INDEX

Explanations

describes quantity or state

The neuron fires on adjective + noun collocations (e.g. “common questions,” “close contact,” “high precision”).

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

in

-2.11

 when

-2.08

 which

-2.00

who

-1.86

if

-1.82

of

-1.75

 whether

-1.53

 therefore

-1.53

 with

-1.48

so

-1.46

POSITIVE LOGITS

 autre

1.70

他の

1.48

༔

1.39

 atractiva

1.36

Debido

1.35

Due

1.29

iczna

1.26

其他

1.26

 cómoda

1.26

Trying

1.24

Activations Density 0.151%