The paper analyzes that while multimodal large language models (MLLMs) offer superior semantic understanding for image generation, this enhanced capability significantly increases safety risks, partic…

The paper proposes AHV-D&S, a novel training-free inference-time safeguard that detects and suppresses risky content in Diffusion Transformers (DiTs) by quantifying token sensitivity across attention…

SafeSteer proposes a localized on-policy distillation method that restricts safety alignment to specific safety tokens, thereby achieving strong safety performance with minimal degradation to general…

This paper introduces TwoHamsters, a new benchmark that rigorously tests Multi-Concept Compositional Unsafety (MCCU) in text-to-image models, demonstrating that current state-of-the-art models and saf…

GLiNER Guard (GLiGuard) introduces a unified, efficient encoder family that simultaneously performs safety classification and PII detection in a single forward pass, offering a practical, low-cost alt…

RiskFlow is a novel framework that generates realistic and safety-critical multi-agent traffic scenarios by reformulating trajectory generation as a single-pass transport problem in the action space.

The paper introduces Safety Bottleneck Regularization (SBR), a novel defense mechanism that anchors LLM safety by constraining the unembedding layer, effectively preventing harmful fine-tuning (HFT) e…

The paper shows that safety failures in low-resource languages are due to a failure in the model's safety decision calibration, not a lack of underlying knowledge, and proposes a recalibration method…

The paper introduces NeWTral, a framework that restores safety alignment to specialized LLM adapters without sacrificing their domain-specific knowledge, achieving a significant reduction in attack su…

This paper introduces the concept of Safety Geometry Collapse, demonstrating that multimodal inputs degrade the safety separation of LLMs, and proposes ReGap, a training-free method that adaptively co…

The paper introduces Cross-Model Neuron Transfer (CNT), a post-hoc method that efficiently transfers safety-oriented functionalities between different large language models by transferring minimal sub…

GLiGuard introduces a compact, schema-conditioned bidirectional encoder that achieves state-of-the-art performance in LLM content moderation across multiple safety dimensions while drastically reducin…

The paper introduces DLM-SWAI, a training-free method that effectively steers diffusion language models (DLMs) toward desired textual styles or properties by biasing the token distribution at each den…

The paper demonstrates that fine-tuning safety guard models on benign data can catastrophically collapse their safety alignment, proposing Fisher-Weighted Safety Subspace Regularization (FW-SSR) to ac…

The paper introduces SafeRedirect, a system-level defense that prevents frontier LLMs from generating harmful content during legitimate tasks that structurally require it, significantly reducing unsaf…

The paper proposes a novel zeroth-order optimization framework to enhance the robustness of LLM safety alignment, showing that few refinement steps can significantly improve safety while maintaining u…

MASCing is a novel framework that enables flexible, non-retraining reconfiguration of Mixture-of-Experts (MoE) models for specific safety objectives by applying activation steering masks to control ex…

The paper introduces PaSBench-Video, a comprehensive streaming video benchmark designed to rigorously test multimodal LLMs' ability to issue proactive safety warnings, finding that current models stru…

The paper introduces RefusalGuard, a novel fine-tuning framework that preserves the geometric structure of safety-relevant representations in LLMs, thereby mitigating the degradation of refusal behavi…

The paper argues that shallow safety alignment in LLMs is due to autoregressive consistency, a mechanism that allows small harmful inputs to redirect the model's generation to unsafe outputs, necessit…