Fetching the paper…

Let's Think Dot by Dot: Hidden Computation in Transformer Language Models · Around