Never Train from Scratch: Fair Comparison of Long-Sequence Models Requires Data-Driven Priors

Ido Amos, Jonathan Berant, Ankit Gupta

BibTeX
@inproceedings{amos2024never,
  author = {Ido Amos and Jonathan Berant and Ankit Gupta},
  title = {{Never Train from Scratch: Fair Comparison of Long-Sequence Models Requires Data-Driven Priors}},
  booktitle = {International Conference on Learning Representations},
  year = {2024},
  url = {https://openreview.net/forum?id=PdaPky8MUn},
}
Figure