I set up the system as a product filter, like a buyer. It can find the sales website of the desired product according to the customer’s budget and needs, and make personal recommendations about the price and product. I fixed the maximum length at 256 bytes. When I set the temperature to 0.5, I found the reply to be more concise, without enough details and suggestions. At this point, it does not offer any substantive and practical recommendations.

After setting the temperature at 1.55 this time, I found that it replied with “temperature”. Compared to the previous lower temperature, its response becomes more detailed. In this manual test, I found that the system I set is a very personal choice, so Open AI may not allow me to get an accurate and unique answer. However, its reply has included suggestions of discount types and price suggestions. At this temperature, the response I got was much more similar to the answer I was looking for.