一种用于大语言模型护栏的双假设推理框架
A Dual-Hypothesis Reasoning Framework for LLM Guardrails
浏览论文内容
中文总结 AI 辅助
该研究提出ARBITER大语言模型护栏框架，采用双假设推理和多组件监督微调方法，用成本效益高的策略生成推理痕迹及基于LoRA微调，性能超现有方法，还能为不安全决策提供解释，在安全审核基准实验中表现出色。
中文摘要 AI 辅助
我们提出了ARBITER，一种新颖的大语言模型护栏框架，它引入了两个关键思想：（i）双假设推理，一种用于大语言模型护栏的推理方法，在做出安全决策之前明确考虑提示的安全和不安全解释；（ii）多组件监督微调（MC-SFT），一种基于推理的护栏的结构化训练损失，将大语言模型输出分解为逻辑组件，并根据其重要性加权。现有基于推理的护栏通常依赖于昂贵的程序。相比之下，ARBITER使用具有成本效益的自我生成策略来生成推理痕迹，并基于LoRA进行参数高效微调，同时性能优于这些昂贵方法。此外，ARBITER为不安全决策提供可靠的证据短语解释。在三个安全审核基准上的实验表明，ARBITER优于现有的基于推理和非推理的护栏基线，在域外评估中有明显提升。
英文摘要
We propose ARBITER, a novel LLM guardrail framework that introduces two key ideas: (i) dual-hypothesis reasoning, a reasoning method for LLM guardrails that explicitly considers both safe and unsafe interpretations of a prompt before making a safety decision, and (ii) multi-component supervised fine-tuning (MC-SFT), a structured training loss for reasoning-based guardrails that decomposes LLM outputs into logical components and weights them according to their importance. Existing reasoning-based guardrails often rely on expensive procedures, such as generating reasoning traces using larger or closed-source teacher models and applying full-parameter fine-tuning. In contrast, ARBITER uses a cost-effective self-generation strategy for reasoning traces and LoRA-based parameter-efficient fine-tuning while still achieving better performance than these expensive approaches. Additionally, ARBITER provides faithful evidence-phrase explanations for unsafe decisions, enabling a more transparent and interpretable guardrail method. Experiments on three safety moderation benchmarks show that ARBITER outperforms existing reasoning-based and non-reasoning guardrail baselines, with clear gains in out-of-domain evaluations.
发表机构
- University of Arizona（亚利桑那大学）
机构由 AI 辅助整理，请以论文原文为准。
