@misc{indiciae1e4df1d89ae5,
  title = {Straight to the Gradient: Learning to Use Novel Tokens for Neural Text Generation},
  author = {Xiang Lin and Simeng Han and Shafiq Joty},
  year = {2021},
  url = {https://arxiv.org/abs/2106.07207},
  note = {Source identifier: 2106.07207}
}