Reinforcement Learning with Attention That Works: A Self-Supervised Approacharxiv.org·8 pts·physicsyogi·0