Xingjun Ma

ShadowFPT: Backdooring Federated Prompt Tuning via Shadow Triggers

RedDarwin: Evolving Jailbreak Skills against Frontier LLMs

VEX-Bench: Benchmarking Verification Complexity of LLM-Generated Misinformation

HazardAuditor: From Executable Threats to Safer Computer-Use Agents

Adversarial Attacks for Good: A Survey of Proactive Protection across the Visual Content Lifecycle

OpenART: Scaling Agent Red Teaming via Open-Ended Environment Evolution

Safety in Self-Evolving LLM Agent Systems: Threats, Amplification, and Case Studies

Constitutional On-Policy Safe Distillation

SentGuard: Sentence-Level Streaming Guardrails for Large Language Models

BraveGuard: From Open-World Threats to Safer Computer-Use Agents

MESA: Improving MoE Safety Alignment via Decentralized Expertise

AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security

When Medical Safety Alignment Fails: A Benchmark for Evaluating LLMs on High-Risk Medical Queries

Adversarial Orthogonal Disentanglement for LVLM Hallucination Mitigation

Towards Context-Invariant Safety Alignment for Large Language Models

A Survey of Large Audio Language Models: Generalization, Trustworthiness, and Outlook

TAME: Test-Time Adversarial Prompt Tuning via Mixture-of-Experts for Vision-Language Models

DarkLLM: Learning Language-Driven Adversarial Attacks with Large Language Models

BabelSafe: A Policy-Grounded Multilingual Safety Benchmark for LLMs

HazardArena: Evaluating Semantic Safety in Vision-Language-Action Models

What Does a Sharing Question Add? Auditing LLM Survey Scores for Misinformation

Steering the Verifiability of Multimodal AI Hallucinations

AgentHazard: A Benchmark for Evaluating Harmful Behavior in Computer-Use Agents

Safety in Embodied AI: A Survey of Risks, Attacks, and Defenses

OOD-MMSafe: Advancing MLLM Safety from Harmful Intent to Hidden Consequences

Backdoor4Good: Benchmarking Beneficial Uses of Backdoors in LLMs

Internal Safety Collapse in Frontier Large Language Models

RA-Det: Towards Universal Detection of AI-Generated Images via Robustness Asymmetry

Toward Universal and Transferable Jailbreak Attacks on Vision-Language Models

Just Ask: Curious Code Agents Reveal System Prompts in Frontier LLMs

A Safety Report on GPT-5.2, Gemini 3 Pro, Qwen3-VL, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5

BackdoorAgent: A Unified Framework for Backdoor Attacks on LLM-based Agents

BackdoorVLM: A Benchmark for Backdoor Attacks and Defenses on Vision-Language Models

AutoBackdoor: Automating Backdoor Attacks via LLM Agents

Evolve the Method, Not the Prompts: Evolutionary Synthesis of Jailbreak Attacks on LLMs

AttackVLA: Benchmarking Adversarial and Backdoor Attacks on Vision-Language-Action Models