INDEX

Explanations

numbers followed by punctuation

The neuron activates on numeric tokens (e.g. reference numerals, figure numbers, part labels) in technical or patent‐style text.

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

 élar

-1.94

 Chaque

-1.91

 situaciones

-1.73

 chacune

-1.67

ほうがいい

-1.66

-1.64

頃は

-1.63

 rassemb

-1.63

 coû

-1.61

 véhic

-1.61

POSITIVE LOGITS

in

2.56

or

2.13

and

1.88

on

1.87

for

1.84

 which

1.61

 that

1.52

}$,

1.52

to

1.45

\&

1.41

Activations Density 0.144%