Trust Functions: Near-Lossless Weak-to-Strong Generalization by Learning When to Trust the Weak Teacher Paper • 2606.01000 • Published May 31 • 6