LatentGuard: Efficient and Inspectable Latent Reasoning for LLM Safeguards

AI 中文解读

核心亮点：这项研究让AI安全审查既能“想得快”又能“看得懂”，在保持高效的同时解决了此前技术难以兼顾的透明性问题。通俗解读：以前给AI装“安全护栏”，需要它每句话都解释理由，虽然安全但特别费算力。后来有人想出让AI“心里想但不写出来”的办法，省了成本，可又让人担心它“黑箱操作”。这次的新方案像给AI配了个“秘密笔记本”——平时直接给答案，只有需要检查时才翻开笔记，既快又能随时抽查。实际影响：以后你用AI写邮件、问医疗建议时，背后的安全审查会更快，响应不卡顿，同时平台能随时验证AI没“胡思乱想”。对企业来说，部署成本大幅降低，意味着更多免费AI工具可能出现，而普通用户也能更放心地使用AI处理敏感问题，不用担心它“暗藏风险”。

Reasoning-based guard models improve LLM safeguards, but decoding explicit rationales for every interaction makes them costly to deploy. Although latent-reasoning methods reduce token generation by moving reasoning into continuous states, they remain underexplored for safety moderation and lack an inspection interface for deployment. In this paper, we propose LatentGuard, an efficient and inspectable safeguard framework that brings continuous latent reasoning to guard models. LatentGuard uses a staged curriculum to progressively compress task-aligned textual rationales into compact latent states, enabling safety verdicts to be predicted directly from continuous representations. To preserve inspectability, an isolated auxiliary decoder generates compact audit artifacts on demand, keeping rationale generation off the standard inference path. Experiments show that LatentGuard-8B improves mean weighted F1 from 83.95 to 84.91 over GuardReasoner-8B, while reducing critical-path reasoning cost from 268.56 generated rationale tokens to 1.60 latent reasoning tokens. Its audit decoder achieves an audit utility score of 85.75, demonstrating an efficient and inspectable path toward deployable LLM safeguards.