用思维链+规则奖励，让多模态广告审核更准更可信
BLM-Guard: Explainable Multimodal Ad Moderation with Chain-of-Thought and Policy-Aligned Rewards
- 引入思维链推理与规则驱动的标注数据生成
- 在真实短视频广告上准确率显著超越基线
- 适合需要可解释性审核的平台内容安全团队
短视频平台如今充斥着大量多模态广告，其欺骗性视觉、语音与字幕内容对审核提出更高要求。本文提出BLM-Guard，一种面向商业广告的内容审计框架，融合思维链推理、基于规则的政策原则及批评者引导的奖励机制。通过规则驱动的ICoT数据合成管道生成结构化场景描述、推理链条与标签，大幅降低人工标注成本。强化学习使用复合奖励优化模型，平衡因果一致性与政策合规性。多任务架构同时建模单模态操纵（如夸张图像）与跨模态不一致（如字幕与语音偏差），提升鲁棒性。在真实短视频广告数据集上的实验表明，BLM-Guard在准确率、一致性与泛化能力上均优于强基线。
原文摘要 · Abstract (English)
Short-video platforms now host vast multimodal ads whose deceptive visuals, speech and subtitles demand finer-grained, policy-driven moderation than community safety filters. We present BLM-Guard, a content-audit framework for commercial ads that fuses Chain-of-Thought reasoning with rule-based policy principles and a critic-guided reward. A rule-driven ICoT data-synthesis pipeline jump-starts training by generating structured scene descriptions, reasoning chains and labels, cutting annotation costs. Reinforcement learning then refines the model using a composite reward balancing causal coherence with policy adherence. A multitask architecture models intra-modal manipulations (e.g., exaggerated imagery) and cross-modal mismatches (e.g., subtitle-speech drift), boosting robustness. Experiments on real short-video ads show BLM-Guard surpasses strong baselines in accuracy, consistency and generalization.
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。