INDEX

Explanations

like as if

The neuron strongly activates on content words inside comparative (“as if”/“as though”) clauses—i.e. the words that follow those simile/likening phrases.

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

of

-1.68

and

-1.64

 when

-1.39

including

-1.26

both

-1.24

Когда

-1.24

ibia

-1.22

 quando

-1.22

while

-1.21

 probabilmente

-1.19

POSITIVE LOGITS

 this

1.73

all

1.63

 only

1.55

 besonders

1.48

 other

1.43

 just

1.27

at

1.26

 turistas

1.25

 welches

1.24

Apothe

1.23

Activations Density 0.028%