 Probe before You Talk: Towards Black-box Defense against Backdoor Unalignment for Large Language ModelsInternational Conference on Learning Representations (ICLR), 2025 |
 ROSA: Finding Backdoors with FuzzingInternational Conference on Software Engineering (ICSE), 2025 |
 Comet: Accelerating Private Inference for Large Language Model by Predicting Activation SparsityIEEE Symposium on Security and Privacy (S&P), 2025 |
 Breaking Free from MMI: A New Frontier in Rationalization by Probing Input UtilizationInternational Conference on Learning Representations (ICLR), 2025 |
 Re-Imagining Multimodal Instruction Tuning: A Representation ViewInternational Conference on Learning Representations (ICLR), 2025 |
 Multi-Target Federated Backdoor Attack Based on Feature AggregationPattern Recognition (Pattern Recogn.), 2025 |
 REFINE: Inversion-Free Backdoor Defense via Model ReprogrammingInternational Conference on Learning Representations (ICLR), 2025 |
