Wikipedia
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Wikipedia 500 soft-deleted vectors
20.73ROUGE-L
4
Wikipedia English (test)
640Bits per Token (BPT)
4
Wikipedia
43.7Recall@1
4
Wikipedia
1.59ABPC
4
Wikipedia Cyrillic-rich subsets (test)
2.38Russian (ru)
4
Wikipedia
83.21F1-Macro (BERT)
4
Wikipedia
52.6R@1
4
Wikipedia
1,170TPR @ 1% FPR
4
Wikipedia auto (test)
81.7SARI
4
Wikipedia Politics domain articles (test)
4.357Out-of-sample Predictive Error
3
Wikipedia Computing domain articles (test)
5.063Out-of-sample Predictive Error
3
Wikipedia Math domain articles (test)
4.401Out-of-sample Predictive Error
3
Wikipedia Finance domain articles (test)
4.674Out-of-Sample Predictive Error
3
Wikipedia n = 500
20.7ROUGE-L
3
Wikipedia -> Code -> Chat
82.1Accuracy
3
Wikipedia Welsh (test)
22.1Perplexity (PPL)
3
Wikipedia Yoruba (test)
7.7Perplexity (PPL)
3
Wikipedia Tamil (test)
3Perplexity (PPL)
3
Wikipedia in-distribution (held-out)
0.41R1
3
Wikipedia (held-out set)
0.72NMI
3
Wikipedia TW5 2010-2015
39F1 Score
3
Wikipedia TW4 2010-2015
37F1 Score
3
Wikipedia 2010-2015 (TW3)
46F1
3
Wikipedia TW2 2010-2015
58F1 Score
3
Wikipedia 2010-2015 (TW1)
58F1 Score
3