Problems with chatbot response

Question

Hi. I am new to building a chatbot and I am having problems with the response the chat is returning to me. In the image you will see a sample of what I am asking, and in general, it has come up with several questions where I enter: “what is the second law of thermodynamics” and the answer is [“second law of thermodynamics”]. This behavior does not always happen, it also happens that when asked about the documents with which the chatbot was built, it returns the corresponding answer. I have been searching about it, but I have not found yet a satisfactory answer to solve the question. Has anyone had the same problem? How did you solve it? Any idea what could be happening?

Thanks

Configuration:

• Azure OpenAI model name and version: gpt-35-turbo-16k, version 0613
• Is chat history enabled? Yes
• Are you using data? If so, what data source? (e.g. Azure AI Search, Azure CosmosDB Mongo vCore, etc) AzureAI Search

Answer 1

Juan David Rendón Luna Greetings & Welcome to Microsoft Q&A forum!

I understand your frustration with the model generating irrelevant or inaccurate response. Unfortunately, AI models may sometimes generate irrelevant or inaccurate content.

One possible solution is to provide more context or information in your prompt to help the model generate a more relevant response. For example, instead of asking "what is the second law of thermodynamics", you could provide more context by asking "can you explain the second law of thermodynamics in simple terms?" or "what are the implications of the second law of thermodynamics for energy systems?".

There are some steps you can take to improve the quality of the generated content:

• Make sure to provide clear and specific instructions to the model on what you want it to generate. This includes providing specific keywords, phrases, and topics that you want the model to focus on.
• Try breaking down the article into smaller sections or paragraphs and providing more detailed instructions for each section. This can help the model focus on specific topics and generate more relevant content.
• Always review and edit the generated content to ensure that it meets your standards and requirements. This can help you catch any irrelevant or inaccurate content and make necessary changes.
• If you're still not satisfied with the quality of the generated content, you may want to try using a different AI model. Try experimenting with different models to find the one that works best for your needs.

Also, check Prompt engineering techniques, Learn how to work with the GPT-35-Turbo and GPT-4 models and Recommended settings for more suggestions.

Do let me know if that helps or have any other queries.

If the response helped, please do click Accept Answer and Yes for was this answer helpful.

Doing so would help other community members with similar issue identify the solution. I highly appreciate your contribution to the community.