incorporating multi-modal large models for joint encoding of text, images, and time-series data, logic for credit limit prediction and model selection, constructing and dynamically allocating credit ...