Large language and foundation models in predictive plant breeding: Capabilities, evidence, and challenges.

Direct evidence for text-based large language models in plant breeding is concentrated in retrieval, question answering, and evidence organization, while molecular prediction relies mainly on biological foundation models.

• Why it matters: Plant breeding now combines genomic, phenotypic, environmental, and multi-omics data with fast-growing literature, creating a need to know which AI tools have been validated versus which remain unproven.
• What they did: The authors review LLMs, retrieval-augmented generation, knowledge graphs, and Bio-FMs, then run a systematic evidence audit to separate direct demonstrations from conclusions dependent on adjacent evidence.
• The result: LLM-informed genomic prediction, parental selection, and breeding design are prospective, and adoption will require provenance, benchmarking, and validation matched to each output, from computation through field trials.

AI summary of the abstract · may contain errors · not medical advice · read the paper.