Training Production Language Models without Memorizing User Data

Ramaswamy, Swaroop, Thakkar, Om, Mathews, Rajiv · arXiv · 2020