Reinforcement Learning for Dynamic Spectrum Allocation in Next-Generation Optical Networks. JDIAS 2026, 1 (3).