List your company
ResearchFormerly WizWand

Question Generation on SQuAD (BLEU-4)

0.496BLEU-4M
0.138240.231120.3240.41688Sep 22, 2022
Updated 8mo ago

Evaluation results

MethodLinks
Mmode=upperbound2022.090.496
bi-gram + APS + round-triptype=ensemble2022.090.401
bi-gram + round-triptype=ensemble2022.090.4
tri-gram + APS + round-triptype=ensemble2022.090.4
tri-gram + round-triptype=ensemble2022.090.398
APS + round-triptype=ensemble2022.090.397
round-tripselection_criteria=QA semantic equivalence2022.090.392
bi-gram + APStype=ensemble2022.090.384
tri-gram + APStype=ensemble2022.090.383
bi-gramselection_criteria=n-gram similarity, n=22022.090.382
tri-gramselection_criteria=n-gram similarity, n=32022.090.38
averaged prompt score (APS)selection_criteria=Prompt-based Score2022.090.38
overall prompt score (OPS)selection_criteria=Prompt-based Score2022.090.373
Mgmode=greedy2022.090.372
Msmode=sample avg2022.090.359
ERNIE-GEN Largetraining=fine-tuned, author=Xiao et al., 20212022.090.254
UniLM v2 Basetraining=fine-tuned, author=Bao et al., 20202022.090.244
UniLM Largetraining=fine-tuned, author=Bao et al., 20202022.090.228
Msmode=lowerbound2022.090.225
Zhang and Bansaltraining=fine-tuned2022.090.184
Du and Cardietraining=fine-tuned2022.090.152