List your company
ResearchFormerly WizWand

Continual Learning on decaNLP (WOZ, QA-SRL, SST, WikiSQL, SQuAD reversed order) (test)

78.2Average ScoreLAMOL
43.3652.40561.4570.495Sep 7, 2019
Updated 8mo ago

Evaluation results

MethodLinks
LAMOLsampling_strategy=REAL, gamma=0.22019.0978.2
LAMOLsampling_strategy=TASK, gamma=0.22019.0976.9
Multitaskedtraining_mode=Simultaneous training2019.0976.6
LAMOLsampling_strategy=REAL, gamma=0.052019.0975.9
LAMOLsampling_strategy=TASK, gamma=0.052019.0975.3
LAMOLsampling_strategy=GEN, gamma=0.22019.0973
LAMOLsampling_strategy=GEN, gamma=0.052019.0963.2
Fine-tunedtraining_mode=Sequential fine-tuning2019.0945.4
MAStraining_mode=Regularization-based2019.0944.7