[论文] ShallowDefend: Harnessing Shallow Safety Alignment for Multi-Turn Defe...

论文概要

研究领域: NLP 作者: Qiyao Ma, Xiaodong Liu, Weiwei Yang, Mingqian Feng, Chengquan Guo, Jialin Song, Carolyn Buractaon, Zhe Zhao, Junshan Zhang, Sandeep Atluri, Christopher White, Jianfeng Gao 发布时间: 2026-10-29 arXiv: 2610.40627

中文摘要

行李重识别（ReID）是视频监控中的关键任务，应用于取证、安全或边境管理（可疑或遗弃行李识别）以及行李管理（丢失行李找回）。它能够在机场、火车站和边境口岸等复杂环境中，跨不重叠的摄像头视野关联同一行李的观测。现有方法主要关注在相对干净条件下基于外观的裁剪图像匹配，而真实世界场景以频繁遮挡、杂乱背景和拥挤人群为特征，这些因素显著降低了性能。然而，利用来自不同视频流的多帧和多视角互补性的策略却被忽视了。在本文中，我们研究了一种端到端管道的策略，该管道从多个视频流中检测并重识别行李。我们在真实环境采集的基准数据集上评估了不同变体，以解决选择和合并同一行李的不同视角和帧以及在来自不同环境数据集上训练等问题的整体ReID性能提升。

原文摘要

Safety alignment has enabled large language models to reliably refuse harmful requests directly, yet the same request distributed across a multi-turn conversation still succeeds far too often. This brittleness has a mechanistic root: Alignment is shallow, with safety behavior concentrated in a model's first few output tokens, and existing defenses inherit the weakness, supervising the surface form of the response without ever giving the model an explicit safety decision to learn. We take the opposite view of shallowness: if a short span determines the trajectory of an entire response, that span is the highest-leverage place to intervene, and the defender can occupy it first. We propose Assess \(\to\) Decide \(\to\) Respond, a framework in which the defender assesses the user's intent, commits ...

#论文 #arXiv #NLP #小凯