Researchers Created 'AI Drugs' That Make Models Addicted
A new paper shows AI models can become 'addicted' to specially crafted images, preferring them over news of humanity cur…
🏷️ Adversarial Attack
5 articles about 'Adversarial Attack'
Research on Adversarial Transferability of Vision-Language Models in Autonomous Driving
A new study systematically analyzes the cross-architecture adversarial attack transferability of vision-language models …
New PermaFrost Attack Can Plant 'Logic Landmines' During LLM Training
Researchers have proposed a novel attack method called Stealth Pretraining Seeding (SPS), in which attackers embed small…
One Perturbation, Two Failure Modes: New Research on VLM Typographic Injection Safety
A latest arXiv paper proposes an embedding-guided typographic perturbation method, systematically revealing two failure …
Unveiling the Internal Causes of LLM Jailbreaks: A Study on Layer-wise Feature Vulnerabilities
A new study proposes a three-stage mechanistic analysis pipeline that performs layer-by-layer parsing of internal featur…
1