INDEX

Explanations

facial expressions and features

The neuron tracks how far into the text you are, gradually increasing its activation for tokens appearing later in a segment.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ぃ

-0.87

leyen

-0.82

wea

-0.81

IPH

-0.80

Commissioner

-0.79

 offres

-0.79

 Gambar

-0.78

捞

-0.77

toarele

-0.77

iseaux

-0.77

POSITIVE LOGITS

 eyes

1.39

 smile

1.26

 facial

1.16

 mouth

1.09

 smiling

1.05

 ears

1.03

 expression

1.02

 eyebrows

0.96

 nose

0.93

 Smile

0.91

Activations Density 0.022%