AI RESEARCH
Skill-Aligned Annotation for Reliable Evaluation in Text-to-Image Generation
arXiv CS.CV
•
ArXi:2605.13223v1 Announce Type: new Text-to-image (T2I) generation has advanced rapidly, making reliable evaluation critical as performance differences between models narrow. Existing evaluation practices typically apply uniform annotation mechanisms, such as Likert-scale or binary question answering (BQA), across heterogeneous evaluation skills, despite fundamental differences in their nature. In this work, we revisit T2I evaluation through the lens of skill-aligned annotation, where annotation strategies reflect the underlying characteristics of each evaluation skill.