Search papers, labs, and topics across Lattice.
4
0
9
6
A 32B model in Marathi performs like a 10B model in English, revealing how model size and reasoning can bridge language resource gaps in cross-lingual transfer.
Achieving state-of-the-art results for Danish with a model that uses only permissible data challenges the notion that larger datasets are necessary for competitive performance.
Text embeddings might be leading you astray: they're substantially less reliable than human experts in capturing semantic nuances critical for policy analysis and AI use case understanding.
Audio models can't do it all: a comprehensive benchmark reveals stark performance trade-offs between acoustic and linguistic understanding, with no single model dominating across diverse audio tasks.
