Welcome to the otter world, by Camille Molinier & Can textual similarity metrics “replace” tests?

Abstract

“Welcome to the otter world”, by Camille Molinier I recently joined the team as a PhD student. This presentation will provide a brief overview of my academic journey, including my educational background, relevant internship experiences, and some personal projects. I will then delve into the core of my PhD thesis, introducing the intriguing field of federated learning. Finally, on a lighter note, I’ll share some personal anecdotes and interests to give you a better overview of who I am.

“Can textual similarity metrics “replace” tests?", by Sergiu Mocano Technological advancements in code-specialized Large Language Models has created a partial shift in the way code is being produced. Some private companies (sadly) have replaced developers with AI in order to cut costs. Although fully relying on AI when producing software products is not a reasonable practice, such a strategy could be employed in smaller-scale, more focused tasks, such as “code diversification”. However, AI is not robust and the generated code must be tested. Instead of testing all the generated variants (which would be costly in terms of time and energy resources), we experimented with textual-similarity metrics as a pre-filtering mechanism for generated code. In this talk we will present our experimental protocol and the obtained results that allow us establish the relevance of textual-similarity metrics as a test proxy.rview of my progresses so far.