一作ACL 2026Poster· Natural Language Processing
Visual Self-Fulfilling Alignment: Shaping Safety-Oriented Personas via Threat-Related Images
Qishun Yang · Shu Yang · Lijie Hu · Di Wang
摘要
Multimodal large language models (MLLMs) face safety misalignment where visual inputs enable harmful outputs. Existing methods require explicit safety labels or contrastive data, yet threat-related concepts are concrete and visually depictable, while safety concepts like helpfulness are abstract...