INDEX

Explanations

French, German, Russian, Welsh, Japanese, Iranian

This neuron activates on non-English word fragments—i.e. tokens that are part of foreign‐language 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

是個

-1.68

 inond

-1.63

Andere

-1.52

𖧵

-1.51

を与える

-1.46

ほんと

-1.44

少しずつ

-1.44

 наших

-1.44

 verhind

-1.43

 humeur

-1.39

POSITIVE LOGITS

</em>

2.03

 about

1.67

StoryboardSegue

1.59

ấ

1.56

1.41

for

1.40

You

1.34

 less

1.34

死者

1.34

how

1.32

Activations Density 0.120%