INDEX

Explanations

label prefix

The neuron strongly activates on numeric tokens (numbers and digits).

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

 çünkü

-1.26

 pauvreté

-1.22

哪怕

-1.20

或者是

-1.15

 anyways

-1.13

 simpel

-1.09

但却

-1.07

或是

-1.07

}\|

-1.06

$>

-1.05

POSITIVE LOGITS

 geographical

1.45

 geographic

1.38

 female

1.34

地说

1.28

colare

1.28

 soñar

1.27

 school

1.24

用

1.24

حاد

1.23

 немец

1.23

Activations Density 0.004%