This paper proposes a multi-scale equilibrium model that shows that implicit models can potentially compete with explicit models. Their work built on top of a previous proof-of-concept and solve many technical difficulties to achieve competitive performance on challenging image datasets. The paper is clear and the method is reproducible.