Automated testing of a code is an essential part of software development. 
Unit testing is a subset of automated testing, which aims to cover the
smallest meaningful code unit with tests.  However, writing unit tests is
often repetitive and time-consuming, which sometimes leads to improper test
coverage or missing test cases.  With rising capabilities of Large Language
Models (LLM), Generative AI, and their usage for generating code, it is
important to be aware of potential issues related to generating tests.

As part of this thesis, the student is expected to analyze the state of the
art in automated testing and identify relevant research and best practices. 
A particular focus should be given to evaluating tests from a qualitative
perspective and selecting metrics that can be automated.  One specific
example is mutation testing, where the expectation is that after modifying
the code, at least one test should fail. 

The student should investigate the feasibility of automating such metrics and propose improved ones that more accurately reflect the quality of the generated tests, thereby enhancing their reliability. These methods will be implemented and evaluated on the Python part of Rosetta dataset using selected large language models (LLMs).