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