Jindřich Libovický

LLMs exhibit a stark performance disparity in mathematical reasoning, with underrepresented languages lagging significantly behind their high-resource counterparts.

Language identification systems falter dramatically when faced with cousin languages and orthographic noise, revealing critical gaps in current approaches.

Forget expensive human annotations: LLMs can reliably generate synthetic data to validate NLP evaluation metrics, even outperforming human agreement in some multilingual tasks.

LLMs can classify endangered languages almost as well as high-resource ones, but still struggle to generate text in these languages fluently, even with parallel training data.