Wikipedia
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Wikipedia 30% ratio (train)
82.58Micro-F1
8
Wikipedia (20% train ratio)
82.08Micro-F1
8
Wikipedia (10% train ratio)
81.04Micro-F1
8
Wikipedia
0.744AUROC
7
Wikipedia GPT2-XL generations, +60L,600 attack
98Accuracy (1% FPR)
7
Wikipedia GPT2-XL generations +60L attack
99Accuracy (at 1% FPR)
7
Wikipedia OPT-13B generations (Original)
100Accuracy (1% FPR)
7
Wikipedia GPT2-XL generations (Original)
100Accuracy (1% FPR)
7
Wikipedia languagemodel_wiki_noref_v128k_11k
6.01Subword Perplexity
7
WIKIPEDIA (full)
73.79Semantic Accuracy
7
Wikipedia French (test)
24.25Lexical Divergence
7
Wikipedia Chinese (test)
25.91Lexical Divergence
7
Wikipedia English (test)
23.07Lexical Divergence
7
Wikipedia English (BPE 16k, context 256) (val)
16.69Validation Perplexity (PPL)
6
Wikipedia Politics (test)
19.04Average F1
6
Wikipedia Computing (test)
25.27Avg F1 Score
6
Wikipedia Math (test)
26.6Average F1 Score
6
Wikipedia Finance (test)
23.2Average F1
6
wikipedia
96.4AUC
6
Wikipedia
52.6Familiarity TPR
6
Wikipedia Chinese Easy (test)
66.1Vocabulary Coverage
6
Wikipedia Chinese Total (test)
80.2Vocabulary Coverage
6
Wikipedia Korean Easy (test)
52.9Vocabulary Coverage
6
Wikipedia Korean Total (test)
70.4Vocabulary Coverage
6
Wikipedia Japanese Easy (test)
60.4Vocabulary Coverage
6