Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

ConflictBank

Benchmarks

Task NameDataset NameSOTA ResultTrend
Question AnsweringConflictBank
Accuracy80.2
44
Post-hoc CorrectionConflictBank (100 atomic claims)
ROUGE89
15
Question AnsweringConflictBank
EM55
12
Question AnsweringConflictBank (Held-out)
Accuracy64.3
8
Showing 4 of 4 rows