The paper examines the safety of small language models (SLMs) deployed on resource‑constrained, on‑device platforms, focusing on the LLaMA‑2‑7B‑Chat model. It shows that safety‑sensitive behavior is concentrated in a sparse subset of parameters, particularly the MLP down_proj layer, and that altering only 0.19% of these weights can significantly affect safety metrics while leaving overall accuracy largely unchanged. The study suggests that targeted fault analysis and selective integrity protection could improve safety for on‑device, agentic language models.
arXiv:2607. 01859v1 Announce Type: new Abstract: Safety training for large language models (LLMs) is conducted predominantly in English, leaving uncertain how well safety mechanisms generalize to low-resource languages and mixed-language code-switching.
arXiv:2608.29936v1 Announce Type: cross Abstract: Safety alignment of large language models (LLMs) degrades across languages, yet the internal mechanism driving this asymmetry remains poorly understo...
Safety training for large language models (LLMs) is conducted predominantly in English, leaving uncertain how well safety mechanisms generalize to low-resource languages and mixed-language code-switching. We show that this creates an epistemic gap in which models confidently generate harmful responses for inputs that fall outside the distribution of their safety training.
The paper investigates whether large language models (LLMs) can internally detect harmful content, bypassing external guardrails that add latency and computational cost. By extracting activations from LLaMA‑3.1‑8B and training lightweight MLP probes, the authors achieve high F1 scores (99%, 83%, and 84%) on WildJailbreak, Beavertails, and AEGIS 2.0 benchmarks, rivaling much larger guard models while reducing overhead. This suggests that internal state monitoring can provide efficient safety checks for resource‑constrained, time‑critical deployments.
arXiv:2607. 11475v1 Announce Type: new Abstract: Safety alignment in large language models can be fragile under fine-tuning, as even benign task adaptation may increase harmful compliance.
arXiv:2605. 17173v2 Announce Type: replace-cross Abstract: Large language models exhibit safety degradation in non-English languages.
The paper investigates whether existing AI safety benchmarks, designed for large language models, are suitable for evaluating small language models (SLMs). By testing five benchmark suites on 26 open‑source SLMs with a unified scoring rubric, the authors find that ambiguous judgments dominate, especially for complex prompts and certain architectures. This ambiguity, linked to factors like lexical density and output perplexity, undermines the reliability of aggregate leaderboards and reveals a confound between model capability and perceived safety.
arXiv:2604. 07223v2 Announce Type: replace-cross Abstract: As large language models (LLMs) evolve from static chatbots into autonomous agents, the primary vulnerability surface shifts from final outputs to intermediate execution traces.
arXiv:2602. 16835v2 Announce Type: replace-cross Abstract: Safety alignment is essential for the responsible deployment of Large Language Models (LLMs).
arXiv:2501. 14940v4 Announce Type: replace-cross Abstract: Aligning large language models (LLMs) with human values is essential for their safe deployment and widespread adoption.
The paper introduces MMSAFE, a multi-layer framework designed to identify safety-degrading data in multilingual large language models. It shows that safety signals are distributed across multiple layers and only partially shared across languages, unlike the single-layer assumption used in monolingual settings. Experiments demonstrate that MMSAFE reduces harmful-response rates by 60% compared to random filtering and outperforms the best single-layer baseline across various models, languages, and safety benchmarks.
arXiv:2606. 09388v1 Announce Type: new Abstract: Deploying safe large language models (LLMs) on resource-constrained edge devices presents a critical challenge: while dual-model systems combining LLMs with guard models provide effective safety guarantees, their substantial memory and computational demands make them prohibitively expensive for on-device deployment.