INDEX

Explanations

words describing concrete things

This neuron is triggered by short, content-word phrases expressing a relationship (e.g. “center of web,” “dog chasing a car,” “paints their own favorite picture,” “run alongside the speeding vehicle”).

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

 również

-1.23

 MULTI

-1.23

despite

-1.23

originally

-1.16

 pere

-1.13

-1.10

顔が

-1.09

Azi

-1.09

↵↵

-1.09

 نیز

-1.09

POSITIVE LOGITS

not

1.64

 only

1.42

Other

1.36

Ингредиенты

1.35

ร้าน

1.33

 other

1.29

six

1.29

ׅ

1.28

 superbly

1.25

 mildly

1.23

Activations Density 0.167%