Complex Abilities of LLMs

Andrea Esuli

Director of Research

Giovanni Puccetti

Andrea Pedrotti

Large Language Models are powerful tools with remarkable potential across a wide range of applications. However, despite their impressive capabilities, they often fall short in tasks that require nuanced reasoning and critical thinking.

For this reason, our research focuses on systematically evaluating LLMs using carefully designed benchmarks that stress their reasoning abilities and uncover their limitations, and improve them accordingly. In recent years, we have assessed the performance of LLMs in domains such as mathematical reasoning, complex decision-making, and multimodal tasks requiring fine-grained alignemnt between perceptual and textual modalities. These evaluations help shed light on the true capabilities and boundaries of current generative models.

Our group is also actively involved in the development of Machine-Generated Text detection methods. As generative models become increasingly pervasive, our goal is to provide tools that can distinguish between human- and machine-generated content, ensuring transparency and enabling responsible use of AI technologies.