The paper investigates how diffusion models, even after fine‑tuning to unlearn harmful concepts, still retain a linear residual of those concepts in token embeddings. It introduces SubAttack, a jailbreak method that extracts this residual by learning orthogonal attack tokens composed of interpretable textual elements, and SubDefense, a plug‑and‑play defense that projects out the same subspace to suppress the residual. Experiments show SubAttack outperforms prior attacks in power and transferability, while SubDefense offers stronger robustness and better preservation of safe generation quality.
arXiv:2606. 04027v1 Announce Type: cross Abstract: Diffusion large language models (dLLMs) generate text by iteratively denoising partially masked sequences under bidirectional context, exposing a safety surface distinct from autoregressive LLMs.
arXiv:2512. 14751v3 Announce Type: replace-cross Abstract: Finetuning pretrained large language models (LLMs) has become the standard paradigm for developing downstream applications.
The paper introduces Dynamic Semantic Steering (DSS), a training‑free, inference‑time defense for robust concept erasure in text‑to‑image diffusion models. DSS models local semantic neighborhoods geometrically, automatically identifies benign semantic anchors, and applies context‑aware, constrained feature correction using cross‑attention signals. Experiments show DSS achieves an average erasure rate of 91.0%, outperforming prior defenses while reducing semantic drift and preserving generation fidelity.
arXiv:2604. 23130v2 Announce Type: replace-cross Abstract: Jailbreak attacks expose a persistent failure mode in safety-aligned LLMs: models can be pushed into harmful behavior, but the internal representations enabling this shift remain poorly localized.
arXiv:2607. 23496v1 Announce Type: new Abstract: Safety-aligned large language models are trained to refuse harmful requests, yet embedding the same requests in particular scenarios can bypass their safeguards.
arXiv:2607. 17779v1 Announce Type: new Abstract: Text-to-Image (T2I) generative models have achieved remarkable progress in synthesizing high-quality visual content, yet they remain vulnerable to adversarial misuse, particularly in generating Not-Safe-For-Work (NSFW) images.
The paper proposes a framework that explains why jailbreak attacks succeed against diffusion-based large language models (dLLMs) by viewing safety alignment as shaping a denoising energy landscape. It identifies two attack strategies—obscuring the query’s safety disposition at initialization or forcing the denoising path across an energy barrier mid‑trajectory—and introduces three training‑free detection signals that monitor initial safety disposition and kinetic energy in complementary subspaces. Experiments on several dense and sparse dLLMs show that these signals complement each other, and any attack that evades detection also fails to produce harmful content.
arXiv:2609.26185v1 Announce Type: cross Abstract: Large Language Models (LLMs) demonstrate impressive capabilities across many applications but remain vulnerable to jailbreak attacks, which elicit ha...
arXiv:2606. 25182v1 Announce Type: cross Abstract: Jailbreak attacks reveal a persistent weakness in aligned Large Language Models: carefully crafted prompts can elicit policy-violating responses despite safety training.
The paper introduces AVCE, a method for verifiable concept erasure in diffusion models that audits latent representations to identify vulnerable directions and then edits cross‑attention and self‑attention projections in closed form. By fine‑tuning with pathway‑level auditing losses and orthogonal gradient projection, AVCE reduces attack success rates by 5.07× and improves auditing scores by 3.84× compared to the best baseline while maintaining generation quality.
Safety-aligned large language models are trained to refuse harmful requests, yet embedding the same requests in particular scenarios can bypass their safeguards. Existing red-teaming methods empirically identify effective scenarios through observed attack outcomes, but why particular scenarios weaken refusal remains mechanistically unclear.
arXiv:2510.17904v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are widely used because they process structures, syntax and code well, but this same ability also makes them par...