Abstract: Compute-in-memory (CiM)-based binary neural network (CiM-BNN) accelerators marry the benefits of CiM and ultra-low precision quantization, making them highly suitable for edge computing.