Preference alignment on diffusion models: A comprehensive survey for image generation and editing

Puntos clave

• Preference alignment in diffusion models aids in better image generation and editing outcomes, enhancing user satisfaction.
• Key insights reveal that aligning preferences significantly improves the performance of image generation techniques.
• Comprehensive survey assesses various diffusion models, focusing on how preference alignment influences their effectiveness.
• Findings suggest that better alignment with user preferences may lead to more intuitive and effective image editing tools.