INDEX

Explanations

to

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

æĸ¹åĲĳ

-0.63

 direction

-0.63

direction

-0.55

 directions

-0.52

Direction

-0.51

-direction

-0.50

çļĦæĸ¹åĲĳ

-0.49

 Direction

-0.48

	direction

-0.48

_DIRECTION

-0.46

POSITIVE LOGITS

ä¸Ģè§Ī

0.29

åķĨ

0.27

é©¹

0.27

çļĦåľŁåľ°

0.26

_acc

0.25

é¹ĺ

0.25

unce

0.25

 reconc

0.25

estr

0.24

Ã§e

0.24

Activations Density 0.906%