QwenSafe: 通过偏好对齐的视觉语言模型实现多模态内容评级描述识别
QwenSafe: Multimodal Content Rating Description Identification via Preference-Aligned VLMs
AI总结：
本文提出QwenSafe，一种通过联合推理应用元数据和截图来自动识别苹果定义的内容评级描述（CRDs）的视觉语言模型，通过引入metadata2CRD数据构建管道和直接偏好优化（DPO）提升模型预测准确性，实验结果显示QwenSafe在二元CRD分类中显著优于现有模型。
摘要：
移动应用市场要求开发者披露标准化的内容评级描述（CRDs）以告知用户潜在敏感或受限制的内容。确保这些披露的准确性和一致性仍然具有挑战性，因为应用内容的多模态性质跨越了文本描述和视觉界面。在本文中，我们提出了QwenSafe，一种视觉语言模型（VLM），旨在通过联合推理应用元数据和截图自动识别苹果定义的CRDs。为了使该任务能够扩展训练，我们引入了metadata2CRD数据构建管道，通过结合应用描述、截图和正式描述定义来合成描述对齐的问题-答案对。我们通过监督微调后直接偏好优化（DPO）调整Qwen3-VL-8B，以使模型预测与视觉和文本模态的描述特定证据和解释对齐。我们在12个苹果定义的内容评级描述上评估QwenSafe，并将其与最先进的视觉语言模型进行比较，包括Qwen3-VL、LLaVA-1.6和Gemini-2.5-Flash。QwenSafe在二元CRD分类中始终优于所有基线模型，分别在正类召回率上实现了111.8%、36.1%和2.1%的提升。我们的结果表明，描述意识的多模态对齐显著提高了自动化内容分类，并突显了视觉语言模型在支持移动应用市场中可扩展和一致的内容评级方面的潜力。
英文摘要：
Mobile app marketplaces require developers to disclose standardized content rating descriptors (CRDs) to inform users about potentially sensitive or restricted content. Ensuring the accuracy and consistency of these disclosures remains challenging due to the multimodal nature of app content, which spans textual descriptions and visual interfaces. In this paper, we present QwenSafe, a Vision-Language Model (VLM) designed to automatically identify the presence of Apple-defined CRDs by jointly reasoning over app metadata and screenshots. To enable scalable training for this task, we introduce metadata2CRD, a data-construction pipeline that synthesizes descriptor-aligned question-answer pairs by combining app descriptions, screenshots, and formal descriptor definitions. We adapt Qwen3-VL-8B using supervised fine-tuning followed by Direct Preference Optimization (DPO) to align model predictions with descriptor-specific evidence and explanations across visual and textual modalities. We evaluate QwenSafe on 12 Apple-defined content rating descriptors and compare it against state-of-the-art vision-language models, including Qwen3-VL, LLaVA-1.6, and Gemini-2.5-Flash. QwenSafe consistently outperforms all baselines in binary CRD classification, achieving improvements in positive-class recall of 111.8%, 36.1%, and 2.1%, respectively. Our results demonstrate that descriptor-aware multimodal alignment substantially improves automated content classification and highlights the potential of vision-language models to support scalable and consistent content rating in mobile app marketplaces.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
