For example, we can create a dialogue system that can give technical and scientific answers to questions. We can tell the model how to behave through explicit instructions. This application scenario is sometimes called "role prompting".
Now, let’s make it give more understandable answers.
I think we've made some progress and you can continue to improve it. If you give the model more examples, you might get better results.
Code Generation
Another effective application area of large language models is code generation. Copilot is a good example. You can perform code generation tasks by giving it some effective prompt words. Let's look at the following example:
Output:
You see, we don't even need to specify a programming venezuela mobile database language.
Now, let’s take this a step further. The following example will show how powerful cue words can make large language models.
Pretty cool, right? In this example, we provided information about our database schema and asked it to generate valid MySQL queries.
reasoning
At present, reasoning tasks are the biggest challenge for large language models. The most exciting thing about reasoning tasks is that they can enable various complex applications to be born from large language models.
At present, mathematical reasoning tasks have made some progress. Performing reasoning tasks is still difficult for current large language models, so more advanced prompt word engineering techniques are needed. We will introduce these advanced techniques in the following guides. Now, let's use a few basic examples to demonstrate arithmetic functions.