pdf bibBenchmarking down-scaled (not so large) pre-trained language modelsMatthias Aßenmacher | Patrick Schulze | Christian HeumannProceedings of the 17th Conference on Natural Language Processing (KONVENS 2021)