However previous work has not systematically evaluated the efficacy of different pretraining language distributions across model scales. |
However previous work has not systematically evaluated the efficacy of different pretraining language distributions across model scales. |