<p>Genomic Selection (GS) is transforming the fields of plant and animal breeding by enabling selection based on genetic predictions without requiring direct phenotypic measurements. However, achieving practical implementation of GS is hindered by the challenge of obtaining high prediction accuracy, often compromised by experimental noise. To address these challenges, researchers have developed numerous strategies to enhance the efficiency of GS. This paper investigates the application of an innovative transfer learning method for improving GS performance. In this study, transfer learning (TL) was employed to leverage information from a proxy environment to improve predictions in a target environment. Using twelve original multi-environment data sets, we compared models with and without transfer learning based on four metrics: (i) Pearson’s correlation (COR), (ii) normalized root mean square error (NRMSE), and proportion of shared lines in the top (iii) 10 and (iv) 20% of lines (PM_10 and PM_20). The results provide compelling evidence of the benefits of transfer learning, demonstrating significant improvements in the four metrics evaluated and in all proportions of testing evaluated compared to models without transfer learning. These findings highlight the potential of transfer learning as a powerful tool to enhance predictive accuracy in genomic selection.</p>