OpenWebText
Benchmarks
TaskDatasetSOTA resultTrendResults
OpenWebText 10k
11.502Perplexity
3
OpenWebText GPT-2 (train)
3.41Loss
3
openwebtext
3.09Perplexity (PPL)
3
OpenWebText Quality
88.2Balanced Accuracy
3
OpenWebText Mainstream
92.4Balanced Accuracy
3
OpenWebText Politics
95.6Balanced Accuracy
3
OpenWebText AI subset
95.7Balanced Acc
3
OpenWebText Climate
96.7Balanced Acc
3
OpenWebText (OWT)
5.7Tie (%)
2
OpenWebText (OWT) (test)
0.85MAUVE
2