Never Train from Scratch: Fair Comparison of Long-Sequence Models Requires Data-Driven Priors (Extended Abstract).

Ido Amos, Jonathan Berant, Ankit Gupta

Browse the full IJCAI paper archive.