让内容审核能说出敏感行为在哪、谁参与、发生了什么。
SenBen: Sensitive Scene Graphs for Explainable Content Moderation
- 用场景图+敏感标签构建首个大规模可解释审核数据集
- 模型在敏感内容召回率上提升6.4个百分点，速度超快内存少
- 适合需要透明化审核的平台和研究者
内容审核系统虽能判断图像安全与否，但缺乏空间定位与可解释性，无法说明检测到的敏感行为涉及何人、何处及具体内容。本文提出首个大规模敏感内容场景图基准——SenBen，涵盖157部电影中的13,999帧，标注了25类物体、28种属性（含疼痛、恐惧、攻击、痛苦等情感状态）及14种谓词，并设置16个敏感标签覆盖5大类别。通过多任务蒸馏方法，将前沿视觉语言模型压缩为仅241M参数的小型学生模型，采用基于后缀的对象身份识别、词汇感知召回损失（VAR Loss）及异步损失解耦的Query2Label标签头，有效缓解自回归生成中的词汇不平衡问题，使模型在SenBen召回率上比标准交叉熵训练高出6.4个百分点。在具象化场景图评估中，该学生模型性能超越所有评测的VLMs（除Gemini系列）及所有商用安全API，且在物体检测与图文生成任务上均达到最高得分，推理速度达7.6倍更快，显存占用减少16倍。
原文摘要 · Abstract (English)
Content moderation systems classify images as safe or unsafe but lack spatial grounding and interpretability: they cannot explain what sensitive behavior was detected, who is involved, or where it occurs. We introduce the Sensitive Benchmark (SenBen), the first large-scale scene graph benchmark for sensitive content, comprising 13,999 frames from 157 movies annotated with Visual Genome-style scene graphs (25 object classes, 28 attributes including affective states such as pain, fear, aggression, and distress, 14 predicates) and 16 sensitivity tags across 5 categories. We distill a frontier VLM into a compact 241M student model using a multi-task recipe that addresses vocabulary imbalance in autoregressive scene graph generation through suffix-based object identity, Vocabulary-Aware Recall (VAR) Loss, and a decoupled Query2Label tag head with asymmetric loss, yielding a +6.4 percentage point improvement in SenBen Recall over standard cross-entropy training. On grounded scene graph metrics, our student model outperforms all evaluated VLMs except Gemini models and all commercial safety APIs, while achieving the highest object detection and captioning scores across all models, at $7.6\times$ faster inference and $16\times$ less GPU memory.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。