Multi-armed bandits for resource efficient, online optimization of language model pre-training: the use case of dynamic masking.

Iigo Urteaga, Moulay-Zadane Dradia, Tomer Lancewicki, Shahram Khadivi

Browse the full ACL paper archive.