Search: “safety”

arXiv:2609.02082v1 Announce Type: cross Abstract: Visual modality enhances the capabilities of multimodal large language models (MLLMs) but also introduces a safety concern: a benign textual query may

arXiv:2608.22490v1 Announce Type: new Abstract: Safety alignment helps models adhere to human values, but it often reduces response utility. We ask a critical but understudied question: Does safety al

arXiv:2608.20053v1 Announce Type: new Abstract: The integration of Artificial Intelligence (AI) in safety-critical aviation systems presents significant challenges for certification and deployment. Av

arXiv:2608.11705v1 Announce Type: new Abstract: Aligned large language models (LLMs) are expected to exhibit safety behavior based on the content of the user request: they should refuse unsafe request

arXiv:2608.29936v1 Announce Type: new Abstract: Safety alignment of large language models (LLMs) degrades across languages, yet the internal mechanism driving this asymmetry remains poorly understood.

arXiv:2601.21297v2 Announce Type: replace Abstract: We introduce Deep QP Safety Filter, a fully data-driven safety layer for black-box dynamical systems. Our method learns a Quadratic-Program (QP) saf

arXiv:2608.23959v1 Announce Type: cross Abstract: Safety alignment in large language models (LLMs) remains brittle against a growing spectrum of attacks. Jailbreak attacks bypass safety mechanisms thr

arXiv:2609.22144v1 Announce Type: new Abstract: Preserving safety alignment during large language models fine-tuning is critical, however, recent studies have demonstrated that even benign fine-tuning

arXiv:2609.01210v1 Announce Type: cross Abstract: Safety benchmarks for large language models often assess the risk of a user query, although the outcome of question answering depends on whether the r

arXiv:2506.07356v3 Announce Type: replace Abstract: While Finetuning-as-a-Service (FaaS) enables customization of Large Language Models (LLMs) using user data, this service is vulnerable to safety deg

arXiv:2607.28685v1 Announce Type: new Abstract: Agent-safety benchmarks measure different behaviors, and their scores get quoted interchangeably as an agent's safety. We treat four of them (R-Judge, I

arXiv:2603.07445v2 Announce Type: replace Abstract: Large language models (LLMs) often require fine-tuning (FT) to perform well on downstream tasks, but FT can induce safety-alignment drift even when

arXiv:2604.23863v1 Announce Type: new Abstract: Autonomous systems are increasingly deployed in real-world environments, where they must achieve high performance while maintaining safety under state a

arXiv:2603.00338v2 Announce Type: replace Abstract: This paper presents a general end-to-end framework for constructing robust and reliable layered safety filters that can be leveraged to perform dyna

arXiv:2608.09095v1 Announce Type: new Abstract: Uncovering the internal mechanisms underlying the safety capabilities of large language models (LLMs) is crucial for developing trustworthy artificial i

arXiv:2604.00310v2 Announce Type: replace-cross Abstract: Multimodal large-language models (MLLMs) often experience degraded safety alignment when harmful queries exploit cross-modal interactions. Mod

arXiv:2610.01801v1 Announce Type: new Abstract: Can response safety be scored by cosine similarity to the mean embedding of known-safe responses? A recent sleeper-agent detector proposes exactly this

arXiv:2606.02562v1 Announce Type: cross Abstract: Autonomous robots that interact with people must make safe and efficient decisions under human-induced uncertainty, such as their preferences, goals,

arXiv:2510.20955v2 Announce Type: replace Abstract: Safe decision-making algorithms for control of mobile robots often require the existence of feedback to verify the safety of proposed actions. This

arXiv:2605.15239v1 Announce Type: new Abstract: Safety alignment often improves robustness to harmful queries at the cost of reasoning ability, a tradeoff known as the safety tax. A common cause is di

arXiv:2609.27312v1 Announce Type: cross Abstract: Robots deployed for competitive tasks must outmaneuver their opponents without sacrificing safety. Existing approaches, including safe reinforcement l

arXiv:2608.18628v1 Announce Type: cross Abstract: Aligned vision-language models (VLMs) are designed to balance grounded visual reasoning with safe generation behavior. However, we observe a striking

arXiv:2608.12025v1 Announce Type: cross Abstract: Medical devices are becoming more software-intensive, connected, and AI-enabled. Their development requires risk-management evidence aligned with ISO

arXiv:2605.12974v1 Announce Type: new Abstract: In this work, we study how to ensure probabilistic safety for nonlinear systems under distributional ambiguity. Our approach builds on a backup-based sa

arXiv:2610.07532v1 Announce Type: cross Abstract: LLMs have advanced rapidly, raising growing concerns about their safety. Recent work has proposed approaches to detect and defend against attacks incl

arXiv:2510.18082v3 Announce Type: replace Abstract: Recent advances in reinforcement learning (RL) enable its use on increasingly complex tasks, but the lack of formal safety guarantees still limits i

arXiv:2609.04482v1 Announce Type: new Abstract: Safety alignment is usually posed as a topic-level question: is this subject harmful? Deployments ask a narrower one. A civics tutor and a public-sector

arXiv:2608.18778v1 Announce Type: cross Abstract: Despite rapid technological advances, the societal acceptability of autonomous vehicles (AVs) remains limited by psychological barriers that extend be

arXiv:2608.18131v1 Announce Type: new Abstract: Current safety alignment training for Large Language Models (LLMs) are heavily English-centric. When such safety filters fail for non-English languages,

arXiv:2506.08473v4 Announce Type: replace Abstract: Fine-tuning large language models (LLMs) improves performance but introduces critical safety vulnerabilities: even minimal harmful data can severely

arXiv:2503.11832v5 Announce Type: replace Abstract: Recent vision language models (VLMs) have made remarkable strides in generative modeling with multimodal inputs, particularly text and images. Howev

arXiv:2603.13502v2 Announce Type: replace Abstract: Wirelessly-connected robotic systems empower robots with real-time intelligence by leveraging remote computing resources for decision-making. Howeve

arXiv:2604.13325v1 Announce Type: new Abstract: Safety filters provide a practical approach for enforcing safety constraints in autonomous systems. While learning-based tools scale to high-dimensional

arXiv:2610.09000v1 Announce Type: cross Abstract: As small language models (SLMs) are increasingly deployed on resource-constrained and on-device platforms, including as components of agentic systems,

arXiv:2610.07774v1 Announce Type: new Abstract: Vision-language models (VLMs) face compositional safety risks where harmful intent emerges from the interaction between visual and textual inputs. As mi

arXiv:2610.04297v1 Announce Type: cross Abstract: Latent safety filters certify safety on a learned low-dimensional representation of the state, enabling constraints that resist analytic description.

arXiv:2609.36434v1 Announce Type: new Abstract: Context tokens in a transformer-based language model can be absorbed into the model's weights as a multiplicative operator. We study this operator in th

arXiv:2609.18328v1 Announce Type: new Abstract: Recent advances in LLMs and VLMs have enabled safety systems to reason beyond simple risk patterns toward more contextual and semantic safety concerns.

arXiv:2609.06934v1 Announce Type: new Abstract: Post-hoc safety training (RLHF, DPO) is the dominant way to align large language models, yet jailbreaks (Zou et al., 2023b), fine-tuning attacks (Qi et

arXiv:2609.01455v1 Announce Type: cross Abstract: Benign fine-tuning severely weakens the safety alignment of large language models (LLMs), so we study why refusal behavior is so fragile. While prior

arXiv:2608.20820v1 Announce Type: new Abstract: Large language models (LLMs) are vulnerable to multi-turn jailbreak attacks that progressively manipulate conversation context. Existing certified robus

arXiv:2607.07420v1 Announce Type: new Abstract: Safety for generalist robots is usually discussed in terms of motion or dialogue. We argue a third question is missing: should the robot take its first

arXiv:2505.18979v2 Announce Type: replace Abstract: Text-to-image (T2I) models can generate not-safe-for-work (NSFW) content, motivating multi-stage safety pipelines with both text and image filters.

arXiv:2610.12420v1 Announce Type: new Abstract: Reinforcement learning in safety-critical domains requires maximizing task performance while strictly adhering to safety constraints. Existing safe rein

arXiv:2609.18515v1 Announce Type: new Abstract: Safety alignment teaches large language models (LLMs) to recognize harmful requests and reject risky instructions. Yet aligned models can fail when harm

arXiv:2609.04021v1 Announce Type: new Abstract: Evaluating large language models (LLMs) in safety-critical, physics-governed environments requires more than accuracy-based metrics, because predictions