allenai/specter

SPECTER is pre-trained on a powerful signal of document-level relatedness: the citation graph . Unlike existing pretrained language models, SPECTER can be easily applied to downstream applications without task-specific fine-tuning . The model can be used to generate document level embedding of documents and can be applied to other applications without fine-tuning . The paper is published by Arman Cohan, Sergey Feldman, Iz Beltagy, Doug Downey, Daniel S. Weld and Daniel C. Weld at the Open University of New York University, New York City, NY, NYE, NJE, NYNU, NJNU and NYE University of Cambridge University, NY,