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

Language Modeling on WikiText-103 50w x 2048 (test)

7.59Perplexity

fp16

-1,008.54645,850.374312,709.29519,568.2157May 26, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.05
7.59
2026.05
8.83
2026.05
9.03
2026.05
9.69
2026.05
98.4
2026.05
109
2026.05
197
2026.05
10,668
2026.05
12,738
2026.05
13,350
2026.05
17,661
2026.05
25,411