文本到视频模型的推理时概念抑制和以视频为中心的评估
Inference-Time Concept Suppression and Video-Centric Evaluation for Text-to-Video Models
查看机构详情
- University of Science and Technology of China（中国科学技术大学）
机构由 AI 辅助整理，请以论文原文为准。
浏览论文内容
中文总结 AI 辅助
研究文本到视频模型中可控去除目标概念的问题，提出无训练推理时框架SIRUS，通过定位目标相关提示证据抑制目标表达，引入视频导向评估框架，在多个概念上实验表现出色，能在不同骨干上泛化。
中文摘要 AI 辅助
文本到视频（T2V）生成器能合成逼真且时间连贯的视频，但可控地从生成器中去除目标概念仍很困难。与文本到图像概念擦除不同，T2V的遗忘必须在保留非目标主体、动作、场景和时间结构的同时抑制可能跨帧持续的目标概念。我们提出了SIRUS，一个用于概念级T2V遗忘的无训练推理时框架。给定目标概念的文本别名，SIRUS定位与目标相关的提示证据并在采样期间抑制目标表达，而不更新文本编码器或去噪网络。我们还引入了一个面向视频的T2V遗忘评估框架，使用视频级失败标准、帧级残差统计、配对保留分析、基于VBench的质量诊断和部署开销测量，分别测量目标遗忘、非目标保留、视频质量、越狱鲁棒性和效率。在CogVideoX上的五个安全、对象和风格概念上，SIRUS平均遗忘成功率达到70.4%，平均帧命中率达到25.7%，而VideoEraser分别为44.4%/47.2%，同时将平均VBench质量下降从-0.043降低到-0.016，在所有评估的基线中实现了最强的遗忘-质量权衡。在Wan2.2上的迁移实验进一步表明SIRUS能在现代T2V骨干上进行泛化。
英文摘要
Text-to-video (T2V) generators can synthesize realistic and temporally coherent videos, but controllably removing a target concept from a generator remains difficult. Unlike text-to-image concept erasure, T2V unlearning must suppress a target concept that may persist across frames while preserving non-target subjects, actions, scenes, and temporal structure. We propose \textbf{SIRUS}, a training-free inference-time framework for concept-level T2V unlearning. Given textual aliases of a target concept, SIRUS localizes target-related prompt evidence and suppresses target expression during sampling, without updating the text encoder or denoising network. We further introduce a video-oriented evaluation framework for T2V unlearning that separately measures target forgetting, non-target preservation, video quality, jailbreak robustness, and efficiency, using video-level failure criteria, frame-level residue statistics, paired preservation analysis, VBench-based quality diagnostics, and deployment overhead measurement. Across five safety, object, and style concepts on CogVideoX, SIRUS reaches 70.4\% average forgetting success and 25.7\% average frame hit, compared with 44.4\% / 47.2\% for VideoEraser, while reducing the average VBench quality drop from -0.043 to -0.016, yielding the strongest forgetting-quality trade-off among fully evaluated baselines. Transfer experiments on Wan2.2 further suggest that SIRUS generalizes across modern T2V backbones.
