@misc{indiciae95165c25d1db,
  title = {Jailbreaking Text-to-Image Models Through Cracks: Navigating Heterogeneous Safety Filters via Multi-Agent Debate},
  author = {Kaiyan Wen and Shijie Zhang and Lu Yu and Guangdong Bai},
  year = {2026},
  url = {https://arxiv.org/abs/2609.01168},
  note = {Source identifier: 2609.01168}
}