Author

Jacob Steinhardt

University of California SystemORCID 0000-0002-0257-3860Unclaimed ORCID profile · follow to get notified of new preprints

Biology Disciplines

Click to filter author papers

Bioinformatics· 1

Signature Assays, Organisms & Methods

Click to filter · or open tag

Research Focus:benchmarking· 1 Organism:multiple species· 1 Research Focus:protein design· 1 Assay:protein sequence· 1 Research Focus:protein structure prediction· 1 Approach:simulation· 1 Approach:statistical model· 1 Approach:transformer / foundation model· 1

Frequent Co-Authors

Frances DingiD· 1 shared

Activity by Year

2024-2024 · peak 1/yr

1 paper

Recent·Most cited

BioinformaticsbioRxivPreprintMar 7, 2024

✦ In-depth digest

Protein language models are biased by unequal sequence sampling across the tree of life

Frances Ding, Jacob Steinhardt

Likelihoods from Progen2 and ESM2 protein language models are systematically higher for sequences from some species, tracking evolution-weighted database sequence counts, and pull likelihood-guided designs away from extremophile properties.

57 citations

multiple speciesprotein sequence

𝕏 PostBluesky