PulseAugur
EN
LIVE 05:43:21

New CLEAR framework enhances LLM safety without utility loss

By PulseAugur Editorial · [2 sources] ·

Researchers have developed CLEAR, a new framework for improving the safety of large language models without sacrificing their utility. CLEAR uses a hidden-state gate to dynamically adjust a safety adapter, allowing it to reduce harmful outputs while preserving performance on benign inputs. Experiments show CLEAR significantly reduces harmful response rates on benchmarks like HarmBench, while maintaining or even improving performance on tasks such as GSM8K, particularly when applied to models like LLaMA-3-8B-Instruct. AI

IMPACT This method could improve the safety-utility trade-off in LLM alignment, leading to more robust and reliable AI systems.

RANK_REASON The cluster contains a research paper detailing a new method for LLM safety alignment. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Hugging Face Daily Papers →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New CLEAR framework enhances LLM safety without utility loss

COVERAGE [2]

  1. arXiv cs.AI TIER_1 English(EN) · Chengxiao Wang, Enyi Jiang, Xiaojing Liao, Sanmi Koyejo ·

    CLEAR: Continuous Latent Adapter Routing for Utility-Preserving LLM Safety Alignment

    arXiv:2608.21278v1 Announce Type: new Abstract: Improving the safety of large language models (LLMs) often comes at the expense of utility, as globally applied safety tuning may affect model responses to both harmful and benign inputs. We propose \textbf{C}ontinuous \textbf{L}at\…

  2. Hugging Face Daily Papers TIER_1 English(EN) ·

    CLEAR: Continuous Latent Adapter Routing for Utility-Preserving LLM Safety Alignment

    CLEAR uses a hidden-state gate to continuously modulate a safety low-rank adapter, improving LLM safety while preserving utility on benign inputs.

RELATED ENTITIES

RELATED TOPICS