Submitted by Benhao Huang 7 Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning Locus Lab 38 3