Zhiwei Jiang
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Harmful Responses Are Jailbreak Templates: Transferring Autoregressive Jailbreaks to Diffusion Language Models
According to the abstract, a preprint presents RMT-Jailbreak, a method that masks tokens of a harmful response to prompt diffusion language models into reconstructing harmful content. The authors say it carries over existing jailbreak attacks from conventional models to diffusion models.