Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

PG19

Benchmarks

Task NameDataset NameSOTA ResultTrend
Long-context Input (Summarization)PG19
TPT6.12
20
Language ModelingPG19 tokens (test)
Bits per Token3.49
14
Language ModelingPG19 bytes (test)
Bits Per Token0.935
14
Language ModelingPG19 T5 (val)
PPLX15.31
10
Lossless text compressionPG19
Compression Ratio (bits)6.77
5
Showing 5 of 5 rows