Submitted by Benhao Huang 7 Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning Locus Lab 40 3