List your company
ResearchFormerly WizWand

Lifelong Language Learning on Five Tasks (SQuAD, WikiSQL, SST, QA-SRL, WOZ) GPT-2 Base (test)

76.6Average ScoreMultitasked
48.41655.73363.0570.367Sep 7, 2019
Updated 8mo ago

Evaluation results

MethodLinks
MultitaskedDescription=All tasks are trained simultaneously2019.0976.6
LAMOLSampling ratio (gamma)=0.2, Replay strategy=REAL2019.0976
LAMOLSampling ratio (gamma)=0.05, Replay strategy=REAL2019.0974.5
LAMOLSampling ratio (gamma)=0.1, Replay strategy=TASK2019.0974.3
LAMOLSampling ratio (gamma)=0.1, Replay strategy=GEN2019.0973.1
LAMOLSampling ratio (gamma)=0.05, Replay strategy=TASK2019.0971.5
LAMOLSampling ratio (gamma)=0.05, Replay strategy=GEN2019.0969.6
Fine-tunedDescription=Directly fine-tuned on the stream of tasks2019.0951.5
MASMethod Type=Regularization-based2019.0949.5