GPN-Star: phylogeny-aware genomic language model hits SOTA on variant effect prediction
pastramimachine · x · 2026-09-10
Yun S. Song's team published GPN-Star in Nature (open access), a genomic language model with a phylogeny-aware architecture that explicitly leverages whole-genome alignments and species trees.
- Motivation: NLP-adapted genomic LMs need large sizes and compute yet still trail classical evolutionary models on prediction tasks.
- Method: species-tree and alignment representations trained across vertebrate, mammal and primate timescales.
- Results: state-of-the-art variant effect prediction in both coding and non-coding regions, with a biologically grounded, smaller model.
A notable AI-for-Science result blending evolutionary priors with language modeling.
Related event: GPN-Star hits Nature, beating 40B Evo 2 with 200M parameters(3 posts)→
More from Research
- HarnessMM cuts agent tokens per task by up to 88% while raising success rates — omarsar0 · 2026-10-26
- Ben Recht: The world needs a truly open language model, and it's easier to build than you think — beenwrekt · 2026-10-26
- New Blog Breaks Black-Box Distillation Into Five Practical Steps — chrisrohlf · 2026-10-26
- DetectorScope turns black-box AI text detectors into auditable ones with sparse autoencoders — _akhaliq · 2026-10-26
- OPST paper shows LLMs self-improve with zero human data, guided only by topic descriptions — _akhaliq · 2026-10-26
- Tencent's HiSTEVE 1.0: hierarchical VLA agent plays Minecraft with native keyboard and mouse, hits 85% SOTA — _akhaliq · 2026-10-26