#machine learning
Sep 2026
CompassPlay: Rewarding the Proposer for Where It Moves the Solver
CompassPlay is introduced, a self-play method that rewards the proposer through gradient alignment that draws on a first-order approximation to learning progress and scores each eligible task without additional solver training.
Sophia Xiao Pu, Xi-Meng Sun, Jiang Liu et al.
· arXiv.org · 0 citations