INDEX

Explanations

relates to a method

This neuron is primarily detecting the token “to” (the common preposition “to” as it appears in phrases like “relates to,” “and to,” etc.).

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

 will

-1.65

and

-1.40

but

-1.37

so

-1.37

FOR

-1.30

 .....

-1.23

 Through

-1.22

 але

-1.20

 Getting

-1.19

 ....

-1.18

POSITIVE LOGITS

similar

1.36

nke

1.34

 heilige

1.34

kennen

1.30

probably

1.30

 parfums

1.27

 exprimer

1.27

 similar

1.27

giu

1.27

perhaps

1.26

Activations Density 0.456%