Atlas
Sign in

Benchmarks

45 benchmarks in 8 domains

Single-cell

2 benchmarks · zero-shot · 4 models
Protocol
Benchmarks: 2 of 2

Mean rank and coverage are recomputed over the benchmarks ticked here. None ticked means all.

Reset

4 models

 Ranked from 2 of 2 benchmarks (a third, at least 2)
# Model ↕ Mean rank ▲ Coverage ↕ Cell type annotation (macro F1) ↑ Macro F1 (zero-shot) ↕ Single-cell batch integration (scIB) ↑ Cell-type ASW (zero-shot) ↕
1 single-cell/scgpt overlap 0.33 2/2 0.688 (1) * 0.553 (3) *
2 single-cell/uce overlap 0.42 2/2 0.688 (2) * 0.557 (2) *
3 single-cell/geneformer 1.00 2/2 0.562 (3) 0.549 (4)
Low coverage: below 2 of 2, or nothing to compare with. Listed, not ranked.
— single-cell/scimilarity overlap 0.00 1/2 · 0.624 (1) *
Reference floors: trivial baselines on the same items, not hub models, unranked.
— hvg-pca-cells floor 0.645 0.574

Each cell is the model's value on that benchmark's board, with its rank there in brackets, as the benchmark's own leaderboard ranks it (its primary zero-shot metric, at its default version); hover for the 95% interval, n and source, click to open that board. A * marks a declared train/test overlap. Rows are ordered by mean normalised rank, (rank − 1) / (board size − 1) averaged over the benchmarks the model has a result on (a board of one entry does not count): ranks, never values, so no average mixes units. Filtering hides rows without re-ranking anyone.