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

FAKEWIKI

Benchmarks

Task NameDataset NameSOTA ResultTrend
Evidence RetrievalFAKEWIKI Indirect
Recall@1021.6
26
Evidence RetrievalFAKEWIKI NoiseInjection
Recall@1063.5
26
Evidence RetrievalFAKEWIKI RolePlay
Recall@1063.8
26
Evidence RetrievalFAKEWIKI Obfuscate
Recall@1059.5
26
Evidence RetrievalFAKEWIKI Clean
Recall@1078.1
26
Training Data AttributionFAKEWIKI Indirect
Recall@1017.7
3
Training Data AttributionFAKEWIKI NoiseInjection
Recall@1059.2
3
Training Data AttributionFAKEWIKI RolePlay
Recall@1062.5
3
Training Data AttributionFAKEWIKI Obfuscate
Recall@1044.4
3
Training Data AttributionFAKEWIKI Clean
Recall@1077.2
3
Showing 10 of 10 rows