Search papers, labs, and topics across Lattice.
Johns Hopkins University
1
0
2
4
Forget scaling laws: a specialized 8B parameter translation model can outperform a 70B general-purpose LLM on 1,600 languages.
