INDEX

Explanations

easy, hard, no surprise, wonder, important, fair

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

nad

-0.09

Mey

-0.09

aby

-0.08

ibri

-0.08

 Curl

-0.08

 fours

-0.08

Teh

-0.08

mie

-0.08

-checkbox

-0.07

moz

-0.07

POSITIVE LOGITS

 wonder

0.42

 unsur

0.33

 Wonder

0.28

no

0.26

 understandably

0.23

Wonder

0.23

 understandable

0.22

 wondered

0.21

 wonders

0.21

 Wonderland

0.20

Activations Density 0.108%

No Known Activations