ICML 2026From Rashomon Theory to PRAXIS: Efficient Decision Tree Rashomon Sets
81. ICML 2026From Talking to Singing: A New Challenge for Audio-Visual Deepfake Detection
82. ICML 2026Frontier Models Can Take Actions at Low Probabilities
83. ICML 2026Furina: Fragmented Uncertainty-Driven Refusal Instability Attack
84. ICML 2026FUSE: Full‑spectrum Unlearnable Examples via Spectral Equalization
85. ICML 2026FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing
86. ICML 2026Gaming Consensus: Coordinated Manipulation in Crowdsourced Fact-Checking
87. ICML 2026GKD-Recruiter: Jointly Modeling Social and Task Heterogeneity for Spatial Crowdsourcing via Graph Knowledge Distillation
88. ICML 2026Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance
89. ICML 2026Guaranteed Optimal Compositional Explanations for Neurons
90. ICML 2026Hidden in Plain Sight -- Class Competition Focuses Attribution Maps
91. ICML 2026Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio
92. ICML 2026How does information access affect LLM monitors' ability to detect sabotage?
93. ICML 2026How to Avoid Debate: Scalable AI Safety via Doubly-Efficient Interactive Proofs
94. ICML 2026Hugging Carbon: Quantifying the Training Carbon Emissions of AI Models at Scale
95. ICML 2026Implicit Safety Alignment from Crowd Preferences
96. ICML 2026In-Training Defenses Against Emergent Misalignment in Language Models
97. ICML 2026Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models
98. ICML 2026Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit
99. ICML 2026Introspection Adapters: Training LLMs to Report Their Learned Behaviors
100. ICML 2026Is Data Shapley Not Better than Random in Data Selection? Ask NASH

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。