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

Language Modeling on WikiText and LAMBADA

7.3WikiText Perplexity

Llama3 8B

-56.408373.621803.651,233.679Sep 26, 2025Nov 2, 2025Dec 9, 2025Jan 15, 2026Feb 21, 2026Mar 30, 2026May 7, 2026
Updated 26d ago

Evaluation Results

MethodLinks
2025.09
7.33.1
2025.09
113.7
2026.02
124.6
2025.09
124.6
2026.02
125.7
2026.02
154.7
2025.09
163.4
2026.02
184.9
2026.02
184.4
2025.09
184.9
2025.09
184.1
2026.05
18.0314.87
2026.05
18.3715.17
2026.05
18.5116.12
2026.05
18.9917.9
2026.05
19.1418.2
2026.05
19.2416.83
2025.09
204.3
2025.09
204.2
2026.02
216.4
2025.09
216.4
2026.02
2114
2026.02
236.3
2025.09
236.3
2025.09
236.5
2026.02
245.9
2026.02
2521
2026.02
2711
2025.09
2711
2026.05
31.5141.62
2026.05
31.7346.19
2026.05
31.9643.44
2026.05
32.145.63
2026.05
32.854.36
2025.09
338.7
2026.05
33.4560.42
2026.02
3524
2025.09
3518
2026.02
3615
2025.09
3615
2026.02
3632
2025.09
4111
2026.02
4336
2025.09
4336
2025.09
459.2
2026.02
5584
2026.02
5941
2026.02
6435
2026.02
7688
2026.02
130190
2025.09
15061
2026.02
170170
2025.09
180120
2026.02
2802,200
2026.02
290660
2025.09
5501,300
2026.02
5902,500
2026.02
8009,200
2026.02
1,60033,000