Hi everyone. Why does the gpt-35-turbo-16k model make a lot of mistakes when I ask it to return the answer in a specific structure and the values of that structure? What would be the best way to request this structure? Input : { "model": "gpt-35-turbo-16k", "messages": [ { "role": "syst…

Hi there and welcome to the Forum! Could you perhaps describe in greater detail what you are trying to achieve? I’m happy to try and help but it’s not yet fully clear to me from the system and user messages. Thanks!

hello, I wanted to have more consistency in the keys: has_context, user_answer and suggestions. he makes a lot of mistakes with these keys and there are times when the key doesn’t even come

I’ve had a lot of success just using typescript as a schema hint. this has worked well since davinci. Answer in json using the following schema: { user_answer: string; // em português, traduza se necessário suggestions: string[]; // max 3 items } That said, your instructions seem a little con…

Like this? @Diet Question: what routines are available? . Please respond completely with as much detail as you can. Answer in Portuguese. user_answer: Your answer, suggestions: Array of a maximum of 3 positions with simple sentences that guide the user to continue processing the routine,has_contex…

Hi everyone. How can I make this chat conversation prompt more assertive? Leave it in the entire format in json, xml, etc… { "model": "gpt-35-turbo-16k", "messages": [ { "role": "system", "content": "Consider the context: The available routines are Software Analysis (analyse-software) a…

Try this restructuring of your system message and insertion of documentation: { “model”: “gpt-3.5-turbo”, “messages”: [ { “role”: “system”, “content”: “You are a helpful programming and language AI expert assistant. Your task is to answer the user’s questions and provide suggestions based on t…

Thanks for the help @_j , Now with this structure you gave it always returns in the desired format, but checking whether the question is in context is very inconsistent and makes a lot of mistakes. Do you know anything to solve this? Thanks in advance!

Specify a response structure for the gpt-35-turbo-16k

Prompting

SomebodySysop January 30, 2024, 5:57am 8

This is how I did it using XML, and it improved the responses dramatically: API Prompt for gpt-3.5-turbo-16k - #12 by SomebodySysop

Topic		Replies	Views
Assistant api don't access the database Prompting gpt-4 , chatgpt , api , assistants-api	3	483	July 4, 2024
How "force" the API to really follow all the instructions? API	41	15288	August 20, 2023
I want to get json format response which I can pass using gpt-4 model. Also I want to give my prompt to get json data Prompting gpt-4	14	21832	October 26, 2023
Any tips on prompting Assistant API to return the entire response in JSON? API	4	5067	May 3, 2024
Inconsistent and invalid JSON response API	7	6930	December 16, 2023

Specify a response structure for the gpt-35-turbo-16k

Related topics