Reading Notes | Exploring and Predicting Transferability across NLP Tasks

[Semantic Scholar] – [Code] – [Tweet] – [Video] – [Website] – [Slide]

Change Logs:

  • 2023-09-16: First draft. This paper appears at ACL 2020.
  • Data selection strategy for best transfer learning performance.

Reference

  1. [1811.01088] Sentence Encoders on STILTs: Supplementary Training on Intermediate Labeled-data Tasks (Phang et al)
  2. Identifying beneficial task relations for multi-task learning in deep neural networks (Bingel & Søgaard, EACL 2017)