neural-architecture
Q-Compass architecture — RL-grounded sequence mixing for text, vision, audio, world modeling, and cross-field tasks
