The performance of my installed repository falls significantly short of the results reported in the paper.

[MilkDargon]

test_DE-EN-Twain_Mark-Tom_Sawyer.jsonl RESULTS: {'recall_strict': 0.793027360988526, 'recall_lax': 0.9324801412180053, 'precision_strict': 0.8157537347215935, 'precision_lax': 0.9334540516070621, 'f1_strict': 0.8042300263072767, 'f1_lax': 0.9329668422499139}
The results of the paper report are 'f1_strict': 0.956