A prompt engineer is creating a few-shot example to teach a language model how to solve a word problem. The goal is to make the model's reasoning process, especially the arithmetic, as clear and explicit as possible for the model to learn from. Given the problem: 'A warehouse had 500 boxes. On Monday, they received a shipment of 150 new boxes. On Tuesday, they shipped out 75 boxes. How many boxes are left?', which of the following reasoning chains best achieves the goal by clearly demarcating the calculation steps?

Applying Calculation Annotation to a Reasoning Chain

A prompt engineer is crafting a few-shot demonstration to teach a language model how to solve multi-step financial problems. The goal is to create a clear, replicable reasoning pattern that the model can follow. Below are two versions of the reasoning chain for the same problem.

Problem: Calculate the total amount after 2 years if $1000 is invested at an annual interest rate of 5%, compounded annually.

Version A: The principal is $1000 and the rate is 5%. After the first year, the interest is 1000 * 0.05 = 50. The new principal is 1000 + 50 = 1050. After the second year, the interest is 1050 * 0.05 = 52.50. The final amount is 1050 + 52.50 = 1102.50. The final answer is $1102.50.

Version B: The principal is $1000 and the rate is 5%. For Year 1, we calculate the interest: ≪1000 * 0.05 = 50≫. The new principal is the original principal plus the interest: ≪1000 + 50 = 1050≫. For Year 2, we calculate the new interest: ≪1050 * 0.05 = 52.50≫. The final amount is the Year 1 principal plus the Year 2 interest: ≪1050 + 52.50 = 1102.50≫. The final answer is $1102.50.

Which version is a more effective demonstration for the language model, and why?