Wikipedia
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Wikipedia Japanese Total (test)
76Vocabulary Coverage
6
Wikipedia English Easy (test)
66.9Vocabulary Coverage
6
Wikipedia English Total (test)
81.6Vocabulary Coverage
6
Wikipedia GPT-3.5 davinci-003 generations +60L 600 attack
49.5Accuracy (at 1% FPR)
6
Wikipedia
-0.122Best FID+
6
wikipedia
0.05CRPS Sum
6
Wikipedia GPT-3.5 davinci-003 generations 60L
65.9Detection Accuracy (1% FPR)
6
Wikipedia GPT-3.5 davinci-003 generations (Original)
1Accuracy (1% FPR)
6
Wikipedia (50% test)
0.996AUC
6
Wikipedia
3.62Simplicity
6
Wikipedia events 2024-2025 (test)
3.78Standing Q_t^bg
5
Wikipedia
159.1Tail-Patch Absolute Score (k=1)
5
Wikipedia
2.47AUPRC (%)
5
Mini-Wikipedia RAG-Thief attack
49.6CRR
5
Mini-Wikipedia (Pirates attack)
78.4CRR
5
Wikipedia (45 country-specific domains, 89 queries) (test)
0.9Comprehensiveness
5
Wikipedia in-domain, in-KB (test)
80.05Accuracy (All)
5
Wikipedia standard (test)
87.3AUC
5
Wikipedia in-house
86.5Precision
5
WIKIPEDIA
26.74Error Rate
5
WIKIPEDIA triplets (test)
0.01Runtime (min)
5
Wikipedia IR (test)
71.8Recall@1 (Top 10001)
5
Wikipedia ONLY-B sentences split
4.05Average Score
5
Wikipedia (ONLY-A sentences split only by UNSUP)
2.42Average Score
5
Wikipedia ALL-B (sentences split)
3.85Average Score
5