INDEX

Explanations

finalized, decided, set

The neuron strongly activates on past‐participle or adjectival forms ending in “-ed” or “-ized.”

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ถ้า

-1.61

彞

-1.54

 cuando

-1.52

liwości

-1.45

ért

-1.41

︠

-1.40

uldu

-1.36

Drupal

-1.35

 специальные

-1.34

 comunes

-1.34

POSITIVE LOGITS

 that

1.80

1.59

for

1.50

を獲得

1.45

後日

1.44

ꗃ

1.36

 However

1.36

 this

1.35

 gemens

1.35

as

1.30

Activations Density 0.044%