INDEX

Explanations

change

The main thing this neuron does is detect the word “change” (in its exact form).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

nak

1.36

 ancol

1.33

cak

1.32

nj

1.30

ु

1.27

nym

1.21

nutrition

1.21

cous

1.21

cooked

1.20

Phu

1.20

POSITIVE LOGITS

tır

1.60

던

1.59

ти

1.54

ною

1.41

tı

1.37

하는

1.31

이를

1.30

alım

1.29

tir

1.28

tions

1.28

Activations Density 0.061%