Marek Šuppa

Machine translation benchmarks have functionally saturated, but pairing human-authored failure cases with deterministic verification rules reveals critical multimodal blind spots that automated metrics consistently miss.

Large multilingual models outperform Slovak-specific embeddings, but new compact models offer a local deployment solution without sacrificing performance.

LLMs can extract better keyphrases in morphologically rich languages like Slovak because they are less sensitive to surface-form variations that stymie traditional methods.