INDEX

Explanations

starts with Pe

The neuron responds to the letter‐pair “Pe,” i.e. it activates on tokens beginning with “Pe.”

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

↵↵

-4.34

-3.30

as

-3.08

-3.06

 been

-3.05

to

-2.77

ang

-2.73

with

-2.67

we

-2.66

eds

-2.58

POSITIVE LOGITS

3.41

 dwindling

2.75

 infamous

2.64

基本的

2.55

 tattered

2.53

Ꭳ

2.52

 навіть

2.47

いかがでしょうか

2.39

店内は

2.38

dáms

2.34

Activations Density 0.003%