Advancing Polish Language Modeling through Tokenizer Optimization in the Bielik v3 7B and 11B Series

Open paper

Benchmark	Model	Rank	Results
common-sense-reasoning-on-winogrande	Bielik-11B-v3.0-Instruct	–	Accuracy: 77.19
sentence-completion-on-hellaswag	Bielik-11B-v3.0-Instruct	–	Accuracy: 81.96