Loopback Transformer: Towards a Unified View of Residual Looped Transformers A paper titled "Loopback Transformer: Towards a Unified View of Residual Looped Transformers" analyzes how residual connections store representations across repeated block applications in looped Transformer architectures, which reduce parameter count by reusing shared blocks. The work argues that parameter sharing alone does not determine how the results of repeated computation are reused, and it adopts a depth-memory view of residual connections to examine which representation is retained after each block application. Looped Transformers reduce parameter count by repeatedly applying shared blocks, but parameter sharing alone does not determine how the results of repeated computation are reused. We take the depth-memory view of residual connections and analyze which representation is stored after each block applic