LEAP: Linear Explainable Attention in Parallel for causal language modeling with O(1) path length, and O(1) inference
Transformer based Multiple Instance Learning
