understood

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

 understood

-1.62

understood

-1.24

évaluateur

-1.13

ItemBackground

-1.13

Efq

-1.12

 Majefty

-1.09

StructEnd

-1.08

InjectAttribute

-1.07

Искәрмәләр

-1.01

 ſtate

-1.01

POSITIVE LOGITS

to

0.59

0.54

0.52

on

0.52

 recently

0.52

and

0.51

as

0.51

 five

0.50

in

0.48

Activations Density 0.063%