TASTE: A Designer-Annotated Multi-Dimensional Preference Dataset for AI-Generated Graphic Design
arXiv:2605.20731v1 Announce Type: cross Abstract: Text-to-image models produce graphic design at production scale, but their supervision comes from photo-style preference data with a single overall verdict per comparison. Designers evaluate along several distinct axes, including typography, visual hierarchy, color harmony, layout, and brief fidelity, and a single label collapses them. We release TASTE (Typography, Aesthetics, Spatial, Tone, Etc.): ten professional designers ranked outputs from four current text-to-image models on nine criteria across two disjoint cohorts, yielding 1,600 rating
The proliferation of advanced text-to-image models necessitates more nuanced evaluation methodologies to push beyond basic, single-verdict preferences, reflecting current bottlenecks in AI graphic design quality control.
This dataset introduces a multi-dimensional approach to evaluating AI-generated graphic design, moving beyond simplistic 'good vs bad' to detailed criteria, which is crucial for advancing AI's capabilities in creative fields.
AI models will now have access to granular feedback on design elements like typography and color harmony, enabling more sophisticated training and potentially higher quality, more stylistically aligned outputs.
- · AI graphic design platforms
- · Generative AI researchers
- · Designers leveraging AI tools
- · Creative agencies
- · Generic text-to-image models lacking granular control
- · Human designers solely competing on speed of basic output
AI-generated graphic design outputs will become significantly more refined and aligned with professional design principles.
This improved quality will accelerate the adoption of AI tools in professional design workflows, displacing some entry-level design tasks.
The definition of 'design' may evolve to focus more on strategic oversight, prompt engineering, and curation, rather than manual execution, as AI handles high-volume creative production.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI