1 paper · 1 filter
Chenghao Xiao, Zhuoxu Huang, Danlu Chen +7
Pretrained language models are long known to be subpar in capturing sentence and document-level semantics. Though heavily investigated, transferring perturbation-based methods from…