depth-scalability

The paper identifies two causes why logic gate networks fail to benefit from increased depth and proposes Input-Anchored Logic Gate Networks (IALGNs) that condition each layer on original inputs, achieving consistent depth-accuracy improvements beyond 100 layers.