ICML 2026 accepted papers on Adversarial robustness and security
116 of the 6,341 papers accepted at ICML 2026 (1.8%) are on Adversarial robustness and security, compared with 2.7% at ICML 2025. They fall into 4 topics.
See ICML 2026 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Attacks on text and graphs · 55 papers
- A Coin Flip for Safety: LLM Judges Fail to Reliably Measure Adversarial Robustness
- AliMark: Enhancing Robustness of Sentence-Level Watermarking Against Text Paraphrasing
- Causes and Consequences of Representational Similarity in Machine Learning Models
- CREDIT: Certified Ownership Verification of Deep Neural Networks Against Model Extraction Attacks
- Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment
- Efficient Preference Poisoning Attack on Offline RLHF
- Eyes-on-Me: Scalable RAG Poisoning through Transferable Attention-Steering Attractors
- From Similarity to Vulnerability: Key Collision Attack on LLM Semantic Caching
- Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance
- Harnessing Non-Adversarial Robustness in Large Language Models
- Jailbreak Foundry: From Papers to Runnable Attacks for Reproducible Benchmarking
- Localize and Neutralize: Gradient-Guided Token Suppression Against Visual Prompt Injection Attack
and 43 more in the interactive map.
Certified and empirical robustness · 31 papers
- Adversarially Robust Approximate Furthest Neighbor
- Benign Overfitting in Adversarial Training for Vision Transformers
- BiRQA: Bidirectional Robust Quality Assessment for Images
- Calibrating Uncertainty for Zero-Shot Adversarial CLIP
- Contrastive Spectral Rectification: Test-Time Defense towards Zero-shot Adversarial Robustness of CLIP
- Dual-branch Robust Unlearnable Examples
- Higher-Order Certified Robustness for Regression
- Improving Adversarial Robustness of Attribution via Implicit Regularization
- Improving ML Attacks on LWE with Data Repetition and Stepwise Regression
- Mind Your Margin and Boundary: Are Your Distilled Datasets Truly Robust?
- MoCo-EA: Exploiting Adversarial Mode Connectivity for Efficient Evolutionary Attacks
- Non-Parametric Probabilistic Robustness: A Conservative Risk Estimator under Unknown Perturbation Distributions
and 19 more in the interactive map.
Backdoor and poisoning · 21 papers
- Angel or Demon: Investigating the Plasticity Interventions' Impact on Backdoor Threats in Deep Reinforcement Learning
- Anti-Backdoor Coreset Selection via Cumulative Entropy
- BYORn: Bootstrap Your Own Responses to Defend Large Vision-Language Models Against Backdoor Attacks
- DF-LoGiT: Data-Free Logic-Gated Backdoor Attacks in Vision Transformers
- Fox in the Henhouse: Supply-Chain Backdoor Attacks Against Reinforcement Learning
- From Parameters to Feature Space: Task Arithmetic for Backdoor Mitigation in Model Merging
- From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMs
- Improving the Sensitivity of Backdoor Detectors via Class Subspace Orthogonalization
- Poison with Style: A Practical Poisoning Attack on Code Large Language Models
- Rapid Poison: Practical Poisoning Attacks Against the Rapid Response Framework
- Semantic-level Backdoor Attack against Text-to-Image Diffusion Models
- SkillTrojan: Backdoor Attacks on Skill-Based Agent Systems
and 9 more in the interactive map.
Black-box and transfer attacks · 9 papers
- ``Someone Hid It!'': Query-Agnostic Black-Box Attacks on LLM-Based Retrieval
- Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models
- Rethinking Forgery Attacks on Semantic Watermarks in Black-Box Settings: A Geometric Distortion Perspective
- STARE: Step-wise Temporal Alignment and Red-teaming Engine for Multi-modal Toxicity Attack
- Adversarial Attacks and Robust Training for Hypergraph Neural Networks
- Dissecting the Safety Circuit: Neuronal Intervention for Transferable Adversarial Attacks on VLMs
- Hearing Without Noticing? Attention-Aware Stealthy Black-Box Adversarial Audio Attacks
- Low-Rank and Sparsity Are All You Need: Exploring Robust Hierarchical Latent Subspaces for Transferable Adversarial Attack
- Ripple Perturbations Through Structure: Likelihood-Constrained Adversarial Attacks on Heterogeneous Tabular Data