Atlas
Sign in

Benchmarks

45 benchmarks in 8 domains

RNA

1 benchmark · zero-shot · 11 models
Protocol
Benchmarks: 1 of 1

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

Reset

11 models

 Ranked from 1 of 1 benchmarks (a third, at least 2)
# Model ↕ Mean rank ▲ Coverage ↕ Protein abundance from coding sequence (E. coli) ↑ Spearman ρ (zero-shot pseudo-log-likelihood) ↕
1 rna-ml/nv-codonfm-encodon-1b-v1 overlap 0.00 1/1 0.552 (1) *
2 rna-ml/nv-codonfm-encodon-80m-v1 overlap 0.10 1/1 0.431 (2) *
3 genomics/caduceus-ps-seqlen-131k-d-model-256-n-layer-16 0.20 1/1 0.349 (3)
4 rna-ml/rnafm 0.30 1/1 0.344 (4)
5 genomics/caduceus-ph-seqlen-131k-d-model-256-n-layer-16 0.40 1/1 0.304 (5)
6 rna-ml/rinalmo-giga 0.50 1/1 0.207 (6)
7 rna-ml/rinalmo-mega 0.60 1/1 0.188 (7)
8 genomics/dnabert-2-117m 0.70 1/1 0.184 (8)
9 rna-ml/rinalmo-micro 0.80 1/1 0.160 (9)
10 genomics/hyenadna-medium-450k-seqlen-hf 0.90 1/1 −0.022 (10)
11 genomics/hyenadna-tiny-1k-seqlen-hf 1.00 1/1 −0.069 (11)

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.