TASTE supplies designer ratings across nine criteria for outputs from four text-to-image models, with statistical tests showing moderate agreement and benchmarks where existing scorers reach at most 0.55 macro agreement while a new head reaches 0.611.
Reliability in content analysis: Some common misconceptions and recommendations.Human Communication Research, 30(3):411–433, 2004
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TASTE: A Designer-Annotated Multi-Dimensional Preference Dataset for AI-Generated Graphic Design
TASTE supplies designer ratings across nine criteria for outputs from four text-to-image models, with statistical tests showing moderate agreement and benchmarks where existing scorers reach at most 0.55 macro agreement while a new head reaches 0.611.