Rapidata SVG · Human Preference
Rapidata · People directly compare model-generated vector graphics for visual preference.
What it measures and how
Only Preference Elo for the same static SVG task is used. Alignment, Coherence, and Overall are not counted again. A fixed official dataset version traces results to original human comparisons.
How this evidence is used
Under review: official data and licensing exist, but three rolling DeepSeek aliases lack fixed-version evidence. The whole table is withheld pending evaluator-supplied run versions. Active-source budgets were not reduced.
Evaluation limits & attribution
SVG drawing is not all interface design. Raw data lacks model versions, generation dates, and API fingerprints, so rolling-endpoint changes cannot be established. Update dates cannot identify model versions.
Data license: CC BY 4.0 · Rapidata/svg-benchmark. Retain official attribution and the data link.
Results are published by Rapidata. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.