INDEX

Explanations

have verbs preceding objects

The neuron fires on verbs that describe performing or initiating actions (e.g. “do,” “run,” “remove,” “wait,” “display”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

it

-1.28

 condominio

-1.19

 Désolé

-1.19

++;

-1.18

 بوده

-1.17

</

-1.16

 дозво

-1.16

蹌

-1.15

 tanken

-1.15

これも

-1.13

POSITIVE LOGITS

my

1.78

 then

1.66

be

1.44

抔

1.38

to

1.36

 instead

1.34

 основных

1.33

 inmediatamente

1.31

来看看

1.30

我去

1.30

Activations Density 0.036%