对抗鲁棒论文趋势 2023–2026

主要结论

• 对抗鲁棒论文从 2023 年的 313 篇增长到 2026 年的 317 篇，三年间扩大到约 1.0 倍。
• NeurIPS 是发对抗鲁棒论文最多的会议，2023–2026 年共 422 篇，占四个会议总量的 33%。
• 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 13 篇增至 2026 年的 56 篇。
• 在「安全与可信」的 5 个主题里，它的 2026 年论文数排第 1；同组增长最快的水印是 9.3 倍，对抗鲁棒是 1.0 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 2026A Little Robustness Is All You Need: Leveraging Predictions for Contextual Optimization
• ICML 2026Bias in Zeroth-Order Normal Estimation for Decision-Based Attacks
• ICLR 2026CERTIFIED VS. EMPIRICAL ADVERSARIAL ROBUSTNESS VIA HYBRID CONVOLUTIONS WITH ATTENTION STOCHASTICITY
• CVPR 2026A Combination of Noise and Bilateral Filters Achieve Supralinear and
• NeurIPS 2025A Unified Approach to Submodular Maximization Under Noise
• ICML 2025Adversarial Perturbations Are Formed by Iteratively Learning Linear Combinations of the Right Singular Vectors of the Adversarial Jacobian
• ICLR 2025$\sigma$-zero: Gradient-based Optimization of $\ell_0$-norm Adversarial Examples
• CVPR 2025A Unified, Resilient, and Explainable Adversarial Patch Detector
• NeurIPS 2024A StrongREJECT for Empty Jailbreaks
• ICML 2024Adversarial Attacks on Combinatorial Multi-Armed Bandits
• ICLR 2024An LLM can Fool Itself: A Prompt-Based Adversarial Attack
• CVPR 2024Defense Against Adversarial Attacks on No-Reference Image Quality Models with Gradient Norm Regularization
• NeurIPS 2023A Closer Look at the Robustness of Contrastive Language-Image Pre-Training (CLIP)
• ICML 2023A Law of Robustness beyond Isoperimetry
• ICLR 2023A law of adversarial risk, interpolation, and label noise
• CVPR 2023Adversarial Normalization: I Can Visualize Everything (ICE)
• NeurIPS 2026A Regularization-Based Approach to Public Belief State Search for Adversarial Games
• ICML 2026CaliDist: Calibrating Large Language Models via Behavioral Robustness to Distraction
• ICLR 2026Conformal Robustness Control: A New Strategy for Robust Decision
• CVPR 2026A Provable Energy-Guided Test-Time Defense Boosting Adversarial Robustness of Large

全部对抗鲁棒论文

共 93 篇，第 2 / 2 页

51. ICML 2026On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs
52. ICML 2026On the Adversarial Robustness of Large Vision-Language Models under Visual Token Compression
53. ICML 2026On the Existence of Consistent Adversarial Attacks in High-Dimensional Linear Classification
54. ICML 2026On the Robustness of Langevin Dynamics to Score Function Error
55. ICML 2026Optimal Attention Temperature Improves the Robustness of In-Context Learning under Distribution Shift in High Dimensions
56. ICML 2026Posterior Mismatch Matters: Adversarial Training for Long-Tailed Robustness
57. ICML 2026Proactive Defense Benchmark against Deepfake Generation
58. ICML 2026Probabilistic Robustness Certificates against Adversarial Attacks
59. ICML 2026Provably Efficient Policy-Reward Co-Pretraining for Adversarial Imitation Learning
60. ICML 2026PSMix: Robust Point Cloud Recognition through Spectral Domain Mixing
61. ICML 2026RADAR: Defending RAG Dynamically against Retrieval Corruption
62. ICML 2026Regret Minimization With a Crowd of Awakening Experts
63. ICML 2026Reward Under Attack: Analyzing the Robustness and Hackability of Process Reward Models
64. ICML 2026Ripple Perturbations Through Structure: Likelihood-Constrained Adversarial Attacks on Heterogeneous Tabular Data
65. ICML 2026Robust Bayesian Optimisation with Unbounded Corruptions
66. ICML 2026Robust Linear Dueling Bandits with Post-serving Context under Unknown Delays and Adversarial Corruptions
67. ICML 2026Robust Vision-Language Models via Manifold-Adversarial Adapters
68. ICML 2026Robustness of Mixtures of Experts to Feature Noise
69. ICML 2026Safety-Efficacy Trade Off: Robustness against Data-Poisoning
70. ICML 2026Scalable GANs with Transformers
71. ICML 2026Security--Fidelity Tradeoffs: The Hidden Cost of Prompt Injection Defense
72. ICML 2026Segment Anything with Robust Uncertainty-Accuracy Correlation
73. ICML 2026Self-Calibrated Consistency can Fight Back for Adversarial Robustness in Vision-Language Models
74. ICML 2026Semantic Router: On the Feasibility of Hijacking MLLMs via a Single Adversarial Perturbation
75. ICML 2026Ski Rental with Distributional Predictions of Unknown Quality
76. ICML 2026SORA: Free Second-Order Attacks in Fast Adversarial Training
77. ICML 2026SS-TPT: Stability and Suitability-Guided Test-Time Prompt Tuning for Adversarially Robust Vision-Language Models
78. ICML 2026Statistically Undetectable Backdoors in Deep Neural Networks
79. ICML 2026Steering Beyond the Support: Adversarial Training on Unsupervised Jailbroken Activation Simulation
80. ICML 2026Stochastic Order Learning: An Approach to Rank Estimation Using Noisy Data
81. ICML 2026Suppress and Diversify: Refining Robust Pathways for Corruption Robustness
82. ICML 2026SWE-ABS: Adversarial Benchmark Strengthening Exposes Inflated Success Rates on Test-based Benchmark
83. ICML 2026The Latent Guardian: Defending Collaborative Perception via Feature-Level Consistency Verification
84. ICML 2026Toward Understanding Adversarial Distillation: Why Robust Teachers Fail
85. ICML 2026Towards Fine-Grained Robustness: Attention-Guided Test-Time Prompt Tuning for Vision-Language Models
86. ICML 2026Towards Optimal Robustness in Learning-Augmented Paging
87. ICML 2026Towards Understanding Generalization of Federated Adversarial Learning: Perspective of Algorithmic Stability
88. ICML 2026Training with Honeypots: Reshaping How LLMs Fail Under Adversarial Attacks
89. ICML 2026Training-Free Adversarial Robustness in Computational MRI
90. ICML 2026Two Modalities Are Better Than One: Efficient Adversarial Purification via Multimodal Diffusion Models
91. ICML 2026Unifying Adversarial Robustness and Training Across Text Scoring Models
92. ICML 2026When Can You Poison Rewards? A Tight Characterization of Reward Poisoning in Linear MDPs
93. ICML 2026You Don't Protect if You Don't Expect: Breaking the Key Assumption behind CLIP's Test-Time Defenses

常见问题

2023–2026 年四大会议一共有多少篇对抗鲁棒论文？

共 1,291 篇，按年份分别是 2023 年 313 篇、2024 年 334 篇、2025 年 327 篇、2026 年 317 篇。

NeurIPS 2026 有多少篇对抗鲁棒论文？

NeurIPS 2026 有 81 篇对抗鲁棒论文，上一届是 110 篇。

ICLR 2026 有多少篇对抗鲁棒论文？

ICLR 2026 有 94 篇对抗鲁棒论文，上一届是 88 篇。

ICML 2026 有多少篇对抗鲁棒论文？

ICML 2026 有 93 篇对抗鲁棒论文，上一届是 85 篇。

CVPR 2026 有多少篇对抗鲁棒论文？

CVPR 2026 有 49 篇对抗鲁棒论文，上一届是 44 篇。

哪个会议的对抗鲁棒论文最多？

NeurIPS，2023–2026 年共 422 篇。