INDEX

Explanations

many of X

The neuron fires on standalone numeric tokens (e.g. “680,” “672,” “664,” etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 וכן

-1.51

hever

-1.44

 with

-1.40

 mengumpulkan

-1.40

 various

-1.39

tya

-1.39

 mengenakan

-1.35

zit

-1.34

)=>{

-1.34

its

-1.34

POSITIVE LOGITS

are

1.77

 have

1.37

1.32

 were

1.32

of

1.27

 actuels

1.27

が

1.27

人は

1.26

 Wiking

1.20

Activations Density 0.025%