output
20132021
most citedBERT and PALs: Projected Attention Layers for Efficient Adaptation in Multi-Task Learning

113 citations

Showing cs.LGShow all

3 papers · 1 filter