The proposed brain-inspired CHT Soft Rule with Sigmoid Decay Density (CHTss) achieves comparable even better performance compared to fully connected models across various tasks, enabling high sparsity in Transformers and LLMs.