INDEX

Explanations

Russian words and phrases

The neuron selectively activates on Cyrillic word fragments—that is, tokens drawn from Russian words.

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

of

-2.23

ТЕЛЬ

-1.37

 سایر

-1.31

acy

-1.30

 تمامی

-1.25

 skóra

-1.24

ğine

-1.20

 sonrası

-1.18

 this

-1.18

 here

-1.16

POSITIVE LOGITS

 will

1.84

 also

1.65

 Tropfen

1.36

 weißem

1.34

これも

1.33

 sebagainya

1.30

眥

1.29

 färg

1.25

 then

1.23

 feinen

1.22

Activations Density 0.018%