Science ❯ Computer Science ❯ Machine Learning ❯ Large Language Models
A non‑peer‑reviewed study credits selective neuron activation with large efficiency gains on long inputs.
