INDEX

Explanations

one versus the other

This neuron fires strongly on words that refer to alternatives or comparisons—particularly “other” (and similar relational tokens like “next”).

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

 ſame

-1.37

 utilisent

-1.34

cæ

-1.31

mische

-1.30

 ordre

-1.30

髫

-1.28

 acheter

-1.27

UNDE

-1.24

DOOR

-1.23

'_

-1.23

POSITIVE LOGITS

 with

2.17

 when

2.11

but

1.98

to

1.81

 without

1.75

one

1.72

 which

1.66

 from

1.58

 that

1.56

 during

1.52

Activations Density 0.024%