Enhancing Cross-Prompt Transferability in Vision-Language Models through Contextual Injection of Target Tokens
Primary LanguageJupyter NotebookMIT LicenseMIT