fig6

Figure 6. Conceptual summary of challenges and opportunities for MF learning in materials design. (A) Challenges: key barriers include data scarcity and imbalance between HF and LF datasets, limited transferability across material domains, the difficulty of interpreting black-box models, and integration with automated laboratories; (B) Opportunities: promising directions involve physics-informed MF models that embed domain knowledge, generative approaches for data augmentation, multi-objective optimization frameworks, standardized open datasets of HF and LF data, and coupling with autonomous discovery platforms. Together, these challenges and opportunities define a roadmap for advancing MF learning as a central tool in next-generation materials discovery.