The abstract is read at the publisher; the summary is JClub's.
TabPFN-3.5, a pretrained tabular foundation model using marker genotypes without genetics-specific tuning, ranked first on 11 of 18 plant traits and bested the median rank.
- Why it matters: Genomic prediction is usually built on assumptions about genotype–phenotype structure, so it is unclear whether a general-purpose learner can match specialized GP methods without domain modeling.
- What they did: The model was benchmarked on 18 traits across six plant species against 18 published GP methods, then compared on the same splits by rerunning GBLUP, RKHS, and elastic-net.
- The result: TabPFN-3.5 was significantly behind only spruce wood density, ahead on 12 of 18, and showed significant gains on all three soybean traits and rice flowering time; in leave-families-out soybean CV it matched GBLUP and RKHS.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.