After fine-tune the result are not as expected

Prompting
1

I ha ve created a JSONL which contains few FAQ kind of prompt and completion.

{"prompt":"What is the primary purpose of using netDocShare?","completion":"To easily integrate NetDocuments with Microsoft SharePoint, Microsoft Teams, and One Drive for Business."}
{"prompt":"Is netDocShare flexible to use?","completion":"Yes, netDocShare has a simple user interface that can be configured by non-technical users for business users."}
{"prompt":"What is netDocShare?","completion":"netDocShare is an innovative solution designed to Live View, Edit, Co-author, Add, Search, Sync, Attach and Provision NetDocuments content in Microsoft Teams, Microsoft SharePoint, and OneDrive for Business (incl. OneNote sync with NetDocuments)."}

The base model I used is Davinci. Once the fine tune completed it gave me the fine-tuned model.
I am once I use the fine-tuned model, I get the below result.

The code:

               var request2 = new HttpRequestMessage(HttpMethod.Post, $"{OpenaiAPIURI}/completions");
                parameters = new
                {
                    model = "davinci:ft-klst-inc-2023-03-01-15-31-36",
                    prompt = message,
                    max_tokens = 500,
                    temperature = 0,
                    top_p = 1,
                    frequency_penalty = 0.2,
                    best_of = 1
                };

Result:

image
image1148×742 69 KB

can anyone let me know why the bot returning an expected result?