Two papers on reinforcement learning and agent optimization accepted at ICLR 2026.