Find a model for
the work you make.
Explore real preferences by creative task. Inspect the evidence, compare uncertainty, and choose a setup to try.
0 models in this cohort. Each judgment may compare several models; per-model counts are not added into a global vote total.
Evidence starts with real comparisons.
No models have ranking evidence in this cohort yet. Judge a community comparison or choose a different creative task.
Explore blind comparisonsHow to read these rankings
Scores describe relative preferences within the selected cohort. Overlapping confidence intervals do not establish a clear winner. Group choices produce several pairwise comparisons, but count as one human judgment. A provisional label means more independent people and prompts are needed.
Community-authored metadata comes from creators. Verified provenance requires separate trusted validation. Sample and test choices are excluded.