Quantum-floor compression: Achieving GPT-4 capability at 1/120th the model size [pdf]oroboroslab.github.io·2 pts·cybertax·2