I have written code that can provide answers using the Chat GPT API. While doing so, I realized that my model was not giving me the correct answers because the prompts I was providing were not clear to the Chat GPT model. To address this issue, I have written the code below, but it doesn’t achieve the desired level of accuracy. Is there a better way to solve this problem?"
for phrase in phrases:
print(“-”*100)
print(“Input_phrase: “, phrase)
print(”-”*100)
para_phrases = parrot.augment(input_phrase=phrase)
for para_phrase in para_phrases:
print(para_phrase)
If OpenAI language models can’t understand the question, then lesser standalone models aren’t going to close the comprehension gap.
A prompt technique to annoy people (or just yourself?):
“Before answering a question, you always examine a question closely, discern its context within the frame of other recent user input and AI assistant responses, and then rephrase the question to the user to be sure that AI has full comprehension of the user desires to be fulfilled before proceeding.”
Using the technique as a custom ChatGPT instruction:
Creating a chatbot that can provide accurate answers from given data. Currently, GPT-3.5 Turbo can’t give accurate answers unless it’s provided with a clear question. However, users often don’t phrase their questions correctly, so rephrasing questions is crucial.
After rephrasing questions, getting an appropriate prompt or question is also essential. This is the main problem I’m trying to solve.
I hope you understand what I’m aiming to achieve. Can anyone provide suggestions for solving these problems?
