MuPHI: 通过语义基础奖励优化学习隐式多模态有害推理
MuPHI: Learning Implicit Multimodal Harm Reasoning via Semantically Grounded Reward Optimization
AI总结：
针对视觉语言模型在隐式跨模态有害语义推理上的不足，提出MuPHI数据集和MuPHIRM训练框架，通过多视角奖励优化联合语义学习，提升有害检测与推理质量及分布外鲁棒性。
摘要：
理解看似良性的图像-文本对之间交互如何产生危害，需要超越表面特征的意图感知跨模态推理。现有的视觉语言模型（VLM）擅长对感知线索进行字面推理，但往往无法推导出依赖于隐式、上下文相关推理的有害语义。为了评估VLM在组合性有害检测和推理方面的能力，我们引入了多模态语用有害解释（MuPHI）数据集，其中包含有害编码在微妙多模态线索中的图像-文本对。MuPHI涵盖多种有害类别，并包含用于评估VLM推理链的注释有害理由。为了改进VLM的检测和推理能力，我们提出了MuPHIRM，一种推理增强的训练框架，通过优化多视角奖励来学习联合语义。MuPHIRM提高了VLM的有害检测和推理质量，同时与训练和推理时基线相比，表现出优越的分布外鲁棒性。我们的发现表明，面向推理的奖励优化为构建超越基准特定捷径进行泛化的多模态系统提供了一个有前景的方向。
英文摘要：
Understanding how harm emerges from interaction between otherwise benign image-text pairs requires intent-aware cross-modal reasoning beyond surface-level features. Existing vision-language models (VLMs) excel at literal reasoning over perceptual cues but often fail to derive harmful semantics that rely on implicit, context-dependent reasoning. To evaluate VLMs on compositional harm detection and reasoning, we introduce Multimodal Pragmatic Harm Interpretation (MuPHI), a dataset containing image-text pairs where harm is encoded in subtle multimodal cues. MuPHI spans diverse harm categories and includes annotated harm rationales for assessing VLM reasoning chains. To improve both detection and reasoning in VLMs, we propose MuPHIRM, a reasoning-augmented training framework which learns joint semantics by optimizing multi-perspective rewards. MuPHIRM improves both harm detection and reasoning quality of VLMs while demonstrating superior out-of-distribution robustness compared to both trained and inference-time baselines. Our findings suggest that reasoning-oriented reward optimization offers a promising direction towards building multimodal systems that generalize beyond benchmark-specific shortcuts.
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。
