{"ID":2861584,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.02194","arxiv_id":"2510.02194","title":"UpSafe$^\\circ$C: Upcycling for Controllable Safety in Large Language Models","abstract":"Large Language Models (LLMs) have achieved remarkable progress across a wide range of tasks, but remain vulnerable to safety risks such as harmful content generation and jailbreak attacks. Existing safety techniques -- including external guardrails, inference-time guidance, and post-training alignment -- each face limitations in balancing safety, utility, and controllability. In this work, we propose UpSafe$^\\circ$C, a unified framework for enhancing LLM safety through safety-aware upcycling. Our approach first identifies safety-critical layers and upcycles them into a sparse Mixture-of-Experts (MoE) structure, where the router acts as a soft guardrail that selectively activates original MLPs and added safety experts. We further introduce a two-stage SFT strategy to strengthen safety discrimination while preserving general capabilities. To enable flexible control at inference time, we introduce a safety temperature mechanism, allowing dynamic adjustment of the trade-off between safety and utility. Experiments across multiple benchmarks, base model, and model scales demonstrate that UpSafe$^\\circ$C achieves robust safety improvements against harmful and jailbreak inputs, while maintaining competitive performance on general tasks. Moreover, analysis shows that safety temperature provides fine-grained inference-time control that achieves the Pareto-optimal frontier between utility and safety. Our results highlight a new direction for LLM safety: moving from static alignment toward dynamic, modular, and inference-aware control.","short_abstract":"Large Language Models (LLMs) have achieved remarkable progress across a wide range of tasks, but remain vulnerable to safety risks such as harmful content generation and jailbreak attacks. Existing safety techniques -- including external guardrails, inference-time guidance, and post-training alignment -- each face limi...","url_abs":"https://arxiv.org/abs/2510.02194","url_pdf":"https://arxiv.org/pdf/2510.02194v1","authors":"[\"Yuhao Sun\",\"Zhuoer Xu\",\"Shiwen Cui\",\"Kun Yang\",\"Lingyun Yu\",\"Yongdong Zhang\",\"Hongtao Xie\"]","published":"2025-10-02T16:43:33Z","proceeding":"cs.AI","tasks":"[\"cs.AI\",\"cs.CR\",\"cs.LG\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}