Search papers, labs, and topics across Lattice.
8
0
6
Automated benchmarks can now evaluate MLLMs' music perception skills across diverse modalities, ensuring more reliable assessments than ever before.
LLMs trained on a synthetic corpus can outperform native data benchmarks while using significantly fewer tokens, challenging the assumption that more data always leads to better performance.
A unified encoding for diastematic Gregorian notation recognition leads to a groundbreaking model that outperforms existing methods across multiple datasets.
Discovering 20-30,000 hidden musical documents in vast collections could revolutionize our understanding of musical heritage.
A nearly 30-year effort culminates in a richly annotated Czech language resource that revolutionizes how we analyze and compare NLP tools.
The Prague Dependency Treebank-Consolidated doubles the size and genre diversity of its predecessor, revealing critical differences in dependency structures that challenge the notion of universality in linguistic annotations.
MorfFlex reduces the size of morphological dictionaries by encoding complex patterns, making it a game-changer for NLP in morphologically rich languages.
Synthetic data can dramatically enhance OMR performance on diverse, real-world manuscripts, reducing the need for expensive in-domain annotations.
