TEAR：面向文本到视频模型的时序感知自动化红队测试
TEAR: Temporal-aware Automated Red-teaming for Text-to-Video Models
对做视频生成的人，这篇给出了一个可复用的安全评测范式：把「时序动态」当作攻击面，用两阶段训练加在线偏好学习自动搜索文本无害但触发违规视频的提示词，可直接用来给自己的 T2V 模型做红队体检或构造安全对齐数据。
原文依据 craft textually innocuous prompts that exploit temporal dynamics to elicit policy-violating video output
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论文本到视频（T2V）模型能够合成高质量、时序连贯的动态视频内容，但这种多样化的生成能力也天然引入了严峻的安全挑战。现有的安全评估方法主要针对静态图像与文本生成，不足以刻画视频生成中复杂的时序动态。为此，我们提出了一个时序感知的自动化红队测试框架，命名为 TEAR，该框架旨在专门揭示与 T2V 模型动态时序编排相关的安全风险。TEAR 采用一个经两阶段方法优化的时序感知测试生成器：先进行生成器初始训练，再进行时序感知的在线偏好学习，从而构造出在文本上无害、却利用时序动态诱使模型输出违反政策的视频的提示词。同时引入一个精炼模型，循环往复地提升提示词的隐蔽性与对抗有效性。大量实验评估表明，TEAR 在开源与商用 T2V 系统上均有效，攻击成功率超过 80%，相较此前最佳结果 57% 有显著提升。
英文原文摘要 1156 字符
Text-to-Video (T2V) models are capable of synthesizing high-quality, temporally coherent dynamic video content, but the diverse generation also inherently introduces critical safety challenges. Existing safety evaluation methods, which focus on static image and text generation, are insufficient to capture the complex temporal dynamics in video generation. To address this, we propose a TEmporal-aware Automated Red-teaming framework, named TEAR, an automated framework designed to uncover safety risks specifically linked to the dynamic temporal sequencing of T2V models. TEAR employs a temporal-aware test generator optimized via a two-stage approach: initial generator training and temporal-aware online preference learning, to craft textually innocuous prompts that exploit temporal dynamics to elicit policy-violating video output. And a refine model is adopted to improve the prompt stealthiness and adversarial effectiveness cyclically. Extensive experimental evaluation demonstrates the effectiveness of TEAR across open-source and commercial T2V systems with an over 80% attack success rate, a significant boost from the prior best result of 57%.