All news
2 articles- paperarXiv (systems, data, ML)·23 Oct 20263.4Diffusion LMs find natural-looking black-box attack prompts that trigger overthinking.
- paperarXiv (systems, data, ML)·22 Oct 20263.2
Harmful Responses Are Jailbreak Templates: Transferring Autoregressive Jailbreaks to Diffusion Language Models
Harmful responses work as templates to jailbreak diffusion LMs.