InvDesFlow-AL: active learning-based workflow for inverse design of functional materials

In three linesInvDesFlow-AL combines diffusion and active learning for inverse design of functional materials. The model achieves RMSE 0.0423 Å in crystal structure prediction (+32.96% vs existing methods) and systematically generates low-formation-energy materials. Validation: discovery of Li₂AuH₆ as BCS superconductor at 140 K.

Summary generated by Claude — human-verified