近日,npj Computational Materials刊发了题为《Score-based diffusion models for accurate crystal-structure inpainting and reconstruction of hydrogen positions》的研究成果。该研究创新性地引入了计算机视觉中的图像修复技术,利用分数扩散模型对晶体结构中缺失的氢原子位点进行高精度重建,综合成功率超过97%,为解决材料数据库中氢原子位置缺失这一长期难题提供了高效的AI驱动方案。
Reents, T., Cantarella, A., Bercx, M. et al. Score-based diffusion models for accurate crystal-structure inpainting and reconstruction of hydrogen positions. npj Comput Mater 12, 203 (2026). https://doi.org/10.1038/s41524-026-02090-1
Developers should successfully set up the basic environment for the Kaiwu-PyTorch-Plugin project, run the QBM-VAE sample code, and calculate the correct FID value based on the random seed value provided by the system.
Pass Rewards
10 quotas for 550-qubit real quantum machines with a one-year validity period
Exclusive "Quantum AI Developer" Community Certification Badge
Developer Benefits
Fixed Monthly Benefits: 5 quotas for 550-qubit real quantum machines
Proceed to Assessment
Step 1
Install the environment dependencies for the Kaiwu-PyTorch-Plugin library according to the README instructions