4 ms·Decoupling Reasoning from Observations for Efficient Augmented Language Models2 points by kytazo 3y ago