INDEX

Explanations

drawing attention

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

çŃı

-0.09

unt

-0.09

åĢĻ

-0.09

tod

-0.08

æĹħ

-0.08

Prem

-0.08

 Clair

-0.08

 Prem

-0.08

éĿłè°±

-0.08

 tastes

-0.08

POSITIVE LOGITS

çªģåĩº

0.16

 attention

0.13

 Ð²Ð½Ð¸Ð¼Ð°Ð½Ð¸Ðµ

0.13

çŀ©çĽ®

0.12

çĿĽ

0.12

æ³¨æĦıåĬĽ

0.12

äº®çĤ¹

0.12

attention

0.11

æıĲçĤ¼

0.11

åĩ¸æĺ¾

0.11

Activations Density 0.066%