Search papers, labs, and topics across Lattice.
1
0
3
2
A new 1B-parameter multilingual LLM beats larger models on key tasks, suggesting that focused training can trump scale for instruction following in European languages.