Search papers, labs, and topics across Lattice.
This paper introduces TFMat, a text-conditioned flow-matching framework that enhances crystal structure generation by integrating structured materials language as a semantic prior for the CrystalFlow generator. The method significantly improves one-candidate match rates across multiple benchmarks, achieving a 92.04% match rate on the MP-20 benchmark with just 20 candidates. By leveraging text to guide the generation process, TFMat enables better alignment of element counts and density distributions while maintaining property consistency in the generated structures.
TFMat achieves a remarkable 92.04% match rate in crystal structure prediction by harnessing structured text as a control layer for flow-based generation.
Crystal generators can now propose periodic structures, but their control interfaces remain poorly matched to the mixed descriptors used in materials design. Text provides a compact way to combine composition, symmetry, prototype and property cues, yet it has not been clear whether such information can steer flow-based crystal generation. Here we introduce TFMat, a text-conditioned flow-matching framework that uses structured materials language as a semantic prior for a CrystalFlow generator. Across Perov-5, Carbon-24 and MP-20 crystal structure prediction benchmarks, TFMat improves one-candidate match rates over CrystalFlow and reaches a 92.04% MP-20 match rate with 20 candidates; in de novo generation, it improves element-count and density distribution alignment while retaining coarse property consistency in composition-selected outputs. These results position structured text as an inspectable control layer for translating human-readable materials intent into candidate crystals for downstream simulation and validation.
Tsinghua AI