@misc{indiciae7bd58ddbd16d,
  title = {Positional Artefacts Propagate Through Masked Language Model Embeddings},
  author = {Ziyang Luo and Artur Kulmizev and Xiaoxi Mao},
  year = {2021},
  url = {https://arxiv.org/abs/2011.04393},
  note = {Source identifier: 2011.04393}
}