DyRIS uses dynamic few-shot learning to predict perovskite space groups

The DyRIS framework employs an LLM-agent to predict space groups in double perovskites using diversity-enhanced dynamic few-shot prompting. It uses rule-guided inference to mitigate biases toward dominant, overrepresented structural classes.