Skip to content
Research

Learning to Reason About Implicit Multimodal Harm: From Diagnosis to Generalization.

arXiv:2605.29951v2 Announce Type: replace Abstract: A benign image paired with a benign caption can become harmful only in combination, making detection fundamentally a cross-modal reasoning problem. Existing multimodal harm benchmarks evaluate only classifica

arXiv cs.AIResearch source·