Search papers, labs, and topics across Lattice.
2
0
4
2
Polish language models get a major efficiency boost: Bielik v3 slashes tokenization overhead with a custom vocabulary, outperforming generic tokenizers.
Achieve a 50% inference speedup on a large language model for European languages by compressing it to 7.35B parameters, while retaining 90% of the original 11B parameter model's performance.
