INDEX

Explanations

friend and friends

The main thing this neuron does is detect mentions of “friend” (and its variations like “friends” or “friendships”).

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

 were

-1.81

 étaient

-1.54

You

-1.51

ですので

-1.43

abbix

-1.41



-1.38

ですから

-1.37

 так

-1.36

 fréquemment

-1.36

＄

-1.31

POSITIVE LOGITS

and

2.16

1.56

1.45

 from

1.44

的時候

1.42

גרפיה

1.42

 when

1.41

uk

1.41

琑

1.39

ów

1.38

Activations Density 0.022%