FR: Folded Rationalization with a Unified Encoder Wei Liu 1 Haozhao Wang 1 Jun Wang 2 Ruixuan Li

There are growing concerns over the interpretability of NLP models, especially when language models are being rapidly applied on various critical fields (Lipton, 2016; Du et al., 2019; Xiang et al., 2019; Miller, 2019; Sun et al., 2021).