Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

PG

Benchmarks

Task NameDataset NameSOTA ResultTrend
Language ModelingPG-19
Perplexity5
160
Layout detection30-page PG (test)
Precision96.9
8
Long-Context GenerationPG-19 20K context length
Throughput Speedup (micro)4
6
Long-Context GenerationPG-19 10K context length
Throughput Speedup (micro-averaged)2.92
6
Text recognition30-page PG (test)
CER (%)1.05
4
Reading order30-page PG (test)
Precision98
1
Line detection30-page PG (test)
Precision98.3
1
Showing 7 of 7 rows