INDEX

Explanations

followed by next

This neuron strongly responds to occurrences of the token “next” (in phrases like “next target,” “next sample,” “next results,” “next DIV,” “next Text,” “nextObj,” 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

spricht

-1.51

"":

-1.41

ymas

-1.39

erdings

-1.38

брь

-1.35

pne

-1.34

倻

-1.34

ѭ

-1.34

 asma

-1.33

bindet

-1.31

POSITIVE LOGITS

 next

1.38

 včetně

1.34

of

1.28

 Затем

1.26

into

1.26

 із

1.25

next

1.24

 feel

1.24

1.23

 Теперь

1.20

Activations Density 0.049%