Refractory High Entropy Alloys (RHEAs) are pivotal materials for the US Department of Energy's pursuit of advanced nuclear reactors characterized by enhanced shielding properties and exceptional resistance to high temperatures. The quest to design new RHEAs with improved functional properties involves navigating complex high-dimensional parameter spaces. These encompass the number of constituents, Bravais lattice size dictating atomistic structures, and the myriad disordered arrangements of atoms across lattice sites. However, the computational demands of accurate and comprehensive first-principles calculations hinder the efficient exploration of these parameters. To address this challenge, surrogate models are indispensable for accelerating exploration, but their efficacy relies on robust data collection. We propose a novel methodology concentrating first-principles training data on chemical compositions and disordered atomic structures for binary alloys. Leveraging the algorithmic capabilities of graph neural networks (GNNs) to learn short-range interactions from small environments, we extrapolate the GNN model's learned interactions for binary alloys to predict the formation energy of disordered atomistic structures in ternary alloys. Our approach is exemplified on four datasets describing body-centered cubic (BCC) solid solutions of binary alloys niobium-tantalum [1], niobium-vanadium [2], tantalum-vanadium [3], and ternary alloy niobium-tantalum-vanadium [4]. The GNN model, trained on first-principles data for binaries, demonstrates its predictive power by extrapolating formation energy predictions for ternary alloys. Our methodology mitigates the curse of dimensionality, showcasing the transferability and effectiveness of GNN surrogate models in navigating the intricate parameter space of RHEAs. REFERENCES [1] Lupo Pasini, M., Samolyuk, G., Eisenbach, M., Choi, J. Y., Yin, J. & Yang, Y. NbTa_BCC_SolidSolution_128atoms_VASP6. OSTI.GOV10.13139/OLCF/2222906 (2023). [2] Lupo Pasini, M., Samolyuk, G., Eisenbach, M., Choi, J. Y. Yin, J. & Yang, Y. NbV_BCC_SolidSolution_128atoms_VASP6. OSTI.GOV10.13139/OLCF/2228839 (2023). [3] Lupo Pasini, M., Samolyuk, G., Eisenbach, M., Choi, J. Y. Yin, J. & Yang, Y. TaV_BCC_SolidSolution_128atoms_VASP6. OSTI.GOV10.13139/OLCF/2222910 (2023). [4] Lupo Pasini, M., Samolyuk, G., Eisenbach, M., Choi, J. Y. & Yang, Y. NbTaV_BCC_SolidSolution_128atoms_VASP6. OSTI.GOV10.13139/OLCF/2217644 (2023).