INDEX

Explanations

Ny followed by n, a, or an

The neuron consistently activates on the character sequence “Ny” within tokens (i.e. it spots the bigram “Ny” in names and terms).

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

と

-2.67

を

-2.64

न

-2.55

襌

-2.45

usado

-2.42

 informée

-2.42

 antworte

-2.41

 ersche

-2.38

-2.30

ۥ

-2.30

POSITIVE LOGITS

ls

2.98

It

2.89

 boken

2.58

ization

2.50

᩠

2.41

 необходи

2.39

.’

2.31

).

2.27

</h4>

2.25

 whom

2.23

Activations Density 0.007%