Datasets / genomics / genomic-benchmarks-demo-coding-vs-interg / benchmarks

genomics/genomic-benchmarks-demo-coding-vs-interg

Visibility: PUBLIC

1 benchmark · 14 results. Each benchmark shows the top of its default board (its first primary metric at its default version); ranks are the server's, ties share a rank, and "best" is the best visible result, not a state of the art.

Genomic Benchmarks accuracy genomics/genomic-benchmarks-accuracy

Accuracy (logistic probe, mean of 3 sets) · ↑ higher is better · probe · atlas-full-v1

#	Model	Value [95% CI]	Source	Overlap
1	genomics/caduceus-ps-seqlen-131k-d-model-256-n-layer-16	0.821 [0.817, 0.826] best	reproduced	overlap Train/test overlap declared by the evaluator: genomics/ucsc-hg38
2	genomics/caduceus-ph-seqlen-131k-d-model-256-n-layer-16	0.817 [0.812, 0.821]	reproduced	overlap Train/test overlap declared by the evaluator: genomics/ucsc-hg38
3	genomics/dnabert-2-117m	0.794 [0.789, 0.798]	reproduced	overlap Train/test overlap declared by the evaluator: DNABERT-2 multi-species genomes (135 species)
4	genomics/hyenadna-medium-450k-seqlen-hf	0.775 [0.770, 0.780]	reproduced	overlap Train/test overlap declared by the evaluator: genomics/ucsc-hg38
5	genomics/hyenadna-tiny-1k-seqlen-hf	0.774 [0.769, 0.779]	reproduced	overlap Train/test overlap declared by the evaluator: genomics/ucsc-hg38

5 on this board. Full leaderboard