INDEX

Explanations

Focusing on effectiveness

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 одно

1.00

 одном

0.99

 ஏழு

0.95

并在

0.93

一丝

0.93

 одну

0.92

 одного

0.92

 podendo

0.91

 известно

0.90

Continu

0.89

POSITIVE LOGITS

 বরং

1.36

むしろ

1.35

 justru

1.32

 오히려

1.31

 instead

1.28

 greatest

1.26

 malah

1.19

の方が

1.17

 inkább

1.17

 Rather

1.16

Activations Density 0.495%