Anhui Agricultural University Unveils "Fengshu" 2.0, an AI-Powered Soybean Breeding Model

On August 16, at the 32nd National Soybean Scientific Research and Production Symposium held in Nanning, Guangxi, Professor Wang Xiaobo from Anhui Agricultural University unveiled "Fengshu" 2.0, a vertical large language model dedicated to soybean research. Developed jointly by Anhui Agricultural University and the Institute of Crop Science at the Chinese Academy of Agricultural Sciences, "Fengshu" is a generative AI-powered breeding platform designed for the entire soybean life cycle and has recently undergone an iterative upgrade. "Fengshu" 2.0 integrates diverse data sources, including soybean germplasm resources, genomic and coding sequences, proteins, transcripts, gene expression profiles, disease samples, breeding trial records, and domestic and international literature, to construct a structured knowledge system. This new version introduces a multi-model collaborative analysis mechanism and embeds interfaces to certain general-purpose large models, enabling it to comprehensively evaluate consensus, discrepancies, and evidence completeness across model outputs—thereby enhancing the reliability of expert conclusions. The model features six functional modules—"Soypedia," "SoyMolecular," "SoyLiterature," "SoyDisease," "SoyPhenotype," and "SoyBreeding"—covering a multidimensional data ecosystem that includes knowledge graphs and literature data, genomic and multi-omics data, phenotypic data, and breeding practice data. Built upon over 10 million Chinese characters of domain-specific text, more than 10,000 scientific publications and patent records, the model incorporates a soybean-specific knowledge graph containing 20,000 entities and 100,000 relationships. It also integrates whole-genome resequencing data from 8,000 accessions, approximately 40,000 soybean germplasm resources, and phenotypic data from over 3,000 soybean varieties.