Causality-Induced Positional Encoding for Transformer-Based Representation Learning of Non-Sequential Features | Read Paper on Bytez