Xiaojun Jia

MIDAS: Multi-Image Dispersion and Semantic Reconstruction for Jailbreaking MLLMs

TISR: Breaking Vision–Language Models via Text–Image Segmentation and Reassembly

Untargeted Jailbreak Attack

Text Modality Oriented Image Feature Extraction for Detecting Diffusion-based DeepFake

Obscure but Effective: Classical Chinese Jailbreak Prompt Optimization via Bio-Inspired Search

Inverse Reinforcement Learning with Dynamic Reward Scaling for LLM Alignment

Reasoning as an Attack Surface: Adaptive Evolutionary CoT Jailbreaks for LLMs

Dynamic Target Attack

PersGuard: Preventing Malicious Personalization in Text-to-Image Diffusion via Model Backdoors

Unified Surgical World Model for Structured Understanding, Long-Horizon Prediction, and Fine-Grained Generation

SAGE-CoT: Self-Adaptive Generated Chain-of-Thought for Jailbreaking

RobustIT: Adapter-Centric and Attack-Agnostic Anti-Backdoor Instruction Tuning

Dynamic $k$-shot In-Context Learning

Strata-Sword: A Hierarchical Safety Evaluation towards LLMs based on Reasoning Complexity of Jailbreak Instructions

DAMA: Data- and Model-aware Alignment of Multi-modal LLMs

Adversarial Attacks against Closed-Source MLLMs via Feature Optimal Alignment

Cannot See the Forest for the Trees: Invoking Heuristics and Biases to Elicit Irrational Choices of LLMs

Improved Techniques for Optimization-Based Jailbreaking on Large Language Models

OT-Attack: Enhancing Adversarial Transferability of Vision-Language Models via Optimal Transport Optimization

SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG

CleanerCLIP: Fine-grained Counterfactual Semantic Augmentation for Backdoor Defense in Contrastive Learning

LLM Jailbreak Detection for (Almost) Free!

No Access, No Safety: Free Lunch Adversarial Attacks on Black-box NLP Models