ConsisGuard：在LLM护栏中对齐安全审议与策略执行
ConsisGuard: Aligning Safety Deliberation with Policy Enforcement in LLM Guardrails
AI总结：
提出ConsisGuard框架，通过策略到决策轨迹蒸馏和功能耦合对齐，解决基于推理的LLM护栏中审议与执行之间的不一致问题，提升安全检测性能并减少策略执行失败。
摘要：
基于推理的LLM护栏通过在做出最终决策前生成明确理由来改进安全审核。然而，它们的理由并不总是导致忠实的执行：模型可能在推理中识别出有害意图，但仍然预测安全标签，或者在没有策略依据的情况下发布不安全决策。我们将这种安全关键性失败模式识别为审议到执行的差距。与一般的思维链忠实性不同，护栏可靠性要求策略执行一致性：生成的推理应基于安全策略，最终决策应由该推理蕴含。我们提出ConsisGuard，一个用于基于推理的LLM护栏的一致性感知框架。ConsisGuard执行策略到决策轨迹蒸馏和功能耦合对齐，对齐安全审议与决策执行之间的内部耦合。在提示和响应有害性检测基准上的实验表明，ConsisGuard在减少策略执行失败的同时提高了检测性能。这些结果表明，可靠的基于推理的护栏需要准确忠实地执行安全策略。
英文摘要：
Reasoning-based LLM guardrails improve safety moderation by generating explicit rationales before issuing final decisions. However, their rationales do not always lead to faithful enforcement: a model may recognize a harmful intent in its reasoning but still predict a safe label, or issue an unsafe decision without policy-grounded justification. We identify this safety-critical failure mode as the deliberation-to-enforcement gap. Unlike general chain-of-thought faithfulness, guardrail reliability requires policy execution consistency: the generated reasoning should be grounded in the safety policy, and the final decision should be entailed by that reasoning. We propose ConsisGuard, a consistency-aware framework for reasoning-based LLM guardrails. ConsisGuard performs Policy-to-Decision Trajectory Distillation and Functional Coupling Alignment, aligning the internal coupling between safety deliberation and decision enforcement. Experiments on prompt and response harmfulness detection benchmarks show that ConsisGuard improves detection performance while reducing policy execution failures. These results suggest that reliable reasoning-based guardrails require accurate faithful execution of safety policies.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
