Boundary Coverage principle of unit-tested variable range

[george-polevoy]

Is your feature request related to a problem? Please describe.

When generating tests for a range of variables, AI does not pay attention to the Boundary Coverage of variables needed to be tested, and it paid attention only to Formula Shape Coverage.
For example, a method accepts a random variable [0, 1) (inclusive zero, exclusive upper bound) in double precision floating point numbers. The test cases it generated:
0.25, 0.75, 0.5.

Describe the solution you'd like

In this case, it should understand the range of variables and provide boundary variables first.
I would expect 0 and 1 - double.Epsilon at the very least, and only then add additional in-between variables.
That is basics of unit testing and I expect a unit-testing superpower to understand those principles and communicate it to the model in form of prompts.

Describe alternatives you've considered
Writing this in prompts manually.