INDEX

Explanations

The neuron flags linking‐verb constructions (forms of “to be” and “become”) used in abstract or existential statements.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 olmayan

0.63

 ikke

0.59

 zowel

0.59

 Despite

0.58

 thus

0.57

 así

0.57

 centenary

0.57

 クリスマス

0.57

 विरो

0.56

ristmas

0.55

POSITIVE LOGITS

 гораздо

1.71

 훨씬

1.68

 намного

1.39

むしろ

1.37

 justru

1.35

 MUCH

1.30

раздо

1.27

훨

1.27

 বরং

1.23

 invece

1.20

Activations Density 0.511%