Diffusion LLMs as Targets and Adversaries: Mechanistic Safety Exploits

Name: Diffusion LLMs as Targets and Adversaries: Mechanistic Safety Exploits
Uploaded: 2026-08-11T01:38:46.458944+00:00
Duration: 1568 s
Description: The episode discusses a paper on diffusion LLMs, showing safety alignment is sparse and transferable. Hosts explain how pruning a few safety neurons raises attack success dramatically, and how an offline black-box jailbreak framework uses diffusion models to generate prompts that transfer to other models, highlighting structural fragility in current alignment.

The episode discusses a paper on diffusion LLMs, showing safety alignment is sparse and transferable. Hosts explain how pruning a few safety neurons raises attack success dramatically, and how an offline black-box jailbreak framework uses diffusion models to generate prompts that transfer to other models, highlighting structural fragility in current alignment.

Open the full episode