Search papers, labs, and topics across Lattice.
Engineering Group
1
0
3
This Italian LLM punches way above its weight, matching the performance of models trained on 6-10x more data while using only 3B active parameters during inference.
