coverage

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 uninsured

-0.12

 briefly

-0.11

 purely

-0.10

 glaring

-0.10

 neither

-0.10

 deliberately

-0.10

oki

-0.09

 perish

-0.09

sav

-0.09

POSITIVE LOGITS

 cover

0.19

 covering

0.18

 adequately

0.16

 coverage

0.15

 Cover

0.15

covering

0.15

-cover

0.14

 covers

0.14

coverage

0.13

cover

0.12

Activations Density 0.025%

No Known Activations