Atlas
Sign in

Benchmarks

45 benchmarks in 8 domains

RNA

2 benchmarks · probe · 11 models
Protocol
Benchmarks: 2 of 2

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

Reset

11 models

 Ranked from 2 of 2 benchmarks (a third, at least 2)
# Model ↕ Mean rank ▲ Coverage ↕ Protein abundance from coding sequence (E. coli) ↑ Spearman ρ (ridge probe) ↕ RNA secondary structure (F1) ↑ Base-pair F1 (2-D pair head) ↕
1 rna-ml/rinalmo-mega overlap 0.20 2/2 0.585 (5) 0.585 (1) *
2 rna-ml/rinalmo-giga overlap 0.27 2/2 0.631 (3) 0.584 (2) *
3 rna-ml/rnafm overlap 0.63 2/2 0.543 (7) 0.580 (3) *
4 rna-ml/rinalmo-micro overlap 0.75 2/2 0.565 (6) 0.545 (4) *
Low coverage: below 2 of 2, or nothing to compare with. Listed, not ranked.
— rna-ml/nv-codonfm-encodon-1b-v1 overlap 0.00 1/2 0.688 (1) * ·
— rna-ml/nv-codonfm-encodon-80m-v1 overlap 0.10 1/2 0.666 (2) * ·
— genomics/caduceus-ps-seqlen-131k-d-model-256-n-layer-16 0.30 1/2 0.600 (4) ·
— genomics/dnabert-2-117m 0.70 1/2 0.496 (8) ·
— genomics/caduceus-ph-seqlen-131k-d-model-256-n-layer-16 0.80 1/2 0.495 (9) ·
— genomics/hyenadna-tiny-1k-seqlen-hf 0.90 1/2 0.486 (10) ·
— genomics/hyenadna-medium-450k-seqlen-hf 1.00 1/2 0.391 (11) ·
Reference floors: trivial baselines on the same items, not hub models, unranked.
— kmer-dna floor 0.513 ·
— onehot-window-rna floor · 0.517

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 probe 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.