BiSECT

The BiSECT dataset is a benchmark for the Split and Rephrase task, consisting of bitexts.

BibTex: