PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection

Open in new window