INDEX

Explanations

Japanese names and company names

The neuron flags Japanese‐language terms and morpheme labels (e.g. “和,” “熟語,” or romanized pieces like “wago”/“kango”) in an otherwise English text.

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

 continúas

-2.80

इटम

-2.61



-2.58

 añadido

-2.56

 asegurado

-2.52

 också

-2.42

AMIENTO

-2.36

 señalado

-2.33

archiwizowane

-2.31

 trabajado

-2.31

POSITIVE LOGITS

</sup>

2.80

2.67

for

2.64

2.59

2.38

but

2.38

на

2.33

2.30

">(</

2.23

Activations Density 0.002%