class OscarWu:
role = "Red Team Otaku" # yes, both things
def __init__(self):
self.status = "jailbreaking"
def motto(self):
return "I trace failure."
def next_target(self):
return "the weakest assumption"|
Internal Safety Collapse in Frontier Large Language Models, with ISC-Bench, JailbreakArena, templates, tutorials, and live cases. |
Toward Universal and Transferable Jailbreak Attacks on Vision-Language Models. |
|
Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety. |
Risks, attacks, and defenses across the full embodied AI pipeline, covering 480+ papers. |
|
An integrated safety evaluation of frontier models across language, vision-language, and image generation. |
|