Part 5/8:
To illustrate the effectiveness of Chain of Draft, let's consider its performance against Chain of Thought using a straightforward mathematical problem:
Question: Jason had 20 lollipops. He gave Denny some lollipops. Now Jason has 12 lollipops. How many lollipops did Jason give to Denny?
In a traditional CoT approach, the AI might articulate its reasoning step-by-step:
Jason starts with 20 lollipops.
After giving some to Denny, he has 12 left.
Therefore, the number of lollipops given is 20 - 12, which equals 8.
While this approach arrives at the correct answer, it also generates a lengthy and somewhat superfluous explanation that raises computational costs and latency.
Using Chain of Draft, the model succinctly captures the essence:
- 20 - x = 12.