Professor at UFC; PhD in Informatics at @dipucrio; BSc degree in Software Engineering from @UfcQuixada in cooperation with the @unirome1.
- Tweets163
- Following462
- Followers279
- Likes946
ALT Background: Large Language Models (LLMs) have demonstrated strong performance across a variety of code-understanding tasks, leading many to believe that they can reason about program semantics. However, existing evaluations primarily focus on single-language settings or rely on synthetically generated code, raising concerns about whether current results reflect true semantic understanding. Aims: We investigate whether LLMs can accurately judge functional equivalence across different programming languages in human-written code, a setting that requires deeper reasoning beyond superficial similarity. Method: We introduce PolyHuman, a dataset of human-written programs in CPP, Java, and Python. Using this dataset, we evaluate intra- and inter-language equivalence detection across open-weight and proprietary LLMs, selecting GPT-o4-mini as a representative model to assess stability. We then manually analyze 81 cases of systematic disagreement in which models incorrectly judge functional equiv
ALT Four images: First figure on top displays four people talking to each other. On the bottom left, 5 people interacting, with the subtitle "Communication Dynamics". In the bottom center, two people talking to each other, with the subtitle "Discussion Content". And at the right corner, a figura of 8 people helping each other to climb a small squared mountain, with the subtitle "Organizational Dynamics".
