Institute of Formal and Applied Linguistics

Less than five years after the invention of masked language modeling changed the mainstream methodology natural language processing (NLP), we are experiencing another paradigm change. Large generative language models (LLMs) with prompt engineering and in-context learning drastically reduce the amount of labeled data needed for NLP and change how we formulate tasks. In monolingual setups, many of the tricks from the previous time work well with LLMs, however, in cross-lingual tasks, it does seem to be the case. In this talk, we will discuss this unusual situation and speculate what it means for not only our future research.

*** The talk will be delivered in person (MFF UK, Malostranské nám. 25, 4th floor, room S1) and will be streamed via Zoom. For details how to join the Zoom meeting, please write to sevcikova et ufal.mff.cuni.cz ***