Prompt templates in your paper did not align with your code.

[SpeeeedLee]

Take PIQA for example, in your paper:

Please choose the correct solution to the question: [QUESTION]
Solution1: [SOLUTION_1]
Solution2: [SOLUTION_2]
Answer format: solution1/solution2
the correct answer is [ANSWER]

but in your code:

Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

Instruction:

Please choose the correct solution to the question: [QUESTION]
Solution1: [SOLUTION_1]
Solution2: [SOLUTION_2]

Response:

the correct answer is [ANSWER]

Since you are fine-tuning base model (not instruction-tuned), It makes quite a huge difference, please help me clarify this, thanks!